Show More
@@ -1,480 +1,480 b'' | |||
|
1 | 1 | # debugcommands.py - debug logic for remotefilelog |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2013 Facebook, Inc. |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | import os |
|
9 | 9 | import zlib |
|
10 | 10 | |
|
11 | 11 | from mercurial.node import ( |
|
12 | 12 | bin, |
|
13 | 13 | hex, |
|
14 | 14 | sha1nodeconstants, |
|
15 | 15 | short, |
|
16 | 16 | ) |
|
17 | 17 | from mercurial.i18n import _ |
|
18 | 18 | from mercurial.pycompat import open |
|
19 | 19 | from mercurial import ( |
|
20 | 20 | error, |
|
21 | 21 | filelog, |
|
22 | 22 | lock as lockmod, |
|
23 | 23 | pycompat, |
|
24 | 24 | revlog, |
|
25 | 25 | ) |
|
26 | 26 | from mercurial.utils import hashutil |
|
27 | 27 | from . import ( |
|
28 | 28 | constants, |
|
29 | 29 | datapack, |
|
30 | 30 | fileserverclient, |
|
31 | 31 | historypack, |
|
32 | 32 | repack, |
|
33 | 33 | shallowutil, |
|
34 | 34 | ) |
|
35 | 35 | |
|
36 | 36 | |
|
37 | 37 | def debugremotefilelog(ui, path, **opts) -> None: |
|
38 | 38 | decompress = opts.get('decompress') |
|
39 | 39 | |
|
40 | 40 | size, firstnode, mapping = parsefileblob(path, decompress) |
|
41 | 41 | |
|
42 | 42 | ui.status(_(b"size: %d bytes\n") % size) |
|
43 | 43 | ui.status(_(b"path: %s \n") % path) |
|
44 | 44 | ui.status(_(b"key: %s \n") % (short(firstnode))) |
|
45 | 45 | ui.status(_(b"\n")) |
|
46 | 46 | ui.status( |
|
47 | 47 | _(b"%12s => %12s %13s %13s %12s\n") |
|
48 | 48 | % (b"node", b"p1", b"p2", b"linknode", b"copyfrom") |
|
49 | 49 | ) |
|
50 | 50 | |
|
51 | 51 | queue = [firstnode] |
|
52 | 52 | while queue: |
|
53 | 53 | node = queue.pop(0) |
|
54 | 54 | p1, p2, linknode, copyfrom = mapping[node] |
|
55 | 55 | ui.status( |
|
56 | 56 | _(b"%s => %s %s %s %s\n") |
|
57 | 57 | % (short(node), short(p1), short(p2), short(linknode), copyfrom) |
|
58 | 58 | ) |
|
59 | 59 | if p1 != sha1nodeconstants.nullid: |
|
60 | 60 | queue.append(p1) |
|
61 | 61 | if p2 != sha1nodeconstants.nullid: |
|
62 | 62 | queue.append(p2) |
|
63 | 63 | |
|
64 | 64 | |
|
65 | def buildtemprevlog(repo, file): | |
|
65 | def buildtemprevlog(repo, file) -> filelog.FileLog: | |
|
66 | 66 | # get filename key |
|
67 | 67 | filekey = hex(hashutil.sha1(file).digest()) |
|
68 | 68 | filedir = os.path.join(repo.path, b'store/data', filekey) |
|
69 | 69 | |
|
70 | 70 | # sort all entries based on linkrev |
|
71 | 71 | fctxs = [] |
|
72 | 72 | for filenode in os.listdir(filedir): |
|
73 | 73 | if b'_old' not in filenode: |
|
74 | 74 | fctxs.append(repo.filectx(file, fileid=bin(filenode))) |
|
75 | 75 | |
|
76 | 76 | fctxs = sorted(fctxs, key=lambda x: x.linkrev()) |
|
77 | 77 | |
|
78 | 78 | # add to revlog |
|
79 | 79 | temppath = repo.sjoin(b'data/temprevlog.i') |
|
80 | 80 | if os.path.exists(temppath): |
|
81 | 81 | os.remove(temppath) |
|
82 | 82 | r = filelog.filelog(repo.svfs, b'temprevlog') |
|
83 | 83 | |
|
84 | 84 | class faket: |
|
85 | 85 | def add(self, a, b, c): |
|
86 | 86 | pass |
|
87 | 87 | |
|
88 | 88 | t = faket() |
|
89 | 89 | for fctx in fctxs: |
|
90 | 90 | if fctx.node() not in repo: |
|
91 | 91 | continue |
|
92 | 92 | |
|
93 | 93 | p = fctx.filelog().parents(fctx.filenode()) |
|
94 | 94 | meta = {} |
|
95 | 95 | if fctx.renamed(): |
|
96 | 96 | meta[b'copy'] = fctx.renamed()[0] |
|
97 | 97 | meta[b'copyrev'] = hex(fctx.renamed()[1]) |
|
98 | 98 | |
|
99 | 99 | r.add(fctx.data(), meta, t, fctx.linkrev(), p[0], p[1]) |
|
100 | 100 | |
|
101 | 101 | return r |
|
102 | 102 | |
|
103 | 103 | |
|
104 | 104 | def debugindex(orig, ui, repo, file_=None, **opts): |
|
105 | 105 | """dump the contents of an index file""" |
|
106 | 106 | if ( |
|
107 | 107 | opts.get('changelog') |
|
108 | 108 | or opts.get('manifest') |
|
109 | 109 | or opts.get('dir') |
|
110 | 110 | or not shallowutil.isenabled(repo) |
|
111 | 111 | or not repo.shallowmatch(file_) |
|
112 | 112 | ): |
|
113 | 113 | return orig(ui, repo, file_, **opts) |
|
114 | 114 | |
|
115 | 115 | r = buildtemprevlog(repo, file_) |
|
116 | 116 | |
|
117 | 117 | # debugindex like normal |
|
118 | 118 | format = opts.get('format', 0) |
|
119 | 119 | if format not in (0, 1): |
|
120 | 120 | raise error.Abort(_(b"unknown format %d") % format) |
|
121 | 121 | |
|
122 | 122 | generaldelta = r.get_revlog()._format_flags & revlog.FLAG_GENERALDELTA |
|
123 | 123 | if generaldelta: |
|
124 | 124 | basehdr = b' delta' |
|
125 | 125 | else: |
|
126 | 126 | basehdr = b' base' |
|
127 | 127 | |
|
128 | 128 | if format == 0: |
|
129 | 129 | ui.write( |
|
130 | 130 | ( |
|
131 | 131 | b" rev offset length " + basehdr + b" linkrev" |
|
132 | 132 | b" nodeid p1 p2\n" |
|
133 | 133 | ) |
|
134 | 134 | ) |
|
135 | 135 | elif format == 1: |
|
136 | 136 | ui.write( |
|
137 | 137 | ( |
|
138 | 138 | b" rev flag offset length" |
|
139 | 139 | b" size " + basehdr + b" link p1 p2" |
|
140 | 140 | b" nodeid\n" |
|
141 | 141 | ) |
|
142 | 142 | ) |
|
143 | 143 | |
|
144 | 144 | for i in r: |
|
145 | 145 | node = r.node(i) |
|
146 | 146 | if generaldelta: |
|
147 | 147 | base = r.get_revlog().deltaparent(i) |
|
148 | 148 | else: |
|
149 | 149 | base = r.get_revlog().chainbase(i) |
|
150 | 150 | if format == 0: |
|
151 | 151 | try: |
|
152 | 152 | pp = r.parents(node) |
|
153 | 153 | except Exception: |
|
154 | 154 | pp = [repo.nullid, repo.nullid] |
|
155 | 155 | ui.write( |
|
156 | 156 | b"% 6d % 9d % 7d % 6d % 7d %s %s %s\n" |
|
157 | 157 | % ( |
|
158 | 158 | i, |
|
159 | 159 | r.get_revlog().start(i), |
|
160 | 160 | r.get_revlog().length(i), |
|
161 | 161 | base, |
|
162 | 162 | r.linkrev(i), |
|
163 | 163 | short(node), |
|
164 | 164 | short(pp[0]), |
|
165 | 165 | short(pp[1]), |
|
166 | 166 | ) |
|
167 | 167 | ) |
|
168 | 168 | elif format == 1: |
|
169 | 169 | pr = r.parentrevs(i) |
|
170 | 170 | ui.write( |
|
171 | 171 | b"% 6d %04x % 8d % 8d % 8d % 6d % 6d % 6d % 6d %s\n" |
|
172 | 172 | % ( |
|
173 | 173 | i, |
|
174 | 174 | r.get_revlog().flags(i), |
|
175 | 175 | r.get_revlog().start(i), |
|
176 | 176 | r.get_revlog().length(i), |
|
177 | 177 | r.get_revlog().rawsize(i), |
|
178 | 178 | base, |
|
179 | 179 | r.linkrev(i), |
|
180 | 180 | pr[0], |
|
181 | 181 | pr[1], |
|
182 | 182 | short(node), |
|
183 | 183 | ) |
|
184 | 184 | ) |
|
185 | 185 | |
|
186 | 186 | |
|
187 | 187 | def debugindexdot(orig, ui, repo, file_): |
|
188 | 188 | """dump an index DAG as a graphviz dot file""" |
|
189 | 189 | if not shallowutil.isenabled(repo): |
|
190 | 190 | return orig(ui, repo, file_) |
|
191 | 191 | |
|
192 | 192 | r = buildtemprevlog(repo, os.path.basename(file_)[:-2]) |
|
193 | 193 | |
|
194 | 194 | ui.writenoi18n(b"digraph G {\n") |
|
195 | 195 | for i in r: |
|
196 | 196 | node = r.node(i) |
|
197 | 197 | pp = r.parents(node) |
|
198 | 198 | ui.write(b"\t%d -> %d\n" % (r.rev(pp[0]), i)) |
|
199 | 199 | if pp[1] != repo.nullid: |
|
200 | 200 | ui.write(b"\t%d -> %d\n" % (r.rev(pp[1]), i)) |
|
201 | 201 | ui.write(b"}\n") |
|
202 | 202 | |
|
203 | 203 | |
|
204 | 204 | def verifyremotefilelog(ui, path, **opts): |
|
205 | 205 | decompress = opts.get('decompress') |
|
206 | 206 | |
|
207 | 207 | for root, dirs, files in os.walk(path): |
|
208 | 208 | for file in files: |
|
209 | 209 | if file == b"repos": |
|
210 | 210 | continue |
|
211 | 211 | filepath = os.path.join(root, file) |
|
212 | 212 | size, firstnode, mapping = parsefileblob(filepath, decompress) |
|
213 | 213 | for p1, p2, linknode, copyfrom in mapping.values(): |
|
214 | 214 | if linknode == sha1nodeconstants.nullid: |
|
215 | 215 | actualpath = os.path.relpath(root, path) |
|
216 | 216 | key = fileserverclient.getcachekey( |
|
217 | 217 | b"reponame", actualpath, file |
|
218 | 218 | ) |
|
219 | 219 | ui.status( |
|
220 | 220 | b"%s %s\n" % (key, os.path.relpath(filepath, path)) |
|
221 | 221 | ) |
|
222 | 222 | |
|
223 | 223 | |
|
224 | 224 | def _decompressblob(raw): |
|
225 | 225 | return zlib.decompress(raw) |
|
226 | 226 | |
|
227 | 227 | |
|
228 | 228 | def parsefileblob(path, decompress): |
|
229 | 229 | f = open(path, b"rb") |
|
230 | 230 | try: |
|
231 | 231 | raw = f.read() |
|
232 | 232 | finally: |
|
233 | 233 | f.close() |
|
234 | 234 | |
|
235 | 235 | if decompress: |
|
236 | 236 | raw = _decompressblob(raw) |
|
237 | 237 | |
|
238 | 238 | offset, size, flags = shallowutil.parsesizeflags(raw) |
|
239 | 239 | start = offset + size |
|
240 | 240 | |
|
241 | 241 | firstnode = None |
|
242 | 242 | |
|
243 | 243 | mapping = {} |
|
244 | 244 | while start < len(raw): |
|
245 | 245 | divider = raw.index(b'\0', start + 80) |
|
246 | 246 | |
|
247 | 247 | currentnode = raw[start : (start + 20)] |
|
248 | 248 | if not firstnode: |
|
249 | 249 | firstnode = currentnode |
|
250 | 250 | |
|
251 | 251 | p1 = raw[(start + 20) : (start + 40)] |
|
252 | 252 | p2 = raw[(start + 40) : (start + 60)] |
|
253 | 253 | linknode = raw[(start + 60) : (start + 80)] |
|
254 | 254 | copyfrom = raw[(start + 80) : divider] |
|
255 | 255 | |
|
256 | 256 | mapping[currentnode] = (p1, p2, linknode, copyfrom) |
|
257 | 257 | start = divider + 1 |
|
258 | 258 | |
|
259 | 259 | return size, firstnode, mapping |
|
260 | 260 | |
|
261 | 261 | |
|
262 | 262 | def debugdatapack(ui, *paths, **opts): |
|
263 | 263 | for path in paths: |
|
264 | 264 | if b'.data' in path: |
|
265 | 265 | path = path[: path.index(b'.data')] |
|
266 | 266 | ui.write(b"%s:\n" % path) |
|
267 | 267 | dpack = datapack.datapack(path) |
|
268 | 268 | node = opts.get('node') |
|
269 | 269 | if node: |
|
270 | 270 | deltachain = dpack.getdeltachain(b'', bin(node)) |
|
271 | 271 | dumpdeltachain(ui, deltachain, **opts) |
|
272 | 272 | return |
|
273 | 273 | |
|
274 | 274 | if opts.get('long'): |
|
275 | 275 | hashformatter = hex |
|
276 | 276 | hashlen = 42 |
|
277 | 277 | else: |
|
278 | 278 | hashformatter = short |
|
279 | 279 | hashlen = 14 |
|
280 | 280 | |
|
281 | 281 | lastfilename = None |
|
282 | 282 | totaldeltasize = 0 |
|
283 | 283 | totalblobsize = 0 |
|
284 | 284 | |
|
285 | 285 | def printtotals(): |
|
286 | 286 | if lastfilename is not None: |
|
287 | 287 | ui.write(b"\n") |
|
288 | 288 | if not totaldeltasize or not totalblobsize: |
|
289 | 289 | return |
|
290 | 290 | difference = totalblobsize - totaldeltasize |
|
291 | 291 | deltastr = b"%0.1f%% %s" % ( |
|
292 | 292 | (100.0 * abs(difference) / totalblobsize), |
|
293 | 293 | (b"smaller" if difference > 0 else b"bigger"), |
|
294 | 294 | ) |
|
295 | 295 | |
|
296 | 296 | ui.writenoi18n( |
|
297 | 297 | b"Total:%s%s %s (%s)\n" |
|
298 | 298 | % ( |
|
299 | 299 | b"".ljust(2 * hashlen - len(b"Total:")), |
|
300 | 300 | (b'%d' % totaldeltasize).ljust(12), |
|
301 | 301 | (b'%d' % totalblobsize).ljust(9), |
|
302 | 302 | deltastr, |
|
303 | 303 | ) |
|
304 | 304 | ) |
|
305 | 305 | |
|
306 | 306 | bases = {} |
|
307 | 307 | nodes = set() |
|
308 | 308 | failures = 0 |
|
309 | 309 | for filename, node, deltabase, deltalen in dpack.iterentries(): |
|
310 | 310 | bases[node] = deltabase |
|
311 | 311 | if node in nodes: |
|
312 | 312 | ui.write((b"Bad entry: %s appears twice\n" % short(node))) |
|
313 | 313 | failures += 1 |
|
314 | 314 | nodes.add(node) |
|
315 | 315 | if filename != lastfilename: |
|
316 | 316 | printtotals() |
|
317 | 317 | name = b'(empty name)' if filename == b'' else filename |
|
318 | 318 | ui.write(b"%s:\n" % name) |
|
319 | 319 | ui.write( |
|
320 | 320 | b"%s%s%s%s\n" |
|
321 | 321 | % ( |
|
322 | 322 | b"Node".ljust(hashlen), |
|
323 | 323 | b"Delta Base".ljust(hashlen), |
|
324 | 324 | b"Delta Length".ljust(14), |
|
325 | 325 | b"Blob Size".ljust(9), |
|
326 | 326 | ) |
|
327 | 327 | ) |
|
328 | 328 | lastfilename = filename |
|
329 | 329 | totalblobsize = 0 |
|
330 | 330 | totaldeltasize = 0 |
|
331 | 331 | |
|
332 | 332 | # Metadata could be missing, in which case it will be an empty dict. |
|
333 | 333 | meta = dpack.getmeta(filename, node) |
|
334 | 334 | if constants.METAKEYSIZE in meta: |
|
335 | 335 | blobsize = meta[constants.METAKEYSIZE] |
|
336 | 336 | totaldeltasize += deltalen |
|
337 | 337 | totalblobsize += blobsize |
|
338 | 338 | else: |
|
339 | 339 | blobsize = b"(missing)" |
|
340 | 340 | ui.write( |
|
341 | 341 | b"%s %s %s%s\n" |
|
342 | 342 | % ( |
|
343 | 343 | hashformatter(node), |
|
344 | 344 | hashformatter(deltabase), |
|
345 | 345 | (b'%d' % deltalen).ljust(14), |
|
346 | 346 | pycompat.bytestr(blobsize), |
|
347 | 347 | ) |
|
348 | 348 | ) |
|
349 | 349 | |
|
350 | 350 | if filename is not None: |
|
351 | 351 | printtotals() |
|
352 | 352 | |
|
353 | 353 | failures += _sanitycheck(ui, set(nodes), bases) |
|
354 | 354 | if failures > 1: |
|
355 | 355 | ui.warn((b"%d failures\n" % failures)) |
|
356 | 356 | return 1 |
|
357 | 357 | |
|
358 | 358 | |
|
359 | 359 | def _sanitycheck(ui, nodes, bases): |
|
360 | 360 | """ |
|
361 | 361 | Does some basic sanity checking on a packfiles with ``nodes`` ``bases`` (a |
|
362 | 362 | mapping of node->base): |
|
363 | 363 | |
|
364 | 364 | - Each deltabase must itself be a node elsewhere in the pack |
|
365 | 365 | - There must be no cycles |
|
366 | 366 | """ |
|
367 | 367 | failures = 0 |
|
368 | 368 | for node in nodes: |
|
369 | 369 | seen = set() |
|
370 | 370 | current = node |
|
371 | 371 | deltabase = bases[current] |
|
372 | 372 | |
|
373 | 373 | while deltabase != sha1nodeconstants.nullid: |
|
374 | 374 | if deltabase not in nodes: |
|
375 | 375 | ui.warn( |
|
376 | 376 | ( |
|
377 | 377 | b"Bad entry: %s has an unknown deltabase (%s)\n" |
|
378 | 378 | % (short(node), short(deltabase)) |
|
379 | 379 | ) |
|
380 | 380 | ) |
|
381 | 381 | failures += 1 |
|
382 | 382 | break |
|
383 | 383 | |
|
384 | 384 | if deltabase in seen: |
|
385 | 385 | ui.warn( |
|
386 | 386 | ( |
|
387 | 387 | b"Bad entry: %s has a cycle (at %s)\n" |
|
388 | 388 | % (short(node), short(deltabase)) |
|
389 | 389 | ) |
|
390 | 390 | ) |
|
391 | 391 | failures += 1 |
|
392 | 392 | break |
|
393 | 393 | |
|
394 | 394 | current = deltabase |
|
395 | 395 | seen.add(current) |
|
396 | 396 | deltabase = bases[current] |
|
397 | 397 | # Since ``node`` begins a valid chain, reset/memoize its base to nullid |
|
398 | 398 | # so we don't traverse it again. |
|
399 | 399 | bases[node] = sha1nodeconstants.nullid |
|
400 | 400 | return failures |
|
401 | 401 | |
|
402 | 402 | |
|
403 | 403 | def dumpdeltachain(ui, deltachain, **opts): |
|
404 | 404 | hashformatter = hex |
|
405 | 405 | hashlen = 40 |
|
406 | 406 | |
|
407 | 407 | lastfilename = None |
|
408 | 408 | for filename, node, filename, deltabasenode, delta in deltachain: |
|
409 | 409 | if filename != lastfilename: |
|
410 | 410 | ui.write(b"\n%s\n" % filename) |
|
411 | 411 | lastfilename = filename |
|
412 | 412 | ui.write( |
|
413 | 413 | b"%s %s %s %s\n" |
|
414 | 414 | % ( |
|
415 | 415 | b"Node".ljust(hashlen), |
|
416 | 416 | b"Delta Base".ljust(hashlen), |
|
417 | 417 | b"Delta SHA1".ljust(hashlen), |
|
418 | 418 | b"Delta Length".ljust(6), |
|
419 | 419 | ) |
|
420 | 420 | ) |
|
421 | 421 | |
|
422 | 422 | ui.write( |
|
423 | 423 | b"%s %s %s %d\n" |
|
424 | 424 | % ( |
|
425 | 425 | hashformatter(node), |
|
426 | 426 | hashformatter(deltabasenode), |
|
427 | 427 | hex(hashutil.sha1(delta).digest()), |
|
428 | 428 | len(delta), |
|
429 | 429 | ) |
|
430 | 430 | ) |
|
431 | 431 | |
|
432 | 432 | |
|
433 | 433 | def debughistorypack(ui, path): |
|
434 | 434 | if b'.hist' in path: |
|
435 | 435 | path = path[: path.index(b'.hist')] |
|
436 | 436 | hpack = historypack.historypack(path) |
|
437 | 437 | |
|
438 | 438 | lastfilename = None |
|
439 | 439 | for entry in hpack.iterentries(): |
|
440 | 440 | filename, node, p1node, p2node, linknode, copyfrom = entry |
|
441 | 441 | if filename != lastfilename: |
|
442 | 442 | ui.write(b"\n%s\n" % filename) |
|
443 | 443 | ui.write( |
|
444 | 444 | b"%s%s%s%s%s\n" |
|
445 | 445 | % ( |
|
446 | 446 | b"Node".ljust(14), |
|
447 | 447 | b"P1 Node".ljust(14), |
|
448 | 448 | b"P2 Node".ljust(14), |
|
449 | 449 | b"Link Node".ljust(14), |
|
450 | 450 | b"Copy From", |
|
451 | 451 | ) |
|
452 | 452 | ) |
|
453 | 453 | lastfilename = filename |
|
454 | 454 | ui.write( |
|
455 | 455 | b"%s %s %s %s %s\n" |
|
456 | 456 | % ( |
|
457 | 457 | short(node), |
|
458 | 458 | short(p1node), |
|
459 | 459 | short(p2node), |
|
460 | 460 | short(linknode), |
|
461 | 461 | copyfrom, |
|
462 | 462 | ) |
|
463 | 463 | ) |
|
464 | 464 | |
|
465 | 465 | |
|
466 | 466 | def debugwaitonrepack(repo): |
|
467 | 467 | with lockmod.lock(repack.repacklockvfs(repo), b"repacklock", timeout=-1): |
|
468 | 468 | return |
|
469 | 469 | |
|
470 | 470 | |
|
471 | 471 | def debugwaitonprefetch(repo): |
|
472 | 472 | with repo._lock( |
|
473 | 473 | repo.svfs, |
|
474 | 474 | b"prefetchlock", |
|
475 | 475 | True, |
|
476 | 476 | None, |
|
477 | 477 | None, |
|
478 | 478 | _(b'prefetching in %s') % repo.origroot, |
|
479 | 479 | ): |
|
480 | 480 | pass |
@@ -1,472 +1,476 b'' | |||
|
1 | 1 | # remotefilelog.py - filelog implementation where filelog history is stored |
|
2 | 2 | # remotely |
|
3 | 3 | # |
|
4 | 4 | # Copyright 2013 Facebook, Inc. |
|
5 | 5 | # |
|
6 | 6 | # This software may be used and distributed according to the terms of the |
|
7 | 7 | # GNU General Public License version 2 or any later version. |
|
8 | 8 | |
|
9 | 9 | import collections |
|
10 | 10 | |
|
11 | from typing import ( | |
|
12 | Iterator, | |
|
13 | ) | |
|
14 | ||
|
11 | 15 | from mercurial.node import bin |
|
12 | 16 | from mercurial.i18n import _ |
|
13 | 17 | from mercurial import ( |
|
14 | 18 | ancestor, |
|
15 | 19 | error, |
|
16 | 20 | mdiff, |
|
17 | 21 | revlog, |
|
18 | 22 | ) |
|
19 | 23 | from mercurial.utils import storageutil |
|
20 | 24 | from mercurial.revlogutils import flagutil |
|
21 | 25 | |
|
22 | 26 | from . import ( |
|
23 | 27 | constants, |
|
24 | 28 | shallowutil, |
|
25 | 29 | ) |
|
26 | 30 | |
|
27 | 31 | |
|
28 | 32 | class remotefilelognodemap: |
|
29 | 33 | def __init__(self, filename, store): |
|
30 | 34 | self._filename = filename |
|
31 | 35 | self._store = store |
|
32 | 36 | |
|
33 | 37 | def __contains__(self, node): |
|
34 | 38 | missing = self._store.getmissing([(self._filename, node)]) |
|
35 | 39 | return not bool(missing) |
|
36 | 40 | |
|
37 | 41 | def __get__(self, node): |
|
38 | 42 | if node not in self: |
|
39 | 43 | raise KeyError(node) |
|
40 | 44 | return node |
|
41 | 45 | |
|
42 | 46 | |
|
43 | 47 | class remotefilelog: |
|
44 | 48 | _flagserrorclass = error.RevlogError |
|
45 | 49 | |
|
46 | 50 | def __init__(self, opener, path, repo): |
|
47 | 51 | self.opener = opener |
|
48 | 52 | self.filename = path |
|
49 | 53 | self.repo = repo |
|
50 | 54 | self.nodemap = remotefilelognodemap(self.filename, repo.contentstore) |
|
51 | 55 | |
|
52 | 56 | self.version = 1 |
|
53 | 57 | |
|
54 | 58 | self._flagprocessors = dict(flagutil.flagprocessors) |
|
55 | 59 | |
|
56 | 60 | def read(self, node): |
|
57 | 61 | """returns the file contents at this node""" |
|
58 | 62 | t = self.revision(node) |
|
59 | 63 | if not t.startswith(b'\1\n'): |
|
60 | 64 | return t |
|
61 | 65 | s = t.index(b'\1\n', 2) |
|
62 | 66 | return t[s + 2 :] |
|
63 | 67 | |
|
64 | 68 | def add(self, text, meta, transaction, linknode, p1=None, p2=None): |
|
65 | 69 | # hash with the metadata, like in vanilla filelogs |
|
66 | 70 | hashtext = shallowutil.createrevlogtext( |
|
67 | 71 | text, meta.get(b'copy'), meta.get(b'copyrev') |
|
68 | 72 | ) |
|
69 | 73 | node = storageutil.hashrevisionsha1(hashtext, p1, p2) |
|
70 | 74 | return self.addrevision( |
|
71 | 75 | hashtext, transaction, linknode, p1, p2, node=node |
|
72 | 76 | ) |
|
73 | 77 | |
|
74 | 78 | def _createfileblob(self, text, meta, flags, p1, p2, node, linknode): |
|
75 | 79 | # text passed to "_createfileblob" does not include filelog metadata |
|
76 | 80 | header = shallowutil.buildfileblobheader(len(text), flags) |
|
77 | 81 | data = b"%s\0%s" % (header, text) |
|
78 | 82 | |
|
79 | 83 | realp1 = p1 |
|
80 | 84 | copyfrom = b"" |
|
81 | 85 | if meta and b'copy' in meta: |
|
82 | 86 | copyfrom = meta[b'copy'] |
|
83 | 87 | realp1 = bin(meta[b'copyrev']) |
|
84 | 88 | |
|
85 | 89 | data += b"%s%s%s%s%s\0" % (node, realp1, p2, linknode, copyfrom) |
|
86 | 90 | |
|
87 | 91 | visited = set() |
|
88 | 92 | |
|
89 | 93 | pancestors = {} |
|
90 | 94 | queue = [] |
|
91 | 95 | if realp1 != self.repo.nullid: |
|
92 | 96 | p1flog = self |
|
93 | 97 | if copyfrom: |
|
94 | 98 | p1flog = remotefilelog(self.opener, copyfrom, self.repo) |
|
95 | 99 | |
|
96 | 100 | pancestors.update(p1flog.ancestormap(realp1)) |
|
97 | 101 | queue.append(realp1) |
|
98 | 102 | visited.add(realp1) |
|
99 | 103 | if p2 != self.repo.nullid: |
|
100 | 104 | pancestors.update(self.ancestormap(p2)) |
|
101 | 105 | queue.append(p2) |
|
102 | 106 | visited.add(p2) |
|
103 | 107 | |
|
104 | 108 | ancestortext = b"" |
|
105 | 109 | |
|
106 | 110 | # add the ancestors in topological order |
|
107 | 111 | while queue: |
|
108 | 112 | c = queue.pop(0) |
|
109 | 113 | pa1, pa2, ancestorlinknode, pacopyfrom = pancestors[c] |
|
110 | 114 | |
|
111 | 115 | pacopyfrom = pacopyfrom or b'' |
|
112 | 116 | ancestortext += b"%s%s%s%s%s\0" % ( |
|
113 | 117 | c, |
|
114 | 118 | pa1, |
|
115 | 119 | pa2, |
|
116 | 120 | ancestorlinknode, |
|
117 | 121 | pacopyfrom, |
|
118 | 122 | ) |
|
119 | 123 | |
|
120 | 124 | if pa1 != self.repo.nullid and pa1 not in visited: |
|
121 | 125 | queue.append(pa1) |
|
122 | 126 | visited.add(pa1) |
|
123 | 127 | if pa2 != self.repo.nullid and pa2 not in visited: |
|
124 | 128 | queue.append(pa2) |
|
125 | 129 | visited.add(pa2) |
|
126 | 130 | |
|
127 | 131 | data += ancestortext |
|
128 | 132 | |
|
129 | 133 | return data |
|
130 | 134 | |
|
131 | 135 | def addrevision( |
|
132 | 136 | self, |
|
133 | 137 | text, |
|
134 | 138 | transaction, |
|
135 | 139 | linknode, |
|
136 | 140 | p1, |
|
137 | 141 | p2, |
|
138 | 142 | cachedelta=None, |
|
139 | 143 | node=None, |
|
140 | 144 | flags=revlog.REVIDX_DEFAULT_FLAGS, |
|
141 | 145 | sidedata=None, |
|
142 | 146 | ): |
|
143 | 147 | # text passed to "addrevision" includes hg filelog metadata header |
|
144 | 148 | if node is None: |
|
145 | 149 | node = storageutil.hashrevisionsha1(text, p1, p2) |
|
146 | 150 | |
|
147 | 151 | meta, metaoffset = storageutil.parsemeta(text) |
|
148 | 152 | rawtext, validatehash = flagutil.processflagswrite( |
|
149 | 153 | self, |
|
150 | 154 | text, |
|
151 | 155 | flags, |
|
152 | 156 | ) |
|
153 | 157 | return self.addrawrevision( |
|
154 | 158 | rawtext, |
|
155 | 159 | transaction, |
|
156 | 160 | linknode, |
|
157 | 161 | p1, |
|
158 | 162 | p2, |
|
159 | 163 | node, |
|
160 | 164 | flags, |
|
161 | 165 | cachedelta, |
|
162 | 166 | _metatuple=(meta, metaoffset), |
|
163 | 167 | ) |
|
164 | 168 | |
|
165 | 169 | def addrawrevision( |
|
166 | 170 | self, |
|
167 | 171 | rawtext, |
|
168 | 172 | transaction, |
|
169 | 173 | linknode, |
|
170 | 174 | p1, |
|
171 | 175 | p2, |
|
172 | 176 | node, |
|
173 | 177 | flags, |
|
174 | 178 | cachedelta=None, |
|
175 | 179 | _metatuple=None, |
|
176 | 180 | ): |
|
177 | 181 | if _metatuple: |
|
178 | 182 | # _metatuple: used by "addrevision" internally by remotefilelog |
|
179 | 183 | # meta was parsed confidently |
|
180 | 184 | meta, metaoffset = _metatuple |
|
181 | 185 | else: |
|
182 | 186 | # not from self.addrevision, but something else (repo._filecommit) |
|
183 | 187 | # calls addrawrevision directly. remotefilelog needs to get and |
|
184 | 188 | # strip filelog metadata. |
|
185 | 189 | # we don't have confidence about whether rawtext contains filelog |
|
186 | 190 | # metadata or not (flag processor could replace it), so we just |
|
187 | 191 | # parse it as best-effort. |
|
188 | 192 | # in LFS (flags != 0)'s case, the best way is to call LFS code to |
|
189 | 193 | # get the meta information, instead of storageutil.parsemeta. |
|
190 | 194 | meta, metaoffset = storageutil.parsemeta(rawtext) |
|
191 | 195 | if flags != 0: |
|
192 | 196 | # when flags != 0, be conservative and do not mangle rawtext, since |
|
193 | 197 | # a read flag processor expects the text not being mangled at all. |
|
194 | 198 | metaoffset = 0 |
|
195 | 199 | if metaoffset: |
|
196 | 200 | # remotefilelog fileblob stores copy metadata in its ancestortext, |
|
197 | 201 | # not its main blob. so we need to remove filelog metadata |
|
198 | 202 | # (containing copy information) from text. |
|
199 | 203 | blobtext = rawtext[metaoffset:] |
|
200 | 204 | else: |
|
201 | 205 | blobtext = rawtext |
|
202 | 206 | data = self._createfileblob( |
|
203 | 207 | blobtext, meta, flags, p1, p2, node, linknode |
|
204 | 208 | ) |
|
205 | 209 | self.repo.contentstore.addremotefilelognode(self.filename, node, data) |
|
206 | 210 | |
|
207 | 211 | return node |
|
208 | 212 | |
|
209 | 213 | def renamed(self, node): |
|
210 | 214 | ancestors = self.repo.metadatastore.getancestors(self.filename, node) |
|
211 | 215 | p1, p2, linknode, copyfrom = ancestors[node] |
|
212 | 216 | if copyfrom: |
|
213 | 217 | return (copyfrom, p1) |
|
214 | 218 | |
|
215 | 219 | return False |
|
216 | 220 | |
|
217 | 221 | def size(self, node): |
|
218 | 222 | """return the size of a given revision""" |
|
219 | 223 | return len(self.read(node)) |
|
220 | 224 | |
|
221 | 225 | rawsize = size |
|
222 | 226 | |
|
223 | 227 | def cmp(self, node, text): |
|
224 | 228 | """compare text with a given file revision |
|
225 | 229 | |
|
226 | 230 | returns True if text is different than what is stored. |
|
227 | 231 | """ |
|
228 | 232 | |
|
229 | 233 | if node == self.repo.nullid: |
|
230 | 234 | return True |
|
231 | 235 | |
|
232 | 236 | nodetext = self.read(node) |
|
233 | 237 | return nodetext != text |
|
234 | 238 | |
|
235 | 239 | def __nonzero__(self): |
|
236 | 240 | return True |
|
237 | 241 | |
|
238 | 242 | __bool__ = __nonzero__ |
|
239 | 243 | |
|
240 | 244 | def __len__(self): |
|
241 | 245 | if self.filename in (b'.hgtags', b'.hgsub', b'.hgsubstate'): |
|
242 | 246 | # Global tag and subrepository support require access to the |
|
243 | 247 | # file history for various performance sensitive operations. |
|
244 | 248 | # excludepattern should be used for repositories depending on |
|
245 | 249 | # those features to fallback to regular filelog. |
|
246 | 250 | return 0 |
|
247 | 251 | |
|
248 | 252 | raise RuntimeError(b"len not supported") |
|
249 | 253 | |
|
250 | 254 | def heads(self): |
|
251 | 255 | # Fake heads of the filelog to satisfy hgweb. |
|
252 | 256 | return [] |
|
253 | 257 | |
|
254 | 258 | def empty(self): |
|
255 | 259 | return False |
|
256 | 260 | |
|
257 | 261 | def flags(self, node): |
|
258 | 262 | if isinstance(node, int): |
|
259 | 263 | raise error.ProgrammingError( |
|
260 | 264 | b'remotefilelog does not accept integer rev for flags' |
|
261 | 265 | ) |
|
262 | 266 | store = self.repo.contentstore |
|
263 | 267 | return store.getmeta(self.filename, node).get(constants.METAKEYFLAG, 0) |
|
264 | 268 | |
|
265 | 269 | def parents(self, node): |
|
266 | 270 | if node == self.repo.nullid: |
|
267 | 271 | return self.repo.nullid, self.repo.nullid |
|
268 | 272 | |
|
269 | 273 | ancestormap = self.repo.metadatastore.getancestors(self.filename, node) |
|
270 | 274 | p1, p2, linknode, copyfrom = ancestormap[node] |
|
271 | 275 | if copyfrom: |
|
272 | 276 | p1 = self.repo.nullid |
|
273 | 277 | |
|
274 | 278 | return p1, p2 |
|
275 | 279 | |
|
276 | 280 | def parentrevs(self, rev): |
|
277 | 281 | # TODO(augie): this is a node and should be a rev, but for now |
|
278 | 282 | # nothing in core seems to actually break. |
|
279 | 283 | return self.parents(rev) |
|
280 | 284 | |
|
281 | 285 | def linknode(self, node): |
|
282 | 286 | ancestormap = self.repo.metadatastore.getancestors(self.filename, node) |
|
283 | 287 | p1, p2, linknode, copyfrom = ancestormap[node] |
|
284 | 288 | return linknode |
|
285 | 289 | |
|
286 | 290 | def linkrev(self, node): |
|
287 | 291 | return self.repo.unfiltered().changelog.rev(self.linknode(node)) |
|
288 | 292 | |
|
289 | 293 | def emitrevisions( |
|
290 | 294 | self, |
|
291 | 295 | nodes, |
|
292 | 296 | nodesorder=None, |
|
293 | 297 | revisiondata=False, |
|
294 | 298 | assumehaveparentrevisions=False, |
|
295 | 299 | deltaprevious=False, |
|
296 | 300 | deltamode=None, |
|
297 | 301 | sidedata_helpers=None, |
|
298 | 302 | debug_info=None, |
|
299 | ): | |
|
303 | ) -> Iterator[revlog.RevLogRevisionDelta]: | |
|
300 | 304 | # we don't use any of these parameters here |
|
301 | 305 | del nodesorder, revisiondata, assumehaveparentrevisions, deltaprevious |
|
302 | 306 | del deltamode |
|
303 | 307 | prevnode = None |
|
304 | 308 | for node in nodes: |
|
305 | 309 | p1, p2 = self.parents(node) |
|
306 | 310 | if prevnode is None: |
|
307 | 311 | basenode = prevnode = p1 |
|
308 | 312 | if basenode == node: |
|
309 | 313 | basenode = self.repo.nullid |
|
310 | 314 | if basenode != self.repo.nullid: |
|
311 | 315 | revision = None |
|
312 | 316 | delta = self.revdiff(basenode, node) |
|
313 | 317 | else: |
|
314 | 318 | revision = self.rawdata(node) |
|
315 | 319 | delta = None |
|
316 | 320 | yield revlog.revlogrevisiondelta( |
|
317 | 321 | node=node, |
|
318 | 322 | p1node=p1, |
|
319 | 323 | p2node=p2, |
|
320 | 324 | linknode=self.linknode(node), |
|
321 | 325 | basenode=basenode, |
|
322 | 326 | flags=self.flags(node), |
|
323 | 327 | baserevisionsize=None, |
|
324 | 328 | revision=revision, |
|
325 | 329 | delta=delta, |
|
326 | 330 | # Sidedata is not supported yet |
|
327 | 331 | sidedata=None, |
|
328 | 332 | # Protocol flags are not used yet |
|
329 | 333 | protocol_flags=0, |
|
330 | 334 | ) |
|
331 | 335 | |
|
332 | 336 | def revdiff(self, node1, node2): |
|
333 | 337 | return mdiff.textdiff(self.rawdata(node1), self.rawdata(node2)) |
|
334 | 338 | |
|
335 | 339 | def lookup(self, node): |
|
336 | 340 | if len(node) == 40: |
|
337 | 341 | node = bin(node) |
|
338 | 342 | if len(node) != 20: |
|
339 | 343 | raise error.LookupError( |
|
340 | 344 | node, self.filename, _(b'invalid lookup input') |
|
341 | 345 | ) |
|
342 | 346 | |
|
343 | 347 | return node |
|
344 | 348 | |
|
345 | 349 | def rev(self, node): |
|
346 | 350 | # This is a hack to make TortoiseHG work. |
|
347 | 351 | return node |
|
348 | 352 | |
|
349 | 353 | def node(self, rev): |
|
350 | 354 | # This is a hack. |
|
351 | 355 | if isinstance(rev, int): |
|
352 | 356 | raise error.ProgrammingError( |
|
353 | 357 | b'remotefilelog does not convert integer rev to node' |
|
354 | 358 | ) |
|
355 | 359 | return rev |
|
356 | 360 | |
|
357 | 361 | def revision(self, node, raw=False): |
|
358 | 362 | """returns the revlog contents at this node. |
|
359 | 363 | this includes the meta data traditionally included in file revlogs. |
|
360 | 364 | this is generally only used for bundling and communicating with vanilla |
|
361 | 365 | hg clients. |
|
362 | 366 | """ |
|
363 | 367 | if node == self.repo.nullid: |
|
364 | 368 | return b"" |
|
365 | 369 | if len(node) != 20: |
|
366 | 370 | raise error.LookupError( |
|
367 | 371 | node, self.filename, _(b'invalid revision input') |
|
368 | 372 | ) |
|
369 | 373 | if ( |
|
370 | 374 | node == self.repo.nodeconstants.wdirid |
|
371 | 375 | or node in self.repo.nodeconstants.wdirfilenodeids |
|
372 | 376 | ): |
|
373 | 377 | raise error.WdirUnsupported |
|
374 | 378 | |
|
375 | 379 | store = self.repo.contentstore |
|
376 | 380 | rawtext = store.get(self.filename, node) |
|
377 | 381 | if raw: |
|
378 | 382 | return rawtext |
|
379 | 383 | flags = store.getmeta(self.filename, node).get(constants.METAKEYFLAG, 0) |
|
380 | 384 | if flags == 0: |
|
381 | 385 | return rawtext |
|
382 | 386 | return flagutil.processflagsread(self, rawtext, flags)[0] |
|
383 | 387 | |
|
384 | 388 | def rawdata(self, node): |
|
385 | 389 | return self.revision(node, raw=False) |
|
386 | 390 | |
|
387 | 391 | def ancestormap(self, node): |
|
388 | 392 | return self.repo.metadatastore.getancestors(self.filename, node) |
|
389 | 393 | |
|
390 | 394 | def ancestor(self, a, b): |
|
391 | 395 | if a == self.repo.nullid or b == self.repo.nullid: |
|
392 | 396 | return self.repo.nullid |
|
393 | 397 | |
|
394 | 398 | revmap, parentfunc = self._buildrevgraph(a, b) |
|
395 | 399 | nodemap = {v: k for (k, v) in revmap.items()} |
|
396 | 400 | |
|
397 | 401 | ancs = ancestor.ancestors(parentfunc, revmap[a], revmap[b]) |
|
398 | 402 | if ancs: |
|
399 | 403 | # choose a consistent winner when there's a tie |
|
400 | 404 | return min(map(nodemap.__getitem__, ancs)) |
|
401 | 405 | return self.repo.nullid |
|
402 | 406 | |
|
403 | 407 | def commonancestorsheads(self, a, b): |
|
404 | 408 | """calculate all the heads of the common ancestors of nodes a and b""" |
|
405 | 409 | |
|
406 | 410 | if a == self.repo.nullid or b == self.repo.nullid: |
|
407 | 411 | return self.repo.nullid |
|
408 | 412 | |
|
409 | 413 | revmap, parentfunc = self._buildrevgraph(a, b) |
|
410 | 414 | nodemap = {v: k for (k, v) in revmap.items()} |
|
411 | 415 | |
|
412 | 416 | ancs = ancestor.commonancestorsheads(parentfunc, revmap[a], revmap[b]) |
|
413 | 417 | return map(nodemap.__getitem__, ancs) |
|
414 | 418 | |
|
415 | 419 | def _buildrevgraph(self, a, b): |
|
416 | 420 | """Builds a numeric revision graph for the given two nodes. |
|
417 | 421 | Returns a node->rev map and a rev->[revs] parent function. |
|
418 | 422 | """ |
|
419 | 423 | amap = self.ancestormap(a) |
|
420 | 424 | bmap = self.ancestormap(b) |
|
421 | 425 | |
|
422 | 426 | # Union the two maps |
|
423 | 427 | parentsmap = collections.defaultdict(list) |
|
424 | 428 | allparents = set() |
|
425 | 429 | for mapping in (amap, bmap): |
|
426 | 430 | for node, pdata in mapping.items(): |
|
427 | 431 | parents = parentsmap[node] |
|
428 | 432 | p1, p2, linknode, copyfrom = pdata |
|
429 | 433 | # Don't follow renames (copyfrom). |
|
430 | 434 | # remotefilectx.ancestor does that. |
|
431 | 435 | if p1 != self.repo.nullid and not copyfrom: |
|
432 | 436 | parents.append(p1) |
|
433 | 437 | allparents.add(p1) |
|
434 | 438 | if p2 != self.repo.nullid: |
|
435 | 439 | parents.append(p2) |
|
436 | 440 | allparents.add(p2) |
|
437 | 441 | |
|
438 | 442 | # Breadth first traversal to build linkrev graph |
|
439 | 443 | parentrevs = collections.defaultdict(list) |
|
440 | 444 | revmap = {} |
|
441 | 445 | queue = collections.deque( |
|
442 | 446 | ((None, n) for n in parentsmap if n not in allparents) |
|
443 | 447 | ) |
|
444 | 448 | while queue: |
|
445 | 449 | prevrev, current = queue.pop() |
|
446 | 450 | if current in revmap: |
|
447 | 451 | if prevrev: |
|
448 | 452 | parentrevs[prevrev].append(revmap[current]) |
|
449 | 453 | continue |
|
450 | 454 | |
|
451 | 455 | # Assign linkrevs in reverse order, so start at |
|
452 | 456 | # len(parentsmap) and work backwards. |
|
453 | 457 | currentrev = len(parentsmap) - len(revmap) - 1 |
|
454 | 458 | revmap[current] = currentrev |
|
455 | 459 | |
|
456 | 460 | if prevrev: |
|
457 | 461 | parentrevs[prevrev].append(currentrev) |
|
458 | 462 | |
|
459 | 463 | for parent in parentsmap.get(current): |
|
460 | 464 | queue.appendleft((currentrev, parent)) |
|
461 | 465 | |
|
462 | 466 | return revmap, parentrevs.__getitem__ |
|
463 | 467 | |
|
464 | 468 | def strip(self, minlink, transaction): |
|
465 | 469 | pass |
|
466 | 470 | |
|
467 | 471 | # misc unused things |
|
468 | 472 | def files(self): |
|
469 | 473 | return [] |
|
470 | 474 | |
|
471 | 475 | def checksize(self): |
|
472 | 476 | return 0, 0 |
@@ -1,319 +1,328 b'' | |||
|
1 | 1 | # filelog.py - file history class for mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2005-2007 Olivia Mackall <olivia@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | |
|
9 | 9 | import typing |
|
10 | 10 | |
|
11 | from typing import ( | |
|
12 | Iterable, | |
|
13 | Iterator, | |
|
14 | ) | |
|
15 | ||
|
11 | 16 | from .i18n import _ |
|
12 | 17 | from .node import nullrev |
|
13 | 18 | from . import ( |
|
14 | 19 | error, |
|
15 | 20 | revlog, |
|
16 | 21 | ) |
|
17 | 22 | from .interfaces import ( |
|
18 | 23 | repository, |
|
19 | 24 | util as interfaceutil, |
|
20 | 25 | ) |
|
21 | 26 | from .utils import storageutil |
|
22 | 27 | from .revlogutils import ( |
|
23 | 28 | constants as revlog_constants, |
|
24 | 29 | rewrite, |
|
25 | 30 | ) |
|
26 | 31 | |
|
27 | 32 | |
|
28 | 33 | class FileLog: |
|
34 | _revlog: revlog.revlog | |
|
35 | nullid: bytes | |
|
36 | _fix_issue6528: bool | |
|
37 | ||
|
29 | 38 | def __init__(self, opener, path, try_split=False): |
|
30 | 39 | self._revlog = revlog.revlog( |
|
31 | 40 | opener, |
|
32 | 41 | # XXX should use the unencoded path |
|
33 | 42 | target=(revlog_constants.KIND_FILELOG, path), |
|
34 | 43 | radix=b'/'.join((b'data', path)), |
|
35 | 44 | censorable=True, |
|
36 | 45 | canonical_parent_order=False, # see comment in revlog.py |
|
37 | 46 | try_split=try_split, |
|
38 | 47 | ) |
|
39 | 48 | # Full name of the user visible file, relative to the repository root. |
|
40 | 49 | # Used by LFS. |
|
41 | 50 | self._revlog.filename = path |
|
42 | 51 | self.nullid = self._revlog.nullid |
|
43 | 52 | opts = opener.options |
|
44 | 53 | self._fix_issue6528 = opts.get(b'issue6528.fix-incoming', True) |
|
45 | 54 | |
|
46 | def get_revlog(self): | |
|
55 | def get_revlog(self) -> revlog.revlog: | |
|
47 | 56 | """return an actual revlog instance if any |
|
48 | 57 | |
|
49 | 58 | This exist because a lot of code leverage the fact the underlying |
|
50 | 59 | storage is a revlog for optimization, so giving simple way to access |
|
51 | 60 | the revlog instance helps such code. |
|
52 | 61 | """ |
|
53 | 62 | return self._revlog |
|
54 | 63 | |
|
55 | def __len__(self): | |
|
64 | def __len__(self) -> int: | |
|
56 | 65 | return len(self._revlog) |
|
57 | 66 | |
|
58 | def __iter__(self): | |
|
67 | def __iter__(self) -> Iterator[int]: | |
|
59 | 68 | return self._revlog.__iter__() |
|
60 | 69 | |
|
61 | 70 | def hasnode(self, node): |
|
62 | 71 | if node in (self.nullid, nullrev): |
|
63 | 72 | return False |
|
64 | 73 | |
|
65 | 74 | try: |
|
66 | 75 | self._revlog.rev(node) |
|
67 | 76 | return True |
|
68 | 77 | except (TypeError, ValueError, IndexError, error.LookupError): |
|
69 | 78 | return False |
|
70 | 79 | |
|
71 | 80 | def revs(self, start=0, stop=None): |
|
72 | 81 | return self._revlog.revs(start=start, stop=stop) |
|
73 | 82 | |
|
74 | 83 | def parents(self, node): |
|
75 | 84 | return self._revlog.parents(node) |
|
76 | 85 | |
|
77 | 86 | def parentrevs(self, rev): |
|
78 | 87 | return self._revlog.parentrevs(rev) |
|
79 | 88 | |
|
80 | 89 | def rev(self, node): |
|
81 | 90 | return self._revlog.rev(node) |
|
82 | 91 | |
|
83 | 92 | def node(self, rev): |
|
84 | 93 | return self._revlog.node(rev) |
|
85 | 94 | |
|
86 | 95 | def lookup(self, node): |
|
87 | 96 | return storageutil.fileidlookup( |
|
88 | 97 | self._revlog, node, self._revlog.display_id |
|
89 | 98 | ) |
|
90 | 99 | |
|
91 | 100 | def linkrev(self, rev): |
|
92 | 101 | return self._revlog.linkrev(rev) |
|
93 | 102 | |
|
94 | 103 | def commonancestorsheads(self, node1, node2): |
|
95 | 104 | return self._revlog.commonancestorsheads(node1, node2) |
|
96 | 105 | |
|
97 | 106 | # Used by dagop.blockdescendants(). |
|
98 | 107 | def descendants(self, revs): |
|
99 | 108 | return self._revlog.descendants(revs) |
|
100 | 109 | |
|
101 | 110 | def heads(self, start=None, stop=None): |
|
102 | 111 | return self._revlog.heads(start, stop) |
|
103 | 112 | |
|
104 | 113 | # Used by hgweb, children extension. |
|
105 | 114 | def children(self, node): |
|
106 | 115 | return self._revlog.children(node) |
|
107 | 116 | |
|
108 | 117 | def iscensored(self, rev): |
|
109 | 118 | return self._revlog.iscensored(rev) |
|
110 | 119 | |
|
111 | 120 | def revision(self, node): |
|
112 | 121 | return self._revlog.revision(node) |
|
113 | 122 | |
|
114 | 123 | def rawdata(self, node): |
|
115 | 124 | return self._revlog.rawdata(node) |
|
116 | 125 | |
|
117 | 126 | def emitrevisions( |
|
118 | 127 | self, |
|
119 | 128 | nodes, |
|
120 | 129 | nodesorder=None, |
|
121 | 130 | revisiondata=False, |
|
122 | 131 | assumehaveparentrevisions=False, |
|
123 | 132 | deltamode=repository.CG_DELTAMODE_STD, |
|
124 | 133 | sidedata_helpers=None, |
|
125 | 134 | debug_info=None, |
|
126 | 135 | ): |
|
127 | 136 | return self._revlog.emitrevisions( |
|
128 | 137 | nodes, |
|
129 | 138 | nodesorder=nodesorder, |
|
130 | 139 | revisiondata=revisiondata, |
|
131 | 140 | assumehaveparentrevisions=assumehaveparentrevisions, |
|
132 | 141 | deltamode=deltamode, |
|
133 | 142 | sidedata_helpers=sidedata_helpers, |
|
134 | 143 | debug_info=debug_info, |
|
135 | 144 | ) |
|
136 | 145 | |
|
137 | 146 | def addrevision( |
|
138 | 147 | self, |
|
139 | 148 | revisiondata, |
|
140 | 149 | transaction, |
|
141 | 150 | linkrev, |
|
142 | 151 | p1, |
|
143 | 152 | p2, |
|
144 | 153 | node=None, |
|
145 | 154 | flags=revlog.REVIDX_DEFAULT_FLAGS, |
|
146 | 155 | cachedelta=None, |
|
147 | 156 | ): |
|
148 | 157 | return self._revlog.addrevision( |
|
149 | 158 | revisiondata, |
|
150 | 159 | transaction, |
|
151 | 160 | linkrev, |
|
152 | 161 | p1, |
|
153 | 162 | p2, |
|
154 | 163 | node=node, |
|
155 | 164 | flags=flags, |
|
156 | 165 | cachedelta=cachedelta, |
|
157 | 166 | ) |
|
158 | 167 | |
|
159 | 168 | def addgroup( |
|
160 | 169 | self, |
|
161 | 170 | deltas, |
|
162 | 171 | linkmapper, |
|
163 | 172 | transaction, |
|
164 | 173 | addrevisioncb=None, |
|
165 | 174 | duplicaterevisioncb=None, |
|
166 | 175 | maybemissingparents=False, |
|
167 | 176 | debug_info=None, |
|
168 | 177 | delta_base_reuse_policy=None, |
|
169 | 178 | ): |
|
170 | 179 | if maybemissingparents: |
|
171 | 180 | raise error.Abort( |
|
172 | 181 | _( |
|
173 | 182 | b'revlog storage does not support missing ' |
|
174 | 183 | b'parents write mode' |
|
175 | 184 | ) |
|
176 | 185 | ) |
|
177 | 186 | |
|
178 | 187 | with self._revlog._writing(transaction): |
|
179 | 188 | if self._fix_issue6528: |
|
180 | 189 | deltas = rewrite.filter_delta_issue6528(self._revlog, deltas) |
|
181 | 190 | |
|
182 | 191 | return self._revlog.addgroup( |
|
183 | 192 | deltas, |
|
184 | 193 | linkmapper, |
|
185 | 194 | transaction, |
|
186 | 195 | addrevisioncb=addrevisioncb, |
|
187 | 196 | duplicaterevisioncb=duplicaterevisioncb, |
|
188 | 197 | debug_info=debug_info, |
|
189 | 198 | delta_base_reuse_policy=delta_base_reuse_policy, |
|
190 | 199 | ) |
|
191 | 200 | |
|
192 | 201 | def getstrippoint(self, minlink): |
|
193 | 202 | return self._revlog.getstrippoint(minlink) |
|
194 | 203 | |
|
195 | 204 | def strip(self, minlink, transaction): |
|
196 | 205 | return self._revlog.strip(minlink, transaction) |
|
197 | 206 | |
|
198 | 207 | def censorrevision(self, tr, node, tombstone=b''): |
|
199 | 208 | return self._revlog.censorrevision(tr, node, tombstone=tombstone) |
|
200 | 209 | |
|
201 | 210 | def files(self): |
|
202 | 211 | return self._revlog.files() |
|
203 | 212 | |
|
204 | 213 | def read(self, node): |
|
205 | 214 | return storageutil.filtermetadata(self.revision(node)) |
|
206 | 215 | |
|
207 | 216 | def add(self, text, meta, transaction, link, p1=None, p2=None): |
|
208 | 217 | if meta or text.startswith(b'\1\n'): |
|
209 | 218 | text = storageutil.packmeta(meta, text) |
|
210 | 219 | rev = self.addrevision(text, transaction, link, p1, p2) |
|
211 | 220 | return self.node(rev) |
|
212 | 221 | |
|
213 | 222 | def renamed(self, node): |
|
214 | 223 | return storageutil.filerevisioncopied(self, node) |
|
215 | 224 | |
|
216 | 225 | def size(self, rev): |
|
217 | 226 | """return the size of a given revision""" |
|
218 | 227 | |
|
219 | 228 | # for revisions with renames, we have to go the slow way |
|
220 | 229 | node = self.node(rev) |
|
221 | 230 | if self.iscensored(rev): |
|
222 | 231 | return 0 |
|
223 | 232 | if self.renamed(node): |
|
224 | 233 | return len(self.read(node)) |
|
225 | 234 | |
|
226 | 235 | # XXX if self.read(node).startswith("\1\n"), this returns (size+4) |
|
227 | 236 | # XXX See also basefilectx.cmp. |
|
228 | 237 | return self._revlog.size(rev) |
|
229 | 238 | |
|
230 | 239 | def cmp(self, node, text): |
|
231 | 240 | """compare text with a given file revision |
|
232 | 241 | |
|
233 | 242 | returns True if text is different than what is stored. |
|
234 | 243 | """ |
|
235 | 244 | return not storageutil.filedataequivalent(self, node, text) |
|
236 | 245 | |
|
237 | def verifyintegrity(self, state): | |
|
246 | def verifyintegrity(self, state) -> Iterable[revlog.RevLogProblem]: | |
|
238 | 247 | return self._revlog.verifyintegrity(state) |
|
239 | 248 | |
|
240 | 249 | def storageinfo( |
|
241 | 250 | self, |
|
242 | 251 | exclusivefiles=False, |
|
243 | 252 | sharedfiles=False, |
|
244 | 253 | revisionscount=False, |
|
245 | 254 | trackedsize=False, |
|
246 | 255 | storedsize=False, |
|
247 | 256 | ): |
|
248 | 257 | return self._revlog.storageinfo( |
|
249 | 258 | exclusivefiles=exclusivefiles, |
|
250 | 259 | sharedfiles=sharedfiles, |
|
251 | 260 | revisionscount=revisionscount, |
|
252 | 261 | trackedsize=trackedsize, |
|
253 | 262 | storedsize=storedsize, |
|
254 | 263 | ) |
|
255 | 264 | |
|
256 | 265 | # Used by repo upgrade. |
|
257 | 266 | def clone(self, tr, destrevlog, **kwargs): |
|
258 | 267 | if not isinstance(destrevlog, filelog): |
|
259 | 268 | msg = b'expected filelog to clone(), not %r' |
|
260 | 269 | msg %= destrevlog |
|
261 | 270 | raise error.ProgrammingError(msg) |
|
262 | 271 | |
|
263 | 272 | return self._revlog.clone(tr, destrevlog._revlog, **kwargs) |
|
264 | 273 | |
|
265 | 274 | |
|
266 | 275 | filelog = interfaceutil.implementer(repository.ifilestorage)(FileLog) |
|
267 | 276 | |
|
268 | 277 | if typing.TYPE_CHECKING: |
|
269 | 278 | filelog = FileLog |
|
270 | 279 | |
|
271 | 280 | |
|
272 | 281 | class narrowfilelog(filelog): |
|
273 | 282 | """Filelog variation to be used with narrow stores.""" |
|
274 | 283 | |
|
275 | 284 | def __init__(self, opener, path, narrowmatch, try_split=False): |
|
276 | 285 | super(narrowfilelog, self).__init__(opener, path, try_split=try_split) |
|
277 | 286 | self._narrowmatch = narrowmatch |
|
278 | 287 | |
|
279 | 288 | def renamed(self, node): |
|
280 | 289 | res = super(narrowfilelog, self).renamed(node) |
|
281 | 290 | |
|
282 | 291 | # Renames that come from outside the narrowspec are problematic |
|
283 | 292 | # because we may lack the base text for the rename. This can result |
|
284 | 293 | # in code attempting to walk the ancestry or compute a diff |
|
285 | 294 | # encountering a missing revision. We address this by silently |
|
286 | 295 | # removing rename metadata if the source file is outside the |
|
287 | 296 | # narrow spec. |
|
288 | 297 | # |
|
289 | 298 | # A better solution would be to see if the base revision is available, |
|
290 | 299 | # rather than assuming it isn't. |
|
291 | 300 | # |
|
292 | 301 | # An even better solution would be to teach all consumers of rename |
|
293 | 302 | # metadata that the base revision may not be available. |
|
294 | 303 | # |
|
295 | 304 | # TODO consider better ways of doing this. |
|
296 | 305 | if res and not self._narrowmatch(res[0]): |
|
297 | 306 | return None |
|
298 | 307 | |
|
299 | 308 | return res |
|
300 | 309 | |
|
301 | 310 | def size(self, rev): |
|
302 | 311 | # Because we have a custom renamed() that may lie, we need to call |
|
303 | 312 | # the base renamed() to report accurate results. |
|
304 | 313 | node = self.node(rev) |
|
305 | 314 | if super(narrowfilelog, self).renamed(node): |
|
306 | 315 | return len(self.read(node)) |
|
307 | 316 | else: |
|
308 | 317 | return super(narrowfilelog, self).size(rev) |
|
309 | 318 | |
|
310 | 319 | def cmp(self, node, text): |
|
311 | 320 | # We don't call `super` because narrow parents can be buggy in case of a |
|
312 | 321 | # ambiguous dirstate. Always take the slow path until there is a better |
|
313 | 322 | # fix, see issue6150. |
|
314 | 323 | |
|
315 | 324 | # Censored files compare against the empty file. |
|
316 | 325 | if self.iscensored(self.rev(node)): |
|
317 | 326 | return text != b'' |
|
318 | 327 | |
|
319 | 328 | return self.read(node) != text |
@@ -1,2780 +1,2797 b'' | |||
|
1 | 1 | # manifest.py - manifest revision class for mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2005-2007 Olivia Mackall <olivia@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | |
|
9 | 9 | import heapq |
|
10 | 10 | import itertools |
|
11 | 11 | import struct |
|
12 | 12 | import typing |
|
13 | 13 | import weakref |
|
14 | 14 | |
|
15 | 15 | from typing import ( |
|
16 | 16 | ByteString, |
|
17 | 17 | Callable, |
|
18 | 18 | Collection, |
|
19 | 19 | Dict, |
|
20 | 20 | Iterable, |
|
21 | 21 | Iterator, |
|
22 | 22 | List, |
|
23 | 23 | Optional, |
|
24 | 24 | Set, |
|
25 | 25 | Tuple, |
|
26 | 26 | Union, |
|
27 | 27 | cast, |
|
28 | 28 | ) |
|
29 | 29 | |
|
30 | 30 | from .i18n import _ |
|
31 | 31 | from .node import ( |
|
32 | 32 | bin, |
|
33 | 33 | hex, |
|
34 | 34 | nullrev, |
|
35 | 35 | ) |
|
36 | 36 | from . import ( |
|
37 | 37 | encoding, |
|
38 | 38 | error, |
|
39 | 39 | match as matchmod, |
|
40 | 40 | mdiff, |
|
41 | 41 | pathutil, |
|
42 | 42 | policy, |
|
43 | 43 | pycompat, |
|
44 | 44 | revlog, |
|
45 | 45 | util, |
|
46 | 46 | ) |
|
47 | 47 | from .interfaces import ( |
|
48 | 48 | repository, |
|
49 | 49 | util as interfaceutil, |
|
50 | 50 | ) |
|
51 | 51 | from .revlogutils import ( |
|
52 | 52 | constants as revlog_constants, |
|
53 | 53 | ) |
|
54 | 54 | |
|
55 | 55 | parsers = policy.importmod('parsers') |
|
56 | 56 | propertycache = util.propertycache |
|
57 | 57 | |
|
58 | 58 | # Allow tests to more easily test the alternate path in manifestdict.fastdelta() |
|
59 | 59 | FASTDELTA_TEXTDIFF_THRESHOLD = 1000 |
|
60 | 60 | |
|
61 | 61 | |
|
62 | 62 | def _parse(nodelen, data: bytes): |
|
63 | 63 | # This method does a little bit of excessive-looking |
|
64 | 64 | # precondition checking. This is so that the behavior of this |
|
65 | 65 | # class exactly matches its C counterpart to try and help |
|
66 | 66 | # prevent surprise breakage for anyone that develops against |
|
67 | 67 | # the pure version. |
|
68 | 68 | if data and data[-1:] != b'\n': |
|
69 | 69 | raise ValueError(b'Manifest did not end in a newline.') |
|
70 | 70 | prev = None |
|
71 | 71 | for l in data.splitlines(): |
|
72 | 72 | if prev is not None and prev > l: |
|
73 | 73 | raise ValueError(b'Manifest lines not in sorted order.') |
|
74 | 74 | prev = l |
|
75 | 75 | f, n = l.split(b'\0') |
|
76 | 76 | nl = len(n) |
|
77 | 77 | flags = n[-1:] |
|
78 | 78 | if flags in _manifestflags: |
|
79 | 79 | n = n[:-1] |
|
80 | 80 | nl -= 1 |
|
81 | 81 | else: |
|
82 | 82 | flags = b'' |
|
83 | 83 | if nl != 2 * nodelen: |
|
84 | 84 | raise ValueError(b'Invalid manifest line') |
|
85 | 85 | |
|
86 | 86 | yield f, bin(n), flags |
|
87 | 87 | |
|
88 | 88 | |
|
89 | 89 | def _text(it): |
|
90 | 90 | files = [] |
|
91 | 91 | lines = [] |
|
92 | 92 | for f, n, fl in it: |
|
93 | 93 | files.append(f) |
|
94 | 94 | # if this is changed to support newlines in filenames, |
|
95 | 95 | # be sure to check the templates/ dir again (especially *-raw.tmpl) |
|
96 | 96 | lines.append(b"%s\0%s%s\n" % (f, hex(n), fl)) |
|
97 | 97 | |
|
98 | 98 | _checkforbidden(files) |
|
99 | 99 | return b''.join(lines) |
|
100 | 100 | |
|
101 | 101 | |
|
102 | 102 | class lazymanifestiter: |
|
103 | 103 | def __init__(self, lm: '_LazyManifest') -> None: |
|
104 | 104 | self.pos = 0 |
|
105 | 105 | self.lm = lm |
|
106 | 106 | |
|
107 | 107 | def __iter__(self) -> 'lazymanifestiter': |
|
108 | 108 | return self |
|
109 | 109 | |
|
110 | 110 | def next(self) -> bytes: |
|
111 | 111 | try: |
|
112 | 112 | data, pos = self.lm._get(self.pos) |
|
113 | 113 | except IndexError: |
|
114 | 114 | raise StopIteration |
|
115 | 115 | if pos == -1: |
|
116 | 116 | assert isinstance(data, tuple) |
|
117 | 117 | self.pos += 1 |
|
118 | 118 | return data[0] |
|
119 | 119 | assert isinstance(data, bytes) |
|
120 | 120 | self.pos += 1 |
|
121 | 121 | zeropos = data.find(b'\x00', pos) |
|
122 | 122 | return data[pos:zeropos] |
|
123 | 123 | |
|
124 | 124 | __next__ = next |
|
125 | 125 | |
|
126 | 126 | |
|
127 | 127 | class lazymanifestiterentries: |
|
128 | 128 | def __init__(self, lm: '_LazyManifest') -> None: |
|
129 | 129 | self.lm = lm |
|
130 | 130 | self.pos = 0 |
|
131 | 131 | |
|
132 | 132 | def __iter__(self) -> 'lazymanifestiterentries': |
|
133 | 133 | return self |
|
134 | 134 | |
|
135 | 135 | def next(self) -> Tuple[bytes, bytes, bytes]: |
|
136 | 136 | try: |
|
137 | 137 | data, pos = self.lm._get(self.pos) |
|
138 | 138 | except IndexError: |
|
139 | 139 | raise StopIteration |
|
140 | 140 | if pos == -1: |
|
141 | 141 | assert isinstance(data, tuple) |
|
142 | 142 | self.pos += 1 |
|
143 | 143 | return data |
|
144 | 144 | assert isinstance(data, bytes) |
|
145 | 145 | zeropos = data.find(b'\x00', pos) |
|
146 | 146 | nlpos = data.find(b'\n', pos) |
|
147 | 147 | if zeropos == -1 or nlpos == -1 or nlpos < zeropos: |
|
148 | 148 | raise error.StorageError(b'Invalid manifest line') |
|
149 | 149 | flags = data[nlpos - 1 : nlpos] |
|
150 | 150 | if flags in _manifestflags: |
|
151 | 151 | hlen = nlpos - zeropos - 2 |
|
152 | 152 | else: |
|
153 | 153 | hlen = nlpos - zeropos - 1 |
|
154 | 154 | flags = b'' |
|
155 | 155 | if hlen != 2 * self.lm._nodelen: |
|
156 | 156 | raise error.StorageError(b'Invalid manifest line') |
|
157 | 157 | hashval = unhexlify( |
|
158 | 158 | data, self.lm.extrainfo[self.pos], zeropos + 1, hlen |
|
159 | 159 | ) |
|
160 | 160 | self.pos += 1 |
|
161 | 161 | return (data[pos:zeropos], hashval, flags) |
|
162 | 162 | |
|
163 | 163 | __next__ = next |
|
164 | 164 | |
|
165 | 165 | |
|
166 | 166 | def unhexlify(data: bytes, extra: int, pos, length: int): |
|
167 | 167 | s = bin(data[pos : pos + length]) |
|
168 | 168 | if extra: |
|
169 | 169 | s += bytes([extra & 0xFF]) |
|
170 | 170 | return s |
|
171 | 171 | |
|
172 | 172 | |
|
173 | 173 | def _cmp(a, b): |
|
174 | 174 | return (a > b) - (a < b) |
|
175 | 175 | |
|
176 | 176 | |
|
177 | 177 | _manifestflags = {b'', b'l', b't', b'x'} |
|
178 | 178 | |
|
179 | 179 | |
|
180 | 180 | class _LazyManifest: |
|
181 | 181 | """A pure python manifest backed by a byte string. It is supplimented with |
|
182 | 182 | internal lists as it is modified, until it is compacted back to a pure byte |
|
183 | 183 | string. |
|
184 | 184 | |
|
185 | 185 | ``data`` is the initial manifest data. |
|
186 | 186 | |
|
187 | 187 | ``positions`` is a list of offsets, one per manifest entry. Positive |
|
188 | 188 | values are offsets into ``data``, negative values are offsets into the |
|
189 | 189 | ``extradata`` list. When an entry is removed, its entry is dropped from |
|
190 | 190 | ``positions``. The values are encoded such that when walking the list and |
|
191 | 191 | indexing into ``data`` or ``extradata`` as appropriate, the entries are |
|
192 | 192 | sorted by filename. |
|
193 | 193 | |
|
194 | 194 | ``extradata`` is a list of (key, hash, flags) for entries that were added or |
|
195 | 195 | modified since the manifest was created or compacted. |
|
196 | 196 | """ |
|
197 | 197 | |
|
198 | 198 | def __init__( |
|
199 | 199 | self, |
|
200 | 200 | nodelen: int, |
|
201 | 201 | data: bytes, |
|
202 | 202 | positions=None, |
|
203 | 203 | extrainfo=None, |
|
204 | 204 | extradata=None, |
|
205 | 205 | hasremovals: bool = False, |
|
206 | 206 | ): |
|
207 | 207 | self._nodelen = nodelen |
|
208 | 208 | if positions is None: |
|
209 | 209 | self.positions = self.findlines(data) |
|
210 | 210 | self.extrainfo = [0] * len(self.positions) |
|
211 | 211 | self.data = data |
|
212 | 212 | self.extradata = [] |
|
213 | 213 | self.hasremovals = False |
|
214 | 214 | else: |
|
215 | 215 | self.positions = positions[:] |
|
216 | 216 | self.extrainfo = extrainfo[:] |
|
217 | 217 | self.extradata = extradata[:] |
|
218 | 218 | self.data = data |
|
219 | 219 | self.hasremovals = hasremovals |
|
220 | 220 | |
|
221 | 221 | def findlines(self, data: bytes) -> List[int]: |
|
222 | 222 | if not data: |
|
223 | 223 | return [] |
|
224 | 224 | pos = data.find(b"\n") |
|
225 | 225 | if pos == -1 or data[-1:] != b'\n': |
|
226 | 226 | raise ValueError(b"Manifest did not end in a newline.") |
|
227 | 227 | positions = [0] |
|
228 | 228 | prev = data[: data.find(b'\x00')] |
|
229 | 229 | while pos < len(data) - 1 and pos != -1: |
|
230 | 230 | positions.append(pos + 1) |
|
231 | 231 | nexts = data[pos + 1 : data.find(b'\x00', pos + 1)] |
|
232 | 232 | if nexts < prev: |
|
233 | 233 | raise ValueError(b"Manifest lines not in sorted order.") |
|
234 | 234 | prev = nexts |
|
235 | 235 | pos = data.find(b"\n", pos + 1) |
|
236 | 236 | return positions |
|
237 | 237 | |
|
238 | 238 | def _get( |
|
239 | 239 | self, index: int |
|
240 | 240 | ) -> Tuple[Union[bytes, Tuple[bytes, bytes, bytes]], int]: |
|
241 | 241 | # get the position encoded in pos: |
|
242 | 242 | # positive number is an index in 'data' |
|
243 | 243 | # negative number is in extrapieces |
|
244 | 244 | pos = self.positions[index] |
|
245 | 245 | if pos >= 0: |
|
246 | 246 | return self.data, pos |
|
247 | 247 | return self.extradata[-pos - 1], -1 |
|
248 | 248 | |
|
249 | 249 | def _getkey(self, pos) -> bytes: |
|
250 | 250 | if pos >= 0: |
|
251 | 251 | return self.data[pos : self.data.find(b'\x00', pos + 1)] |
|
252 | 252 | return self.extradata[-pos - 1][0] |
|
253 | 253 | |
|
254 | 254 | def bsearch(self, key: bytes) -> int: |
|
255 | 255 | first = 0 |
|
256 | 256 | last = len(self.positions) - 1 |
|
257 | 257 | |
|
258 | 258 | while first <= last: |
|
259 | 259 | midpoint = (first + last) // 2 |
|
260 | 260 | nextpos = self.positions[midpoint] |
|
261 | 261 | candidate = self._getkey(nextpos) |
|
262 | 262 | r = _cmp(key, candidate) |
|
263 | 263 | if r == 0: |
|
264 | 264 | return midpoint |
|
265 | 265 | else: |
|
266 | 266 | if r < 0: |
|
267 | 267 | last = midpoint - 1 |
|
268 | 268 | else: |
|
269 | 269 | first = midpoint + 1 |
|
270 | 270 | return -1 |
|
271 | 271 | |
|
272 | 272 | def bsearch2(self, key: bytes) -> Tuple[int, bool]: |
|
273 | 273 | # same as the above, but will always return the position |
|
274 | 274 | # done for performance reasons |
|
275 | 275 | first = 0 |
|
276 | 276 | last = len(self.positions) - 1 |
|
277 | 277 | |
|
278 | 278 | while first <= last: |
|
279 | 279 | midpoint = (first + last) // 2 |
|
280 | 280 | nextpos = self.positions[midpoint] |
|
281 | 281 | candidate = self._getkey(nextpos) |
|
282 | 282 | r = _cmp(key, candidate) |
|
283 | 283 | if r == 0: |
|
284 | 284 | return (midpoint, True) |
|
285 | 285 | else: |
|
286 | 286 | if r < 0: |
|
287 | 287 | last = midpoint - 1 |
|
288 | 288 | else: |
|
289 | 289 | first = midpoint + 1 |
|
290 | 290 | return (first, False) |
|
291 | 291 | |
|
292 | 292 | def __contains__(self, key: bytes) -> bool: |
|
293 | 293 | return self.bsearch(key) != -1 |
|
294 | 294 | |
|
295 | 295 | def __getitem__(self, key: bytes) -> Tuple[bytes, bytes]: |
|
296 | 296 | if not isinstance(key, bytes): |
|
297 | 297 | raise TypeError(b"getitem: manifest keys must be a bytes.") |
|
298 | 298 | needle = self.bsearch(key) |
|
299 | 299 | if needle == -1: |
|
300 | 300 | raise KeyError |
|
301 | 301 | data, pos = self._get(needle) |
|
302 | 302 | if pos == -1: |
|
303 | 303 | assert isinstance(data, tuple) |
|
304 | 304 | return (data[1], data[2]) |
|
305 | 305 | |
|
306 | 306 | assert isinstance(data, bytes) |
|
307 | 307 | zeropos = data.find(b'\x00', pos) |
|
308 | 308 | nlpos = data.find(b'\n', zeropos) |
|
309 | 309 | assert 0 <= needle <= len(self.positions) |
|
310 | 310 | assert len(self.extrainfo) == len(self.positions) |
|
311 | 311 | if zeropos == -1 or nlpos == -1 or nlpos < zeropos: |
|
312 | 312 | raise error.StorageError(b'Invalid manifest line') |
|
313 | 313 | hlen = nlpos - zeropos - 1 |
|
314 | 314 | flags = data[nlpos - 1 : nlpos] |
|
315 | 315 | if flags in _manifestflags: |
|
316 | 316 | hlen -= 1 |
|
317 | 317 | else: |
|
318 | 318 | flags = b'' |
|
319 | 319 | if hlen != 2 * self._nodelen: |
|
320 | 320 | raise error.StorageError(b'Invalid manifest line') |
|
321 | 321 | hashval = unhexlify(data, self.extrainfo[needle], zeropos + 1, hlen) |
|
322 | 322 | return (hashval, flags) |
|
323 | 323 | |
|
324 | 324 | def __delitem__(self, key: bytes) -> None: |
|
325 | 325 | needle, found = self.bsearch2(key) |
|
326 | 326 | if not found: |
|
327 | 327 | raise KeyError |
|
328 | 328 | cur = self.positions[needle] |
|
329 | 329 | self.positions = self.positions[:needle] + self.positions[needle + 1 :] |
|
330 | 330 | self.extrainfo = self.extrainfo[:needle] + self.extrainfo[needle + 1 :] |
|
331 | 331 | if cur >= 0: |
|
332 | 332 | # This does NOT unsort the list as far as the search functions are |
|
333 | 333 | # concerned, as they only examine lines mapped by self.positions. |
|
334 | 334 | self.data = self.data[:cur] + b'\x00' + self.data[cur + 1 :] |
|
335 | 335 | self.hasremovals = True |
|
336 | 336 | |
|
337 | 337 | def __setitem__(self, key: bytes, value: Tuple[bytes, bytes]): |
|
338 | 338 | if not isinstance(key, bytes): |
|
339 | 339 | raise TypeError(b"setitem: manifest keys must be a byte string.") |
|
340 | 340 | if not isinstance(value, tuple) or len(value) != 2: |
|
341 | 341 | raise TypeError( |
|
342 | 342 | b"Manifest values must be a tuple of (node, flags)." |
|
343 | 343 | ) |
|
344 | 344 | hashval = value[0] |
|
345 | 345 | if not isinstance(hashval, bytes) or len(hashval) not in (20, 32): |
|
346 | 346 | raise TypeError(b"node must be a 20-byte or 32-byte byte string") |
|
347 | 347 | flags = value[1] |
|
348 | 348 | if not isinstance(flags, bytes) or len(flags) > 1: |
|
349 | 349 | raise TypeError(b"flags must a 0 or 1 byte string, got %r", flags) |
|
350 | 350 | needle, found = self.bsearch2(key) |
|
351 | 351 | if found: |
|
352 | 352 | # put the item |
|
353 | 353 | pos = self.positions[needle] |
|
354 | 354 | if pos < 0: |
|
355 | 355 | self.extradata[-pos - 1] = (key, hashval, value[1]) |
|
356 | 356 | else: |
|
357 | 357 | # just don't bother |
|
358 | 358 | self.extradata.append((key, hashval, value[1])) |
|
359 | 359 | self.positions[needle] = -len(self.extradata) |
|
360 | 360 | else: |
|
361 | 361 | # not found, put it in with extra positions |
|
362 | 362 | self.extradata.append((key, hashval, value[1])) |
|
363 | 363 | self.positions = ( |
|
364 | 364 | self.positions[:needle] |
|
365 | 365 | + [-len(self.extradata)] |
|
366 | 366 | + self.positions[needle:] |
|
367 | 367 | ) |
|
368 | 368 | self.extrainfo = ( |
|
369 | 369 | self.extrainfo[:needle] + [0] + self.extrainfo[needle:] |
|
370 | 370 | ) |
|
371 | 371 | |
|
372 | 372 | def copy(self) -> '_LazyManifest': |
|
373 | 373 | # XXX call _compact like in C? |
|
374 | 374 | return _lazymanifest( |
|
375 | 375 | self._nodelen, |
|
376 | 376 | self.data, |
|
377 | 377 | self.positions, |
|
378 | 378 | self.extrainfo, |
|
379 | 379 | self.extradata, |
|
380 | 380 | self.hasremovals, |
|
381 | 381 | ) |
|
382 | 382 | |
|
383 | 383 | def _compact(self) -> None: |
|
384 | 384 | # hopefully not called TOO often |
|
385 | 385 | if len(self.extradata) == 0 and not self.hasremovals: |
|
386 | 386 | return |
|
387 | 387 | l = [] |
|
388 | 388 | i = 0 |
|
389 | 389 | offset = 0 |
|
390 | 390 | self.extrainfo = [0] * len(self.positions) |
|
391 | 391 | while i < len(self.positions): |
|
392 | 392 | if self.positions[i] >= 0: |
|
393 | 393 | cur = self.positions[i] |
|
394 | 394 | last_cut = cur |
|
395 | 395 | |
|
396 | 396 | # Collect all contiguous entries in the buffer at the current |
|
397 | 397 | # offset, breaking out only for added/modified items held in |
|
398 | 398 | # extradata, or a deleted line prior to the next position. |
|
399 | 399 | while True: |
|
400 | 400 | self.positions[i] = offset |
|
401 | 401 | i += 1 |
|
402 | 402 | if i == len(self.positions) or self.positions[i] < 0: |
|
403 | 403 | break |
|
404 | 404 | |
|
405 | 405 | # A removed file has no positions[] entry, but does have an |
|
406 | 406 | # overwritten first byte. Break out and find the end of the |
|
407 | 407 | # current good entry/entries if there is a removed file |
|
408 | 408 | # before the next position. |
|
409 | 409 | if ( |
|
410 | 410 | self.hasremovals |
|
411 | 411 | and self.data.find(b'\n\x00', cur, self.positions[i]) |
|
412 | 412 | != -1 |
|
413 | 413 | ): |
|
414 | 414 | break |
|
415 | 415 | |
|
416 | 416 | offset += self.positions[i] - cur |
|
417 | 417 | cur = self.positions[i] |
|
418 | 418 | end_cut = self.data.find(b'\n', cur) |
|
419 | 419 | if end_cut != -1: |
|
420 | 420 | end_cut += 1 |
|
421 | 421 | offset += end_cut - cur |
|
422 | 422 | l.append(self.data[last_cut:end_cut]) |
|
423 | 423 | else: |
|
424 | 424 | while i < len(self.positions) and self.positions[i] < 0: |
|
425 | 425 | cur = self.positions[i] |
|
426 | 426 | t = self.extradata[-cur - 1] |
|
427 | 427 | l.append(self._pack(t)) |
|
428 | 428 | self.positions[i] = offset |
|
429 | 429 | # Hashes are either 20 bytes (old sha1s) or 32 |
|
430 | 430 | # bytes (new non-sha1). |
|
431 | 431 | hlen = 20 |
|
432 | 432 | if len(t[1]) > 25: |
|
433 | 433 | hlen = 32 |
|
434 | 434 | if len(t[1]) > hlen: |
|
435 | 435 | self.extrainfo[i] = ord(t[1][hlen + 1]) |
|
436 | 436 | offset += len(l[-1]) |
|
437 | 437 | i += 1 |
|
438 | 438 | self.data = b''.join(l) |
|
439 | 439 | self.hasremovals = False |
|
440 | 440 | self.extradata = [] |
|
441 | 441 | |
|
442 | 442 | def _pack(self, d: Tuple[bytes, bytes, bytes]) -> bytes: |
|
443 | 443 | n = d[1] |
|
444 | 444 | assert len(n) in (20, 32) |
|
445 | 445 | return d[0] + b'\x00' + hex(n) + d[2] + b'\n' |
|
446 | 446 | |
|
447 | 447 | def text(self) -> ByteString: |
|
448 | 448 | self._compact() |
|
449 | 449 | return self.data |
|
450 | 450 | |
|
451 | 451 | def diff( |
|
452 | 452 | self, m2: '_LazyManifest', clean: bool = False |
|
453 | 453 | ) -> Dict[ |
|
454 | 454 | bytes, |
|
455 | 455 | Optional[ |
|
456 | 456 | Tuple[Tuple[Optional[bytes], bytes], Tuple[Optional[bytes], bytes]] |
|
457 | 457 | ], |
|
458 | 458 | ]: |
|
459 | 459 | '''Finds changes between the current manifest and m2.''' |
|
460 | 460 | # XXX think whether efficiency matters here |
|
461 | 461 | diff = {} |
|
462 | 462 | |
|
463 | 463 | for fn, e1, flags in self.iterentries(): |
|
464 | 464 | if fn not in m2: |
|
465 | 465 | diff[fn] = (e1, flags), (None, b'') |
|
466 | 466 | else: |
|
467 | 467 | e2 = m2[fn] |
|
468 | 468 | if (e1, flags) != e2: |
|
469 | 469 | diff[fn] = (e1, flags), e2 |
|
470 | 470 | elif clean: |
|
471 | 471 | diff[fn] = None |
|
472 | 472 | |
|
473 | 473 | for fn, e2, flags in m2.iterentries(): |
|
474 | 474 | if fn not in self: |
|
475 | 475 | diff[fn] = (None, b''), (e2, flags) |
|
476 | 476 | |
|
477 | 477 | return diff |
|
478 | 478 | |
|
479 | 479 | def iterentries(self) -> lazymanifestiterentries: |
|
480 | 480 | return lazymanifestiterentries(self) |
|
481 | 481 | |
|
482 | 482 | def iterkeys(self) -> lazymanifestiter: |
|
483 | 483 | return lazymanifestiter(self) |
|
484 | 484 | |
|
485 | 485 | def __iter__(self) -> lazymanifestiter: |
|
486 | 486 | return lazymanifestiter(self) |
|
487 | 487 | |
|
488 | 488 | def __len__(self) -> int: |
|
489 | 489 | return len(self.positions) |
|
490 | 490 | |
|
491 | 491 | def filtercopy(self, filterfn: Callable[[bytes], bool]) -> '_LazyManifest': |
|
492 | 492 | # XXX should be optimized |
|
493 | 493 | c = _lazymanifest(self._nodelen, b'') |
|
494 | 494 | for f, n, fl in self.iterentries(): |
|
495 | 495 | if filterfn(f): |
|
496 | 496 | c[f] = n, fl |
|
497 | 497 | return c |
|
498 | 498 | |
|
499 | 499 | |
|
500 | 500 | try: |
|
501 | 501 | _lazymanifest = parsers.lazymanifest |
|
502 | 502 | except AttributeError: |
|
503 | 503 | _lazymanifest = _LazyManifest |
|
504 | 504 | |
|
505 | 505 | |
|
506 | 506 | class ManifestDict: |
|
507 | 507 | def __init__(self, nodelen: int, data: ByteString = b''): |
|
508 | 508 | self._nodelen = nodelen |
|
509 | 509 | self._lm = _lazymanifest(nodelen, data) |
|
510 | 510 | |
|
511 | 511 | def __getitem__(self, key: bytes) -> bytes: |
|
512 | 512 | return self._lm[key][0] |
|
513 | 513 | |
|
514 | 514 | def find(self, key: bytes) -> Tuple[bytes, bytes]: |
|
515 | 515 | return self._lm[key] |
|
516 | 516 | |
|
517 | 517 | def __len__(self) -> int: |
|
518 | 518 | return len(self._lm) |
|
519 | 519 | |
|
520 | 520 | def __nonzero__(self) -> bool: |
|
521 | 521 | # nonzero is covered by the __len__ function, but implementing it here |
|
522 | 522 | # makes it easier for extensions to override. |
|
523 | 523 | return len(self._lm) != 0 |
|
524 | 524 | |
|
525 | 525 | __bool__ = __nonzero__ |
|
526 | 526 | |
|
527 | 527 | def set(self, key: bytes, node: bytes, flags: bytes) -> None: |
|
528 | 528 | self._lm[key] = node, flags |
|
529 | 529 | |
|
530 | 530 | def __setitem__(self, key: bytes, node: bytes) -> None: |
|
531 | 531 | self._lm[key] = node, self.flags(key) |
|
532 | 532 | |
|
533 | 533 | def __contains__(self, key: bytes) -> bool: |
|
534 | 534 | if key is None: |
|
535 | 535 | return False |
|
536 | 536 | return key in self._lm |
|
537 | 537 | |
|
538 | 538 | def __delitem__(self, key: bytes) -> None: |
|
539 | 539 | del self._lm[key] |
|
540 | 540 | |
|
541 | 541 | def __iter__(self) -> Iterator[bytes]: |
|
542 | 542 | return self._lm.__iter__() |
|
543 | 543 | |
|
544 | 544 | def iterkeys(self) -> Iterator[bytes]: |
|
545 | 545 | return self._lm.iterkeys() |
|
546 | 546 | |
|
547 | 547 | def keys(self) -> List[bytes]: |
|
548 | 548 | return list(self.iterkeys()) |
|
549 | 549 | |
|
550 | 550 | def filesnotin(self, m2, match=None) -> Set[bytes]: |
|
551 | 551 | '''Set of files in this manifest that are not in the other''' |
|
552 | 552 | if match is not None: |
|
553 | 553 | match = matchmod.badmatch(match, lambda path, msg: None) |
|
554 | 554 | sm2 = set(m2.walk(match)) |
|
555 | 555 | return {f for f in self.walk(match) if f not in sm2} |
|
556 | 556 | return {f for f in self if f not in m2} |
|
557 | 557 | |
|
558 | 558 | @propertycache |
|
559 | 559 | def _dirs(self) -> pathutil.dirs: |
|
560 | 560 | return pathutil.dirs(self) |
|
561 | 561 | |
|
562 | 562 | def dirs(self) -> pathutil.dirs: |
|
563 | 563 | return self._dirs |
|
564 | 564 | |
|
565 | 565 | def hasdir(self, dir: bytes) -> bool: |
|
566 | 566 | return dir in self._dirs |
|
567 | 567 | |
|
568 | 568 | def _filesfastpath(self, match: matchmod.basematcher) -> bool: |
|
569 | 569 | """Checks whether we can correctly and quickly iterate over matcher |
|
570 | 570 | files instead of over manifest files.""" |
|
571 | 571 | files = match.files() |
|
572 | 572 | return len(files) < 100 and ( |
|
573 | 573 | match.isexact() |
|
574 | 574 | or (match.prefix() and all(fn in self for fn in files)) |
|
575 | 575 | ) |
|
576 | 576 | |
|
577 | 577 | def walk(self, match: matchmod.basematcher) -> Iterator[bytes]: |
|
578 | 578 | """Generates matching file names. |
|
579 | 579 | |
|
580 | 580 | Equivalent to manifest.matches(match).iterkeys(), but without creating |
|
581 | 581 | an entirely new manifest. |
|
582 | 582 | |
|
583 | 583 | It also reports nonexistent files by marking them bad with match.bad(). |
|
584 | 584 | """ |
|
585 | 585 | if match.always(): |
|
586 | 586 | for f in iter(self): |
|
587 | 587 | yield f |
|
588 | 588 | return |
|
589 | 589 | |
|
590 | 590 | fset = set(match.files()) |
|
591 | 591 | |
|
592 | 592 | # avoid the entire walk if we're only looking for specific files |
|
593 | 593 | if self._filesfastpath(match): |
|
594 | 594 | for fn in sorted(fset): |
|
595 | 595 | if fn in self: |
|
596 | 596 | yield fn |
|
597 | 597 | return |
|
598 | 598 | |
|
599 | 599 | for fn in self: |
|
600 | 600 | if fn in fset: |
|
601 | 601 | # specified pattern is the exact name |
|
602 | 602 | fset.remove(fn) |
|
603 | 603 | if match(fn): |
|
604 | 604 | yield fn |
|
605 | 605 | |
|
606 | 606 | # for dirstate.walk, files=[''] means "walk the whole tree". |
|
607 | 607 | # follow that here, too |
|
608 | 608 | fset.discard(b'') |
|
609 | 609 | |
|
610 | 610 | for fn in sorted(fset): |
|
611 | 611 | if not self.hasdir(fn): |
|
612 | 612 | match.bad(fn, None) |
|
613 | 613 | |
|
614 | 614 | def _matches(self, match: matchmod.basematcher) -> 'ManifestDict': |
|
615 | 615 | '''generate a new manifest filtered by the match argument''' |
|
616 | 616 | if match.always(): |
|
617 | 617 | return self.copy() |
|
618 | 618 | |
|
619 | 619 | if self._filesfastpath(match): |
|
620 | 620 | m = manifestdict(self._nodelen) |
|
621 | 621 | lm = self._lm |
|
622 | 622 | for fn in match.files(): |
|
623 | 623 | if fn in lm: |
|
624 | 624 | m._lm[fn] = lm[fn] |
|
625 | 625 | return m |
|
626 | 626 | |
|
627 | 627 | m = manifestdict(self._nodelen) |
|
628 | 628 | m._lm = self._lm.filtercopy(match) |
|
629 | 629 | return m |
|
630 | 630 | |
|
631 | 631 | def diff( |
|
632 | 632 | self, |
|
633 | 633 | m2: 'ManifestDict', |
|
634 | 634 | match: Optional[matchmod.basematcher] = None, |
|
635 | 635 | clean: bool = False, |
|
636 | 636 | ) -> Dict[ |
|
637 | 637 | bytes, |
|
638 | 638 | Optional[ |
|
639 | 639 | Tuple[Tuple[Optional[bytes], bytes], Tuple[Optional[bytes], bytes]] |
|
640 | 640 | ], |
|
641 | 641 | ]: |
|
642 | 642 | """Finds changes between the current manifest and m2. |
|
643 | 643 | |
|
644 | 644 | Args: |
|
645 | 645 | m2: the manifest to which this manifest should be compared. |
|
646 | 646 | clean: if true, include files unchanged between these manifests |
|
647 | 647 | with a None value in the returned dictionary. |
|
648 | 648 | |
|
649 | 649 | The result is returned as a dict with filename as key and |
|
650 | 650 | values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the |
|
651 | 651 | nodeid in the current/other manifest and fl1/fl2 is the flag |
|
652 | 652 | in the current/other manifest. Where the file does not exist, |
|
653 | 653 | the nodeid will be None and the flags will be the empty |
|
654 | 654 | string. |
|
655 | 655 | """ |
|
656 | 656 | if match: |
|
657 | 657 | m1 = self._matches(match) |
|
658 | 658 | m2 = m2._matches(match) |
|
659 | 659 | return m1.diff(m2, clean=clean) |
|
660 | 660 | return self._lm.diff(m2._lm, clean) |
|
661 | 661 | |
|
662 | 662 | def setflag(self, key: bytes, flag: bytes) -> None: |
|
663 | 663 | if flag not in _manifestflags: |
|
664 | 664 | raise TypeError(b"Invalid manifest flag set.") |
|
665 | 665 | self._lm[key] = self[key], flag |
|
666 | 666 | |
|
667 | 667 | def get(self, key: bytes, default=None) -> Optional[bytes]: |
|
668 | 668 | try: |
|
669 | 669 | return self._lm[key][0] |
|
670 | 670 | except KeyError: |
|
671 | 671 | return default |
|
672 | 672 | |
|
673 | 673 | def flags(self, key: bytes) -> bytes: |
|
674 | 674 | try: |
|
675 | 675 | return self._lm[key][1] |
|
676 | 676 | except KeyError: |
|
677 | 677 | return b'' |
|
678 | 678 | |
|
679 | 679 | def copy(self) -> 'ManifestDict': |
|
680 | 680 | c = manifestdict(self._nodelen) |
|
681 | 681 | c._lm = self._lm.copy() |
|
682 | 682 | return c |
|
683 | 683 | |
|
684 | 684 | def items(self) -> Iterator[Tuple[bytes, bytes]]: |
|
685 | 685 | return (x[:2] for x in self._lm.iterentries()) |
|
686 | 686 | |
|
687 | 687 | def iteritems(self) -> Iterator[Tuple[bytes, bytes]]: |
|
688 | 688 | return (x[:2] for x in self._lm.iterentries()) |
|
689 | 689 | |
|
690 | 690 | def iterentries(self) -> Iterator[Tuple[bytes, bytes, bytes]]: |
|
691 | 691 | return self._lm.iterentries() |
|
692 | 692 | |
|
693 | 693 | def text(self) -> ByteString: |
|
694 | 694 | # most likely uses native version |
|
695 | 695 | return self._lm.text() |
|
696 | 696 | |
|
697 | 697 | def fastdelta( |
|
698 | 698 | self, base: ByteString, changes: Iterable[Tuple[bytes, bool]] |
|
699 | 699 | ) -> Tuple[ByteString, ByteString]: |
|
700 | 700 | """Given a base manifest text as a bytearray and a list of changes |
|
701 | 701 | relative to that text, compute a delta that can be used by revlog. |
|
702 | 702 | """ |
|
703 | 703 | delta = [] |
|
704 | 704 | dstart = None |
|
705 | 705 | dend = None |
|
706 | 706 | dline = [b""] |
|
707 | 707 | start = 0 |
|
708 | 708 | # zero copy representation of base as a buffer |
|
709 | 709 | addbuf = util.buffer(base) |
|
710 | 710 | |
|
711 | 711 | changes = list(changes) |
|
712 | 712 | if len(changes) < FASTDELTA_TEXTDIFF_THRESHOLD: |
|
713 | 713 | # start with a readonly loop that finds the offset of |
|
714 | 714 | # each line and creates the deltas |
|
715 | 715 | for f, todelete in changes: |
|
716 | 716 | # bs will either be the index of the item or the insert point |
|
717 | 717 | start, end = _msearch(addbuf, f, start) |
|
718 | 718 | if not todelete: |
|
719 | 719 | h, fl = self._lm[f] |
|
720 | 720 | l = b"%s\0%s%s\n" % (f, hex(h), fl) |
|
721 | 721 | else: |
|
722 | 722 | if start == end: |
|
723 | 723 | # item we want to delete was not found, error out |
|
724 | 724 | raise AssertionError( |
|
725 | 725 | _(b"failed to remove %s from manifest") % f |
|
726 | 726 | ) |
|
727 | 727 | l = b"" |
|
728 | 728 | if dstart is not None and dstart <= start and dend >= start: |
|
729 | 729 | if dend < end: |
|
730 | 730 | dend = end |
|
731 | 731 | if l: |
|
732 | 732 | dline.append(l) |
|
733 | 733 | else: |
|
734 | 734 | if dstart is not None: |
|
735 | 735 | delta.append((dstart, dend, b"".join(dline))) |
|
736 | 736 | dstart = start |
|
737 | 737 | dend = end |
|
738 | 738 | dline = [l] |
|
739 | 739 | |
|
740 | 740 | if dstart is not None: |
|
741 | 741 | delta.append((dstart, dend, b"".join(dline))) |
|
742 | 742 | # apply the delta to the base, and get a delta for addrevision |
|
743 | 743 | deltatext, arraytext = _addlistdelta(base, delta) |
|
744 | 744 | else: |
|
745 | 745 | # For large changes, it's much cheaper to just build the text and |
|
746 | 746 | # diff it. |
|
747 | 747 | arraytext = bytearray(self.text()) |
|
748 | 748 | deltatext = mdiff.textdiff( |
|
749 | 749 | util.buffer(base), util.buffer(arraytext) |
|
750 | 750 | ) |
|
751 | 751 | |
|
752 | 752 | return arraytext, deltatext |
|
753 | 753 | |
|
754 | 754 | |
|
755 | 755 | manifestdict = interfaceutil.implementer(repository.imanifestdict)(ManifestDict) |
|
756 | 756 | |
|
757 | 757 | if typing.TYPE_CHECKING: |
|
758 | 758 | manifestdict = ManifestDict |
|
759 | 759 | |
|
760 | 760 | |
|
761 | 761 | def _msearch( |
|
762 | 762 | m: ByteString, s: bytes, lo: int = 0, hi: Optional[int] = None |
|
763 | 763 | ) -> Tuple[int, int]: |
|
764 | 764 | """return a tuple (start, end) that says where to find s within m. |
|
765 | 765 | |
|
766 | 766 | If the string is found m[start:end] are the line containing |
|
767 | 767 | that string. If start == end the string was not found and |
|
768 | 768 | they indicate the proper sorted insertion point. |
|
769 | 769 | """ |
|
770 | 770 | |
|
771 | 771 | def advance(i: int, c: bytes): |
|
772 | 772 | while i < lenm and m[i : i + 1] != c: |
|
773 | 773 | i += 1 |
|
774 | 774 | return i |
|
775 | 775 | |
|
776 | 776 | if not s: |
|
777 | 777 | return (lo, lo) |
|
778 | 778 | lenm = len(m) |
|
779 | 779 | if not hi: |
|
780 | 780 | hi = lenm |
|
781 | 781 | while lo < hi: |
|
782 | 782 | mid = (lo + hi) // 2 |
|
783 | 783 | start = mid |
|
784 | 784 | while start > 0 and m[start - 1 : start] != b'\n': |
|
785 | 785 | start -= 1 |
|
786 | 786 | end = advance(start, b'\0') |
|
787 | 787 | if bytes(m[start:end]) < s: |
|
788 | 788 | # we know that after the null there are 40 bytes of sha1 |
|
789 | 789 | # this translates to the bisect lo = mid + 1 |
|
790 | 790 | lo = advance(end + 40, b'\n') + 1 |
|
791 | 791 | else: |
|
792 | 792 | # this translates to the bisect hi = mid |
|
793 | 793 | hi = start |
|
794 | 794 | end = advance(lo, b'\0') |
|
795 | 795 | found = m[lo:end] |
|
796 | 796 | if s == found: |
|
797 | 797 | # we know that after the null there are 40 bytes of sha1 |
|
798 | 798 | end = advance(end + 40, b'\n') |
|
799 | 799 | return (lo, end + 1) |
|
800 | 800 | else: |
|
801 | 801 | return (lo, lo) |
|
802 | 802 | |
|
803 | 803 | |
|
804 | 804 | def _checkforbidden(l: Iterable[bytes]) -> None: |
|
805 | 805 | """Check filenames for illegal characters.""" |
|
806 | 806 | for f in l: |
|
807 | 807 | if b'\n' in f or b'\r' in f: |
|
808 | 808 | raise error.StorageError( |
|
809 | 809 | _(b"'\\n' and '\\r' disallowed in filenames: %r") |
|
810 | 810 | % pycompat.bytestr(f) |
|
811 | 811 | ) |
|
812 | 812 | |
|
813 | 813 | |
|
814 | 814 | # apply the changes collected during the bisect loop to our addlist |
|
815 | 815 | # return a delta suitable for addrevision |
|
816 | 816 | def _addlistdelta( |
|
817 | 817 | addlist: ByteString, |
|
818 | 818 | x: Iterable[Tuple[int, int, bytes]], |
|
819 | 819 | ) -> Tuple[bytes, ByteString]: |
|
820 | 820 | # for large addlist arrays, building a new array is cheaper |
|
821 | 821 | # than repeatedly modifying the existing one |
|
822 | 822 | currentposition = 0 |
|
823 | 823 | newaddlist = bytearray() |
|
824 | 824 | |
|
825 | 825 | for start, end, content in x: |
|
826 | 826 | newaddlist += addlist[currentposition:start] |
|
827 | 827 | if content: |
|
828 | 828 | newaddlist += bytearray(content) |
|
829 | 829 | |
|
830 | 830 | currentposition = end |
|
831 | 831 | |
|
832 | 832 | newaddlist += addlist[currentposition:] |
|
833 | 833 | |
|
834 | 834 | deltatext = b"".join( |
|
835 | 835 | struct.pack(b">lll", start, end, len(content)) + content |
|
836 | 836 | for start, end, content in x |
|
837 | 837 | ) |
|
838 | 838 | return deltatext, newaddlist |
|
839 | 839 | |
|
840 | 840 | |
|
841 | 841 | def _splittopdir(f: bytes) -> Tuple[bytes, bytes]: |
|
842 | 842 | if b'/' in f: |
|
843 | 843 | dir, subpath = f.split(b'/', 1) |
|
844 | 844 | return dir + b'/', subpath |
|
845 | 845 | else: |
|
846 | 846 | return b'', f |
|
847 | 847 | |
|
848 | 848 | |
|
849 | 849 | _noop = lambda s: None |
|
850 | 850 | |
|
851 | 851 | |
|
852 | 852 | class TreeManifest: |
|
853 | _dir: bytes | |
|
854 | _dirs: Dict[bytes, 'TreeManifest'] | |
|
855 | _dirty: bool | |
|
856 | _files: Dict[bytes, bytes] | |
|
857 | _flags: Dict[bytes, bytes] | |
|
858 | ||
|
853 | 859 | def __init__(self, nodeconstants, dir: bytes = b'', text: bytes = b''): |
|
854 | 860 | self._dir = dir |
|
855 | 861 | self.nodeconstants = nodeconstants |
|
856 | 862 | self._node = self.nodeconstants.nullid |
|
857 | 863 | self._nodelen = self.nodeconstants.nodelen |
|
858 | 864 | self._loadfunc = _noop |
|
859 | 865 | self._copyfunc = _noop |
|
860 | 866 | self._dirty = False |
|
861 |
self._dirs |
|
|
867 | self._dirs = {} | |
|
862 | 868 | self._lazydirs: Dict[ |
|
863 | 869 | bytes, |
|
864 | 870 | Tuple[bytes, Callable[[bytes, bytes], 'TreeManifest'], bool], |
|
865 | 871 | ] = {} |
|
866 | 872 | # Using _lazymanifest here is a little slower than plain old dicts |
|
867 |
self._files |
|
|
873 | self._files = {} | |
|
868 | 874 | self._flags = {} |
|
869 | 875 | if text: |
|
870 | 876 | |
|
871 | 877 | def readsubtree(subdir, subm): |
|
872 | 878 | raise AssertionError( |
|
873 | 879 | b'treemanifest constructor only accepts flat manifests' |
|
874 | 880 | ) |
|
875 | 881 | |
|
876 | 882 | self.parse(text, readsubtree) |
|
877 | 883 | self._dirty = True # Mark flat manifest dirty after parsing |
|
878 | 884 | |
|
879 | 885 | def _subpath(self, path: bytes) -> bytes: |
|
880 | 886 | return self._dir + path |
|
881 | 887 | |
|
882 | 888 | def _loadalllazy(self) -> None: |
|
883 | 889 | selfdirs = self._dirs |
|
884 | 890 | subpath = self._subpath |
|
885 | 891 | for d, (node, readsubtree, docopy) in self._lazydirs.items(): |
|
886 | 892 | if docopy: |
|
887 | 893 | selfdirs[d] = readsubtree(subpath(d), node).copy() |
|
888 | 894 | else: |
|
889 | 895 | selfdirs[d] = readsubtree(subpath(d), node) |
|
890 | 896 | self._lazydirs.clear() |
|
891 | 897 | |
|
892 | 898 | def _loadlazy(self, d: bytes) -> None: |
|
893 | 899 | v = self._lazydirs.get(d) |
|
894 | 900 | if v is not None: |
|
895 | 901 | node, readsubtree, docopy = v |
|
896 | 902 | if docopy: |
|
897 | 903 | self._dirs[d] = readsubtree(self._subpath(d), node).copy() |
|
898 | 904 | else: |
|
899 | 905 | self._dirs[d] = readsubtree(self._subpath(d), node) |
|
900 | 906 | del self._lazydirs[d] |
|
901 | 907 | |
|
902 | 908 | def _loadchildrensetlazy( |
|
903 | 909 | self, visit: Union[Set[bytes], bytes] |
|
904 | 910 | ) -> Optional[Set[bytes]]: |
|
905 | 911 | if not visit: |
|
906 | 912 | return None |
|
907 | 913 | if visit == b'all' or visit == b'this': |
|
908 | 914 | self._loadalllazy() |
|
909 | 915 | return None |
|
910 | 916 | |
|
911 | 917 | visit = cast(Set[bytes], visit) |
|
912 | 918 | |
|
913 | 919 | loadlazy = self._loadlazy |
|
914 | 920 | for k in visit: |
|
915 | 921 | loadlazy(k + b'/') |
|
916 | 922 | return visit |
|
917 | 923 | |
|
918 | 924 | def _loaddifflazy(self, t1: 'TreeManifest', t2: 'TreeManifest'): |
|
919 | 925 | """load items in t1 and t2 if they're needed for diffing. |
|
920 | 926 | |
|
921 | 927 | The criteria currently is: |
|
922 | 928 | - if it's not present in _lazydirs in either t1 or t2, load it in the |
|
923 | 929 | other (it may already be loaded or it may not exist, doesn't matter) |
|
924 | 930 | - if it's present in _lazydirs in both, compare the nodeid; if it |
|
925 | 931 | differs, load it in both |
|
926 | 932 | """ |
|
927 | 933 | toloadlazy = [] |
|
928 | 934 | for d, v1 in t1._lazydirs.items(): |
|
929 | 935 | v2 = t2._lazydirs.get(d) |
|
930 | 936 | if v2 is None or v2[0] != v1[0]: |
|
931 | 937 | toloadlazy.append(d) |
|
932 | 938 | for d, v1 in t2._lazydirs.items(): |
|
933 | 939 | if d not in t1._lazydirs: |
|
934 | 940 | toloadlazy.append(d) |
|
935 | 941 | |
|
936 | 942 | for d in toloadlazy: |
|
937 | 943 | t1._loadlazy(d) |
|
938 | 944 | t2._loadlazy(d) |
|
939 | 945 | |
|
940 | 946 | def __len__(self) -> int: |
|
941 | 947 | self._load() |
|
942 | 948 | size = len(self._files) |
|
943 | 949 | self._loadalllazy() |
|
944 | 950 | for m in self._dirs.values(): |
|
945 | 951 | size += m.__len__() |
|
946 | 952 | return size |
|
947 | 953 | |
|
948 | 954 | def __nonzero__(self) -> bool: |
|
949 | 955 | # Faster than "__len__() != 0" since it avoids loading sub-manifests |
|
950 | 956 | return not self._isempty() |
|
951 | 957 | |
|
952 | 958 | __bool__ = __nonzero__ |
|
953 | 959 | |
|
954 | 960 | def _isempty(self) -> bool: |
|
955 | 961 | self._load() # for consistency; already loaded by all callers |
|
956 | 962 | # See if we can skip loading everything. |
|
957 | 963 | if self._files or ( |
|
958 | 964 | self._dirs and any(not m._isempty() for m in self._dirs.values()) |
|
959 | 965 | ): |
|
960 | 966 | return False |
|
961 | 967 | self._loadalllazy() |
|
962 | 968 | return not self._dirs or all(m._isempty() for m in self._dirs.values()) |
|
963 | 969 | |
|
964 | 970 | @encoding.strmethod |
|
965 | 971 | def __repr__(self) -> bytes: |
|
966 | 972 | return ( |
|
967 | 973 | b'<treemanifest dir=%s, node=%s, loaded=%r, dirty=%r at 0x%x>' |
|
968 | 974 | % ( |
|
969 | 975 | self._dir, |
|
970 | 976 | hex(self._node), |
|
971 | 977 | bool(self._loadfunc is _noop), |
|
972 | 978 | self._dirty, |
|
973 | 979 | id(self), |
|
974 | 980 | ) |
|
975 | 981 | ) |
|
976 | 982 | |
|
977 | 983 | def dir(self) -> bytes: |
|
978 | 984 | """The directory that this tree manifest represents, including a |
|
979 | 985 | trailing '/'. Empty string for the repo root directory.""" |
|
980 | 986 | return self._dir |
|
981 | 987 | |
|
982 | 988 | def node(self) -> bytes: |
|
983 | 989 | """This node of this instance. nullid for unsaved instances. Should |
|
984 | 990 | be updated when the instance is read or written from a revlog. |
|
985 | 991 | """ |
|
986 | 992 | assert not self._dirty |
|
987 | 993 | return self._node |
|
988 | 994 | |
|
989 | 995 | def setnode(self, node: bytes) -> None: |
|
990 | 996 | self._node = node |
|
991 | 997 | self._dirty = False |
|
992 | 998 | |
|
993 | 999 | def iterentries( |
|
994 | 1000 | self, |
|
995 | 1001 | ) -> Iterator[Tuple[bytes, Union[bytes, 'TreeManifest'], bytes]]: |
|
996 | 1002 | self._load() |
|
997 | 1003 | self._loadalllazy() |
|
998 | 1004 | for p, n in sorted( |
|
999 | 1005 | itertools.chain(self._dirs.items(), self._files.items()) |
|
1000 | 1006 | ): |
|
1001 | 1007 | if p in self._files: |
|
1002 | 1008 | yield self._subpath(p), n, self._flags.get(p, b'') |
|
1003 | 1009 | else: |
|
1004 | 1010 | for x in n.iterentries(): |
|
1005 | 1011 | yield x |
|
1006 | 1012 | |
|
1007 | 1013 | def items(self) -> Iterator[Tuple[bytes, Union[bytes, 'TreeManifest']]]: |
|
1008 | 1014 | self._load() |
|
1009 | 1015 | self._loadalllazy() |
|
1010 | 1016 | for p, n in sorted( |
|
1011 | 1017 | itertools.chain(self._dirs.items(), self._files.items()) |
|
1012 | 1018 | ): |
|
1013 | 1019 | if p in self._files: |
|
1014 | 1020 | yield self._subpath(p), n |
|
1015 | 1021 | else: |
|
1016 | 1022 | for f, sn in n.items(): |
|
1017 | 1023 | yield f, sn |
|
1018 | 1024 | |
|
1019 | 1025 | iteritems = items |
|
1020 | 1026 | |
|
1021 | 1027 | def iterkeys(self) -> Iterator[bytes]: |
|
1022 | 1028 | self._load() |
|
1023 | 1029 | self._loadalllazy() |
|
1024 | 1030 | for p in sorted(itertools.chain(self._dirs, self._files)): |
|
1025 | 1031 | if p in self._files: |
|
1026 | 1032 | yield self._subpath(p) |
|
1027 | 1033 | else: |
|
1028 | 1034 | for f in self._dirs[p]: |
|
1029 | 1035 | yield f |
|
1030 | 1036 | |
|
1031 | 1037 | def keys(self) -> List[bytes]: |
|
1032 | 1038 | return list(self.iterkeys()) |
|
1033 | 1039 | |
|
1034 | 1040 | def __iter__(self) -> Iterator[bytes]: |
|
1035 | 1041 | return self.iterkeys() |
|
1036 | 1042 | |
|
1037 | 1043 | def __contains__(self, f: bytes) -> bool: |
|
1038 | 1044 | if f is None: |
|
1039 | 1045 | return False |
|
1040 | 1046 | self._load() |
|
1041 | 1047 | dir, subpath = _splittopdir(f) |
|
1042 | 1048 | if dir: |
|
1043 | 1049 | self._loadlazy(dir) |
|
1044 | 1050 | |
|
1045 | 1051 | if dir not in self._dirs: |
|
1046 | 1052 | return False |
|
1047 | 1053 | |
|
1048 | 1054 | return self._dirs[dir].__contains__(subpath) |
|
1049 | 1055 | else: |
|
1050 | 1056 | return f in self._files |
|
1051 | 1057 | |
|
1052 | 1058 | def get(self, f: bytes, default: Optional[bytes] = None) -> Optional[bytes]: |
|
1053 | 1059 | self._load() |
|
1054 | 1060 | dir, subpath = _splittopdir(f) |
|
1055 | 1061 | if dir: |
|
1056 | 1062 | self._loadlazy(dir) |
|
1057 | 1063 | |
|
1058 | 1064 | if dir not in self._dirs: |
|
1059 | 1065 | return default |
|
1060 | 1066 | return self._dirs[dir].get(subpath, default) |
|
1061 | 1067 | else: |
|
1062 | 1068 | return self._files.get(f, default) |
|
1063 | 1069 | |
|
1064 | 1070 | def __getitem__(self, f: bytes) -> bytes: |
|
1065 | 1071 | self._load() |
|
1066 | 1072 | dir, subpath = _splittopdir(f) |
|
1067 | 1073 | if dir: |
|
1068 | 1074 | self._loadlazy(dir) |
|
1069 | 1075 | |
|
1070 | 1076 | return self._dirs[dir].__getitem__(subpath) |
|
1071 | 1077 | else: |
|
1072 | 1078 | return self._files[f] |
|
1073 | 1079 | |
|
1074 | 1080 | def flags(self, f: bytes) -> bytes: |
|
1075 | 1081 | self._load() |
|
1076 | 1082 | dir, subpath = _splittopdir(f) |
|
1077 | 1083 | if dir: |
|
1078 | 1084 | self._loadlazy(dir) |
|
1079 | 1085 | |
|
1080 | 1086 | if dir not in self._dirs: |
|
1081 | 1087 | return b'' |
|
1082 | 1088 | return self._dirs[dir].flags(subpath) |
|
1083 | 1089 | else: |
|
1084 | 1090 | if f in self._lazydirs or f in self._dirs: |
|
1085 | 1091 | return b'' |
|
1086 | 1092 | return self._flags.get(f, b'') |
|
1087 | 1093 | |
|
1088 | 1094 | def find(self, f: bytes) -> Tuple[bytes, bytes]: |
|
1089 | 1095 | self._load() |
|
1090 | 1096 | dir, subpath = _splittopdir(f) |
|
1091 | 1097 | if dir: |
|
1092 | 1098 | self._loadlazy(dir) |
|
1093 | 1099 | |
|
1094 | 1100 | return self._dirs[dir].find(subpath) |
|
1095 | 1101 | else: |
|
1096 | 1102 | return self._files[f], self._flags.get(f, b'') |
|
1097 | 1103 | |
|
1098 | 1104 | def __delitem__(self, f: bytes) -> None: |
|
1099 | 1105 | self._load() |
|
1100 | 1106 | dir, subpath = _splittopdir(f) |
|
1101 | 1107 | if dir: |
|
1102 | 1108 | self._loadlazy(dir) |
|
1103 | 1109 | |
|
1104 | 1110 | self._dirs[dir].__delitem__(subpath) |
|
1105 | 1111 | # If the directory is now empty, remove it |
|
1106 | 1112 | if self._dirs[dir]._isempty(): |
|
1107 | 1113 | del self._dirs[dir] |
|
1108 | 1114 | else: |
|
1109 | 1115 | del self._files[f] |
|
1110 | 1116 | if f in self._flags: |
|
1111 | 1117 | del self._flags[f] |
|
1112 | 1118 | self._dirty = True |
|
1113 | 1119 | |
|
1114 | 1120 | def set(self, f: bytes, node: bytes, flags: bytes) -> None: |
|
1115 | 1121 | """Set both the node and the flags for path f.""" |
|
1116 | 1122 | assert node is not None |
|
1117 | 1123 | if flags not in _manifestflags: |
|
1118 | 1124 | raise TypeError(b"Invalid manifest flag set.") |
|
1119 | 1125 | self._load() |
|
1120 | 1126 | dir, subpath = _splittopdir(f) |
|
1121 | 1127 | if dir: |
|
1122 | 1128 | self._loadlazy(dir) |
|
1123 | 1129 | if dir not in self._dirs: |
|
1124 | 1130 | self._dirs[dir] = treemanifest( |
|
1125 | 1131 | self.nodeconstants, self._subpath(dir) |
|
1126 | 1132 | ) |
|
1127 | 1133 | self._dirs[dir].set(subpath, node, flags) |
|
1128 | 1134 | else: |
|
1129 | 1135 | assert len(node) in (20, 32) |
|
1130 | 1136 | self._files[f] = node |
|
1131 | 1137 | self._flags[f] = flags |
|
1132 | 1138 | self._dirty = True |
|
1133 | 1139 | |
|
1134 | 1140 | def __setitem__(self, f: bytes, n: bytes) -> None: |
|
1135 | 1141 | assert n is not None |
|
1136 | 1142 | self._load() |
|
1137 | 1143 | dir, subpath = _splittopdir(f) |
|
1138 | 1144 | if dir: |
|
1139 | 1145 | self._loadlazy(dir) |
|
1140 | 1146 | if dir not in self._dirs: |
|
1141 | 1147 | self._dirs[dir] = treemanifest( |
|
1142 | 1148 | self.nodeconstants, self._subpath(dir) |
|
1143 | 1149 | ) |
|
1144 | 1150 | self._dirs[dir].__setitem__(subpath, n) |
|
1145 | 1151 | else: |
|
1146 | 1152 | # manifest nodes are either 20 bytes or 32 bytes, |
|
1147 | 1153 | # depending on the hash in use. Assert this as historically |
|
1148 | 1154 | # sometimes extra bytes were added. |
|
1149 | 1155 | assert len(n) in (20, 32) |
|
1150 | 1156 | self._files[f] = n |
|
1151 | 1157 | self._dirty = True |
|
1152 | 1158 | |
|
1153 | 1159 | def _load(self) -> None: |
|
1154 | 1160 | if self._loadfunc is not _noop: |
|
1155 | 1161 | lf, self._loadfunc = self._loadfunc, _noop |
|
1156 | 1162 | lf(self) |
|
1157 | 1163 | elif self._copyfunc is not _noop: |
|
1158 | 1164 | cf, self._copyfunc = self._copyfunc, _noop |
|
1159 | 1165 | cf(self) |
|
1160 | 1166 | |
|
1161 | 1167 | def setflag(self, f: bytes, flags: bytes) -> None: |
|
1162 | 1168 | """Set the flags (symlink, executable) for path f.""" |
|
1163 | 1169 | if flags not in _manifestflags: |
|
1164 | 1170 | raise TypeError(b"Invalid manifest flag set.") |
|
1165 | 1171 | self._load() |
|
1166 | 1172 | dir, subpath = _splittopdir(f) |
|
1167 | 1173 | if dir: |
|
1168 | 1174 | self._loadlazy(dir) |
|
1169 | 1175 | if dir not in self._dirs: |
|
1170 | 1176 | self._dirs[dir] = treemanifest( |
|
1171 | 1177 | self.nodeconstants, self._subpath(dir) |
|
1172 | 1178 | ) |
|
1173 | 1179 | self._dirs[dir].setflag(subpath, flags) |
|
1174 | 1180 | else: |
|
1175 | 1181 | self._flags[f] = flags |
|
1176 | 1182 | self._dirty = True |
|
1177 | 1183 | |
|
1178 | 1184 | def copy(self) -> 'TreeManifest': |
|
1179 | 1185 | copy = treemanifest(self.nodeconstants, self._dir) |
|
1180 | 1186 | copy._node = self._node |
|
1181 | 1187 | copy._dirty = self._dirty |
|
1182 | 1188 | if self._copyfunc is _noop: |
|
1183 | 1189 | |
|
1184 | 1190 | def _copyfunc(s): |
|
1185 | 1191 | self._load() |
|
1186 | 1192 | s._lazydirs = { |
|
1187 | 1193 | d: (n, r, True) for d, (n, r, c) in self._lazydirs.items() |
|
1188 | 1194 | } |
|
1189 | 1195 | sdirs = s._dirs |
|
1190 | 1196 | for d, v in self._dirs.items(): |
|
1191 | 1197 | sdirs[d] = v.copy() |
|
1192 | 1198 | s._files = dict.copy(self._files) |
|
1193 | 1199 | s._flags = dict.copy(self._flags) |
|
1194 | 1200 | |
|
1195 | 1201 | if self._loadfunc is _noop: |
|
1196 | 1202 | _copyfunc(copy) |
|
1197 | 1203 | else: |
|
1198 | 1204 | copy._copyfunc = _copyfunc |
|
1199 | 1205 | else: |
|
1200 | 1206 | copy._copyfunc = self._copyfunc |
|
1201 | 1207 | return copy |
|
1202 | 1208 | |
|
1203 | 1209 | def filesnotin( |
|
1204 | 1210 | self, m2: 'TreeManifest', match: Optional[matchmod.basematcher] = None |
|
1205 | 1211 | ) -> Set[bytes]: |
|
1206 | 1212 | '''Set of files in this manifest that are not in the other''' |
|
1207 | 1213 | if match and not match.always(): |
|
1208 | 1214 | m1 = self._matches(match) |
|
1209 | 1215 | m2 = m2._matches(match) |
|
1210 | 1216 | return m1.filesnotin(m2) |
|
1211 | 1217 | |
|
1212 | 1218 | files = set() |
|
1213 | 1219 | |
|
1214 | 1220 | def _filesnotin(t1, t2): |
|
1215 | 1221 | if t1._node == t2._node and not t1._dirty and not t2._dirty: |
|
1216 | 1222 | return |
|
1217 | 1223 | t1._load() |
|
1218 | 1224 | t2._load() |
|
1219 | 1225 | self._loaddifflazy(t1, t2) |
|
1220 | 1226 | for d, m1 in t1._dirs.items(): |
|
1221 | 1227 | if d in t2._dirs: |
|
1222 | 1228 | m2 = t2._dirs[d] |
|
1223 | 1229 | _filesnotin(m1, m2) |
|
1224 | 1230 | else: |
|
1225 | 1231 | files.update(m1.iterkeys()) |
|
1226 | 1232 | |
|
1227 | 1233 | for fn in t1._files: |
|
1228 | 1234 | if fn not in t2._files: |
|
1229 | 1235 | files.add(t1._subpath(fn)) |
|
1230 | 1236 | |
|
1231 | 1237 | _filesnotin(self, m2) |
|
1232 | 1238 | return files |
|
1233 | 1239 | |
|
1234 | 1240 | @propertycache |
|
1235 | 1241 | def _alldirs(self) -> pathutil.dirs: |
|
1236 | 1242 | return pathutil.dirs(self) |
|
1237 | 1243 | |
|
1238 | 1244 | def dirs(self) -> pathutil.dirs: |
|
1239 | 1245 | return self._alldirs |
|
1240 | 1246 | |
|
1241 | 1247 | def hasdir(self, dir: bytes) -> bool: |
|
1242 | 1248 | self._load() |
|
1243 | 1249 | topdir, subdir = _splittopdir(dir) |
|
1244 | 1250 | if topdir: |
|
1245 | 1251 | self._loadlazy(topdir) |
|
1246 | 1252 | if topdir in self._dirs: |
|
1247 | 1253 | return self._dirs[topdir].hasdir(subdir) |
|
1248 | 1254 | return False |
|
1249 | 1255 | dirslash = dir + b'/' |
|
1250 | 1256 | return dirslash in self._dirs or dirslash in self._lazydirs |
|
1251 | 1257 | |
|
1252 | 1258 | def walk(self, match: matchmod.basematcher) -> Iterator[bytes]: |
|
1253 | 1259 | """Generates matching file names. |
|
1254 | 1260 | |
|
1255 | 1261 | It also reports nonexistent files by marking them bad with match.bad(). |
|
1256 | 1262 | """ |
|
1257 | 1263 | if match.always(): |
|
1258 | 1264 | for f in iter(self): |
|
1259 | 1265 | yield f |
|
1260 | 1266 | return |
|
1261 | 1267 | |
|
1262 | 1268 | fset = set(match.files()) |
|
1263 | 1269 | |
|
1264 | 1270 | for fn in self._walk(match): |
|
1265 | 1271 | if fn in fset: |
|
1266 | 1272 | # specified pattern is the exact name |
|
1267 | 1273 | fset.remove(fn) |
|
1268 | 1274 | yield fn |
|
1269 | 1275 | |
|
1270 | 1276 | # for dirstate.walk, files=[''] means "walk the whole tree". |
|
1271 | 1277 | # follow that here, too |
|
1272 | 1278 | fset.discard(b'') |
|
1273 | 1279 | |
|
1274 | 1280 | for fn in sorted(fset): |
|
1275 | 1281 | if not self.hasdir(fn): |
|
1276 | 1282 | match.bad(fn, None) |
|
1277 | 1283 | |
|
1278 | 1284 | def _walk(self, match: matchmod.basematcher) -> Iterator[bytes]: |
|
1279 | 1285 | '''Recursively generates matching file names for walk().''' |
|
1280 | 1286 | visit = match.visitchildrenset(self._dir[:-1]) |
|
1281 | 1287 | if not visit: |
|
1282 | 1288 | return |
|
1283 | 1289 | |
|
1284 | 1290 | # yield this dir's files and walk its submanifests |
|
1285 | 1291 | self._load() |
|
1286 | 1292 | visit = self._loadchildrensetlazy(visit) |
|
1287 | 1293 | for p in sorted(list(self._dirs) + list(self._files)): |
|
1288 | 1294 | if p in self._files: |
|
1289 | 1295 | fullp = self._subpath(p) |
|
1290 | 1296 | if match(fullp): |
|
1291 | 1297 | yield fullp |
|
1292 | 1298 | else: |
|
1293 | 1299 | if not visit or p[:-1] in visit: |
|
1294 | 1300 | for f in self._dirs[p]._walk(match): |
|
1295 | 1301 | yield f |
|
1296 | 1302 | |
|
1297 | 1303 | def _matches(self, match: matchmod.basematcher) -> 'TreeManifest': |
|
1298 | 1304 | """recursively generate a new manifest filtered by the match argument.""" |
|
1299 | 1305 | if match.always(): |
|
1300 | 1306 | return self.copy() |
|
1301 | 1307 | return self._matches_inner(match) |
|
1302 | 1308 | |
|
1303 | 1309 | def _matches_inner(self, match: matchmod.basematcher) -> 'TreeManifest': |
|
1304 | 1310 | if match.always(): |
|
1305 | 1311 | return self.copy() |
|
1306 | 1312 | |
|
1307 | 1313 | visit = match.visitchildrenset(self._dir[:-1]) |
|
1308 | 1314 | if visit == b'all': |
|
1309 | 1315 | return self.copy() |
|
1310 | 1316 | ret = treemanifest(self.nodeconstants, self._dir) |
|
1311 | 1317 | if not visit: |
|
1312 | 1318 | return ret |
|
1313 | 1319 | |
|
1314 | 1320 | self._load() |
|
1315 | 1321 | for fn in self._files: |
|
1316 | 1322 | # While visitchildrenset *usually* lists only subdirs, this is |
|
1317 | 1323 | # actually up to the matcher and may have some files in the set(). |
|
1318 | 1324 | # If visit == 'this', we should obviously look at the files in this |
|
1319 | 1325 | # directory; if visit is a set, and fn is in it, we should inspect |
|
1320 | 1326 | # fn (but no need to inspect things not in the set). |
|
1321 | 1327 | if visit != b'this' and fn not in visit: |
|
1322 | 1328 | continue |
|
1323 | 1329 | fullp = self._subpath(fn) |
|
1324 | 1330 | # visitchildrenset isn't perfect, we still need to call the regular |
|
1325 | 1331 | # matcher code to further filter results. |
|
1326 | 1332 | if not match(fullp): |
|
1327 | 1333 | continue |
|
1328 | 1334 | ret._files[fn] = self._files[fn] |
|
1329 | 1335 | if fn in self._flags: |
|
1330 | 1336 | ret._flags[fn] = self._flags[fn] |
|
1331 | 1337 | |
|
1332 | 1338 | visit = self._loadchildrensetlazy(visit) |
|
1333 | 1339 | for dir, subm in self._dirs.items(): |
|
1334 | 1340 | if visit and dir[:-1] not in visit: |
|
1335 | 1341 | continue |
|
1336 | 1342 | m = subm._matches_inner(match) |
|
1337 | 1343 | if not m._isempty(): |
|
1338 | 1344 | ret._dirs[dir] = m |
|
1339 | 1345 | |
|
1340 | 1346 | if not ret._isempty(): |
|
1341 | 1347 | ret._dirty = True |
|
1342 | 1348 | return ret |
|
1343 | 1349 | |
|
1344 | 1350 | def fastdelta( |
|
1345 | 1351 | self, base: ByteString, changes: Iterable[Tuple[bytes, bool]] |
|
1346 | 1352 | ) -> ByteString: |
|
1347 | 1353 | raise FastdeltaUnavailable() |
|
1348 | 1354 | |
|
1349 | 1355 | def diff( |
|
1350 | 1356 | self, |
|
1351 | 1357 | m2: 'TreeManifest', |
|
1352 | 1358 | match: Optional[matchmod.basematcher] = None, |
|
1353 | 1359 | clean: bool = False, |
|
1354 | 1360 | ) -> Dict[ |
|
1355 | 1361 | bytes, |
|
1356 | 1362 | Optional[ |
|
1357 | 1363 | Tuple[Tuple[Optional[bytes], bytes], Tuple[Optional[bytes], bytes]] |
|
1358 | 1364 | ], |
|
1359 | 1365 | ]: |
|
1360 | 1366 | """Finds changes between the current manifest and m2. |
|
1361 | 1367 | |
|
1362 | 1368 | Args: |
|
1363 | 1369 | m2: the manifest to which this manifest should be compared. |
|
1364 | 1370 | clean: if true, include files unchanged between these manifests |
|
1365 | 1371 | with a None value in the returned dictionary. |
|
1366 | 1372 | |
|
1367 | 1373 | The result is returned as a dict with filename as key and |
|
1368 | 1374 | values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the |
|
1369 | 1375 | nodeid in the current/other manifest and fl1/fl2 is the flag |
|
1370 | 1376 | in the current/other manifest. Where the file does not exist, |
|
1371 | 1377 | the nodeid will be None and the flags will be the empty |
|
1372 | 1378 | string. |
|
1373 | 1379 | """ |
|
1374 | 1380 | if match and not match.always(): |
|
1375 | 1381 | m1 = self._matches(match) |
|
1376 | 1382 | m2 = m2._matches(match) |
|
1377 | 1383 | return m1.diff(m2, clean=clean) |
|
1378 | 1384 | result = {} |
|
1379 | 1385 | emptytree = treemanifest(self.nodeconstants) |
|
1380 | 1386 | |
|
1381 | 1387 | def _iterativediff(t1, t2, stack): |
|
1382 | 1388 | """compares two tree manifests and append new tree-manifests which |
|
1383 | 1389 | needs to be compared to stack""" |
|
1384 | 1390 | if t1._node == t2._node and not t1._dirty and not t2._dirty: |
|
1385 | 1391 | return |
|
1386 | 1392 | t1._load() |
|
1387 | 1393 | t2._load() |
|
1388 | 1394 | self._loaddifflazy(t1, t2) |
|
1389 | 1395 | |
|
1390 | 1396 | for d, m1 in t1._dirs.items(): |
|
1391 | 1397 | m2 = t2._dirs.get(d, emptytree) |
|
1392 | 1398 | stack.append((m1, m2)) |
|
1393 | 1399 | |
|
1394 | 1400 | for d, m2 in t2._dirs.items(): |
|
1395 | 1401 | if d not in t1._dirs: |
|
1396 | 1402 | stack.append((emptytree, m2)) |
|
1397 | 1403 | |
|
1398 | 1404 | for fn, n1 in t1._files.items(): |
|
1399 | 1405 | fl1 = t1._flags.get(fn, b'') |
|
1400 | 1406 | n2 = t2._files.get(fn, None) |
|
1401 | 1407 | fl2 = t2._flags.get(fn, b'') |
|
1402 | 1408 | if n1 != n2 or fl1 != fl2: |
|
1403 | 1409 | result[t1._subpath(fn)] = ((n1, fl1), (n2, fl2)) |
|
1404 | 1410 | elif clean: |
|
1405 | 1411 | result[t1._subpath(fn)] = None |
|
1406 | 1412 | |
|
1407 | 1413 | for fn, n2 in t2._files.items(): |
|
1408 | 1414 | if fn not in t1._files: |
|
1409 | 1415 | fl2 = t2._flags.get(fn, b'') |
|
1410 | 1416 | result[t2._subpath(fn)] = ((None, b''), (n2, fl2)) |
|
1411 | 1417 | |
|
1412 | 1418 | stackls = [] |
|
1413 | 1419 | _iterativediff(self, m2, stackls) |
|
1414 | 1420 | while stackls: |
|
1415 | 1421 | t1, t2 = stackls.pop() |
|
1416 | 1422 | # stackls is populated in the function call |
|
1417 | 1423 | _iterativediff(t1, t2, stackls) |
|
1418 | 1424 | return result |
|
1419 | 1425 | |
|
1420 | 1426 | def unmodifiedsince(self, m2: 'TreeManifest') -> bool: |
|
1421 | 1427 | return not self._dirty and not m2._dirty and self._node == m2._node |
|
1422 | 1428 | |
|
1423 | 1429 | def parse( |
|
1424 | 1430 | self, |
|
1425 | 1431 | text: bytes, |
|
1426 | 1432 | readsubtree: Callable[[bytes, bytes], 'TreeManifest'], |
|
1427 | 1433 | ) -> None: |
|
1428 | 1434 | selflazy = self._lazydirs |
|
1429 | 1435 | for f, n, fl in _parse(self._nodelen, text): |
|
1430 | 1436 | if fl == b't': |
|
1431 | 1437 | f = f + b'/' |
|
1432 | 1438 | # False below means "doesn't need to be copied" and can use the |
|
1433 | 1439 | # cached value from readsubtree directly. |
|
1434 | 1440 | selflazy[f] = (n, readsubtree, False) |
|
1435 | 1441 | elif b'/' in f: |
|
1436 | 1442 | # This is a flat manifest, so use __setitem__ and setflag rather |
|
1437 | 1443 | # than assigning directly to _files and _flags, so we can |
|
1438 | 1444 | # assign a path in a subdirectory, and to mark dirty (compared |
|
1439 | 1445 | # to nullid). |
|
1440 | 1446 | self[f] = n |
|
1441 | 1447 | if fl: |
|
1442 | 1448 | self.setflag(f, fl) |
|
1443 | 1449 | else: |
|
1444 | 1450 | # Assigning to _files and _flags avoids marking as dirty, |
|
1445 | 1451 | # and should be a little faster. |
|
1446 | 1452 | self._files[f] = n |
|
1447 | 1453 | if fl: |
|
1448 | 1454 | self._flags[f] = fl |
|
1449 | 1455 | |
|
1450 | 1456 | def text(self) -> ByteString: |
|
1451 | 1457 | """Get the full data of this manifest as a bytestring.""" |
|
1452 | 1458 | self._load() |
|
1453 | 1459 | return _text(self.iterentries()) |
|
1454 | 1460 | |
|
1455 | 1461 | def dirtext(self) -> ByteString: |
|
1456 | 1462 | """Get the full data of this directory as a bytestring. Make sure that |
|
1457 | 1463 | any submanifests have been written first, so their nodeids are correct. |
|
1458 | 1464 | """ |
|
1459 | 1465 | self._load() |
|
1460 | 1466 | flags = self.flags |
|
1461 | 1467 | lazydirs = [(d[:-1], v[0], b't') for d, v in self._lazydirs.items()] |
|
1462 | 1468 | dirs = [(d[:-1], self._dirs[d]._node, b't') for d in self._dirs] |
|
1463 | 1469 | files = [(f, self._files[f], flags(f)) for f in self._files] |
|
1464 | 1470 | return _text(sorted(dirs + files + lazydirs)) |
|
1465 | 1471 | |
|
1466 | 1472 | def read( |
|
1467 | 1473 | self, |
|
1468 | 1474 | gettext: Callable[[], ByteString], |
|
1469 | 1475 | readsubtree: Callable[[bytes, bytes], 'TreeManifest'], |
|
1470 | 1476 | ) -> None: |
|
1471 | 1477 | def _load_for_read(s): |
|
1472 | 1478 | s.parse(gettext(), readsubtree) |
|
1473 | 1479 | s._dirty = False |
|
1474 | 1480 | |
|
1475 | 1481 | self._loadfunc = _load_for_read |
|
1476 | 1482 | |
|
1477 | 1483 | def writesubtrees( |
|
1478 | 1484 | self, |
|
1479 | 1485 | m1: 'TreeManifest', |
|
1480 | 1486 | m2: 'TreeManifest', |
|
1481 | 1487 | writesubtree: Callable[ |
|
1482 | 1488 | [ |
|
1483 | 1489 | Callable[['TreeManifest'], None], |
|
1484 | 1490 | bytes, |
|
1485 | 1491 | bytes, |
|
1486 | 1492 | matchmod.basematcher, |
|
1487 | 1493 | ], |
|
1488 | 1494 | None, |
|
1489 | 1495 | ], |
|
1490 | 1496 | match: matchmod.basematcher, |
|
1491 | 1497 | ) -> None: |
|
1492 | 1498 | self._load() # for consistency; should never have any effect here |
|
1493 | 1499 | m1._load() |
|
1494 | 1500 | m2._load() |
|
1495 | 1501 | emptytree = treemanifest(self.nodeconstants) |
|
1496 | 1502 | |
|
1497 | 1503 | def getnode(m, d): |
|
1498 | 1504 | ld = m._lazydirs.get(d) |
|
1499 | 1505 | if ld: |
|
1500 | 1506 | return ld[0] |
|
1501 | 1507 | tree = m._dirs.get(d, emptytree) |
|
1502 | 1508 | assert tree is not None # helps pytype |
|
1503 | 1509 | return tree._node |
|
1504 | 1510 | |
|
1505 | 1511 | # let's skip investigating things that `match` says we do not need. |
|
1506 | 1512 | visit = match.visitchildrenset(self._dir[:-1]) |
|
1507 | 1513 | visit = self._loadchildrensetlazy(visit) |
|
1508 | 1514 | if visit == b'this' or visit == b'all': |
|
1509 | 1515 | visit = None |
|
1510 | 1516 | for d, subm in self._dirs.items(): |
|
1511 | 1517 | if visit and d[:-1] not in visit: |
|
1512 | 1518 | continue |
|
1513 | 1519 | subp1 = getnode(m1, d) |
|
1514 | 1520 | subp2 = getnode(m2, d) |
|
1515 | 1521 | if subp1 == self.nodeconstants.nullid: |
|
1516 | 1522 | subp1, subp2 = subp2, subp1 |
|
1517 | 1523 | writesubtree(subm, subp1, subp2, match) |
|
1518 | 1524 | |
|
1519 | 1525 | def walksubtrees( |
|
1520 | 1526 | self, matcher: Optional[matchmod.basematcher] = None |
|
1521 | 1527 | ) -> Iterator['TreeManifest']: |
|
1522 | 1528 | """Returns an iterator of the subtrees of this manifest, including this |
|
1523 | 1529 | manifest itself. |
|
1524 | 1530 | |
|
1525 | 1531 | If `matcher` is provided, it only returns subtrees that match. |
|
1526 | 1532 | """ |
|
1527 | 1533 | if matcher and not matcher.visitdir(self._dir[:-1]): |
|
1528 | 1534 | return |
|
1529 | 1535 | if not matcher or matcher(self._dir[:-1]): |
|
1530 | 1536 | yield self |
|
1531 | 1537 | |
|
1532 | 1538 | self._load() |
|
1533 | 1539 | # OPT: use visitchildrenset to avoid loading everything. |
|
1534 | 1540 | self._loadalllazy() |
|
1535 | 1541 | for d, subm in self._dirs.items(): |
|
1536 | 1542 | for subtree in subm.walksubtrees(matcher=matcher): |
|
1537 | 1543 | yield subtree |
|
1538 | 1544 | |
|
1539 | 1545 | |
|
1540 | 1546 | treemanifest = interfaceutil.implementer(repository.imanifestdict)(TreeManifest) |
|
1541 | 1547 | |
|
1542 | 1548 | if typing.TYPE_CHECKING: |
|
1543 | 1549 | treemanifest = TreeManifest |
|
1544 | 1550 | |
|
1545 | 1551 | |
|
1546 | 1552 | class manifestfulltextcache(util.lrucachedict): |
|
1547 | 1553 | """File-backed LRU cache for the manifest cache |
|
1548 | 1554 | |
|
1549 | 1555 | File consists of entries, up to EOF: |
|
1550 | 1556 | |
|
1551 | 1557 | - 20 bytes node, 4 bytes length, <length> manifest data |
|
1552 | 1558 | |
|
1553 | 1559 | These are written in reverse cache order (oldest to newest). |
|
1554 | 1560 | |
|
1555 | 1561 | """ |
|
1556 | 1562 | |
|
1557 | 1563 | _file = b'manifestfulltextcache' |
|
1558 | 1564 | |
|
1559 | 1565 | def __init__(self, max): |
|
1560 | 1566 | super(manifestfulltextcache, self).__init__(max) |
|
1561 | 1567 | self._dirty = False |
|
1562 | 1568 | self._read = False |
|
1563 | 1569 | self._opener = None |
|
1564 | 1570 | |
|
1565 | 1571 | def read(self): |
|
1566 | 1572 | if self._read or self._opener is None: |
|
1567 | 1573 | return |
|
1568 | 1574 | |
|
1569 | 1575 | try: |
|
1570 | 1576 | with self._opener(self._file) as fp: |
|
1571 | 1577 | set = super(manifestfulltextcache, self).__setitem__ |
|
1572 | 1578 | # ignore trailing data, this is a cache, corruption is skipped |
|
1573 | 1579 | while True: |
|
1574 | 1580 | # TODO do we need to do work here for sha1 portability? |
|
1575 | 1581 | node = fp.read(20) |
|
1576 | 1582 | if len(node) < 20: |
|
1577 | 1583 | break |
|
1578 | 1584 | try: |
|
1579 | 1585 | size = struct.unpack(b'>L', fp.read(4))[0] |
|
1580 | 1586 | except struct.error: |
|
1581 | 1587 | break |
|
1582 | 1588 | value = bytearray(fp.read(size)) |
|
1583 | 1589 | if len(value) != size: |
|
1584 | 1590 | break |
|
1585 | 1591 | set(node, value) |
|
1586 | 1592 | except IOError: |
|
1587 | 1593 | # the file is allowed to be missing |
|
1588 | 1594 | pass |
|
1589 | 1595 | |
|
1590 | 1596 | self._read = True |
|
1591 | 1597 | self._dirty = False |
|
1592 | 1598 | |
|
1593 | 1599 | def write(self): |
|
1594 | 1600 | if not self._dirty or self._opener is None: |
|
1595 | 1601 | return |
|
1596 | 1602 | # rotate backwards to the first used node |
|
1597 | 1603 | try: |
|
1598 | 1604 | with self._opener( |
|
1599 | 1605 | self._file, b'w', atomictemp=True, checkambig=True |
|
1600 | 1606 | ) as fp: |
|
1601 | 1607 | node = self._head.prev |
|
1602 | 1608 | while True: |
|
1603 | 1609 | if node.key in self._cache: |
|
1604 | 1610 | fp.write(node.key) |
|
1605 | 1611 | fp.write(struct.pack(b'>L', len(node.value))) |
|
1606 | 1612 | fp.write(node.value) |
|
1607 | 1613 | if node is self._head: |
|
1608 | 1614 | break |
|
1609 | 1615 | node = node.prev |
|
1610 | 1616 | except IOError: |
|
1611 | 1617 | # We could not write the cache (eg: permission error) |
|
1612 | 1618 | # the content can be missing. |
|
1613 | 1619 | # |
|
1614 | 1620 | # We could try harder and see if we could recreate a wcache |
|
1615 | 1621 | # directory were we coudl write too. |
|
1616 | 1622 | # |
|
1617 | 1623 | # XXX the error pass silently, having some way to issue an error |
|
1618 | 1624 | # log `ui.log` would be nice. |
|
1619 | 1625 | pass |
|
1620 | 1626 | |
|
1621 | 1627 | def __len__(self): |
|
1622 | 1628 | if not self._read: |
|
1623 | 1629 | self.read() |
|
1624 | 1630 | return super(manifestfulltextcache, self).__len__() |
|
1625 | 1631 | |
|
1626 | 1632 | def __contains__(self, k): |
|
1627 | 1633 | if not self._read: |
|
1628 | 1634 | self.read() |
|
1629 | 1635 | return super(manifestfulltextcache, self).__contains__(k) |
|
1630 | 1636 | |
|
1631 | 1637 | def __iter__(self): |
|
1632 | 1638 | if not self._read: |
|
1633 | 1639 | self.read() |
|
1634 | 1640 | return super(manifestfulltextcache, self).__iter__() |
|
1635 | 1641 | |
|
1636 | 1642 | def __getitem__(self, k): |
|
1637 | 1643 | if not self._read: |
|
1638 | 1644 | self.read() |
|
1639 | 1645 | # the cache lru order can change on read |
|
1640 | 1646 | setdirty = self._cache.get(k) is not self._head |
|
1641 | 1647 | value = super(manifestfulltextcache, self).__getitem__(k) |
|
1642 | 1648 | if setdirty: |
|
1643 | 1649 | self._dirty = True |
|
1644 | 1650 | return value |
|
1645 | 1651 | |
|
1646 | 1652 | def __setitem__(self, k, v): |
|
1647 | 1653 | if not self._read: |
|
1648 | 1654 | self.read() |
|
1649 | 1655 | super(manifestfulltextcache, self).__setitem__(k, v) |
|
1650 | 1656 | self._dirty = True |
|
1651 | 1657 | |
|
1652 | 1658 | def __delitem__(self, k): |
|
1653 | 1659 | if not self._read: |
|
1654 | 1660 | self.read() |
|
1655 | 1661 | super(manifestfulltextcache, self).__delitem__(k) |
|
1656 | 1662 | self._dirty = True |
|
1657 | 1663 | |
|
1658 | 1664 | def get(self, k, default=None): |
|
1659 | 1665 | if not self._read: |
|
1660 | 1666 | self.read() |
|
1661 | 1667 | return super(manifestfulltextcache, self).get(k, default=default) |
|
1662 | 1668 | |
|
1663 | 1669 | def clear(self, clear_persisted_data=False): |
|
1664 | 1670 | super(manifestfulltextcache, self).clear() |
|
1665 | 1671 | if clear_persisted_data: |
|
1666 | 1672 | self._dirty = True |
|
1667 | 1673 | self.write() |
|
1668 | 1674 | self._read = False |
|
1669 | 1675 | |
|
1670 | 1676 | |
|
1671 | 1677 | # and upper bound of what we expect from compression |
|
1672 | 1678 | # (real live value seems to be "3") |
|
1673 | 1679 | MAXCOMPRESSION = 3 |
|
1674 | 1680 | |
|
1675 | 1681 | |
|
1676 | 1682 | class FastdeltaUnavailable(Exception): |
|
1677 | 1683 | """Exception raised when fastdelta isn't usable on a manifest.""" |
|
1678 | 1684 | |
|
1679 | 1685 | |
|
1680 | 1686 | class ManifestRevlog: |
|
1681 | 1687 | """A revlog that stores manifest texts. This is responsible for caching the |
|
1682 | 1688 | full-text manifest contents. |
|
1683 | 1689 | """ |
|
1684 | 1690 | |
|
1685 | 1691 | def __init__( |
|
1686 | 1692 | self, |
|
1687 | 1693 | nodeconstants, |
|
1688 | 1694 | opener, |
|
1689 | 1695 | tree=b'', |
|
1690 | 1696 | dirlogcache=None, |
|
1691 | 1697 | treemanifest=False, |
|
1692 | 1698 | ): |
|
1693 | 1699 | """Constructs a new manifest revlog |
|
1694 | 1700 | |
|
1695 | 1701 | `indexfile` - used by extensions to have two manifests at once, like |
|
1696 | 1702 | when transitioning between flatmanifeset and treemanifests. |
|
1697 | 1703 | |
|
1698 | 1704 | `treemanifest` - used to indicate this is a tree manifest revlog. Opener |
|
1699 | 1705 | options can also be used to make this a tree manifest revlog. The opener |
|
1700 | 1706 | option takes precedence, so if it is set to True, we ignore whatever |
|
1701 | 1707 | value is passed in to the constructor. |
|
1702 | 1708 | """ |
|
1703 | 1709 | self.nodeconstants = nodeconstants |
|
1704 | 1710 | # During normal operations, we expect to deal with not more than four |
|
1705 | 1711 | # revs at a time (such as during commit --amend). When rebasing large |
|
1706 | 1712 | # stacks of commits, the number can go up, hence the config knob below. |
|
1707 | 1713 | cachesize = 4 |
|
1708 | 1714 | optiontreemanifest = False |
|
1709 | 1715 | persistentnodemap = False |
|
1710 | 1716 | opts = getattr(opener, 'options', None) |
|
1711 | 1717 | if opts is not None: |
|
1712 | 1718 | cachesize = opts.get(b'manifestcachesize', cachesize) |
|
1713 | 1719 | optiontreemanifest = opts.get(b'treemanifest', False) |
|
1714 | 1720 | persistentnodemap = opts.get(b'persistent-nodemap', False) |
|
1715 | 1721 | |
|
1716 | 1722 | self._treeondisk = optiontreemanifest or treemanifest |
|
1717 | 1723 | |
|
1718 | 1724 | self._fulltextcache = manifestfulltextcache(cachesize) |
|
1719 | 1725 | |
|
1720 | 1726 | if tree: |
|
1721 | 1727 | assert self._treeondisk, (tree, b'opts is %r' % opts) |
|
1722 | 1728 | |
|
1723 | 1729 | radix = b'00manifest' |
|
1724 | 1730 | if tree: |
|
1725 | 1731 | radix = b"meta/" + tree + radix |
|
1726 | 1732 | |
|
1727 | 1733 | self.tree = tree |
|
1728 | 1734 | |
|
1729 | 1735 | # The dirlogcache is kept on the root manifest log |
|
1730 | 1736 | if tree: |
|
1731 | 1737 | self._dirlogcache = dirlogcache |
|
1732 | 1738 | else: |
|
1733 | 1739 | self._dirlogcache = {b'': self} |
|
1734 | 1740 | |
|
1735 | 1741 | self._revlog = revlog.revlog( |
|
1736 | 1742 | opener, |
|
1737 | 1743 | target=(revlog_constants.KIND_MANIFESTLOG, self.tree), |
|
1738 | 1744 | radix=radix, |
|
1739 | 1745 | # only root indexfile is cached |
|
1740 | 1746 | checkambig=not bool(tree), |
|
1741 | 1747 | mmaplargeindex=True, |
|
1742 | 1748 | upperboundcomp=MAXCOMPRESSION, |
|
1743 | 1749 | persistentnodemap=persistentnodemap, |
|
1744 | 1750 | ) |
|
1745 | 1751 | |
|
1746 | 1752 | self.index = self._revlog.index |
|
1747 | 1753 | |
|
1748 | 1754 | def get_revlog(self): |
|
1749 | 1755 | """return an actual revlog instance if any |
|
1750 | 1756 | |
|
1751 | 1757 | This exist because a lot of code leverage the fact the underlying |
|
1752 | 1758 | storage is a revlog for optimization, so giving simple way to access |
|
1753 | 1759 | the revlog instance helps such code. |
|
1754 | 1760 | """ |
|
1755 | 1761 | return self._revlog |
|
1756 | 1762 | |
|
1757 | 1763 | def _setupmanifestcachehooks(self, repo): |
|
1758 | 1764 | """Persist the manifestfulltextcache on lock release""" |
|
1759 | 1765 | if not hasattr(repo, '_wlockref'): |
|
1760 | 1766 | return |
|
1761 | 1767 | |
|
1762 | 1768 | self._fulltextcache._opener = repo.wcachevfs |
|
1763 | 1769 | if repo._currentlock(repo._wlockref) is None: |
|
1764 | 1770 | return |
|
1765 | 1771 | |
|
1766 | 1772 | reporef = weakref.ref(repo) |
|
1767 | 1773 | manifestrevlogref = weakref.ref(self) |
|
1768 | 1774 | |
|
1769 | 1775 | def persistmanifestcache(success): |
|
1770 | 1776 | # Repo is in an unknown state, do not persist. |
|
1771 | 1777 | if not success: |
|
1772 | 1778 | return |
|
1773 | 1779 | |
|
1774 | 1780 | repo = reporef() |
|
1775 | 1781 | self = manifestrevlogref() |
|
1776 | 1782 | if repo is None or self is None: |
|
1777 | 1783 | return |
|
1778 | 1784 | if repo.manifestlog.getstorage(b'') is not self: |
|
1779 | 1785 | # there's a different manifest in play now, abort |
|
1780 | 1786 | return |
|
1781 | 1787 | self._fulltextcache.write() |
|
1782 | 1788 | |
|
1783 | 1789 | repo._afterlock(persistmanifestcache) |
|
1784 | 1790 | |
|
1785 | 1791 | @property |
|
1786 | 1792 | def fulltextcache(self): |
|
1787 | 1793 | return self._fulltextcache |
|
1788 | 1794 | |
|
1789 | 1795 | def clearcaches(self, clear_persisted_data=False): |
|
1790 | 1796 | self._revlog.clearcaches() |
|
1791 | 1797 | self._fulltextcache.clear(clear_persisted_data=clear_persisted_data) |
|
1792 | 1798 | self._dirlogcache = {self.tree: self} |
|
1793 | 1799 | |
|
1794 | 1800 | def dirlog(self, d): |
|
1795 | 1801 | if d: |
|
1796 | 1802 | assert self._treeondisk |
|
1797 | 1803 | if d not in self._dirlogcache: |
|
1798 | 1804 | mfrevlog = manifestrevlog( |
|
1799 | 1805 | self.nodeconstants, |
|
1800 | 1806 | self.opener, |
|
1801 | 1807 | d, |
|
1802 | 1808 | self._dirlogcache, |
|
1803 | 1809 | treemanifest=self._treeondisk, |
|
1804 | 1810 | ) |
|
1805 | 1811 | self._dirlogcache[d] = mfrevlog |
|
1806 | 1812 | return self._dirlogcache[d] |
|
1807 | 1813 | |
|
1808 | 1814 | def add( |
|
1809 | 1815 | self, |
|
1810 | 1816 | m, |
|
1811 | 1817 | transaction, |
|
1812 | 1818 | link, |
|
1813 | 1819 | p1, |
|
1814 | 1820 | p2, |
|
1815 | 1821 | added: Iterable[bytes], |
|
1816 | 1822 | removed: Iterable[bytes], |
|
1817 | 1823 | readtree=None, |
|
1818 | 1824 | match=None, |
|
1819 | 1825 | ): |
|
1820 | 1826 | """add some manifest entry in to the manifest log |
|
1821 | 1827 | |
|
1822 | 1828 | input: |
|
1823 | 1829 | |
|
1824 | 1830 | m: the manifest dict we want to store |
|
1825 | 1831 | transaction: the open transaction |
|
1826 | 1832 | p1: manifest-node of p1 |
|
1827 | 1833 | p2: manifest-node of p2 |
|
1828 | 1834 | added: file added/changed compared to parent |
|
1829 | 1835 | removed: file removed compared to parent |
|
1830 | 1836 | |
|
1831 | 1837 | tree manifest input: |
|
1832 | 1838 | |
|
1833 | 1839 | readtree: a function to read a subtree |
|
1834 | 1840 | match: a filematcher for the subpart of the tree manifest |
|
1835 | 1841 | """ |
|
1836 | 1842 | try: |
|
1837 | 1843 | if p1 not in self.fulltextcache: |
|
1838 | 1844 | raise FastdeltaUnavailable() |
|
1839 | 1845 | # If our first parent is in the manifest cache, we can |
|
1840 | 1846 | # compute a delta here using properties we know about the |
|
1841 | 1847 | # manifest up-front, which may save time later for the |
|
1842 | 1848 | # revlog layer. |
|
1843 | 1849 | |
|
1844 | 1850 | _checkforbidden(added) |
|
1845 | 1851 | # combine the changed lists into one sorted iterator |
|
1846 | 1852 | work = heapq.merge( |
|
1847 | 1853 | [(x, False) for x in sorted(added)], |
|
1848 | 1854 | [(x, True) for x in sorted(removed)], |
|
1849 | 1855 | ) |
|
1850 | 1856 | |
|
1851 | 1857 | arraytext, deltatext = m.fastdelta(self.fulltextcache[p1], work) |
|
1852 | 1858 | cachedelta = self._revlog.rev(p1), deltatext |
|
1853 | 1859 | text = util.buffer(arraytext) |
|
1854 | 1860 | rev = self._revlog.addrevision( |
|
1855 | 1861 | text, transaction, link, p1, p2, cachedelta |
|
1856 | 1862 | ) |
|
1857 | 1863 | n = self._revlog.node(rev) |
|
1858 | 1864 | except FastdeltaUnavailable: |
|
1859 | 1865 | # The first parent manifest isn't already loaded or the |
|
1860 | 1866 | # manifest implementation doesn't support fastdelta, so |
|
1861 | 1867 | # we'll just encode a fulltext of the manifest and pass |
|
1862 | 1868 | # that through to the revlog layer, and let it handle the |
|
1863 | 1869 | # delta process. |
|
1864 | 1870 | if self._treeondisk: |
|
1865 | 1871 | assert readtree, b"readtree must be set for treemanifest writes" |
|
1866 | 1872 | assert match, b"match must be specified for treemanifest writes" |
|
1867 | 1873 | m1 = readtree(self.tree, p1) |
|
1868 | 1874 | m2 = readtree(self.tree, p2) |
|
1869 | 1875 | n = self._addtree( |
|
1870 | 1876 | m, transaction, link, m1, m2, readtree, match=match |
|
1871 | 1877 | ) |
|
1872 | 1878 | arraytext = None |
|
1873 | 1879 | else: |
|
1874 | 1880 | text = m.text() |
|
1875 | 1881 | rev = self._revlog.addrevision(text, transaction, link, p1, p2) |
|
1876 | 1882 | n = self._revlog.node(rev) |
|
1877 | 1883 | arraytext = bytearray(text) |
|
1878 | 1884 | |
|
1879 | 1885 | if arraytext is not None: |
|
1880 | 1886 | self.fulltextcache[n] = arraytext |
|
1881 | 1887 | |
|
1882 | 1888 | return n |
|
1883 | 1889 | |
|
1884 | 1890 | def _addtree(self, m, transaction, link, m1, m2, readtree, match): |
|
1885 | 1891 | # If the manifest is unchanged compared to one parent, |
|
1886 | 1892 | # don't write a new revision |
|
1887 | 1893 | if self.tree != b'' and ( |
|
1888 | 1894 | m.unmodifiedsince(m1) or m.unmodifiedsince(m2) |
|
1889 | 1895 | ): |
|
1890 | 1896 | return m.node() |
|
1891 | 1897 | |
|
1892 | 1898 | def writesubtree(subm, subp1, subp2, match): |
|
1893 | 1899 | sublog = self.dirlog(subm.dir()) |
|
1894 | 1900 | sublog.add( |
|
1895 | 1901 | subm, |
|
1896 | 1902 | transaction, |
|
1897 | 1903 | link, |
|
1898 | 1904 | subp1, |
|
1899 | 1905 | subp2, |
|
1900 | 1906 | None, |
|
1901 | 1907 | None, |
|
1902 | 1908 | readtree=readtree, |
|
1903 | 1909 | match=match, |
|
1904 | 1910 | ) |
|
1905 | 1911 | |
|
1906 | 1912 | m.writesubtrees(m1, m2, writesubtree, match) |
|
1907 | 1913 | text = m.dirtext() |
|
1908 | 1914 | n = None |
|
1909 | 1915 | if self.tree != b'': |
|
1910 | 1916 | # Double-check whether contents are unchanged to one parent |
|
1911 | 1917 | if text == m1.dirtext(): |
|
1912 | 1918 | n = m1.node() |
|
1913 | 1919 | elif text == m2.dirtext(): |
|
1914 | 1920 | n = m2.node() |
|
1915 | 1921 | |
|
1916 | 1922 | if not n: |
|
1917 | 1923 | rev = self._revlog.addrevision( |
|
1918 | 1924 | text, transaction, link, m1.node(), m2.node() |
|
1919 | 1925 | ) |
|
1920 | 1926 | n = self._revlog.node(rev) |
|
1921 | 1927 | |
|
1922 | 1928 | # Save nodeid so parent manifest can calculate its nodeid |
|
1923 | 1929 | m.setnode(n) |
|
1924 | 1930 | return n |
|
1925 | 1931 | |
|
1926 | 1932 | def __len__(self): |
|
1927 | 1933 | return len(self._revlog) |
|
1928 | 1934 | |
|
1929 | 1935 | def __iter__(self): |
|
1930 | 1936 | return self._revlog.__iter__() |
|
1931 | 1937 | |
|
1932 | 1938 | def rev(self, node): |
|
1933 | 1939 | return self._revlog.rev(node) |
|
1934 | 1940 | |
|
1935 | 1941 | def node(self, rev): |
|
1936 | 1942 | return self._revlog.node(rev) |
|
1937 | 1943 | |
|
1938 | 1944 | def lookup(self, value): |
|
1939 | 1945 | return self._revlog.lookup(value) |
|
1940 | 1946 | |
|
1941 | 1947 | def parentrevs(self, rev): |
|
1942 | 1948 | return self._revlog.parentrevs(rev) |
|
1943 | 1949 | |
|
1944 | 1950 | def parents(self, node): |
|
1945 | 1951 | return self._revlog.parents(node) |
|
1946 | 1952 | |
|
1947 | 1953 | def linkrev(self, rev): |
|
1948 | 1954 | return self._revlog.linkrev(rev) |
|
1949 | 1955 | |
|
1950 | 1956 | def checksize(self): |
|
1951 | 1957 | return self._revlog.checksize() |
|
1952 | 1958 | |
|
1953 | 1959 | def revision(self, node): |
|
1954 | 1960 | return self._revlog.revision(node) |
|
1955 | 1961 | |
|
1956 | 1962 | def rawdata(self, node): |
|
1957 | 1963 | return self._revlog.rawdata(node) |
|
1958 | 1964 | |
|
1959 | 1965 | def revdiff(self, rev1, rev2): |
|
1960 | 1966 | return self._revlog.revdiff(rev1, rev2) |
|
1961 | 1967 | |
|
1962 | 1968 | def cmp(self, node, text): |
|
1963 | 1969 | return self._revlog.cmp(node, text) |
|
1964 | 1970 | |
|
1965 | 1971 | def deltaparent(self, rev): |
|
1966 | 1972 | return self._revlog.deltaparent(rev) |
|
1967 | 1973 | |
|
1968 | 1974 | def emitrevisions( |
|
1969 | 1975 | self, |
|
1970 | 1976 | nodes, |
|
1971 | 1977 | nodesorder=None, |
|
1972 | 1978 | revisiondata=False, |
|
1973 | 1979 | assumehaveparentrevisions=False, |
|
1974 | 1980 | deltamode=repository.CG_DELTAMODE_STD, |
|
1975 | 1981 | sidedata_helpers=None, |
|
1976 | 1982 | debug_info=None, |
|
1977 | 1983 | ): |
|
1978 | 1984 | return self._revlog.emitrevisions( |
|
1979 | 1985 | nodes, |
|
1980 | 1986 | nodesorder=nodesorder, |
|
1981 | 1987 | revisiondata=revisiondata, |
|
1982 | 1988 | assumehaveparentrevisions=assumehaveparentrevisions, |
|
1983 | 1989 | deltamode=deltamode, |
|
1984 | 1990 | sidedata_helpers=sidedata_helpers, |
|
1985 | 1991 | debug_info=debug_info, |
|
1986 | 1992 | ) |
|
1987 | 1993 | |
|
1988 | 1994 | def addgroup( |
|
1989 | 1995 | self, |
|
1990 | 1996 | deltas, |
|
1991 | 1997 | linkmapper, |
|
1992 | 1998 | transaction, |
|
1993 | 1999 | alwayscache=False, |
|
1994 | 2000 | addrevisioncb=None, |
|
1995 | 2001 | duplicaterevisioncb=None, |
|
1996 | 2002 | debug_info=None, |
|
1997 | 2003 | delta_base_reuse_policy=None, |
|
1998 | 2004 | ): |
|
1999 | 2005 | return self._revlog.addgroup( |
|
2000 | 2006 | deltas, |
|
2001 | 2007 | linkmapper, |
|
2002 | 2008 | transaction, |
|
2003 | 2009 | alwayscache=alwayscache, |
|
2004 | 2010 | addrevisioncb=addrevisioncb, |
|
2005 | 2011 | duplicaterevisioncb=duplicaterevisioncb, |
|
2006 | 2012 | debug_info=debug_info, |
|
2007 | 2013 | delta_base_reuse_policy=delta_base_reuse_policy, |
|
2008 | 2014 | ) |
|
2009 | 2015 | |
|
2010 | 2016 | def rawsize(self, rev): |
|
2011 | 2017 | return self._revlog.rawsize(rev) |
|
2012 | 2018 | |
|
2013 | 2019 | def getstrippoint(self, minlink): |
|
2014 | 2020 | return self._revlog.getstrippoint(minlink) |
|
2015 | 2021 | |
|
2016 | 2022 | def strip(self, minlink, transaction): |
|
2017 | 2023 | return self._revlog.strip(minlink, transaction) |
|
2018 | 2024 | |
|
2019 | 2025 | def files(self): |
|
2020 | 2026 | return self._revlog.files() |
|
2021 | 2027 | |
|
2022 | 2028 | def clone(self, tr, destrevlog, **kwargs): |
|
2023 | 2029 | if not isinstance(destrevlog, manifestrevlog): |
|
2024 | 2030 | raise error.ProgrammingError(b'expected manifestrevlog to clone()') |
|
2025 | 2031 | |
|
2026 | 2032 | return self._revlog.clone(tr, destrevlog._revlog, **kwargs) |
|
2027 | 2033 | |
|
2028 | 2034 | def storageinfo( |
|
2029 | 2035 | self, |
|
2030 | 2036 | exclusivefiles=False, |
|
2031 | 2037 | sharedfiles=False, |
|
2032 | 2038 | revisionscount=False, |
|
2033 | 2039 | trackedsize=False, |
|
2034 | 2040 | storedsize=False, |
|
2035 | 2041 | ): |
|
2036 | 2042 | return self._revlog.storageinfo( |
|
2037 | 2043 | exclusivefiles=exclusivefiles, |
|
2038 | 2044 | sharedfiles=sharedfiles, |
|
2039 | 2045 | revisionscount=revisionscount, |
|
2040 | 2046 | trackedsize=trackedsize, |
|
2041 | 2047 | storedsize=storedsize, |
|
2042 | 2048 | ) |
|
2043 | 2049 | |
|
2044 | 2050 | @property |
|
2045 | 2051 | def opener(self): |
|
2046 | 2052 | return self._revlog.opener |
|
2047 | 2053 | |
|
2048 | 2054 | @opener.setter |
|
2049 | 2055 | def opener(self, value): |
|
2050 | 2056 | self._revlog.opener = value |
|
2051 | 2057 | |
|
2052 | 2058 | |
|
2053 | 2059 | manifestrevlog = interfaceutil.implementer(repository.imanifeststorage)( |
|
2054 | 2060 | ManifestRevlog |
|
2055 | 2061 | ) |
|
2056 | 2062 | |
|
2057 | 2063 | if typing.TYPE_CHECKING: |
|
2058 | 2064 | manifestrevlog = ManifestRevlog |
|
2059 | 2065 | |
|
2060 | 2066 | AnyManifestCtx = Union['ManifestCtx', 'TreeManifestCtx'] |
|
2061 | 2067 | AnyManifestDict = Union[ManifestDict, TreeManifest] |
|
2062 | 2068 | |
|
2063 | 2069 | |
|
2064 | 2070 | class ManifestLog: |
|
2065 | 2071 | """A collection class representing the collection of manifest snapshots |
|
2066 | 2072 | referenced by commits in the repository. |
|
2067 | 2073 | |
|
2068 | 2074 | In this situation, 'manifest' refers to the abstract concept of a snapshot |
|
2069 | 2075 | of the list of files in the given commit. Consumers of the output of this |
|
2070 | 2076 | class do not care about the implementation details of the actual manifests |
|
2071 | 2077 | they receive (i.e. tree or flat or lazily loaded, etc).""" |
|
2072 | 2078 | |
|
2073 | 2079 | def __init__(self, opener, repo, rootstore, narrowmatch): |
|
2074 | 2080 | self.nodeconstants = repo.nodeconstants |
|
2075 | 2081 | usetreemanifest = False |
|
2076 | 2082 | cachesize = 4 |
|
2077 | 2083 | |
|
2078 | 2084 | opts = getattr(opener, 'options', None) |
|
2079 | 2085 | if opts is not None: |
|
2080 | 2086 | usetreemanifest = opts.get(b'treemanifest', usetreemanifest) |
|
2081 | 2087 | cachesize = opts.get(b'manifestcachesize', cachesize) |
|
2082 | 2088 | |
|
2083 | 2089 | self._treemanifests = usetreemanifest |
|
2084 | 2090 | |
|
2085 | 2091 | self._rootstore = rootstore |
|
2086 | 2092 | self._rootstore._setupmanifestcachehooks(repo) |
|
2087 | 2093 | self._narrowmatch = narrowmatch |
|
2088 | 2094 | |
|
2089 | 2095 | # A cache of the manifestctx or treemanifestctx for each directory |
|
2090 | 2096 | self._dirmancache = {} |
|
2091 | 2097 | self._dirmancache[b''] = util.lrucachedict(cachesize) |
|
2092 | 2098 | |
|
2093 | 2099 | self._cachesize = cachesize |
|
2094 | 2100 | |
|
2095 | 2101 | def __getitem__(self, node): |
|
2096 | 2102 | """Retrieves the manifest instance for the given node. Throws a |
|
2097 | 2103 | LookupError if not found. |
|
2098 | 2104 | """ |
|
2099 | 2105 | return self.get(b'', node) |
|
2100 | 2106 | |
|
2101 | 2107 | @property |
|
2102 | 2108 | def narrowed(self): |
|
2103 | 2109 | return not (self._narrowmatch is None or self._narrowmatch.always()) |
|
2104 | 2110 | |
|
2105 | 2111 | def get( |
|
2106 | 2112 | self, tree: bytes, node: bytes, verify: bool = True |
|
2107 | 2113 | ) -> AnyManifestCtx: |
|
2108 | 2114 | """Retrieves the manifest instance for the given node. Throws a |
|
2109 | 2115 | LookupError if not found. |
|
2110 | 2116 | |
|
2111 | 2117 | `verify` - if True an exception will be thrown if the node is not in |
|
2112 | 2118 | the revlog |
|
2113 | 2119 | """ |
|
2114 | 2120 | if node in self._dirmancache.get(tree, ()): |
|
2115 | 2121 | return self._dirmancache[tree][node] |
|
2116 | 2122 | |
|
2117 | 2123 | if not self._narrowmatch.always(): |
|
2118 | 2124 | if not self._narrowmatch.visitdir(tree[:-1]): |
|
2119 | 2125 | return excludeddirmanifestctx(self.nodeconstants, tree, node) |
|
2120 | 2126 | if tree: |
|
2121 | 2127 | if self._rootstore._treeondisk: |
|
2122 | 2128 | if verify: |
|
2123 | 2129 | # Side-effect is LookupError is raised if node doesn't |
|
2124 | 2130 | # exist. |
|
2125 | 2131 | self.getstorage(tree).rev(node) |
|
2126 | 2132 | |
|
2127 | 2133 | m = treemanifestctx(self, tree, node) |
|
2128 | 2134 | else: |
|
2129 | 2135 | raise error.Abort( |
|
2130 | 2136 | _( |
|
2131 | 2137 | b"cannot ask for manifest directory '%s' in a flat " |
|
2132 | 2138 | b"manifest" |
|
2133 | 2139 | ) |
|
2134 | 2140 | % tree |
|
2135 | 2141 | ) |
|
2136 | 2142 | else: |
|
2137 | 2143 | if verify: |
|
2138 | 2144 | # Side-effect is LookupError is raised if node doesn't exist. |
|
2139 | 2145 | self._rootstore.rev(node) |
|
2140 | 2146 | |
|
2141 | 2147 | if self._treemanifests: |
|
2142 | 2148 | m = treemanifestctx(self, b'', node) |
|
2143 | 2149 | else: |
|
2144 | 2150 | m = manifestctx(self, node) |
|
2145 | 2151 | |
|
2146 | 2152 | if node != self.nodeconstants.nullid: |
|
2147 | 2153 | mancache = self._dirmancache.get(tree) |
|
2148 | 2154 | if not mancache: |
|
2149 | 2155 | mancache = util.lrucachedict(self._cachesize) |
|
2150 | 2156 | self._dirmancache[tree] = mancache |
|
2151 | 2157 | mancache[node] = m |
|
2152 | 2158 | return m |
|
2153 | 2159 | |
|
2154 | 2160 | def getstorage(self, tree): |
|
2155 | 2161 | return self._rootstore.dirlog(tree) |
|
2156 | 2162 | |
|
2157 | 2163 | def clearcaches(self, clear_persisted_data: bool = False) -> None: |
|
2158 | 2164 | self._dirmancache.clear() |
|
2159 | 2165 | self._rootstore.clearcaches(clear_persisted_data=clear_persisted_data) |
|
2160 | 2166 | |
|
2161 | 2167 | def rev(self, node) -> int: |
|
2162 | 2168 | return self._rootstore.rev(node) |
|
2163 | 2169 | |
|
2164 | 2170 | def update_caches(self, transaction) -> None: |
|
2165 | 2171 | return self._rootstore._revlog.update_caches(transaction=transaction) |
|
2166 | 2172 | |
|
2167 | 2173 | |
|
2168 | 2174 | manifestlog = interfaceutil.implementer(repository.imanifestlog)(ManifestLog) |
|
2169 | 2175 | |
|
2170 | 2176 | if typing.TYPE_CHECKING: |
|
2171 | 2177 | manifestlog = ManifestLog |
|
2172 | 2178 | |
|
2173 | 2179 | |
|
2174 | 2180 | class MemManifestCtx: |
|
2181 | _manifestdict: ManifestDict | |
|
2182 | ||
|
2175 | 2183 | def __init__(self, manifestlog): |
|
2176 | 2184 | self._manifestlog = manifestlog |
|
2177 | 2185 | self._manifestdict = manifestdict(manifestlog.nodeconstants.nodelen) |
|
2178 | 2186 | |
|
2179 | 2187 | def _storage(self) -> ManifestRevlog: |
|
2180 | 2188 | return self._manifestlog.getstorage(b'') |
|
2181 | 2189 | |
|
2182 | 2190 | def copy(self) -> 'MemManifestCtx': |
|
2183 | 2191 | memmf = memmanifestctx(self._manifestlog) |
|
2184 | 2192 | memmf._manifestdict = self.read().copy() |
|
2185 | 2193 | return memmf |
|
2186 | 2194 | |
|
2187 | 2195 | def read(self) -> 'ManifestDict': |
|
2188 | 2196 | return self._manifestdict |
|
2189 | 2197 | |
|
2190 | 2198 | def write(self, transaction, link, p1, p2, added, removed, match=None): |
|
2191 | 2199 | return self._storage().add( |
|
2192 | 2200 | self._manifestdict, |
|
2193 | 2201 | transaction, |
|
2194 | 2202 | link, |
|
2195 | 2203 | p1, |
|
2196 | 2204 | p2, |
|
2197 | 2205 | added, |
|
2198 | 2206 | removed, |
|
2199 | 2207 | match=match, |
|
2200 | 2208 | ) |
|
2201 | 2209 | |
|
2202 | 2210 | |
|
2203 | 2211 | memmanifestctx = interfaceutil.implementer( |
|
2204 | 2212 | repository.imanifestrevisionwritable |
|
2205 | 2213 | )(MemManifestCtx) |
|
2206 | 2214 | |
|
2207 | 2215 | if typing.TYPE_CHECKING: |
|
2208 | 2216 | memmanifestctx = MemManifestCtx |
|
2209 | 2217 | |
|
2210 | 2218 | |
|
2211 | 2219 | class ManifestCtx: |
|
2212 | 2220 | """A class representing a single revision of a manifest, including its |
|
2213 | 2221 | contents, its parent revs, and its linkrev. |
|
2214 | 2222 | """ |
|
2215 | 2223 | |
|
2224 | _data: Optional[ManifestDict] | |
|
2225 | ||
|
2216 | 2226 | def __init__(self, manifestlog, node): |
|
2217 | 2227 | self._manifestlog = manifestlog |
|
2218 | 2228 | self._data = None |
|
2219 | 2229 | |
|
2220 | 2230 | self._node = node |
|
2221 | 2231 | |
|
2222 | 2232 | # TODO: We eventually want p1, p2, and linkrev exposed on this class, |
|
2223 | 2233 | # but let's add it later when something needs it and we can load it |
|
2224 | 2234 | # lazily. |
|
2225 | 2235 | # self.p1, self.p2 = store.parents(node) |
|
2226 | 2236 | # rev = store.rev(node) |
|
2227 | 2237 | # self.linkrev = store.linkrev(rev) |
|
2228 | 2238 | |
|
2229 | 2239 | def _storage(self) -> 'ManifestRevlog': |
|
2230 | 2240 | return self._manifestlog.getstorage(b'') |
|
2231 | 2241 | |
|
2232 | 2242 | def node(self) -> bytes: |
|
2233 | 2243 | return self._node |
|
2234 | 2244 | |
|
2235 | 2245 | def copy(self) -> MemManifestCtx: |
|
2236 | 2246 | memmf = memmanifestctx(self._manifestlog) |
|
2237 | 2247 | memmf._manifestdict = self.read().copy() |
|
2238 | 2248 | return memmf |
|
2239 | 2249 | |
|
2240 | 2250 | @propertycache |
|
2241 | 2251 | def parents(self) -> Tuple[bytes, bytes]: |
|
2242 | 2252 | return self._storage().parents(self._node) |
|
2243 | 2253 | |
|
2244 | 2254 | def read(self) -> 'ManifestDict': |
|
2245 | 2255 | if self._data is None: |
|
2246 | 2256 | nc = self._manifestlog.nodeconstants |
|
2247 | 2257 | if self._node == nc.nullid: |
|
2248 | 2258 | self._data = manifestdict(nc.nodelen) |
|
2249 | 2259 | else: |
|
2250 | 2260 | store = self._storage() |
|
2251 | 2261 | if self._node in store.fulltextcache: |
|
2252 | 2262 | text = pycompat.bytestr(store.fulltextcache[self._node]) |
|
2253 | 2263 | else: |
|
2254 | 2264 | text = store.revision(self._node) |
|
2255 | 2265 | arraytext = bytearray(text) |
|
2256 | 2266 | store.fulltextcache[self._node] = arraytext |
|
2257 | 2267 | self._data = manifestdict(nc.nodelen, text) |
|
2258 | 2268 | return self._data |
|
2259 | 2269 | |
|
2260 | 2270 | def readfast(self, shallow: bool = False) -> 'ManifestDict': |
|
2261 | 2271 | """Calls either readdelta or read, based on which would be less work. |
|
2262 | 2272 | readdelta is called if the delta is against the p1, and therefore can be |
|
2263 | 2273 | read quickly. |
|
2264 | 2274 | |
|
2265 | 2275 | If `shallow` is True, nothing changes since this is a flat manifest. |
|
2266 | 2276 | """ |
|
2267 | 2277 | util.nouideprecwarn( |
|
2268 | 2278 | b'"readfast" is deprecated use "read_any_fast_delta" or "read_delta_parents"', |
|
2269 | 2279 | b"6.9", |
|
2270 | 2280 | stacklevel=2, |
|
2271 | 2281 | ) |
|
2272 | 2282 | store = self._storage() |
|
2273 | 2283 | r = store.rev(self._node) |
|
2274 | 2284 | deltaparent = store.deltaparent(r) |
|
2275 | 2285 | if deltaparent != nullrev and deltaparent in store.parentrevs(r): |
|
2276 | 2286 | return self.readdelta() |
|
2277 | 2287 | return self.read() |
|
2278 | 2288 | |
|
2279 | 2289 | def readdelta(self, shallow: bool = False) -> 'ManifestDict': |
|
2280 | 2290 | """Returns a manifest containing just the entries that are present |
|
2281 | 2291 | in this manifest, but not in its p1 manifest. This is efficient to read |
|
2282 | 2292 | if the revlog delta is already p1. |
|
2283 | 2293 | |
|
2284 | 2294 | Changing the value of `shallow` has no effect on flat manifests. |
|
2285 | 2295 | """ |
|
2286 | 2296 | util.nouideprecwarn( |
|
2287 | 2297 | b'"readfast" is deprecated use "read_any_fast_delta" or "read_delta_new_entries"', |
|
2288 | 2298 | b"6.9", |
|
2289 | 2299 | stacklevel=2, |
|
2290 | 2300 | ) |
|
2291 | 2301 | store = self._storage() |
|
2292 | 2302 | r = store.rev(self._node) |
|
2293 | 2303 | d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r)) |
|
2294 | 2304 | return manifestdict(store.nodeconstants.nodelen, d) |
|
2295 | 2305 | |
|
2296 | 2306 | def read_any_fast_delta( |
|
2297 | 2307 | self, |
|
2298 | 2308 | valid_bases: Optional[Collection[int]] = None, |
|
2299 | 2309 | *, |
|
2300 | 2310 | shallow: bool = False, |
|
2301 | 2311 | ) -> Tuple[Optional[int], ManifestDict]: |
|
2302 | 2312 | """see `imanifestrevisionstored` documentation""" |
|
2303 | 2313 | store = self._storage() |
|
2304 | 2314 | r = store.rev(self._node) |
|
2305 | 2315 | deltaparent = store.deltaparent(r) |
|
2306 | 2316 | if valid_bases is None: |
|
2307 | 2317 | # make sure the next check is True |
|
2308 | 2318 | valid_bases = (deltaparent,) |
|
2309 | 2319 | if deltaparent != nullrev and deltaparent in valid_bases: |
|
2310 | 2320 | d = mdiff.patchtext(store.revdiff(deltaparent, r)) |
|
2311 | 2321 | return ( |
|
2312 | 2322 | deltaparent, |
|
2313 | 2323 | manifestdict(store.nodeconstants.nodelen, d), |
|
2314 | 2324 | ) |
|
2315 | 2325 | return (None, self.read()) |
|
2316 | 2326 | |
|
2317 | 2327 | def read_delta_parents( |
|
2318 | 2328 | self, |
|
2319 | 2329 | *, |
|
2320 | 2330 | shallow: bool = False, |
|
2321 | 2331 | exact: bool = True, |
|
2322 | 2332 | ) -> ManifestDict: |
|
2323 | 2333 | """see `interface.imanifestrevisionbase` documentations""" |
|
2324 | 2334 | store = self._storage() |
|
2325 | 2335 | r = store.rev(self._node) |
|
2326 | 2336 | deltaparent = store.deltaparent(r) |
|
2327 | 2337 | parents = [p for p in store.parentrevs(r) if p is not nullrev] |
|
2328 | 2338 | if not exact and deltaparent in parents: |
|
2329 | 2339 | d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r)) |
|
2330 | 2340 | return manifestdict(store.nodeconstants.nodelen, d) |
|
2331 | 2341 | elif not exact or len(parents) == 0: |
|
2332 | 2342 | return self.read() |
|
2333 | 2343 | elif len(parents) == 1: |
|
2334 | 2344 | p = parents[0] |
|
2335 | 2345 | d = mdiff.patchtext(store.revdiff(p, r)) |
|
2336 | 2346 | return manifestdict(store.nodeconstants.nodelen, d) |
|
2337 | 2347 | else: |
|
2338 | 2348 | p1, p2 = parents |
|
2339 | 2349 | d1 = mdiff.patchtext(store.revdiff(p1, r)) |
|
2340 | 2350 | d2 = mdiff.patchtext(store.revdiff(p2, r)) |
|
2341 | 2351 | d1 = manifestdict(store.nodeconstants.nodelen, d1) |
|
2342 | 2352 | d2 = manifestdict(store.nodeconstants.nodelen, d2) |
|
2343 | 2353 | md = manifestdict(store.nodeconstants.nodelen) |
|
2344 | 2354 | for f, new_node, new_flag in d1.iterentries(): |
|
2345 | 2355 | if f not in d2: |
|
2346 | 2356 | continue |
|
2347 | 2357 | if new_node is not None: |
|
2348 | 2358 | md.set(f, new_node, new_flag) |
|
2349 | 2359 | return md |
|
2350 | 2360 | |
|
2351 | 2361 | def read_delta_new_entries(self, *, shallow=False) -> ManifestDict: |
|
2352 | 2362 | """see `interface.imanifestrevisionbase` documentations""" |
|
2353 | 2363 | # If we are using narrow, returning a delta against an arbitrary |
|
2354 | 2364 | # changeset might return file outside the narrowspec. This can create |
|
2355 | 2365 | # issue when running validation server side with strict security as |
|
2356 | 2366 | # push from low priviledge usage might be seen as adding new revision |
|
2357 | 2367 | # for files they cannot touch. So we are strict if narrow is involved. |
|
2358 | 2368 | if self._manifestlog.narrowed: |
|
2359 | 2369 | return self.read_delta_parents(shallow=shallow, exact=True) |
|
2360 | 2370 | store = self._storage() |
|
2361 | 2371 | r = store.rev(self._node) |
|
2362 | 2372 | d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r)) |
|
2363 | 2373 | return manifestdict(store.nodeconstants.nodelen, d) |
|
2364 | 2374 | |
|
2365 | 2375 | def find(self, key: bytes) -> Tuple[bytes, bytes]: |
|
2366 | 2376 | return self.read().find(key) |
|
2367 | 2377 | |
|
2368 | 2378 | |
|
2369 | 2379 | manifestctx = interfaceutil.implementer(repository.imanifestrevisionstored)( |
|
2370 | 2380 | ManifestCtx |
|
2371 | 2381 | ) |
|
2372 | 2382 | |
|
2373 | 2383 | if typing.TYPE_CHECKING: |
|
2374 | 2384 | manifestctx = ManifestCtx |
|
2375 | 2385 | |
|
2376 | 2386 | |
|
2377 | 2387 | class MemTreeManifestCtx: |
|
2388 | _treemanifest: TreeManifest | |
|
2389 | ||
|
2378 | 2390 | def __init__(self, manifestlog, dir=b''): |
|
2379 | 2391 | self._manifestlog = manifestlog |
|
2380 | 2392 | self._dir = dir |
|
2381 | 2393 | self._treemanifest = treemanifest(manifestlog.nodeconstants) |
|
2382 | 2394 | |
|
2383 | 2395 | def _storage(self) -> ManifestRevlog: |
|
2384 | 2396 | return self._manifestlog.getstorage(b'') |
|
2385 | 2397 | |
|
2386 | 2398 | def copy(self) -> 'MemTreeManifestCtx': |
|
2387 | 2399 | memmf = memtreemanifestctx(self._manifestlog, dir=self._dir) |
|
2388 | 2400 | memmf._treemanifest = self._treemanifest.copy() |
|
2389 | 2401 | return memmf |
|
2390 | 2402 | |
|
2391 | 2403 | def read(self) -> 'TreeManifest': |
|
2392 | 2404 | return self._treemanifest |
|
2393 | 2405 | |
|
2394 | 2406 | def write(self, transaction, link, p1, p2, added, removed, match=None): |
|
2395 | 2407 | def readtree(dir, node): |
|
2396 | 2408 | return self._manifestlog.get(dir, node).read() |
|
2397 | 2409 | |
|
2398 | 2410 | return self._storage().add( |
|
2399 | 2411 | self._treemanifest, |
|
2400 | 2412 | transaction, |
|
2401 | 2413 | link, |
|
2402 | 2414 | p1, |
|
2403 | 2415 | p2, |
|
2404 | 2416 | added, |
|
2405 | 2417 | removed, |
|
2406 | 2418 | readtree=readtree, |
|
2407 | 2419 | match=match, |
|
2408 | 2420 | ) |
|
2409 | 2421 | |
|
2410 | 2422 | |
|
2411 | 2423 | memtreemanifestctx = interfaceutil.implementer( |
|
2412 | 2424 | repository.imanifestrevisionwritable |
|
2413 | 2425 | )(MemTreeManifestCtx) |
|
2414 | 2426 | |
|
2415 | 2427 | if typing.TYPE_CHECKING: |
|
2416 | 2428 | memtreemanifestctx = MemTreeManifestCtx |
|
2417 | 2429 | |
|
2418 | 2430 | |
|
2419 | 2431 | class TreeManifestCtx: |
|
2432 | _data: Optional[TreeManifest] | |
|
2433 | ||
|
2420 | 2434 | def __init__(self, manifestlog, dir, node): |
|
2421 | 2435 | self._manifestlog = manifestlog |
|
2422 | 2436 | self._dir = dir |
|
2423 | 2437 | self._data = None |
|
2424 | 2438 | |
|
2425 | 2439 | self._node = node |
|
2426 | 2440 | |
|
2427 | 2441 | # TODO: Load p1/p2/linkrev lazily. They need to be lazily loaded so that |
|
2428 | 2442 | # we can instantiate treemanifestctx objects for directories we don't |
|
2429 | 2443 | # have on disk. |
|
2430 | 2444 | # self.p1, self.p2 = store.parents(node) |
|
2431 | 2445 | # rev = store.rev(node) |
|
2432 | 2446 | # self.linkrev = store.linkrev(rev) |
|
2433 | 2447 | |
|
2434 | 2448 | def _storage(self) -> ManifestRevlog: |
|
2435 | 2449 | narrowmatch = self._manifestlog._narrowmatch |
|
2436 | 2450 | if not narrowmatch.always(): |
|
2437 | 2451 | if not narrowmatch.visitdir(self._dir[:-1]): |
|
2438 | 2452 | return excludedmanifestrevlog( |
|
2439 | 2453 | self._manifestlog.nodeconstants, self._dir |
|
2440 | 2454 | ) |
|
2441 | 2455 | return self._manifestlog.getstorage(self._dir) |
|
2442 | 2456 | |
|
2443 | 2457 | def read(self) -> 'TreeManifest': |
|
2444 | 2458 | if self._data is None: |
|
2445 | 2459 | store = self._storage() |
|
2446 | 2460 | if self._node == self._manifestlog.nodeconstants.nullid: |
|
2447 | 2461 | self._data = treemanifest(self._manifestlog.nodeconstants) |
|
2448 | 2462 | # TODO accessing non-public API |
|
2449 | 2463 | elif store._treeondisk: |
|
2450 | 2464 | m = treemanifest(self._manifestlog.nodeconstants, dir=self._dir) |
|
2451 | 2465 | |
|
2452 | 2466 | def gettext(): |
|
2453 | 2467 | return store.revision(self._node) |
|
2454 | 2468 | |
|
2455 | 2469 | def readsubtree(dir, subm): |
|
2456 | 2470 | # Set verify to False since we need to be able to create |
|
2457 | 2471 | # subtrees for trees that don't exist on disk. |
|
2458 | 2472 | return self._manifestlog.get(dir, subm, verify=False).read() |
|
2459 | 2473 | |
|
2460 | 2474 | m.read(gettext, readsubtree) |
|
2461 | 2475 | m.setnode(self._node) |
|
2462 | 2476 | self._data = m |
|
2463 | 2477 | else: |
|
2464 | 2478 | if self._node in store.fulltextcache: |
|
2465 | 2479 | text = pycompat.bytestr(store.fulltextcache[self._node]) |
|
2466 | 2480 | else: |
|
2467 | 2481 | text = store.revision(self._node) |
|
2468 | 2482 | arraytext = bytearray(text) |
|
2469 | 2483 | store.fulltextcache[self._node] = arraytext |
|
2470 | 2484 | self._data = treemanifest( |
|
2471 | 2485 | self._manifestlog.nodeconstants, dir=self._dir, text=text |
|
2472 | 2486 | ) |
|
2473 | 2487 | |
|
2474 | 2488 | return self._data |
|
2475 | 2489 | |
|
2476 | 2490 | def node(self) -> bytes: |
|
2477 | 2491 | return self._node |
|
2478 | 2492 | |
|
2479 | 2493 | def copy(self) -> 'MemTreeManifestCtx': |
|
2480 | 2494 | memmf = memtreemanifestctx(self._manifestlog, dir=self._dir) |
|
2481 | 2495 | memmf._treemanifest = self.read().copy() |
|
2482 | 2496 | return memmf |
|
2483 | 2497 | |
|
2484 | 2498 | @propertycache |
|
2485 | 2499 | def parents(self) -> Tuple[bytes, bytes]: |
|
2486 | 2500 | return self._storage().parents(self._node) |
|
2487 | 2501 | |
|
2488 | 2502 | def readdelta(self, shallow: bool = False) -> AnyManifestDict: |
|
2489 | 2503 | """see `imanifestrevisionstored` documentation""" |
|
2490 | 2504 | util.nouideprecwarn( |
|
2491 | 2505 | b'"readdelta" is deprecated use "read_any_fast_delta" or "read_delta_new_entries"', |
|
2492 | 2506 | b"6.9", |
|
2493 | 2507 | stacklevel=2, |
|
2494 | 2508 | ) |
|
2495 | 2509 | store = self._storage() |
|
2496 | 2510 | if shallow: |
|
2497 | 2511 | r = store.rev(self._node) |
|
2498 | 2512 | d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r)) |
|
2499 | 2513 | return manifestdict(store.nodeconstants.nodelen, d) |
|
2500 | 2514 | else: |
|
2501 | 2515 | # Need to perform a slow delta |
|
2502 | 2516 | r0 = store.deltaparent(store.rev(self._node)) |
|
2503 | 2517 | m0 = self._manifestlog.get(self._dir, store.node(r0)).read() |
|
2504 | 2518 | m1 = self.read() |
|
2505 | 2519 | md = treemanifest(self._manifestlog.nodeconstants, dir=self._dir) |
|
2506 | 2520 | for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).items(): |
|
2507 | 2521 | if n1: |
|
2508 | 2522 | md[f] = n1 |
|
2509 | 2523 | if fl1: |
|
2510 | 2524 | md.setflag(f, fl1) |
|
2511 | 2525 | return md |
|
2512 | 2526 | |
|
2513 | 2527 | def read_any_fast_delta( |
|
2514 | 2528 | self, |
|
2515 | 2529 | valid_bases: Optional[Collection[int]] = None, |
|
2516 | 2530 | *, |
|
2517 | 2531 | shallow: bool = False, |
|
2518 | 2532 | ) -> Tuple[Optional[int], AnyManifestDict]: |
|
2519 | 2533 | """see `imanifestrevisionstored` documentation""" |
|
2520 | 2534 | store = self._storage() |
|
2521 | 2535 | r = store.rev(self._node) |
|
2522 | 2536 | deltaparent = store.deltaparent(r) |
|
2523 | 2537 | |
|
2524 | 2538 | if valid_bases is None: |
|
2525 | 2539 | # make sure the next check is True |
|
2526 | 2540 | valid_bases = (deltaparent,) |
|
2527 | 2541 | can_use_delta = deltaparent != nullrev and deltaparent in valid_bases |
|
2528 | 2542 | |
|
2529 | 2543 | if shallow: |
|
2530 | 2544 | if can_use_delta: |
|
2531 | 2545 | return (deltaparent, self._read_storage_delta_shallow()) |
|
2532 | 2546 | else: |
|
2533 | 2547 | d = store.revision(self._node) |
|
2534 | 2548 | return (None, manifestdict(store.nodeconstants.nodelen, d)) |
|
2535 | 2549 | else: |
|
2536 | 2550 | # note: This use "slow_delta" here is cargo culted from the previous |
|
2537 | 2551 | # implementation. I am not sure it make sense since the goal here is to |
|
2538 | 2552 | # be fast, so why are we computing a delta? On the other hand, tree |
|
2539 | 2553 | # manifest delta as fairly "cheap" and allow for skipping whole part of |
|
2540 | 2554 | # the tree that a full read would access. So it might be a good idea. |
|
2541 | 2555 | # |
|
2542 | 2556 | # If we realize we don't need delta here, we should simply use: |
|
2543 | 2557 | # |
|
2544 | 2558 | # return (None, self.read()) |
|
2545 | 2559 | if can_use_delta: |
|
2546 | 2560 | return (None, self._read_storage_slow_delta(base=deltaparent)) |
|
2547 | 2561 | else: |
|
2548 | 2562 | parents = [ |
|
2549 | 2563 | p |
|
2550 | 2564 | for p in store.parentrevs(r) |
|
2551 | 2565 | if p is not nullrev and p in valid_bases |
|
2552 | 2566 | ] |
|
2553 | 2567 | if parents: |
|
2554 | 2568 | best_base = max(parents) |
|
2555 | 2569 | else: |
|
2556 | 2570 | best_base = max(valid_bases) |
|
2557 | 2571 | return (None, self._read_storage_slow_delta(base=best_base)) |
|
2558 | 2572 | |
|
2559 | 2573 | def _read_storage_delta_shallow(self) -> ManifestDict: |
|
2560 | 2574 | store = self._storage() |
|
2561 | 2575 | r = store.rev(self._node) |
|
2562 | 2576 | d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r)) |
|
2563 | 2577 | return manifestdict(store.nodeconstants.nodelen, d) |
|
2564 | 2578 | |
|
2565 | 2579 | def _read_storage_slow_delta(self, base) -> 'TreeManifest': |
|
2566 | 2580 | store = self._storage() |
|
2567 | 2581 | if base is None: |
|
2568 | 2582 | base = store.deltaparent(store.rev(self._node)) |
|
2569 | 2583 | m0 = self._manifestlog.get(self._dir, store.node(base)).read() |
|
2570 | 2584 | m1 = self.read() |
|
2571 | 2585 | md = treemanifest(self._manifestlog.nodeconstants, dir=self._dir) |
|
2572 | 2586 | for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).items(): |
|
2573 | 2587 | if n1: |
|
2574 | 2588 | md[f] = n1 |
|
2575 | 2589 | if fl1: |
|
2576 | 2590 | md.setflag(f, fl1) |
|
2577 | 2591 | return md |
|
2578 | 2592 | |
|
2579 | 2593 | def read_delta_parents( |
|
2580 | 2594 | self, |
|
2581 | 2595 | *, |
|
2582 | 2596 | shallow: bool = False, |
|
2583 | 2597 | exact: bool = True, |
|
2584 | 2598 | ) -> AnyManifestDict: |
|
2585 | 2599 | """see `interface.imanifestrevisionbase` documentations""" |
|
2586 | 2600 | store = self._storage() |
|
2587 | 2601 | r = store.rev(self._node) |
|
2588 | 2602 | parents = [p for p in store.parentrevs(r) if p is not nullrev] |
|
2589 | 2603 | if not exact: |
|
2590 | 2604 | return self.read_any_fast_delta(parents, shallow=shallow)[1] |
|
2591 | 2605 | elif len(parents) == 0: |
|
2592 | 2606 | if shallow: |
|
2593 | 2607 | d = store.revision(self._node) |
|
2594 | 2608 | return manifestdict(store.nodeconstants.nodelen, d) |
|
2595 | 2609 | else: |
|
2596 | 2610 | return self.read() |
|
2597 | 2611 | elif len(parents) == 1: |
|
2598 | 2612 | p = parents[0] |
|
2599 | 2613 | if shallow: |
|
2600 | 2614 | d = mdiff.patchtext(store.revdiff(p, r)) |
|
2601 | 2615 | return manifestdict(store.nodeconstants.nodelen, d) |
|
2602 | 2616 | else: |
|
2603 | 2617 | return self._read_storage_slow_delta(base=p) |
|
2604 | 2618 | else: |
|
2605 | 2619 | p1, p2 = parents |
|
2606 | 2620 | if shallow: |
|
2607 | 2621 | d1 = mdiff.patchtext(store.revdiff(p1, r)) |
|
2608 | 2622 | d2 = mdiff.patchtext(store.revdiff(p2, r)) |
|
2609 | 2623 | d1 = manifestdict(store.nodeconstants.nodelen, d1) |
|
2610 | 2624 | d2 = manifestdict(store.nodeconstants.nodelen, d2) |
|
2611 | 2625 | md = manifestdict(store.nodeconstants.nodelen) |
|
2612 | 2626 | for f, new_node, new_flag in d1.iterentries(): |
|
2613 | 2627 | if f not in d2: |
|
2614 | 2628 | continue |
|
2615 | 2629 | if new_node is not None: |
|
2616 | 2630 | md.set(f, new_node, new_flag) |
|
2617 | 2631 | return md |
|
2618 | 2632 | else: |
|
2619 | 2633 | m1 = self._manifestlog.get(self._dir, store.node(p1)).read() |
|
2620 | 2634 | m2 = self._manifestlog.get(self._dir, store.node(p2)).read() |
|
2621 | 2635 | mc = self.read() |
|
2622 | 2636 | d1 = m1.diff(mc) |
|
2623 | 2637 | d2 = m2.diff(mc) |
|
2624 | 2638 | md = treemanifest( |
|
2625 | 2639 | self._manifestlog.nodeconstants, |
|
2626 | 2640 | dir=self._dir, |
|
2627 | 2641 | ) |
|
2628 | 2642 | for f, new_node, new_flag in d1.iterentries(): |
|
2629 | 2643 | if f not in d2: |
|
2630 | 2644 | continue |
|
2631 | 2645 | if new_node is not None: |
|
2632 | 2646 | md.set(f, new_node, new_flag) |
|
2633 | 2647 | return md |
|
2634 | 2648 | |
|
2635 | 2649 | def read_delta_new_entries( |
|
2636 | 2650 | self, *, shallow: bool = False |
|
2637 | 2651 | ) -> AnyManifestDict: |
|
2638 | 2652 | """see `interface.imanifestrevisionbase` documentations""" |
|
2639 | 2653 | # If we are using narrow, returning a delta against an arbitrary |
|
2640 | 2654 | # changeset might return file outside the narrowspec. This can create |
|
2641 | 2655 | # issue when running validation server side with strict security as |
|
2642 | 2656 | # push from low priviledge usage might be seen as adding new revision |
|
2643 | 2657 | # for files they cannot touch. So we are strict if narrow is involved. |
|
2644 | 2658 | if self._manifestlog.narrowed: |
|
2645 | 2659 | return self.read_delta_parents(shallow=shallow, exact=True) |
|
2646 | 2660 | # delegate to existing another existing method for simplicity |
|
2647 | 2661 | store = self._storage() |
|
2648 | 2662 | r = store.rev(self._node) |
|
2649 | 2663 | bases = (store.deltaparent(r),) |
|
2650 | 2664 | return self.read_any_fast_delta(bases, shallow=shallow)[1] |
|
2651 | 2665 | |
|
2652 | 2666 | def readfast(self, shallow=False) -> AnyManifestDict: |
|
2653 | 2667 | """Calls either readdelta or read, based on which would be less work. |
|
2654 | 2668 | readdelta is called if the delta is against the p1, and therefore can be |
|
2655 | 2669 | read quickly. |
|
2656 | 2670 | |
|
2657 | 2671 | If `shallow` is True, it only returns the entries from this manifest, |
|
2658 | 2672 | and not any submanifests. |
|
2659 | 2673 | """ |
|
2660 | 2674 | util.nouideprecwarn( |
|
2661 | 2675 | b'"readdelta" is deprecated use "read_any_fast_delta" or "read_delta_parents"', |
|
2662 | 2676 | b"6.9", |
|
2663 | 2677 | stacklevel=2, |
|
2664 | 2678 | ) |
|
2665 | 2679 | store = self._storage() |
|
2666 | 2680 | r = store.rev(self._node) |
|
2667 | 2681 | deltaparent = store.deltaparent(r) |
|
2668 | 2682 | if deltaparent != nullrev and deltaparent in store.parentrevs(r): |
|
2669 | 2683 | return self.readdelta(shallow=shallow) |
|
2670 | 2684 | |
|
2671 | 2685 | if shallow: |
|
2672 | 2686 | return manifestdict( |
|
2673 | 2687 | store.nodeconstants.nodelen, store.revision(self._node) |
|
2674 | 2688 | ) |
|
2675 | 2689 | else: |
|
2676 | 2690 | return self.read() |
|
2677 | 2691 | |
|
2678 | 2692 | def find(self, key: bytes) -> Tuple[bytes, bytes]: |
|
2679 | 2693 | return self.read().find(key) |
|
2680 | 2694 | |
|
2681 | 2695 | |
|
2682 | 2696 | treemanifestctx = interfaceutil.implementer(repository.imanifestrevisionstored)( |
|
2683 | 2697 | TreeManifestCtx |
|
2684 | 2698 | ) |
|
2685 | 2699 | |
|
2686 | 2700 | if typing.TYPE_CHECKING: |
|
2687 | 2701 | treemanifestctx = TreeManifestCtx |
|
2688 | 2702 | |
|
2689 | 2703 | |
|
2690 | 2704 | class excludeddir(treemanifest): |
|
2691 | 2705 | """Stand-in for a directory that is excluded from the repository. |
|
2692 | 2706 | |
|
2693 | 2707 | With narrowing active on a repository that uses treemanifests, |
|
2694 | 2708 | some of the directory revlogs will be excluded from the resulting |
|
2695 | 2709 | clone. This is a huge storage win for clients, but means we need |
|
2696 | 2710 | some sort of pseudo-manifest to surface to internals so we can |
|
2697 | 2711 | detect a merge conflict outside the narrowspec. That's what this |
|
2698 | 2712 | class is: it stands in for a directory whose node is known, but |
|
2699 | 2713 | whose contents are unknown. |
|
2700 | 2714 | """ |
|
2701 | 2715 | |
|
2716 | _files: Dict[bytes, bytes] | |
|
2717 | _flags: Dict[bytes, bytes] | |
|
2718 | ||
|
2702 | 2719 | def __init__(self, nodeconstants, dir, node): |
|
2703 | 2720 | super(excludeddir, self).__init__(nodeconstants, dir) |
|
2704 | 2721 | self._node = node |
|
2705 | 2722 | # Add an empty file, which will be included by iterators and such, |
|
2706 | 2723 | # appearing as the directory itself (i.e. something like "dir/") |
|
2707 | 2724 | self._files[b''] = node |
|
2708 | 2725 | self._flags[b''] = b't' |
|
2709 | 2726 | |
|
2710 | 2727 | # Manifests outside the narrowspec should never be modified, so avoid |
|
2711 | 2728 | # copying. This makes a noticeable difference when there are very many |
|
2712 | 2729 | # directories outside the narrowspec. Also, it makes sense for the copy to |
|
2713 | 2730 | # be of the same type as the original, which would not happen with the |
|
2714 | 2731 | # super type's copy(). |
|
2715 | 2732 | def copy(self): |
|
2716 | 2733 | return self |
|
2717 | 2734 | |
|
2718 | 2735 | |
|
2719 | 2736 | class excludeddirmanifestctx(treemanifestctx): |
|
2720 | 2737 | """context wrapper for excludeddir - see that docstring for rationale""" |
|
2721 | 2738 | |
|
2722 | 2739 | def __init__(self, nodeconstants, dir, node): |
|
2723 | 2740 | self.nodeconstants = nodeconstants |
|
2724 | 2741 | self._dir = dir |
|
2725 | 2742 | self._node = node |
|
2726 | 2743 | |
|
2727 | 2744 | def read(self): |
|
2728 | 2745 | return excludeddir(self.nodeconstants, self._dir, self._node) |
|
2729 | 2746 | |
|
2730 | 2747 | def readfast(self, shallow=False): |
|
2731 | 2748 | # special version of readfast since we don't have underlying storage |
|
2732 | 2749 | return self.read() |
|
2733 | 2750 | |
|
2734 | 2751 | def write(self, *args): |
|
2735 | 2752 | raise error.ProgrammingError( |
|
2736 | 2753 | b'attempt to write manifest from excluded dir %s' % self._dir |
|
2737 | 2754 | ) |
|
2738 | 2755 | |
|
2739 | 2756 | |
|
2740 | 2757 | class excludedmanifestrevlog(manifestrevlog): |
|
2741 | 2758 | """Stand-in for excluded treemanifest revlogs. |
|
2742 | 2759 | |
|
2743 | 2760 | When narrowing is active on a treemanifest repository, we'll have |
|
2744 | 2761 | references to directories we can't see due to the revlog being |
|
2745 | 2762 | skipped. This class exists to conform to the manifestrevlog |
|
2746 | 2763 | interface for those directories and proactively prevent writes to |
|
2747 | 2764 | outside the narrowspec. |
|
2748 | 2765 | """ |
|
2749 | 2766 | |
|
2750 | 2767 | def __init__(self, nodeconstants, dir): |
|
2751 | 2768 | self.nodeconstants = nodeconstants |
|
2752 | 2769 | self._dir = dir |
|
2753 | 2770 | |
|
2754 | 2771 | def __len__(self): |
|
2755 | 2772 | raise error.ProgrammingError( |
|
2756 | 2773 | b'attempt to get length of excluded dir %s' % self._dir |
|
2757 | 2774 | ) |
|
2758 | 2775 | |
|
2759 | 2776 | def rev(self, node): |
|
2760 | 2777 | raise error.ProgrammingError( |
|
2761 | 2778 | b'attempt to get rev from excluded dir %s' % self._dir |
|
2762 | 2779 | ) |
|
2763 | 2780 | |
|
2764 | 2781 | def linkrev(self, node): |
|
2765 | 2782 | raise error.ProgrammingError( |
|
2766 | 2783 | b'attempt to get linkrev from excluded dir %s' % self._dir |
|
2767 | 2784 | ) |
|
2768 | 2785 | |
|
2769 | 2786 | def node(self, rev): |
|
2770 | 2787 | raise error.ProgrammingError( |
|
2771 | 2788 | b'attempt to get node from excluded dir %s' % self._dir |
|
2772 | 2789 | ) |
|
2773 | 2790 | |
|
2774 | 2791 | def add(self, *args, **kwargs): |
|
2775 | 2792 | # We should never write entries in dirlogs outside the narrow clone. |
|
2776 | 2793 | # However, the method still gets called from writesubtree() in |
|
2777 | 2794 | # _addtree(), so we need to handle it. We should possibly make that |
|
2778 | 2795 | # avoid calling add() with a clean manifest (_dirty is always False |
|
2779 | 2796 | # in excludeddir instances). |
|
2780 | 2797 | pass |
@@ -1,4123 +1,4125 b'' | |||
|
1 | 1 | # revlog.py - storage back-end for mercurial |
|
2 | 2 | # coding: utf8 |
|
3 | 3 | # |
|
4 | 4 | # Copyright 2005-2007 Olivia Mackall <olivia@selenic.com> |
|
5 | 5 | # |
|
6 | 6 | # This software may be used and distributed according to the terms of the |
|
7 | 7 | # GNU General Public License version 2 or any later version. |
|
8 | 8 | |
|
9 | 9 | """Storage back-end for Mercurial. |
|
10 | 10 | |
|
11 | 11 | This provides efficient delta storage with O(1) retrieve and append |
|
12 | 12 | and O(changes) merge between branches. |
|
13 | 13 | """ |
|
14 | 14 | |
|
15 | 15 | |
|
16 | 16 | import binascii |
|
17 | 17 | import collections |
|
18 | 18 | import contextlib |
|
19 | 19 | import functools |
|
20 | 20 | import io |
|
21 | 21 | import os |
|
22 | 22 | import struct |
|
23 | 23 | import typing |
|
24 | 24 | import weakref |
|
25 | 25 | import zlib |
|
26 | 26 | |
|
27 | 27 | from typing import ( |
|
28 | Iterable, | |
|
29 | Iterator, | |
|
28 | 30 | Optional, |
|
29 | 31 | Tuple, |
|
30 | 32 | ) |
|
31 | 33 | |
|
32 | 34 | # import stuff from node for others to import from revlog |
|
33 | 35 | from .node import ( |
|
34 | 36 | bin, |
|
35 | 37 | hex, |
|
36 | 38 | nullrev, |
|
37 | 39 | sha1nodeconstants, |
|
38 | 40 | short, |
|
39 | 41 | wdirrev, |
|
40 | 42 | ) |
|
41 | 43 | from .i18n import _ |
|
42 | 44 | from .revlogutils.constants import ( |
|
43 | 45 | ALL_KINDS, |
|
44 | 46 | CHANGELOGV2, |
|
45 | 47 | COMP_MODE_DEFAULT, |
|
46 | 48 | COMP_MODE_INLINE, |
|
47 | 49 | COMP_MODE_PLAIN, |
|
48 | 50 | DELTA_BASE_REUSE_NO, |
|
49 | 51 | DELTA_BASE_REUSE_TRY, |
|
50 | 52 | ENTRY_RANK, |
|
51 | 53 | FEATURES_BY_VERSION, |
|
52 | 54 | FLAG_GENERALDELTA, |
|
53 | 55 | FLAG_INLINE_DATA, |
|
54 | 56 | INDEX_HEADER, |
|
55 | 57 | KIND_CHANGELOG, |
|
56 | 58 | KIND_FILELOG, |
|
57 | 59 | RANK_UNKNOWN, |
|
58 | 60 | REVLOGV0, |
|
59 | 61 | REVLOGV1, |
|
60 | 62 | REVLOGV1_FLAGS, |
|
61 | 63 | REVLOGV2, |
|
62 | 64 | REVLOGV2_FLAGS, |
|
63 | 65 | REVLOG_DEFAULT_FLAGS, |
|
64 | 66 | REVLOG_DEFAULT_FORMAT, |
|
65 | 67 | REVLOG_DEFAULT_VERSION, |
|
66 | 68 | SUPPORTED_FLAGS, |
|
67 | 69 | ) |
|
68 | 70 | from .revlogutils.flagutil import ( |
|
69 | 71 | REVIDX_DEFAULT_FLAGS, |
|
70 | 72 | REVIDX_ELLIPSIS, |
|
71 | 73 | REVIDX_EXTSTORED, |
|
72 | 74 | REVIDX_FLAGS_ORDER, |
|
73 | 75 | REVIDX_HASCOPIESINFO, |
|
74 | 76 | REVIDX_ISCENSORED, |
|
75 | 77 | REVIDX_RAWTEXT_CHANGING_FLAGS, |
|
76 | 78 | ) |
|
77 | 79 | from .thirdparty import attr |
|
78 | 80 | |
|
79 | 81 | # Force pytype to use the non-vendored package |
|
80 | 82 | if typing.TYPE_CHECKING: |
|
81 | 83 | # noinspection PyPackageRequirements |
|
82 | 84 | import attr |
|
83 | 85 | |
|
84 | 86 | from . import ( |
|
85 | 87 | ancestor, |
|
86 | 88 | dagop, |
|
87 | 89 | error, |
|
88 | 90 | mdiff, |
|
89 | 91 | policy, |
|
90 | 92 | pycompat, |
|
91 | 93 | revlogutils, |
|
92 | 94 | templatefilters, |
|
93 | 95 | util, |
|
94 | 96 | vfs as vfsmod, |
|
95 | 97 | ) |
|
96 | 98 | from .interfaces import ( |
|
97 | 99 | repository, |
|
98 | 100 | util as interfaceutil, |
|
99 | 101 | ) |
|
100 | 102 | from .revlogutils import ( |
|
101 | 103 | deltas as deltautil, |
|
102 | 104 | docket as docketutil, |
|
103 | 105 | flagutil, |
|
104 | 106 | nodemap as nodemaputil, |
|
105 | 107 | randomaccessfile, |
|
106 | 108 | revlogv0, |
|
107 | 109 | rewrite, |
|
108 | 110 | sidedata as sidedatautil, |
|
109 | 111 | ) |
|
110 | 112 | from .utils import ( |
|
111 | 113 | storageutil, |
|
112 | 114 | stringutil, |
|
113 | 115 | ) |
|
114 | 116 | |
|
115 | 117 | # blanked usage of all the name to prevent pyflakes constraints |
|
116 | 118 | # We need these name available in the module for extensions. |
|
117 | 119 | |
|
118 | 120 | REVLOGV0 |
|
119 | 121 | REVLOGV1 |
|
120 | 122 | REVLOGV2 |
|
121 | 123 | CHANGELOGV2 |
|
122 | 124 | FLAG_INLINE_DATA |
|
123 | 125 | FLAG_GENERALDELTA |
|
124 | 126 | REVLOG_DEFAULT_FLAGS |
|
125 | 127 | REVLOG_DEFAULT_FORMAT |
|
126 | 128 | REVLOG_DEFAULT_VERSION |
|
127 | 129 | REVLOGV1_FLAGS |
|
128 | 130 | REVLOGV2_FLAGS |
|
129 | 131 | REVIDX_ISCENSORED |
|
130 | 132 | REVIDX_ELLIPSIS |
|
131 | 133 | REVIDX_HASCOPIESINFO |
|
132 | 134 | REVIDX_EXTSTORED |
|
133 | 135 | REVIDX_DEFAULT_FLAGS |
|
134 | 136 | REVIDX_FLAGS_ORDER |
|
135 | 137 | REVIDX_RAWTEXT_CHANGING_FLAGS |
|
136 | 138 | |
|
137 | 139 | parsers = policy.importmod('parsers') |
|
138 | 140 | rustancestor = policy.importrust('ancestor') |
|
139 | 141 | rustdagop = policy.importrust('dagop') |
|
140 | 142 | rustrevlog = policy.importrust('revlog') |
|
141 | 143 | |
|
142 | 144 | # Aliased for performance. |
|
143 | 145 | _zlibdecompress = zlib.decompress |
|
144 | 146 | |
|
145 | 147 | # max size of inline data embedded into a revlog |
|
146 | 148 | _maxinline = 131072 |
|
147 | 149 | |
|
148 | 150 | |
|
149 | 151 | # Flag processors for REVIDX_ELLIPSIS. |
|
150 | 152 | def ellipsisreadprocessor(rl, text): |
|
151 | 153 | return text, False |
|
152 | 154 | |
|
153 | 155 | |
|
154 | 156 | def ellipsiswriteprocessor(rl, text): |
|
155 | 157 | return text, False |
|
156 | 158 | |
|
157 | 159 | |
|
158 | 160 | def ellipsisrawprocessor(rl, text): |
|
159 | 161 | return False |
|
160 | 162 | |
|
161 | 163 | |
|
162 | 164 | ellipsisprocessor = ( |
|
163 | 165 | ellipsisreadprocessor, |
|
164 | 166 | ellipsiswriteprocessor, |
|
165 | 167 | ellipsisrawprocessor, |
|
166 | 168 | ) |
|
167 | 169 | |
|
168 | 170 | |
|
169 | 171 | def _verify_revision(rl, skipflags, state, node): |
|
170 | 172 | """Verify the integrity of the given revlog ``node`` while providing a hook |
|
171 | 173 | point for extensions to influence the operation.""" |
|
172 | 174 | if skipflags: |
|
173 | 175 | state[b'skipread'].add(node) |
|
174 | 176 | else: |
|
175 | 177 | # Side-effect: read content and verify hash. |
|
176 | 178 | rl.revision(node) |
|
177 | 179 | |
|
178 | 180 | |
|
179 | 181 | # True if a fast implementation for persistent-nodemap is available |
|
180 | 182 | # |
|
181 | 183 | # We also consider we have a "fast" implementation in "pure" python because |
|
182 | 184 | # people using pure don't really have performance consideration (and a |
|
183 | 185 | # wheelbarrow of other slowness source) |
|
184 | 186 | HAS_FAST_PERSISTENT_NODEMAP = rustrevlog is not None or hasattr( |
|
185 | 187 | parsers, 'BaseIndexObject' |
|
186 | 188 | ) |
|
187 | 189 | |
|
188 | 190 | |
|
189 | 191 | @attr.s(slots=True) |
|
190 | 192 | class RevLogRevisionDelta: |
|
191 | 193 | node = attr.ib() |
|
192 | 194 | p1node = attr.ib() |
|
193 | 195 | p2node = attr.ib() |
|
194 | 196 | basenode = attr.ib() |
|
195 | 197 | flags = attr.ib() |
|
196 | 198 | baserevisionsize = attr.ib() |
|
197 | 199 | revision = attr.ib() |
|
198 | 200 | delta = attr.ib() |
|
199 | 201 | sidedata = attr.ib() |
|
200 | 202 | protocol_flags = attr.ib() |
|
201 | 203 | linknode = attr.ib(default=None) |
|
202 | 204 | |
|
203 | 205 | |
|
204 | 206 | revlogrevisiondelta = interfaceutil.implementer(repository.irevisiondelta)( |
|
205 | 207 | RevLogRevisionDelta |
|
206 | 208 | ) |
|
207 | 209 | |
|
208 | 210 | if typing.TYPE_CHECKING: |
|
209 | 211 | revlogrevisiondelta = RevLogRevisionDelta |
|
210 | 212 | |
|
211 | 213 | |
|
212 | 214 | @attr.s(frozen=True) |
|
213 | 215 | class RevLogProblem: |
|
214 | 216 | warning = attr.ib(default=None, type=Optional[bytes]) |
|
215 | 217 | error = attr.ib(default=None, type=Optional[bytes]) |
|
216 | 218 | node = attr.ib(default=None, type=Optional[bytes]) |
|
217 | 219 | |
|
218 | 220 | |
|
219 | 221 | revlogproblem = interfaceutil.implementer(repository.iverifyproblem)( |
|
220 | 222 | RevLogProblem |
|
221 | 223 | ) |
|
222 | 224 | |
|
223 | 225 | if typing.TYPE_CHECKING: |
|
224 | 226 | revlogproblem = RevLogProblem |
|
225 | 227 | |
|
226 | 228 | |
|
227 | 229 | def parse_index_v1(data, inline): |
|
228 | 230 | # call the C implementation to parse the index data |
|
229 | 231 | index, cache = parsers.parse_index2(data, inline) |
|
230 | 232 | return index, cache |
|
231 | 233 | |
|
232 | 234 | |
|
233 | 235 | def parse_index_v2(data, inline): |
|
234 | 236 | # call the C implementation to parse the index data |
|
235 | 237 | index, cache = parsers.parse_index2(data, inline, format=REVLOGV2) |
|
236 | 238 | return index, cache |
|
237 | 239 | |
|
238 | 240 | |
|
239 | 241 | def parse_index_cl_v2(data, inline): |
|
240 | 242 | # call the C implementation to parse the index data |
|
241 | 243 | index, cache = parsers.parse_index2(data, inline, format=CHANGELOGV2) |
|
242 | 244 | return index, cache |
|
243 | 245 | |
|
244 | 246 | |
|
245 | 247 | if hasattr(parsers, 'parse_index_devel_nodemap'): |
|
246 | 248 | |
|
247 | 249 | def parse_index_v1_nodemap(data, inline): |
|
248 | 250 | index, cache = parsers.parse_index_devel_nodemap(data, inline) |
|
249 | 251 | return index, cache |
|
250 | 252 | |
|
251 | 253 | else: |
|
252 | 254 | parse_index_v1_nodemap = None |
|
253 | 255 | |
|
254 | 256 | |
|
255 | 257 | def parse_index_v1_rust(data, inline, default_header): |
|
256 | 258 | cache = (0, data) if inline else None |
|
257 | 259 | return rustrevlog.Index(data, default_header), cache |
|
258 | 260 | |
|
259 | 261 | |
|
260 | 262 | # corresponds to uncompressed length of indexformatng (2 gigs, 4-byte |
|
261 | 263 | # signed integer) |
|
262 | 264 | _maxentrysize = 0x7FFFFFFF |
|
263 | 265 | |
|
264 | 266 | FILE_TOO_SHORT_MSG = _( |
|
265 | 267 | b'cannot read from revlog %s;' |
|
266 | 268 | b' expected %d bytes from offset %d, data size is %d' |
|
267 | 269 | ) |
|
268 | 270 | |
|
269 | 271 | hexdigits = b'0123456789abcdefABCDEF' |
|
270 | 272 | |
|
271 | 273 | |
|
272 | 274 | class _Config: |
|
273 | 275 | def copy(self): |
|
274 | 276 | return self.__class__(**self.__dict__) |
|
275 | 277 | |
|
276 | 278 | |
|
277 | 279 | @attr.s() |
|
278 | 280 | class FeatureConfig(_Config): |
|
279 | 281 | """Hold configuration values about the available revlog features""" |
|
280 | 282 | |
|
281 | 283 | # the default compression engine |
|
282 | 284 | compression_engine = attr.ib(default=b'zlib') |
|
283 | 285 | # compression engines options |
|
284 | 286 | compression_engine_options = attr.ib(default=attr.Factory(dict)) |
|
285 | 287 | |
|
286 | 288 | # can we use censor on this revlog |
|
287 | 289 | censorable = attr.ib(default=False) |
|
288 | 290 | # does this revlog use the "side data" feature |
|
289 | 291 | has_side_data = attr.ib(default=False) |
|
290 | 292 | # might remove rank configuration once the computation has no impact |
|
291 | 293 | compute_rank = attr.ib(default=False) |
|
292 | 294 | # parent order is supposed to be semantically irrelevant, so we |
|
293 | 295 | # normally resort parents to ensure that the first parent is non-null, |
|
294 | 296 | # if there is a non-null parent at all. |
|
295 | 297 | # filelog abuses the parent order as flag to mark some instances of |
|
296 | 298 | # meta-encoded files, so allow it to disable this behavior. |
|
297 | 299 | canonical_parent_order = attr.ib(default=False) |
|
298 | 300 | # can ellipsis commit be used |
|
299 | 301 | enable_ellipsis = attr.ib(default=False) |
|
300 | 302 | |
|
301 | 303 | def copy(self): |
|
302 | 304 | new = super().copy() |
|
303 | 305 | new.compression_engine_options = self.compression_engine_options.copy() |
|
304 | 306 | return new |
|
305 | 307 | |
|
306 | 308 | |
|
307 | 309 | @attr.s() |
|
308 | 310 | class DataConfig(_Config): |
|
309 | 311 | """Hold configuration value about how the revlog data are read""" |
|
310 | 312 | |
|
311 | 313 | # should we try to open the "pending" version of the revlog |
|
312 | 314 | try_pending = attr.ib(default=False) |
|
313 | 315 | # should we try to open the "splitted" version of the revlog |
|
314 | 316 | try_split = attr.ib(default=False) |
|
315 | 317 | # When True, indexfile should be opened with checkambig=True at writing, |
|
316 | 318 | # to avoid file stat ambiguity. |
|
317 | 319 | check_ambig = attr.ib(default=False) |
|
318 | 320 | |
|
319 | 321 | # If true, use mmap instead of reading to deal with large index |
|
320 | 322 | mmap_large_index = attr.ib(default=False) |
|
321 | 323 | # how much data is large |
|
322 | 324 | mmap_index_threshold = attr.ib(default=None) |
|
323 | 325 | # How much data to read and cache into the raw revlog data cache. |
|
324 | 326 | chunk_cache_size = attr.ib(default=65536) |
|
325 | 327 | |
|
326 | 328 | # The size of the uncompressed cache compared to the largest revision seen. |
|
327 | 329 | uncompressed_cache_factor = attr.ib(default=None) |
|
328 | 330 | |
|
329 | 331 | # The number of chunk cached |
|
330 | 332 | uncompressed_cache_count = attr.ib(default=None) |
|
331 | 333 | |
|
332 | 334 | # Allow sparse reading of the revlog data |
|
333 | 335 | with_sparse_read = attr.ib(default=False) |
|
334 | 336 | # minimal density of a sparse read chunk |
|
335 | 337 | sr_density_threshold = attr.ib(default=0.50) |
|
336 | 338 | # minimal size of data we skip when performing sparse read |
|
337 | 339 | sr_min_gap_size = attr.ib(default=262144) |
|
338 | 340 | |
|
339 | 341 | # are delta encoded against arbitrary bases. |
|
340 | 342 | generaldelta = attr.ib(default=False) |
|
341 | 343 | |
|
342 | 344 | |
|
343 | 345 | @attr.s() |
|
344 | 346 | class DeltaConfig(_Config): |
|
345 | 347 | """Hold configuration value about how new delta are computed |
|
346 | 348 | |
|
347 | 349 | Some attributes are duplicated from DataConfig to help havign each object |
|
348 | 350 | self contained. |
|
349 | 351 | """ |
|
350 | 352 | |
|
351 | 353 | # can delta be encoded against arbitrary bases. |
|
352 | 354 | general_delta = attr.ib(default=False) |
|
353 | 355 | # Allow sparse writing of the revlog data |
|
354 | 356 | sparse_revlog = attr.ib(default=False) |
|
355 | 357 | # maximum length of a delta chain |
|
356 | 358 | max_chain_len = attr.ib(default=None) |
|
357 | 359 | # Maximum distance between delta chain base start and end |
|
358 | 360 | max_deltachain_span = attr.ib(default=-1) |
|
359 | 361 | # If `upper_bound_comp` is not None, this is the expected maximal gain from |
|
360 | 362 | # compression for the data content. |
|
361 | 363 | upper_bound_comp = attr.ib(default=None) |
|
362 | 364 | # Should we try a delta against both parent |
|
363 | 365 | delta_both_parents = attr.ib(default=True) |
|
364 | 366 | # Test delta base candidate group by chunk of this maximal size. |
|
365 | 367 | candidate_group_chunk_size = attr.ib(default=0) |
|
366 | 368 | # Should we display debug information about delta computation |
|
367 | 369 | debug_delta = attr.ib(default=False) |
|
368 | 370 | # trust incoming delta by default |
|
369 | 371 | lazy_delta = attr.ib(default=True) |
|
370 | 372 | # trust the base of incoming delta by default |
|
371 | 373 | lazy_delta_base = attr.ib(default=False) |
|
372 | 374 | |
|
373 | 375 | |
|
374 | 376 | class _InnerRevlog: |
|
375 | 377 | """An inner layer of the revlog object |
|
376 | 378 | |
|
377 | 379 | That layer exist to be able to delegate some operation to Rust, its |
|
378 | 380 | boundaries are arbitrary and based on what we can delegate to Rust. |
|
379 | 381 | """ |
|
380 | 382 | |
|
381 | 383 | opener: vfsmod.vfs |
|
382 | 384 | |
|
383 | 385 | def __init__( |
|
384 | 386 | self, |
|
385 | 387 | opener: vfsmod.vfs, |
|
386 | 388 | index, |
|
387 | 389 | index_file, |
|
388 | 390 | data_file, |
|
389 | 391 | sidedata_file, |
|
390 | 392 | inline, |
|
391 | 393 | data_config, |
|
392 | 394 | delta_config, |
|
393 | 395 | feature_config, |
|
394 | 396 | chunk_cache, |
|
395 | 397 | default_compression_header, |
|
396 | 398 | ): |
|
397 | 399 | self.opener = opener |
|
398 | 400 | self.index = index |
|
399 | 401 | |
|
400 | 402 | self.index_file = index_file |
|
401 | 403 | self.data_file = data_file |
|
402 | 404 | self.sidedata_file = sidedata_file |
|
403 | 405 | self.inline = inline |
|
404 | 406 | self.data_config = data_config |
|
405 | 407 | self.delta_config = delta_config |
|
406 | 408 | self.feature_config = feature_config |
|
407 | 409 | |
|
408 | 410 | # used during diverted write. |
|
409 | 411 | self._orig_index_file = None |
|
410 | 412 | |
|
411 | 413 | self._default_compression_header = default_compression_header |
|
412 | 414 | |
|
413 | 415 | # index |
|
414 | 416 | |
|
415 | 417 | # 3-tuple of file handles being used for active writing. |
|
416 | 418 | self._writinghandles = None |
|
417 | 419 | |
|
418 | 420 | self._segmentfile = randomaccessfile.randomaccessfile( |
|
419 | 421 | self.opener, |
|
420 | 422 | (self.index_file if self.inline else self.data_file), |
|
421 | 423 | self.data_config.chunk_cache_size, |
|
422 | 424 | chunk_cache, |
|
423 | 425 | ) |
|
424 | 426 | self._segmentfile_sidedata = randomaccessfile.randomaccessfile( |
|
425 | 427 | self.opener, |
|
426 | 428 | self.sidedata_file, |
|
427 | 429 | self.data_config.chunk_cache_size, |
|
428 | 430 | ) |
|
429 | 431 | |
|
430 | 432 | # revlog header -> revlog compressor |
|
431 | 433 | self._decompressors = {} |
|
432 | 434 | # 3-tuple of (node, rev, text) for a raw revision. |
|
433 | 435 | self._revisioncache = None |
|
434 | 436 | |
|
435 | 437 | # cache some uncompressed chunks |
|
436 | 438 | # rev β uncompressed_chunk |
|
437 | 439 | # |
|
438 | 440 | # the max cost is dynamically updated to be proportionnal to the |
|
439 | 441 | # size of revision we actually encounter. |
|
440 | 442 | self._uncompressed_chunk_cache = None |
|
441 | 443 | if self.data_config.uncompressed_cache_factor is not None: |
|
442 | 444 | self._uncompressed_chunk_cache = util.lrucachedict( |
|
443 | 445 | self.data_config.uncompressed_cache_count, |
|
444 | 446 | maxcost=65536, # some arbitrary initial value |
|
445 | 447 | ) |
|
446 | 448 | |
|
447 | 449 | self._delay_buffer = None |
|
448 | 450 | |
|
449 | 451 | def __len__(self): |
|
450 | 452 | return len(self.index) |
|
451 | 453 | |
|
452 | 454 | def clear_cache(self): |
|
453 | 455 | assert not self.is_delaying |
|
454 | 456 | self._revisioncache = None |
|
455 | 457 | if self._uncompressed_chunk_cache is not None: |
|
456 | 458 | self._uncompressed_chunk_cache.clear() |
|
457 | 459 | self._segmentfile.clear_cache() |
|
458 | 460 | self._segmentfile_sidedata.clear_cache() |
|
459 | 461 | |
|
460 | 462 | @property |
|
461 | 463 | def canonical_index_file(self): |
|
462 | 464 | if self._orig_index_file is not None: |
|
463 | 465 | return self._orig_index_file |
|
464 | 466 | return self.index_file |
|
465 | 467 | |
|
466 | 468 | @property |
|
467 | 469 | def is_delaying(self): |
|
468 | 470 | """is the revlog is currently delaying the visibility of written data? |
|
469 | 471 | |
|
470 | 472 | The delaying mechanism can be either in-memory or written on disk in a |
|
471 | 473 | side-file.""" |
|
472 | 474 | return (self._delay_buffer is not None) or ( |
|
473 | 475 | self._orig_index_file is not None |
|
474 | 476 | ) |
|
475 | 477 | |
|
476 | 478 | # Derived from index values. |
|
477 | 479 | |
|
478 | 480 | def start(self, rev): |
|
479 | 481 | """the offset of the data chunk for this revision""" |
|
480 | 482 | return int(self.index[rev][0] >> 16) |
|
481 | 483 | |
|
482 | 484 | def length(self, rev): |
|
483 | 485 | """the length of the data chunk for this revision""" |
|
484 | 486 | return self.index[rev][1] |
|
485 | 487 | |
|
486 | 488 | def end(self, rev): |
|
487 | 489 | """the end of the data chunk for this revision""" |
|
488 | 490 | return self.start(rev) + self.length(rev) |
|
489 | 491 | |
|
490 | 492 | def deltaparent(self, rev): |
|
491 | 493 | """return deltaparent of the given revision""" |
|
492 | 494 | base = self.index[rev][3] |
|
493 | 495 | if base == rev: |
|
494 | 496 | return nullrev |
|
495 | 497 | elif self.delta_config.general_delta: |
|
496 | 498 | return base |
|
497 | 499 | else: |
|
498 | 500 | return rev - 1 |
|
499 | 501 | |
|
500 | 502 | def issnapshot(self, rev): |
|
501 | 503 | """tells whether rev is a snapshot""" |
|
502 | 504 | if not self.delta_config.sparse_revlog: |
|
503 | 505 | return self.deltaparent(rev) == nullrev |
|
504 | 506 | elif hasattr(self.index, 'issnapshot'): |
|
505 | 507 | # directly assign the method to cache the testing and access |
|
506 | 508 | self.issnapshot = self.index.issnapshot |
|
507 | 509 | return self.issnapshot(rev) |
|
508 | 510 | if rev == nullrev: |
|
509 | 511 | return True |
|
510 | 512 | entry = self.index[rev] |
|
511 | 513 | base = entry[3] |
|
512 | 514 | if base == rev: |
|
513 | 515 | return True |
|
514 | 516 | if base == nullrev: |
|
515 | 517 | return True |
|
516 | 518 | p1 = entry[5] |
|
517 | 519 | while self.length(p1) == 0: |
|
518 | 520 | b = self.deltaparent(p1) |
|
519 | 521 | if b == p1: |
|
520 | 522 | break |
|
521 | 523 | p1 = b |
|
522 | 524 | p2 = entry[6] |
|
523 | 525 | while self.length(p2) == 0: |
|
524 | 526 | b = self.deltaparent(p2) |
|
525 | 527 | if b == p2: |
|
526 | 528 | break |
|
527 | 529 | p2 = b |
|
528 | 530 | if base == p1 or base == p2: |
|
529 | 531 | return False |
|
530 | 532 | return self.issnapshot(base) |
|
531 | 533 | |
|
532 | 534 | def _deltachain(self, rev, stoprev=None): |
|
533 | 535 | """Obtain the delta chain for a revision. |
|
534 | 536 | |
|
535 | 537 | ``stoprev`` specifies a revision to stop at. If not specified, we |
|
536 | 538 | stop at the base of the chain. |
|
537 | 539 | |
|
538 | 540 | Returns a 2-tuple of (chain, stopped) where ``chain`` is a list of |
|
539 | 541 | revs in ascending order and ``stopped`` is a bool indicating whether |
|
540 | 542 | ``stoprev`` was hit. |
|
541 | 543 | """ |
|
542 | 544 | generaldelta = self.delta_config.general_delta |
|
543 | 545 | # Try C implementation. |
|
544 | 546 | try: |
|
545 | 547 | return self.index.deltachain(rev, stoprev, generaldelta) |
|
546 | 548 | except AttributeError: |
|
547 | 549 | pass |
|
548 | 550 | |
|
549 | 551 | chain = [] |
|
550 | 552 | |
|
551 | 553 | # Alias to prevent attribute lookup in tight loop. |
|
552 | 554 | index = self.index |
|
553 | 555 | |
|
554 | 556 | iterrev = rev |
|
555 | 557 | e = index[iterrev] |
|
556 | 558 | while iterrev != e[3] and iterrev != stoprev: |
|
557 | 559 | chain.append(iterrev) |
|
558 | 560 | if generaldelta: |
|
559 | 561 | iterrev = e[3] |
|
560 | 562 | else: |
|
561 | 563 | iterrev -= 1 |
|
562 | 564 | e = index[iterrev] |
|
563 | 565 | |
|
564 | 566 | if iterrev == stoprev: |
|
565 | 567 | stopped = True |
|
566 | 568 | else: |
|
567 | 569 | chain.append(iterrev) |
|
568 | 570 | stopped = False |
|
569 | 571 | |
|
570 | 572 | chain.reverse() |
|
571 | 573 | return chain, stopped |
|
572 | 574 | |
|
573 | 575 | @util.propertycache |
|
574 | 576 | def _compressor(self): |
|
575 | 577 | engine = util.compengines[self.feature_config.compression_engine] |
|
576 | 578 | return engine.revlogcompressor( |
|
577 | 579 | self.feature_config.compression_engine_options |
|
578 | 580 | ) |
|
579 | 581 | |
|
580 | 582 | @util.propertycache |
|
581 | 583 | def _decompressor(self): |
|
582 | 584 | """the default decompressor""" |
|
583 | 585 | if self._default_compression_header is None: |
|
584 | 586 | return None |
|
585 | 587 | t = self._default_compression_header |
|
586 | 588 | c = self._get_decompressor(t) |
|
587 | 589 | return c.decompress |
|
588 | 590 | |
|
589 | 591 | def _get_decompressor(self, t: bytes): |
|
590 | 592 | try: |
|
591 | 593 | compressor = self._decompressors[t] |
|
592 | 594 | except KeyError: |
|
593 | 595 | try: |
|
594 | 596 | engine = util.compengines.forrevlogheader(t) |
|
595 | 597 | compressor = engine.revlogcompressor( |
|
596 | 598 | self.feature_config.compression_engine_options |
|
597 | 599 | ) |
|
598 | 600 | self._decompressors[t] = compressor |
|
599 | 601 | except KeyError: |
|
600 | 602 | raise error.RevlogError( |
|
601 | 603 | _(b'unknown compression type %s') % binascii.hexlify(t) |
|
602 | 604 | ) |
|
603 | 605 | return compressor |
|
604 | 606 | |
|
605 | 607 | def compress(self, data: bytes) -> Tuple[bytes, bytes]: |
|
606 | 608 | """Generate a possibly-compressed representation of data.""" |
|
607 | 609 | if not data: |
|
608 | 610 | return b'', data |
|
609 | 611 | |
|
610 | 612 | compressed = self._compressor.compress(data) |
|
611 | 613 | |
|
612 | 614 | if compressed: |
|
613 | 615 | # The revlog compressor added the header in the returned data. |
|
614 | 616 | return b'', compressed |
|
615 | 617 | |
|
616 | 618 | if data[0:1] == b'\0': |
|
617 | 619 | return b'', data |
|
618 | 620 | return b'u', data |
|
619 | 621 | |
|
620 | 622 | def decompress(self, data: bytes): |
|
621 | 623 | """Decompress a revlog chunk. |
|
622 | 624 | |
|
623 | 625 | The chunk is expected to begin with a header identifying the |
|
624 | 626 | format type so it can be routed to an appropriate decompressor. |
|
625 | 627 | """ |
|
626 | 628 | if not data: |
|
627 | 629 | return data |
|
628 | 630 | |
|
629 | 631 | # Revlogs are read much more frequently than they are written and many |
|
630 | 632 | # chunks only take microseconds to decompress, so performance is |
|
631 | 633 | # important here. |
|
632 | 634 | # |
|
633 | 635 | # We can make a few assumptions about revlogs: |
|
634 | 636 | # |
|
635 | 637 | # 1) the majority of chunks will be compressed (as opposed to inline |
|
636 | 638 | # raw data). |
|
637 | 639 | # 2) decompressing *any* data will likely by at least 10x slower than |
|
638 | 640 | # returning raw inline data. |
|
639 | 641 | # 3) we want to prioritize common and officially supported compression |
|
640 | 642 | # engines |
|
641 | 643 | # |
|
642 | 644 | # It follows that we want to optimize for "decompress compressed data |
|
643 | 645 | # when encoded with common and officially supported compression engines" |
|
644 | 646 | # case over "raw data" and "data encoded by less common or non-official |
|
645 | 647 | # compression engines." That is why we have the inline lookup first |
|
646 | 648 | # followed by the compengines lookup. |
|
647 | 649 | # |
|
648 | 650 | # According to `hg perfrevlogchunks`, this is ~0.5% faster for zlib |
|
649 | 651 | # compressed chunks. And this matters for changelog and manifest reads. |
|
650 | 652 | t = data[0:1] |
|
651 | 653 | |
|
652 | 654 | if t == b'x': |
|
653 | 655 | try: |
|
654 | 656 | return _zlibdecompress(data) |
|
655 | 657 | except zlib.error as e: |
|
656 | 658 | raise error.RevlogError( |
|
657 | 659 | _(b'revlog decompress error: %s') |
|
658 | 660 | % stringutil.forcebytestr(e) |
|
659 | 661 | ) |
|
660 | 662 | # '\0' is more common than 'u' so it goes first. |
|
661 | 663 | elif t == b'\0': |
|
662 | 664 | return data |
|
663 | 665 | elif t == b'u': |
|
664 | 666 | return util.buffer(data, 1) |
|
665 | 667 | |
|
666 | 668 | compressor = self._get_decompressor(t) |
|
667 | 669 | |
|
668 | 670 | return compressor.decompress(data) |
|
669 | 671 | |
|
670 | 672 | @contextlib.contextmanager |
|
671 | 673 | def reading(self): |
|
672 | 674 | """Context manager that keeps data and sidedata files open for reading""" |
|
673 | 675 | if len(self.index) == 0: |
|
674 | 676 | yield # nothing to be read |
|
675 | 677 | elif self._delay_buffer is not None and self.inline: |
|
676 | 678 | msg = "revlog with delayed write should not be inline" |
|
677 | 679 | raise error.ProgrammingError(msg) |
|
678 | 680 | else: |
|
679 | 681 | with self._segmentfile.reading(): |
|
680 | 682 | with self._segmentfile_sidedata.reading(): |
|
681 | 683 | yield |
|
682 | 684 | |
|
683 | 685 | @property |
|
684 | 686 | def is_writing(self): |
|
685 | 687 | """True is a writing context is open""" |
|
686 | 688 | return self._writinghandles is not None |
|
687 | 689 | |
|
688 | 690 | @property |
|
689 | 691 | def is_open(self): |
|
690 | 692 | """True if any file handle is being held |
|
691 | 693 | |
|
692 | 694 | Used for assert and debug in the python code""" |
|
693 | 695 | return self._segmentfile.is_open or self._segmentfile_sidedata.is_open |
|
694 | 696 | |
|
695 | 697 | @contextlib.contextmanager |
|
696 | 698 | def writing(self, transaction, data_end=None, sidedata_end=None): |
|
697 | 699 | """Open the revlog files for writing |
|
698 | 700 | |
|
699 | 701 | Add content to a revlog should be done within such context. |
|
700 | 702 | """ |
|
701 | 703 | if self.is_writing: |
|
702 | 704 | yield |
|
703 | 705 | else: |
|
704 | 706 | ifh = dfh = sdfh = None |
|
705 | 707 | try: |
|
706 | 708 | r = len(self.index) |
|
707 | 709 | # opening the data file. |
|
708 | 710 | dsize = 0 |
|
709 | 711 | if r: |
|
710 | 712 | dsize = self.end(r - 1) |
|
711 | 713 | dfh = None |
|
712 | 714 | if not self.inline: |
|
713 | 715 | try: |
|
714 | 716 | dfh = self.opener(self.data_file, mode=b"r+") |
|
715 | 717 | if data_end is None: |
|
716 | 718 | dfh.seek(0, os.SEEK_END) |
|
717 | 719 | else: |
|
718 | 720 | dfh.seek(data_end, os.SEEK_SET) |
|
719 | 721 | except FileNotFoundError: |
|
720 | 722 | dfh = self.opener(self.data_file, mode=b"w+") |
|
721 | 723 | transaction.add(self.data_file, dsize) |
|
722 | 724 | if self.sidedata_file is not None: |
|
723 | 725 | assert sidedata_end is not None |
|
724 | 726 | # revlog-v2 does not inline, help Pytype |
|
725 | 727 | assert dfh is not None |
|
726 | 728 | try: |
|
727 | 729 | sdfh = self.opener(self.sidedata_file, mode=b"r+") |
|
728 | 730 | dfh.seek(sidedata_end, os.SEEK_SET) |
|
729 | 731 | except FileNotFoundError: |
|
730 | 732 | sdfh = self.opener(self.sidedata_file, mode=b"w+") |
|
731 | 733 | transaction.add(self.sidedata_file, sidedata_end) |
|
732 | 734 | |
|
733 | 735 | # opening the index file. |
|
734 | 736 | isize = r * self.index.entry_size |
|
735 | 737 | ifh = self.__index_write_fp() |
|
736 | 738 | if self.inline: |
|
737 | 739 | transaction.add(self.index_file, dsize + isize) |
|
738 | 740 | else: |
|
739 | 741 | transaction.add(self.index_file, isize) |
|
740 | 742 | # exposing all file handle for writing. |
|
741 | 743 | self._writinghandles = (ifh, dfh, sdfh) |
|
742 | 744 | self._segmentfile.writing_handle = ifh if self.inline else dfh |
|
743 | 745 | self._segmentfile_sidedata.writing_handle = sdfh |
|
744 | 746 | yield |
|
745 | 747 | finally: |
|
746 | 748 | self._writinghandles = None |
|
747 | 749 | self._segmentfile.writing_handle = None |
|
748 | 750 | self._segmentfile_sidedata.writing_handle = None |
|
749 | 751 | if dfh is not None: |
|
750 | 752 | dfh.close() |
|
751 | 753 | if sdfh is not None: |
|
752 | 754 | sdfh.close() |
|
753 | 755 | # closing the index file last to avoid exposing referent to |
|
754 | 756 | # potential unflushed data content. |
|
755 | 757 | if ifh is not None: |
|
756 | 758 | ifh.close() |
|
757 | 759 | |
|
758 | 760 | def __index_write_fp(self, index_end=None): |
|
759 | 761 | """internal method to open the index file for writing |
|
760 | 762 | |
|
761 | 763 | You should not use this directly and use `_writing` instead |
|
762 | 764 | """ |
|
763 | 765 | try: |
|
764 | 766 | if self._delay_buffer is None: |
|
765 | 767 | f = self.opener( |
|
766 | 768 | self.index_file, |
|
767 | 769 | mode=b"r+", |
|
768 | 770 | checkambig=self.data_config.check_ambig, |
|
769 | 771 | ) |
|
770 | 772 | else: |
|
771 | 773 | # check_ambig affect we way we open file for writing, however |
|
772 | 774 | # here, we do not actually open a file for writting as write |
|
773 | 775 | # will appened to a delay_buffer. So check_ambig is not |
|
774 | 776 | # meaningful and unneeded here. |
|
775 | 777 | f = randomaccessfile.appender( |
|
776 | 778 | self.opener, self.index_file, b"r+", self._delay_buffer |
|
777 | 779 | ) |
|
778 | 780 | if index_end is None: |
|
779 | 781 | f.seek(0, os.SEEK_END) |
|
780 | 782 | else: |
|
781 | 783 | f.seek(index_end, os.SEEK_SET) |
|
782 | 784 | return f |
|
783 | 785 | except FileNotFoundError: |
|
784 | 786 | if self._delay_buffer is None: |
|
785 | 787 | return self.opener( |
|
786 | 788 | self.index_file, |
|
787 | 789 | mode=b"w+", |
|
788 | 790 | checkambig=self.data_config.check_ambig, |
|
789 | 791 | ) |
|
790 | 792 | else: |
|
791 | 793 | return randomaccessfile.appender( |
|
792 | 794 | self.opener, self.index_file, b"w+", self._delay_buffer |
|
793 | 795 | ) |
|
794 | 796 | |
|
795 | 797 | def __index_new_fp(self): |
|
796 | 798 | """internal method to create a new index file for writing |
|
797 | 799 | |
|
798 | 800 | You should not use this unless you are upgrading from inline revlog |
|
799 | 801 | """ |
|
800 | 802 | return self.opener( |
|
801 | 803 | self.index_file, |
|
802 | 804 | mode=b"w", |
|
803 | 805 | checkambig=self.data_config.check_ambig, |
|
804 | 806 | ) |
|
805 | 807 | |
|
806 | 808 | def split_inline(self, tr, header, new_index_file_path=None): |
|
807 | 809 | """split the data of an inline revlog into an index and a data file""" |
|
808 | 810 | assert self._delay_buffer is None |
|
809 | 811 | existing_handles = False |
|
810 | 812 | if self._writinghandles is not None: |
|
811 | 813 | existing_handles = True |
|
812 | 814 | fp = self._writinghandles[0] |
|
813 | 815 | fp.flush() |
|
814 | 816 | fp.close() |
|
815 | 817 | # We can't use the cached file handle after close(). So prevent |
|
816 | 818 | # its usage. |
|
817 | 819 | self._writinghandles = None |
|
818 | 820 | self._segmentfile.writing_handle = None |
|
819 | 821 | # No need to deal with sidedata writing handle as it is only |
|
820 | 822 | # relevant with revlog-v2 which is never inline, not reaching |
|
821 | 823 | # this code |
|
822 | 824 | |
|
823 | 825 | new_dfh = self.opener(self.data_file, mode=b"w+") |
|
824 | 826 | new_dfh.truncate(0) # drop any potentially existing data |
|
825 | 827 | try: |
|
826 | 828 | with self.reading(): |
|
827 | 829 | for r in range(len(self.index)): |
|
828 | 830 | new_dfh.write(self.get_segment_for_revs(r, r)[1]) |
|
829 | 831 | new_dfh.flush() |
|
830 | 832 | |
|
831 | 833 | if new_index_file_path is not None: |
|
832 | 834 | self.index_file = new_index_file_path |
|
833 | 835 | with self.__index_new_fp() as fp: |
|
834 | 836 | self.inline = False |
|
835 | 837 | for i in range(len(self.index)): |
|
836 | 838 | e = self.index.entry_binary(i) |
|
837 | 839 | if i == 0: |
|
838 | 840 | packed_header = self.index.pack_header(header) |
|
839 | 841 | e = packed_header + e |
|
840 | 842 | fp.write(e) |
|
841 | 843 | |
|
842 | 844 | # If we don't use side-write, the temp file replace the real |
|
843 | 845 | # index when we exit the context manager |
|
844 | 846 | |
|
845 | 847 | self._segmentfile = randomaccessfile.randomaccessfile( |
|
846 | 848 | self.opener, |
|
847 | 849 | self.data_file, |
|
848 | 850 | self.data_config.chunk_cache_size, |
|
849 | 851 | ) |
|
850 | 852 | |
|
851 | 853 | if existing_handles: |
|
852 | 854 | # switched from inline to conventional reopen the index |
|
853 | 855 | ifh = self.__index_write_fp() |
|
854 | 856 | self._writinghandles = (ifh, new_dfh, None) |
|
855 | 857 | self._segmentfile.writing_handle = new_dfh |
|
856 | 858 | new_dfh = None |
|
857 | 859 | # No need to deal with sidedata writing handle as it is only |
|
858 | 860 | # relevant with revlog-v2 which is never inline, not reaching |
|
859 | 861 | # this code |
|
860 | 862 | finally: |
|
861 | 863 | if new_dfh is not None: |
|
862 | 864 | new_dfh.close() |
|
863 | 865 | return self.index_file |
|
864 | 866 | |
|
865 | 867 | def get_segment_for_revs(self, startrev, endrev): |
|
866 | 868 | """Obtain a segment of raw data corresponding to a range of revisions. |
|
867 | 869 | |
|
868 | 870 | Accepts the start and end revisions and an optional already-open |
|
869 | 871 | file handle to be used for reading. If the file handle is read, its |
|
870 | 872 | seek position will not be preserved. |
|
871 | 873 | |
|
872 | 874 | Requests for data may be satisfied by a cache. |
|
873 | 875 | |
|
874 | 876 | Returns a 2-tuple of (offset, data) for the requested range of |
|
875 | 877 | revisions. Offset is the integer offset from the beginning of the |
|
876 | 878 | revlog and data is a str or buffer of the raw byte data. |
|
877 | 879 | |
|
878 | 880 | Callers will need to call ``self.start(rev)`` and ``self.length(rev)`` |
|
879 | 881 | to determine where each revision's data begins and ends. |
|
880 | 882 | |
|
881 | 883 | API: we should consider making this a private part of the InnerRevlog |
|
882 | 884 | at some point. |
|
883 | 885 | """ |
|
884 | 886 | # Inlined self.start(startrev) & self.end(endrev) for perf reasons |
|
885 | 887 | # (functions are expensive). |
|
886 | 888 | index = self.index |
|
887 | 889 | istart = index[startrev] |
|
888 | 890 | start = int(istart[0] >> 16) |
|
889 | 891 | if startrev == endrev: |
|
890 | 892 | end = start + istart[1] |
|
891 | 893 | else: |
|
892 | 894 | iend = index[endrev] |
|
893 | 895 | end = int(iend[0] >> 16) + iend[1] |
|
894 | 896 | |
|
895 | 897 | if self.inline: |
|
896 | 898 | start += (startrev + 1) * self.index.entry_size |
|
897 | 899 | end += (endrev + 1) * self.index.entry_size |
|
898 | 900 | length = end - start |
|
899 | 901 | |
|
900 | 902 | return start, self._segmentfile.read_chunk(start, length) |
|
901 | 903 | |
|
902 | 904 | def _chunk(self, rev): |
|
903 | 905 | """Obtain a single decompressed chunk for a revision. |
|
904 | 906 | |
|
905 | 907 | Accepts an integer revision and an optional already-open file handle |
|
906 | 908 | to be used for reading. If used, the seek position of the file will not |
|
907 | 909 | be preserved. |
|
908 | 910 | |
|
909 | 911 | Returns a str holding uncompressed data for the requested revision. |
|
910 | 912 | """ |
|
911 | 913 | if self._uncompressed_chunk_cache is not None: |
|
912 | 914 | uncomp = self._uncompressed_chunk_cache.get(rev) |
|
913 | 915 | if uncomp is not None: |
|
914 | 916 | return uncomp |
|
915 | 917 | |
|
916 | 918 | compression_mode = self.index[rev][10] |
|
917 | 919 | data = self.get_segment_for_revs(rev, rev)[1] |
|
918 | 920 | if compression_mode == COMP_MODE_PLAIN: |
|
919 | 921 | uncomp = data |
|
920 | 922 | elif compression_mode == COMP_MODE_DEFAULT: |
|
921 | 923 | uncomp = self._decompressor(data) |
|
922 | 924 | elif compression_mode == COMP_MODE_INLINE: |
|
923 | 925 | uncomp = self.decompress(data) |
|
924 | 926 | else: |
|
925 | 927 | msg = b'unknown compression mode %d' |
|
926 | 928 | msg %= compression_mode |
|
927 | 929 | raise error.RevlogError(msg) |
|
928 | 930 | if self._uncompressed_chunk_cache is not None: |
|
929 | 931 | self._uncompressed_chunk_cache.insert(rev, uncomp, cost=len(uncomp)) |
|
930 | 932 | return uncomp |
|
931 | 933 | |
|
932 | 934 | def _chunks(self, revs, targetsize=None): |
|
933 | 935 | """Obtain decompressed chunks for the specified revisions. |
|
934 | 936 | |
|
935 | 937 | Accepts an iterable of numeric revisions that are assumed to be in |
|
936 | 938 | ascending order. Also accepts an optional already-open file handle |
|
937 | 939 | to be used for reading. If used, the seek position of the file will |
|
938 | 940 | not be preserved. |
|
939 | 941 | |
|
940 | 942 | This function is similar to calling ``self._chunk()`` multiple times, |
|
941 | 943 | but is faster. |
|
942 | 944 | |
|
943 | 945 | Returns a list with decompressed data for each requested revision. |
|
944 | 946 | """ |
|
945 | 947 | if not revs: |
|
946 | 948 | return [] |
|
947 | 949 | start = self.start |
|
948 | 950 | length = self.length |
|
949 | 951 | inline = self.inline |
|
950 | 952 | iosize = self.index.entry_size |
|
951 | 953 | buffer = util.buffer |
|
952 | 954 | |
|
953 | 955 | fetched_revs = [] |
|
954 | 956 | fadd = fetched_revs.append |
|
955 | 957 | |
|
956 | 958 | chunks = [] |
|
957 | 959 | ladd = chunks.append |
|
958 | 960 | |
|
959 | 961 | if self._uncompressed_chunk_cache is None: |
|
960 | 962 | fetched_revs = revs |
|
961 | 963 | else: |
|
962 | 964 | for rev in revs: |
|
963 | 965 | cached_value = self._uncompressed_chunk_cache.get(rev) |
|
964 | 966 | if cached_value is None: |
|
965 | 967 | fadd(rev) |
|
966 | 968 | else: |
|
967 | 969 | ladd((rev, cached_value)) |
|
968 | 970 | |
|
969 | 971 | if not fetched_revs: |
|
970 | 972 | slicedchunks = () |
|
971 | 973 | elif not self.data_config.with_sparse_read: |
|
972 | 974 | slicedchunks = (fetched_revs,) |
|
973 | 975 | else: |
|
974 | 976 | slicedchunks = deltautil.slicechunk( |
|
975 | 977 | self, |
|
976 | 978 | fetched_revs, |
|
977 | 979 | targetsize=targetsize, |
|
978 | 980 | ) |
|
979 | 981 | |
|
980 | 982 | for revschunk in slicedchunks: |
|
981 | 983 | firstrev = revschunk[0] |
|
982 | 984 | # Skip trailing revisions with empty diff |
|
983 | 985 | for lastrev in revschunk[::-1]: |
|
984 | 986 | if length(lastrev) != 0: |
|
985 | 987 | break |
|
986 | 988 | |
|
987 | 989 | try: |
|
988 | 990 | offset, data = self.get_segment_for_revs(firstrev, lastrev) |
|
989 | 991 | except OverflowError: |
|
990 | 992 | # issue4215 - we can't cache a run of chunks greater than |
|
991 | 993 | # 2G on Windows |
|
992 | 994 | for rev in revschunk: |
|
993 | 995 | ladd((rev, self._chunk(rev))) |
|
994 | 996 | |
|
995 | 997 | decomp = self.decompress |
|
996 | 998 | # self._decompressor might be None, but will not be used in that case |
|
997 | 999 | def_decomp = self._decompressor |
|
998 | 1000 | for rev in revschunk: |
|
999 | 1001 | chunkstart = start(rev) |
|
1000 | 1002 | if inline: |
|
1001 | 1003 | chunkstart += (rev + 1) * iosize |
|
1002 | 1004 | chunklength = length(rev) |
|
1003 | 1005 | comp_mode = self.index[rev][10] |
|
1004 | 1006 | c = buffer(data, chunkstart - offset, chunklength) |
|
1005 | 1007 | if comp_mode == COMP_MODE_PLAIN: |
|
1006 | 1008 | c = c |
|
1007 | 1009 | elif comp_mode == COMP_MODE_INLINE: |
|
1008 | 1010 | c = decomp(c) |
|
1009 | 1011 | elif comp_mode == COMP_MODE_DEFAULT: |
|
1010 | 1012 | c = def_decomp(c) |
|
1011 | 1013 | else: |
|
1012 | 1014 | msg = b'unknown compression mode %d' |
|
1013 | 1015 | msg %= comp_mode |
|
1014 | 1016 | raise error.RevlogError(msg) |
|
1015 | 1017 | ladd((rev, c)) |
|
1016 | 1018 | if self._uncompressed_chunk_cache is not None: |
|
1017 | 1019 | self._uncompressed_chunk_cache.insert(rev, c, len(c)) |
|
1018 | 1020 | |
|
1019 | 1021 | chunks.sort() |
|
1020 | 1022 | return [x[1] for x in chunks] |
|
1021 | 1023 | |
|
1022 | 1024 | def raw_text(self, node, rev): |
|
1023 | 1025 | """return the possibly unvalidated rawtext for a revision |
|
1024 | 1026 | |
|
1025 | 1027 | returns (rev, rawtext, validated) |
|
1026 | 1028 | """ |
|
1027 | 1029 | |
|
1028 | 1030 | # revision in the cache (could be useful to apply delta) |
|
1029 | 1031 | cachedrev = None |
|
1030 | 1032 | # An intermediate text to apply deltas to |
|
1031 | 1033 | basetext = None |
|
1032 | 1034 | |
|
1033 | 1035 | # Check if we have the entry in cache |
|
1034 | 1036 | # The cache entry looks like (node, rev, rawtext) |
|
1035 | 1037 | if self._revisioncache: |
|
1036 | 1038 | cachedrev = self._revisioncache[1] |
|
1037 | 1039 | |
|
1038 | 1040 | chain, stopped = self._deltachain(rev, stoprev=cachedrev) |
|
1039 | 1041 | if stopped: |
|
1040 | 1042 | basetext = self._revisioncache[2] |
|
1041 | 1043 | |
|
1042 | 1044 | # drop cache to save memory, the caller is expected to |
|
1043 | 1045 | # update self._inner._revisioncache after validating the text |
|
1044 | 1046 | self._revisioncache = None |
|
1045 | 1047 | |
|
1046 | 1048 | targetsize = None |
|
1047 | 1049 | rawsize = self.index[rev][2] |
|
1048 | 1050 | if 0 <= rawsize: |
|
1049 | 1051 | targetsize = 4 * rawsize |
|
1050 | 1052 | |
|
1051 | 1053 | if self._uncompressed_chunk_cache is not None: |
|
1052 | 1054 | # dynamically update the uncompressed_chunk_cache size to the |
|
1053 | 1055 | # largest revision we saw in this revlog. |
|
1054 | 1056 | factor = self.data_config.uncompressed_cache_factor |
|
1055 | 1057 | candidate_size = rawsize * factor |
|
1056 | 1058 | if candidate_size > self._uncompressed_chunk_cache.maxcost: |
|
1057 | 1059 | self._uncompressed_chunk_cache.maxcost = candidate_size |
|
1058 | 1060 | |
|
1059 | 1061 | bins = self._chunks(chain, targetsize=targetsize) |
|
1060 | 1062 | if basetext is None: |
|
1061 | 1063 | basetext = bytes(bins[0]) |
|
1062 | 1064 | bins = bins[1:] |
|
1063 | 1065 | |
|
1064 | 1066 | rawtext = mdiff.patches(basetext, bins) |
|
1065 | 1067 | del basetext # let us have a chance to free memory early |
|
1066 | 1068 | return (rev, rawtext, False) |
|
1067 | 1069 | |
|
1068 | 1070 | def sidedata(self, rev, sidedata_end): |
|
1069 | 1071 | """Return the sidedata for a given revision number.""" |
|
1070 | 1072 | index_entry = self.index[rev] |
|
1071 | 1073 | sidedata_offset = index_entry[8] |
|
1072 | 1074 | sidedata_size = index_entry[9] |
|
1073 | 1075 | |
|
1074 | 1076 | if self.inline: |
|
1075 | 1077 | sidedata_offset += self.index.entry_size * (1 + rev) |
|
1076 | 1078 | if sidedata_size == 0: |
|
1077 | 1079 | return {} |
|
1078 | 1080 | |
|
1079 | 1081 | if sidedata_end < sidedata_offset + sidedata_size: |
|
1080 | 1082 | filename = self.sidedata_file |
|
1081 | 1083 | end = sidedata_end |
|
1082 | 1084 | offset = sidedata_offset |
|
1083 | 1085 | length = sidedata_size |
|
1084 | 1086 | m = FILE_TOO_SHORT_MSG % (filename, length, offset, end) |
|
1085 | 1087 | raise error.RevlogError(m) |
|
1086 | 1088 | |
|
1087 | 1089 | comp_segment = self._segmentfile_sidedata.read_chunk( |
|
1088 | 1090 | sidedata_offset, sidedata_size |
|
1089 | 1091 | ) |
|
1090 | 1092 | |
|
1091 | 1093 | comp = self.index[rev][11] |
|
1092 | 1094 | if comp == COMP_MODE_PLAIN: |
|
1093 | 1095 | segment = comp_segment |
|
1094 | 1096 | elif comp == COMP_MODE_DEFAULT: |
|
1095 | 1097 | segment = self._decompressor(comp_segment) |
|
1096 | 1098 | elif comp == COMP_MODE_INLINE: |
|
1097 | 1099 | segment = self.decompress(comp_segment) |
|
1098 | 1100 | else: |
|
1099 | 1101 | msg = b'unknown compression mode %d' |
|
1100 | 1102 | msg %= comp |
|
1101 | 1103 | raise error.RevlogError(msg) |
|
1102 | 1104 | |
|
1103 | 1105 | sidedata = sidedatautil.deserialize_sidedata(segment) |
|
1104 | 1106 | return sidedata |
|
1105 | 1107 | |
|
1106 | 1108 | def write_entry( |
|
1107 | 1109 | self, |
|
1108 | 1110 | transaction, |
|
1109 | 1111 | entry, |
|
1110 | 1112 | data, |
|
1111 | 1113 | link, |
|
1112 | 1114 | offset, |
|
1113 | 1115 | sidedata, |
|
1114 | 1116 | sidedata_offset, |
|
1115 | 1117 | index_end, |
|
1116 | 1118 | data_end, |
|
1117 | 1119 | sidedata_end, |
|
1118 | 1120 | ): |
|
1119 | 1121 | # Files opened in a+ mode have inconsistent behavior on various |
|
1120 | 1122 | # platforms. Windows requires that a file positioning call be made |
|
1121 | 1123 | # when the file handle transitions between reads and writes. See |
|
1122 | 1124 | # 3686fa2b8eee and the mixedfilemodewrapper in windows.py. On other |
|
1123 | 1125 | # platforms, Python or the platform itself can be buggy. Some versions |
|
1124 | 1126 | # of Solaris have been observed to not append at the end of the file |
|
1125 | 1127 | # if the file was seeked to before the end. See issue4943 for more. |
|
1126 | 1128 | # |
|
1127 | 1129 | # We work around this issue by inserting a seek() before writing. |
|
1128 | 1130 | # Note: This is likely not necessary on Python 3. However, because |
|
1129 | 1131 | # the file handle is reused for reads and may be seeked there, we need |
|
1130 | 1132 | # to be careful before changing this. |
|
1131 | 1133 | if self._writinghandles is None: |
|
1132 | 1134 | msg = b'adding revision outside `revlog._writing` context' |
|
1133 | 1135 | raise error.ProgrammingError(msg) |
|
1134 | 1136 | ifh, dfh, sdfh = self._writinghandles |
|
1135 | 1137 | if index_end is None: |
|
1136 | 1138 | ifh.seek(0, os.SEEK_END) |
|
1137 | 1139 | else: |
|
1138 | 1140 | ifh.seek(index_end, os.SEEK_SET) |
|
1139 | 1141 | if dfh: |
|
1140 | 1142 | if data_end is None: |
|
1141 | 1143 | dfh.seek(0, os.SEEK_END) |
|
1142 | 1144 | else: |
|
1143 | 1145 | dfh.seek(data_end, os.SEEK_SET) |
|
1144 | 1146 | if sdfh: |
|
1145 | 1147 | sdfh.seek(sidedata_end, os.SEEK_SET) |
|
1146 | 1148 | |
|
1147 | 1149 | curr = len(self.index) - 1 |
|
1148 | 1150 | if not self.inline: |
|
1149 | 1151 | transaction.add(self.data_file, offset) |
|
1150 | 1152 | if self.sidedata_file: |
|
1151 | 1153 | transaction.add(self.sidedata_file, sidedata_offset) |
|
1152 | 1154 | transaction.add(self.canonical_index_file, curr * len(entry)) |
|
1153 | 1155 | if data[0]: |
|
1154 | 1156 | dfh.write(data[0]) |
|
1155 | 1157 | dfh.write(data[1]) |
|
1156 | 1158 | if sidedata: |
|
1157 | 1159 | sdfh.write(sidedata) |
|
1158 | 1160 | if self._delay_buffer is None: |
|
1159 | 1161 | ifh.write(entry) |
|
1160 | 1162 | else: |
|
1161 | 1163 | self._delay_buffer.append(entry) |
|
1162 | 1164 | elif self._delay_buffer is not None: |
|
1163 | 1165 | msg = b'invalid delayed write on inline revlog' |
|
1164 | 1166 | raise error.ProgrammingError(msg) |
|
1165 | 1167 | else: |
|
1166 | 1168 | offset += curr * self.index.entry_size |
|
1167 | 1169 | transaction.add(self.canonical_index_file, offset) |
|
1168 | 1170 | assert not sidedata |
|
1169 | 1171 | ifh.write(entry) |
|
1170 | 1172 | ifh.write(data[0]) |
|
1171 | 1173 | ifh.write(data[1]) |
|
1172 | 1174 | return ( |
|
1173 | 1175 | ifh.tell(), |
|
1174 | 1176 | dfh.tell() if dfh else None, |
|
1175 | 1177 | sdfh.tell() if sdfh else None, |
|
1176 | 1178 | ) |
|
1177 | 1179 | |
|
1178 | 1180 | def _divert_index(self): |
|
1179 | 1181 | index_file = self.index_file |
|
1180 | 1182 | # when we encounter a legacy inline-changelog, split it. However it is |
|
1181 | 1183 | # important to use the expected filename for pending content |
|
1182 | 1184 | # (<radix>.a) otherwise hooks won't be seeing the content of the |
|
1183 | 1185 | # pending transaction. |
|
1184 | 1186 | if index_file.endswith(b'.s'): |
|
1185 | 1187 | index_file = self.index_file[:-2] |
|
1186 | 1188 | return index_file + b'.a' |
|
1187 | 1189 | |
|
1188 | 1190 | def delay(self): |
|
1189 | 1191 | assert not self.is_open |
|
1190 | 1192 | if self.inline: |
|
1191 | 1193 | msg = "revlog with delayed write should not be inline" |
|
1192 | 1194 | raise error.ProgrammingError(msg) |
|
1193 | 1195 | if self._delay_buffer is not None or self._orig_index_file is not None: |
|
1194 | 1196 | # delay or divert already in place |
|
1195 | 1197 | return None |
|
1196 | 1198 | elif len(self.index) == 0: |
|
1197 | 1199 | self._orig_index_file = self.index_file |
|
1198 | 1200 | self.index_file = self._divert_index() |
|
1199 | 1201 | assert self._orig_index_file is not None |
|
1200 | 1202 | assert self.index_file is not None |
|
1201 | 1203 | if self.opener.exists(self.index_file): |
|
1202 | 1204 | self.opener.unlink(self.index_file) |
|
1203 | 1205 | return self.index_file |
|
1204 | 1206 | else: |
|
1205 | 1207 | self._delay_buffer = [] |
|
1206 | 1208 | return None |
|
1207 | 1209 | |
|
1208 | 1210 | def write_pending(self): |
|
1209 | 1211 | assert not self.is_open |
|
1210 | 1212 | if self.inline: |
|
1211 | 1213 | msg = "revlog with delayed write should not be inline" |
|
1212 | 1214 | raise error.ProgrammingError(msg) |
|
1213 | 1215 | if self._orig_index_file is not None: |
|
1214 | 1216 | return None, True |
|
1215 | 1217 | any_pending = False |
|
1216 | 1218 | pending_index_file = self._divert_index() |
|
1217 | 1219 | if self.opener.exists(pending_index_file): |
|
1218 | 1220 | self.opener.unlink(pending_index_file) |
|
1219 | 1221 | util.copyfile( |
|
1220 | 1222 | self.opener.join(self.index_file), |
|
1221 | 1223 | self.opener.join(pending_index_file), |
|
1222 | 1224 | ) |
|
1223 | 1225 | if self._delay_buffer: |
|
1224 | 1226 | with self.opener(pending_index_file, b'r+') as ifh: |
|
1225 | 1227 | ifh.seek(0, os.SEEK_END) |
|
1226 | 1228 | ifh.write(b"".join(self._delay_buffer)) |
|
1227 | 1229 | any_pending = True |
|
1228 | 1230 | self._delay_buffer = None |
|
1229 | 1231 | self._orig_index_file = self.index_file |
|
1230 | 1232 | self.index_file = pending_index_file |
|
1231 | 1233 | return self.index_file, any_pending |
|
1232 | 1234 | |
|
1233 | 1235 | def finalize_pending(self): |
|
1234 | 1236 | assert not self.is_open |
|
1235 | 1237 | if self.inline: |
|
1236 | 1238 | msg = "revlog with delayed write should not be inline" |
|
1237 | 1239 | raise error.ProgrammingError(msg) |
|
1238 | 1240 | |
|
1239 | 1241 | delay = self._delay_buffer is not None |
|
1240 | 1242 | divert = self._orig_index_file is not None |
|
1241 | 1243 | |
|
1242 | 1244 | if delay and divert: |
|
1243 | 1245 | assert False, "unreachable" |
|
1244 | 1246 | elif delay: |
|
1245 | 1247 | if self._delay_buffer: |
|
1246 | 1248 | with self.opener(self.index_file, b'r+') as ifh: |
|
1247 | 1249 | ifh.seek(0, os.SEEK_END) |
|
1248 | 1250 | ifh.write(b"".join(self._delay_buffer)) |
|
1249 | 1251 | self._delay_buffer = None |
|
1250 | 1252 | elif divert: |
|
1251 | 1253 | if self.opener.exists(self.index_file): |
|
1252 | 1254 | self.opener.rename( |
|
1253 | 1255 | self.index_file, |
|
1254 | 1256 | self._orig_index_file, |
|
1255 | 1257 | checkambig=True, |
|
1256 | 1258 | ) |
|
1257 | 1259 | self.index_file = self._orig_index_file |
|
1258 | 1260 | self._orig_index_file = None |
|
1259 | 1261 | else: |
|
1260 | 1262 | msg = b"not delay or divert found on this revlog" |
|
1261 | 1263 | raise error.ProgrammingError(msg) |
|
1262 | 1264 | return self.canonical_index_file |
|
1263 | 1265 | |
|
1264 | 1266 | |
|
1265 | 1267 | class revlog: |
|
1266 | 1268 | """ |
|
1267 | 1269 | the underlying revision storage object |
|
1268 | 1270 | |
|
1269 | 1271 | A revlog consists of two parts, an index and the revision data. |
|
1270 | 1272 | |
|
1271 | 1273 | The index is a file with a fixed record size containing |
|
1272 | 1274 | information on each revision, including its nodeid (hash), the |
|
1273 | 1275 | nodeids of its parents, the position and offset of its data within |
|
1274 | 1276 | the data file, and the revision it's based on. Finally, each entry |
|
1275 | 1277 | contains a linkrev entry that can serve as a pointer to external |
|
1276 | 1278 | data. |
|
1277 | 1279 | |
|
1278 | 1280 | The revision data itself is a linear collection of data chunks. |
|
1279 | 1281 | Each chunk represents a revision and is usually represented as a |
|
1280 | 1282 | delta against the previous chunk. To bound lookup time, runs of |
|
1281 | 1283 | deltas are limited to about 2 times the length of the original |
|
1282 | 1284 | version data. This makes retrieval of a version proportional to |
|
1283 | 1285 | its size, or O(1) relative to the number of revisions. |
|
1284 | 1286 | |
|
1285 | 1287 | Both pieces of the revlog are written to in an append-only |
|
1286 | 1288 | fashion, which means we never need to rewrite a file to insert or |
|
1287 | 1289 | remove data, and can use some simple techniques to avoid the need |
|
1288 | 1290 | for locking while reading. |
|
1289 | 1291 | |
|
1290 | 1292 | If checkambig, indexfile is opened with checkambig=True at |
|
1291 | 1293 | writing, to avoid file stat ambiguity. |
|
1292 | 1294 | |
|
1293 | 1295 | If mmaplargeindex is True, and an mmapindexthreshold is set, the |
|
1294 | 1296 | index will be mmapped rather than read if it is larger than the |
|
1295 | 1297 | configured threshold. |
|
1296 | 1298 | |
|
1297 | 1299 | If censorable is True, the revlog can have censored revisions. |
|
1298 | 1300 | |
|
1299 | 1301 | If `upperboundcomp` is not None, this is the expected maximal gain from |
|
1300 | 1302 | compression for the data content. |
|
1301 | 1303 | |
|
1302 | 1304 | `concurrencychecker` is an optional function that receives 3 arguments: a |
|
1303 | 1305 | file handle, a filename, and an expected position. It should check whether |
|
1304 | 1306 | the current position in the file handle is valid, and log/warn/fail (by |
|
1305 | 1307 | raising). |
|
1306 | 1308 | |
|
1307 | 1309 | See mercurial/revlogutils/contants.py for details about the content of an |
|
1308 | 1310 | index entry. |
|
1309 | 1311 | """ |
|
1310 | 1312 | |
|
1311 | 1313 | _flagserrorclass = error.RevlogError |
|
1312 | 1314 | |
|
1313 | 1315 | opener: vfsmod.vfs |
|
1314 | 1316 | |
|
1315 | 1317 | @staticmethod |
|
1316 | 1318 | def is_inline_index(header_bytes): |
|
1317 | 1319 | """Determine if a revlog is inline from the initial bytes of the index""" |
|
1318 | 1320 | if len(header_bytes) == 0: |
|
1319 | 1321 | return True |
|
1320 | 1322 | |
|
1321 | 1323 | header = INDEX_HEADER.unpack(header_bytes)[0] |
|
1322 | 1324 | |
|
1323 | 1325 | _format_flags = header & ~0xFFFF |
|
1324 | 1326 | _format_version = header & 0xFFFF |
|
1325 | 1327 | |
|
1326 | 1328 | features = FEATURES_BY_VERSION[_format_version] |
|
1327 | 1329 | return features[b'inline'](_format_flags) |
|
1328 | 1330 | |
|
1329 | 1331 | _docket_file: Optional[bytes] |
|
1330 | 1332 | |
|
1331 | 1333 | def __init__( |
|
1332 | 1334 | self, |
|
1333 | 1335 | opener: vfsmod.vfs, |
|
1334 | 1336 | target, |
|
1335 | 1337 | radix, |
|
1336 | 1338 | postfix=None, # only exist for `tmpcensored` now |
|
1337 | 1339 | checkambig=False, |
|
1338 | 1340 | mmaplargeindex=False, |
|
1339 | 1341 | censorable=False, |
|
1340 | 1342 | upperboundcomp=None, |
|
1341 | 1343 | persistentnodemap=False, |
|
1342 | 1344 | concurrencychecker=None, |
|
1343 | 1345 | trypending=False, |
|
1344 | 1346 | try_split=False, |
|
1345 | 1347 | canonical_parent_order=True, |
|
1346 | 1348 | data_config=None, |
|
1347 | 1349 | delta_config=None, |
|
1348 | 1350 | feature_config=None, |
|
1349 | 1351 | may_inline=True, # may inline new revlog |
|
1350 | 1352 | ): |
|
1351 | 1353 | """ |
|
1352 | 1354 | create a revlog object |
|
1353 | 1355 | |
|
1354 | 1356 | opener is a function that abstracts the file opening operation |
|
1355 | 1357 | and can be used to implement COW semantics or the like. |
|
1356 | 1358 | |
|
1357 | 1359 | `target`: a (KIND, ID) tuple that identify the content stored in |
|
1358 | 1360 | this revlog. It help the rest of the code to understand what the revlog |
|
1359 | 1361 | is about without having to resort to heuristic and index filename |
|
1360 | 1362 | analysis. Note: that this must be reliably be set by normal code, but |
|
1361 | 1363 | that test, debug, or performance measurement code might not set this to |
|
1362 | 1364 | accurate value. |
|
1363 | 1365 | """ |
|
1364 | 1366 | |
|
1365 | 1367 | self.radix = radix |
|
1366 | 1368 | |
|
1367 | 1369 | self._docket_file = None |
|
1368 | 1370 | self._indexfile = None |
|
1369 | 1371 | self._datafile = None |
|
1370 | 1372 | self._sidedatafile = None |
|
1371 | 1373 | self._nodemap_file = None |
|
1372 | 1374 | self.postfix = postfix |
|
1373 | 1375 | self._trypending = trypending |
|
1374 | 1376 | self._try_split = try_split |
|
1375 | 1377 | self._may_inline = may_inline |
|
1376 | 1378 | self.opener = opener |
|
1377 | 1379 | if persistentnodemap: |
|
1378 | 1380 | self._nodemap_file = nodemaputil.get_nodemap_file(self) |
|
1379 | 1381 | |
|
1380 | 1382 | assert target[0] in ALL_KINDS |
|
1381 | 1383 | assert len(target) == 2 |
|
1382 | 1384 | self.target = target |
|
1383 | 1385 | if feature_config is not None: |
|
1384 | 1386 | self.feature_config = feature_config.copy() |
|
1385 | 1387 | elif b'feature-config' in self.opener.options: |
|
1386 | 1388 | self.feature_config = self.opener.options[b'feature-config'].copy() |
|
1387 | 1389 | else: |
|
1388 | 1390 | self.feature_config = FeatureConfig() |
|
1389 | 1391 | self.feature_config.censorable = censorable |
|
1390 | 1392 | self.feature_config.canonical_parent_order = canonical_parent_order |
|
1391 | 1393 | if data_config is not None: |
|
1392 | 1394 | self.data_config = data_config.copy() |
|
1393 | 1395 | elif b'data-config' in self.opener.options: |
|
1394 | 1396 | self.data_config = self.opener.options[b'data-config'].copy() |
|
1395 | 1397 | else: |
|
1396 | 1398 | self.data_config = DataConfig() |
|
1397 | 1399 | self.data_config.check_ambig = checkambig |
|
1398 | 1400 | self.data_config.mmap_large_index = mmaplargeindex |
|
1399 | 1401 | if delta_config is not None: |
|
1400 | 1402 | self.delta_config = delta_config.copy() |
|
1401 | 1403 | elif b'delta-config' in self.opener.options: |
|
1402 | 1404 | self.delta_config = self.opener.options[b'delta-config'].copy() |
|
1403 | 1405 | else: |
|
1404 | 1406 | self.delta_config = DeltaConfig() |
|
1405 | 1407 | self.delta_config.upper_bound_comp = upperboundcomp |
|
1406 | 1408 | |
|
1407 | 1409 | # Maps rev to chain base rev. |
|
1408 | 1410 | self._chainbasecache = util.lrucachedict(100) |
|
1409 | 1411 | |
|
1410 | 1412 | self.index = None |
|
1411 | 1413 | self._docket = None |
|
1412 | 1414 | self._nodemap_docket = None |
|
1413 | 1415 | # Mapping of partial identifiers to full nodes. |
|
1414 | 1416 | self._pcache = {} |
|
1415 | 1417 | |
|
1416 | 1418 | # other optionnals features |
|
1417 | 1419 | |
|
1418 | 1420 | # Make copy of flag processors so each revlog instance can support |
|
1419 | 1421 | # custom flags. |
|
1420 | 1422 | self._flagprocessors = dict(flagutil.flagprocessors) |
|
1421 | 1423 | # prevent nesting of addgroup |
|
1422 | 1424 | self._adding_group = None |
|
1423 | 1425 | |
|
1424 | 1426 | chunk_cache = self._loadindex() |
|
1425 | 1427 | self._load_inner(chunk_cache) |
|
1426 | 1428 | self._concurrencychecker = concurrencychecker |
|
1427 | 1429 | |
|
1428 | 1430 | def _init_opts(self): |
|
1429 | 1431 | """process options (from above/config) to setup associated default revlog mode |
|
1430 | 1432 | |
|
1431 | 1433 | These values might be affected when actually reading on disk information. |
|
1432 | 1434 | |
|
1433 | 1435 | The relevant values are returned for use in _loadindex(). |
|
1434 | 1436 | |
|
1435 | 1437 | * newversionflags: |
|
1436 | 1438 | version header to use if we need to create a new revlog |
|
1437 | 1439 | |
|
1438 | 1440 | * mmapindexthreshold: |
|
1439 | 1441 | minimal index size for start to use mmap |
|
1440 | 1442 | |
|
1441 | 1443 | * force_nodemap: |
|
1442 | 1444 | force the usage of a "development" version of the nodemap code |
|
1443 | 1445 | """ |
|
1444 | 1446 | opts = self.opener.options |
|
1445 | 1447 | |
|
1446 | 1448 | if b'changelogv2' in opts and self.revlog_kind == KIND_CHANGELOG: |
|
1447 | 1449 | new_header = CHANGELOGV2 |
|
1448 | 1450 | compute_rank = opts.get(b'changelogv2.compute-rank', True) |
|
1449 | 1451 | self.feature_config.compute_rank = compute_rank |
|
1450 | 1452 | elif b'revlogv2' in opts: |
|
1451 | 1453 | new_header = REVLOGV2 |
|
1452 | 1454 | elif b'revlogv1' in opts: |
|
1453 | 1455 | new_header = REVLOGV1 |
|
1454 | 1456 | if self._may_inline: |
|
1455 | 1457 | new_header |= FLAG_INLINE_DATA |
|
1456 | 1458 | if b'generaldelta' in opts: |
|
1457 | 1459 | new_header |= FLAG_GENERALDELTA |
|
1458 | 1460 | elif b'revlogv0' in self.opener.options: |
|
1459 | 1461 | new_header = REVLOGV0 |
|
1460 | 1462 | else: |
|
1461 | 1463 | new_header = REVLOG_DEFAULT_VERSION |
|
1462 | 1464 | |
|
1463 | 1465 | mmapindexthreshold = None |
|
1464 | 1466 | if self.data_config.mmap_large_index: |
|
1465 | 1467 | mmapindexthreshold = self.data_config.mmap_index_threshold |
|
1466 | 1468 | if self.feature_config.enable_ellipsis: |
|
1467 | 1469 | self._flagprocessors[REVIDX_ELLIPSIS] = ellipsisprocessor |
|
1468 | 1470 | |
|
1469 | 1471 | # revlog v0 doesn't have flag processors |
|
1470 | 1472 | for flag, processor in opts.get(b'flagprocessors', {}).items(): |
|
1471 | 1473 | flagutil.insertflagprocessor(flag, processor, self._flagprocessors) |
|
1472 | 1474 | |
|
1473 | 1475 | chunk_cache_size = self.data_config.chunk_cache_size |
|
1474 | 1476 | if chunk_cache_size <= 0: |
|
1475 | 1477 | raise error.RevlogError( |
|
1476 | 1478 | _(b'revlog chunk cache size %r is not greater than 0') |
|
1477 | 1479 | % chunk_cache_size |
|
1478 | 1480 | ) |
|
1479 | 1481 | elif chunk_cache_size & (chunk_cache_size - 1): |
|
1480 | 1482 | raise error.RevlogError( |
|
1481 | 1483 | _(b'revlog chunk cache size %r is not a power of 2') |
|
1482 | 1484 | % chunk_cache_size |
|
1483 | 1485 | ) |
|
1484 | 1486 | force_nodemap = opts.get(b'devel-force-nodemap', False) |
|
1485 | 1487 | return new_header, mmapindexthreshold, force_nodemap |
|
1486 | 1488 | |
|
1487 | 1489 | def _get_data(self, filepath, mmap_threshold, size=None): |
|
1488 | 1490 | """return a file content with or without mmap |
|
1489 | 1491 | |
|
1490 | 1492 | If the file is missing return the empty string""" |
|
1491 | 1493 | try: |
|
1492 | 1494 | with self.opener(filepath) as fp: |
|
1493 | 1495 | if mmap_threshold is not None: |
|
1494 | 1496 | file_size = self.opener.fstat(fp).st_size |
|
1495 | 1497 | if ( |
|
1496 | 1498 | file_size >= mmap_threshold |
|
1497 | 1499 | and self.opener.is_mmap_safe(filepath) |
|
1498 | 1500 | ): |
|
1499 | 1501 | if size is not None: |
|
1500 | 1502 | # avoid potentiel mmap crash |
|
1501 | 1503 | size = min(file_size, size) |
|
1502 | 1504 | # TODO: should .close() to release resources without |
|
1503 | 1505 | # relying on Python GC |
|
1504 | 1506 | if size is None: |
|
1505 | 1507 | return util.buffer(util.mmapread(fp)) |
|
1506 | 1508 | else: |
|
1507 | 1509 | return util.buffer(util.mmapread(fp, size)) |
|
1508 | 1510 | if size is None: |
|
1509 | 1511 | return fp.read() |
|
1510 | 1512 | else: |
|
1511 | 1513 | return fp.read(size) |
|
1512 | 1514 | except FileNotFoundError: |
|
1513 | 1515 | return b'' |
|
1514 | 1516 | |
|
1515 | 1517 | def get_streams(self, max_linkrev, force_inline=False): |
|
1516 | 1518 | """return a list of streams that represent this revlog |
|
1517 | 1519 | |
|
1518 | 1520 | This is used by stream-clone to do bytes to bytes copies of a repository. |
|
1519 | 1521 | |
|
1520 | 1522 | This streams data for all revisions that refer to a changelog revision up |
|
1521 | 1523 | to `max_linkrev`. |
|
1522 | 1524 | |
|
1523 | 1525 | If `force_inline` is set, it enforces that the stream will represent an inline revlog. |
|
1524 | 1526 | |
|
1525 | 1527 | It returns is a list of three-tuple: |
|
1526 | 1528 | |
|
1527 | 1529 | [ |
|
1528 | 1530 | (filename, bytes_stream, stream_size), |
|
1529 | 1531 | β¦ |
|
1530 | 1532 | ] |
|
1531 | 1533 | """ |
|
1532 | 1534 | n = len(self) |
|
1533 | 1535 | index = self.index |
|
1534 | 1536 | while n > 0: |
|
1535 | 1537 | linkrev = index[n - 1][4] |
|
1536 | 1538 | if linkrev < max_linkrev: |
|
1537 | 1539 | break |
|
1538 | 1540 | # note: this loop will rarely go through multiple iterations, since |
|
1539 | 1541 | # it only traverses commits created during the current streaming |
|
1540 | 1542 | # pull operation. |
|
1541 | 1543 | # |
|
1542 | 1544 | # If this become a problem, using a binary search should cap the |
|
1543 | 1545 | # runtime of this. |
|
1544 | 1546 | n = n - 1 |
|
1545 | 1547 | if n == 0: |
|
1546 | 1548 | # no data to send |
|
1547 | 1549 | return [] |
|
1548 | 1550 | index_size = n * index.entry_size |
|
1549 | 1551 | data_size = self.end(n - 1) |
|
1550 | 1552 | |
|
1551 | 1553 | # XXX we might have been split (or stripped) since the object |
|
1552 | 1554 | # initialization, We need to close this race too, but having a way to |
|
1553 | 1555 | # pre-open the file we feed to the revlog and never closing them before |
|
1554 | 1556 | # we are done streaming. |
|
1555 | 1557 | |
|
1556 | 1558 | if self._inline: |
|
1557 | 1559 | |
|
1558 | 1560 | def get_stream(): |
|
1559 | 1561 | with self.opener(self._indexfile, mode=b"r") as fp: |
|
1560 | 1562 | yield None |
|
1561 | 1563 | size = index_size + data_size |
|
1562 | 1564 | if size <= 65536: |
|
1563 | 1565 | yield fp.read(size) |
|
1564 | 1566 | else: |
|
1565 | 1567 | yield from util.filechunkiter(fp, limit=size) |
|
1566 | 1568 | |
|
1567 | 1569 | inline_stream = get_stream() |
|
1568 | 1570 | next(inline_stream) |
|
1569 | 1571 | return [ |
|
1570 | 1572 | (self._indexfile, inline_stream, index_size + data_size), |
|
1571 | 1573 | ] |
|
1572 | 1574 | elif force_inline: |
|
1573 | 1575 | |
|
1574 | 1576 | def get_stream(): |
|
1575 | 1577 | with self.reading(): |
|
1576 | 1578 | yield None |
|
1577 | 1579 | |
|
1578 | 1580 | for rev in range(n): |
|
1579 | 1581 | idx = self.index.entry_binary(rev) |
|
1580 | 1582 | if rev == 0 and self._docket is None: |
|
1581 | 1583 | # re-inject the inline flag |
|
1582 | 1584 | header = self._format_flags |
|
1583 | 1585 | header |= self._format_version |
|
1584 | 1586 | header |= FLAG_INLINE_DATA |
|
1585 | 1587 | header = self.index.pack_header(header) |
|
1586 | 1588 | idx = header + idx |
|
1587 | 1589 | yield idx |
|
1588 | 1590 | yield self._inner.get_segment_for_revs(rev, rev)[1] |
|
1589 | 1591 | |
|
1590 | 1592 | inline_stream = get_stream() |
|
1591 | 1593 | next(inline_stream) |
|
1592 | 1594 | return [ |
|
1593 | 1595 | (self._indexfile, inline_stream, index_size + data_size), |
|
1594 | 1596 | ] |
|
1595 | 1597 | else: |
|
1596 | 1598 | |
|
1597 | 1599 | def get_index_stream(): |
|
1598 | 1600 | with self.opener(self._indexfile, mode=b"r") as fp: |
|
1599 | 1601 | yield None |
|
1600 | 1602 | if index_size <= 65536: |
|
1601 | 1603 | yield fp.read(index_size) |
|
1602 | 1604 | else: |
|
1603 | 1605 | yield from util.filechunkiter(fp, limit=index_size) |
|
1604 | 1606 | |
|
1605 | 1607 | def get_data_stream(): |
|
1606 | 1608 | with self._datafp() as fp: |
|
1607 | 1609 | yield None |
|
1608 | 1610 | if data_size <= 65536: |
|
1609 | 1611 | yield fp.read(data_size) |
|
1610 | 1612 | else: |
|
1611 | 1613 | yield from util.filechunkiter(fp, limit=data_size) |
|
1612 | 1614 | |
|
1613 | 1615 | index_stream = get_index_stream() |
|
1614 | 1616 | next(index_stream) |
|
1615 | 1617 | data_stream = get_data_stream() |
|
1616 | 1618 | next(data_stream) |
|
1617 | 1619 | return [ |
|
1618 | 1620 | (self._datafile, data_stream, data_size), |
|
1619 | 1621 | (self._indexfile, index_stream, index_size), |
|
1620 | 1622 | ] |
|
1621 | 1623 | |
|
1622 | 1624 | def _loadindex(self, docket=None): |
|
1623 | 1625 | new_header, mmapindexthreshold, force_nodemap = self._init_opts() |
|
1624 | 1626 | |
|
1625 | 1627 | if self.postfix is not None: |
|
1626 | 1628 | entry_point = b'%s.i.%s' % (self.radix, self.postfix) |
|
1627 | 1629 | elif self._trypending and self.opener.exists(b'%s.i.a' % self.radix): |
|
1628 | 1630 | entry_point = b'%s.i.a' % self.radix |
|
1629 | 1631 | elif self._try_split and self.opener.exists(self._split_index_file): |
|
1630 | 1632 | entry_point = self._split_index_file |
|
1631 | 1633 | else: |
|
1632 | 1634 | entry_point = b'%s.i' % self.radix |
|
1633 | 1635 | |
|
1634 | 1636 | if docket is not None: |
|
1635 | 1637 | self._docket = docket |
|
1636 | 1638 | self._docket_file = entry_point |
|
1637 | 1639 | else: |
|
1638 | 1640 | self._initempty = True |
|
1639 | 1641 | entry_data = self._get_data(entry_point, mmapindexthreshold) |
|
1640 | 1642 | if len(entry_data) > 0: |
|
1641 | 1643 | header = INDEX_HEADER.unpack(entry_data[:4])[0] |
|
1642 | 1644 | self._initempty = False |
|
1643 | 1645 | else: |
|
1644 | 1646 | header = new_header |
|
1645 | 1647 | |
|
1646 | 1648 | self._format_flags = header & ~0xFFFF |
|
1647 | 1649 | self._format_version = header & 0xFFFF |
|
1648 | 1650 | |
|
1649 | 1651 | supported_flags = SUPPORTED_FLAGS.get(self._format_version) |
|
1650 | 1652 | if supported_flags is None: |
|
1651 | 1653 | msg = _(b'unknown version (%d) in revlog %s') |
|
1652 | 1654 | msg %= (self._format_version, self.display_id) |
|
1653 | 1655 | raise error.RevlogError(msg) |
|
1654 | 1656 | elif self._format_flags & ~supported_flags: |
|
1655 | 1657 | msg = _(b'unknown flags (%#04x) in version %d revlog %s') |
|
1656 | 1658 | display_flag = self._format_flags >> 16 |
|
1657 | 1659 | msg %= (display_flag, self._format_version, self.display_id) |
|
1658 | 1660 | raise error.RevlogError(msg) |
|
1659 | 1661 | |
|
1660 | 1662 | features = FEATURES_BY_VERSION[self._format_version] |
|
1661 | 1663 | self._inline = features[b'inline'](self._format_flags) |
|
1662 | 1664 | self.delta_config.general_delta = features[b'generaldelta']( |
|
1663 | 1665 | self._format_flags |
|
1664 | 1666 | ) |
|
1665 | 1667 | self.feature_config.has_side_data = features[b'sidedata'] |
|
1666 | 1668 | |
|
1667 | 1669 | if not features[b'docket']: |
|
1668 | 1670 | self._indexfile = entry_point |
|
1669 | 1671 | index_data = entry_data |
|
1670 | 1672 | else: |
|
1671 | 1673 | self._docket_file = entry_point |
|
1672 | 1674 | if self._initempty: |
|
1673 | 1675 | self._docket = docketutil.default_docket(self, header) |
|
1674 | 1676 | else: |
|
1675 | 1677 | self._docket = docketutil.parse_docket( |
|
1676 | 1678 | self, entry_data, use_pending=self._trypending |
|
1677 | 1679 | ) |
|
1678 | 1680 | |
|
1679 | 1681 | if self._docket is not None: |
|
1680 | 1682 | self._indexfile = self._docket.index_filepath() |
|
1681 | 1683 | index_data = b'' |
|
1682 | 1684 | index_size = self._docket.index_end |
|
1683 | 1685 | if index_size > 0: |
|
1684 | 1686 | index_data = self._get_data( |
|
1685 | 1687 | self._indexfile, mmapindexthreshold, size=index_size |
|
1686 | 1688 | ) |
|
1687 | 1689 | if len(index_data) < index_size: |
|
1688 | 1690 | msg = _(b'too few index data for %s: got %d, expected %d') |
|
1689 | 1691 | msg %= (self.display_id, len(index_data), index_size) |
|
1690 | 1692 | raise error.RevlogError(msg) |
|
1691 | 1693 | |
|
1692 | 1694 | self._inline = False |
|
1693 | 1695 | # generaldelta implied by version 2 revlogs. |
|
1694 | 1696 | self.delta_config.general_delta = True |
|
1695 | 1697 | # the logic for persistent nodemap will be dealt with within the |
|
1696 | 1698 | # main docket, so disable it for now. |
|
1697 | 1699 | self._nodemap_file = None |
|
1698 | 1700 | |
|
1699 | 1701 | if self._docket is not None: |
|
1700 | 1702 | self._datafile = self._docket.data_filepath() |
|
1701 | 1703 | self._sidedatafile = self._docket.sidedata_filepath() |
|
1702 | 1704 | elif self.postfix is None: |
|
1703 | 1705 | self._datafile = b'%s.d' % self.radix |
|
1704 | 1706 | else: |
|
1705 | 1707 | self._datafile = b'%s.d.%s' % (self.radix, self.postfix) |
|
1706 | 1708 | |
|
1707 | 1709 | self.nodeconstants = sha1nodeconstants |
|
1708 | 1710 | self.nullid = self.nodeconstants.nullid |
|
1709 | 1711 | |
|
1710 | 1712 | # sparse-revlog can't be on without general-delta (issue6056) |
|
1711 | 1713 | if not self.delta_config.general_delta: |
|
1712 | 1714 | self.delta_config.sparse_revlog = False |
|
1713 | 1715 | |
|
1714 | 1716 | self._storedeltachains = True |
|
1715 | 1717 | |
|
1716 | 1718 | devel_nodemap = ( |
|
1717 | 1719 | self._nodemap_file |
|
1718 | 1720 | and force_nodemap |
|
1719 | 1721 | and parse_index_v1_nodemap is not None |
|
1720 | 1722 | ) |
|
1721 | 1723 | |
|
1722 | 1724 | use_rust_index = False |
|
1723 | 1725 | if rustrevlog is not None and self._nodemap_file is not None: |
|
1724 | 1726 | # we would like to use the rust_index in all case, especially |
|
1725 | 1727 | # because it is necessary for AncestorsIterator and LazyAncestors |
|
1726 | 1728 | # since the 6.7 cycle. |
|
1727 | 1729 | # |
|
1728 | 1730 | # However, the performance impact of inconditionnaly building the |
|
1729 | 1731 | # nodemap is currently a problem for non-persistent nodemap |
|
1730 | 1732 | # repository. |
|
1731 | 1733 | use_rust_index = True |
|
1732 | 1734 | |
|
1733 | 1735 | self._parse_index = parse_index_v1 |
|
1734 | 1736 | if self._format_version == REVLOGV0: |
|
1735 | 1737 | self._parse_index = revlogv0.parse_index_v0 |
|
1736 | 1738 | elif self._format_version == REVLOGV2: |
|
1737 | 1739 | self._parse_index = parse_index_v2 |
|
1738 | 1740 | elif self._format_version == CHANGELOGV2: |
|
1739 | 1741 | self._parse_index = parse_index_cl_v2 |
|
1740 | 1742 | elif devel_nodemap: |
|
1741 | 1743 | self._parse_index = parse_index_v1_nodemap |
|
1742 | 1744 | elif use_rust_index: |
|
1743 | 1745 | self._parse_index = functools.partial( |
|
1744 | 1746 | parse_index_v1_rust, default_header=new_header |
|
1745 | 1747 | ) |
|
1746 | 1748 | try: |
|
1747 | 1749 | d = self._parse_index(index_data, self._inline) |
|
1748 | 1750 | index, chunkcache = d |
|
1749 | 1751 | use_nodemap = ( |
|
1750 | 1752 | not self._inline |
|
1751 | 1753 | and self._nodemap_file is not None |
|
1752 | 1754 | and hasattr(index, 'update_nodemap_data') |
|
1753 | 1755 | ) |
|
1754 | 1756 | if use_nodemap: |
|
1755 | 1757 | nodemap_data = nodemaputil.persisted_data(self) |
|
1756 | 1758 | if nodemap_data is not None: |
|
1757 | 1759 | docket = nodemap_data[0] |
|
1758 | 1760 | if ( |
|
1759 | 1761 | len(d[0]) > docket.tip_rev |
|
1760 | 1762 | and d[0][docket.tip_rev][7] == docket.tip_node |
|
1761 | 1763 | ): |
|
1762 | 1764 | # no changelog tampering |
|
1763 | 1765 | self._nodemap_docket = docket |
|
1764 | 1766 | index.update_nodemap_data(*nodemap_data) |
|
1765 | 1767 | except (ValueError, IndexError): |
|
1766 | 1768 | raise error.RevlogError( |
|
1767 | 1769 | _(b"index %s is corrupted") % self.display_id |
|
1768 | 1770 | ) |
|
1769 | 1771 | self.index = index |
|
1770 | 1772 | # revnum -> (chain-length, sum-delta-length) |
|
1771 | 1773 | self._chaininfocache = util.lrucachedict(500) |
|
1772 | 1774 | |
|
1773 | 1775 | return chunkcache |
|
1774 | 1776 | |
|
1775 | 1777 | def _load_inner(self, chunk_cache): |
|
1776 | 1778 | if self._docket is None: |
|
1777 | 1779 | default_compression_header = None |
|
1778 | 1780 | else: |
|
1779 | 1781 | default_compression_header = self._docket.default_compression_header |
|
1780 | 1782 | |
|
1781 | 1783 | self._inner = _InnerRevlog( |
|
1782 | 1784 | opener=self.opener, |
|
1783 | 1785 | index=self.index, |
|
1784 | 1786 | index_file=self._indexfile, |
|
1785 | 1787 | data_file=self._datafile, |
|
1786 | 1788 | sidedata_file=self._sidedatafile, |
|
1787 | 1789 | inline=self._inline, |
|
1788 | 1790 | data_config=self.data_config, |
|
1789 | 1791 | delta_config=self.delta_config, |
|
1790 | 1792 | feature_config=self.feature_config, |
|
1791 | 1793 | chunk_cache=chunk_cache, |
|
1792 | 1794 | default_compression_header=default_compression_header, |
|
1793 | 1795 | ) |
|
1794 | 1796 | |
|
1795 | 1797 | def get_revlog(self): |
|
1796 | 1798 | """simple function to mirror API of other not-really-revlog API""" |
|
1797 | 1799 | return self |
|
1798 | 1800 | |
|
1799 | 1801 | @util.propertycache |
|
1800 | 1802 | def revlog_kind(self): |
|
1801 | 1803 | return self.target[0] |
|
1802 | 1804 | |
|
1803 | 1805 | @util.propertycache |
|
1804 | 1806 | def display_id(self): |
|
1805 | 1807 | """The public facing "ID" of the revlog that we use in message""" |
|
1806 | 1808 | if self.revlog_kind == KIND_FILELOG: |
|
1807 | 1809 | # Reference the file without the "data/" prefix, so it is familiar |
|
1808 | 1810 | # to the user. |
|
1809 | 1811 | return self.target[1] |
|
1810 | 1812 | else: |
|
1811 | 1813 | return self.radix |
|
1812 | 1814 | |
|
1813 | 1815 | def _datafp(self, mode=b'r'): |
|
1814 | 1816 | """file object for the revlog's data file""" |
|
1815 | 1817 | return self.opener(self._datafile, mode=mode) |
|
1816 | 1818 | |
|
1817 | 1819 | def tiprev(self): |
|
1818 | 1820 | return len(self.index) - 1 |
|
1819 | 1821 | |
|
1820 | 1822 | def tip(self): |
|
1821 | 1823 | return self.node(self.tiprev()) |
|
1822 | 1824 | |
|
1823 | 1825 | def __contains__(self, rev): |
|
1824 | 1826 | return 0 <= rev < len(self) |
|
1825 | 1827 | |
|
1826 | 1828 | def __len__(self): |
|
1827 | 1829 | return len(self.index) |
|
1828 | 1830 | |
|
1829 | def __iter__(self): | |
|
1831 | def __iter__(self) -> Iterator[int]: | |
|
1830 | 1832 | return iter(range(len(self))) |
|
1831 | 1833 | |
|
1832 | 1834 | def revs(self, start=0, stop=None): |
|
1833 | 1835 | """iterate over all rev in this revlog (from start to stop)""" |
|
1834 | 1836 | return storageutil.iterrevs(len(self), start=start, stop=stop) |
|
1835 | 1837 | |
|
1836 | 1838 | def hasnode(self, node): |
|
1837 | 1839 | try: |
|
1838 | 1840 | self.rev(node) |
|
1839 | 1841 | return True |
|
1840 | 1842 | except KeyError: |
|
1841 | 1843 | return False |
|
1842 | 1844 | |
|
1843 | 1845 | def _candelta(self, baserev, rev): |
|
1844 | 1846 | """whether two revisions (baserev, rev) can be delta-ed or not""" |
|
1845 | 1847 | # Disable delta if either rev requires a content-changing flag |
|
1846 | 1848 | # processor (ex. LFS). This is because such flag processor can alter |
|
1847 | 1849 | # the rawtext content that the delta will be based on, and two clients |
|
1848 | 1850 | # could have a same revlog node with different flags (i.e. different |
|
1849 | 1851 | # rawtext contents) and the delta could be incompatible. |
|
1850 | 1852 | if (self.flags(baserev) & REVIDX_RAWTEXT_CHANGING_FLAGS) or ( |
|
1851 | 1853 | self.flags(rev) & REVIDX_RAWTEXT_CHANGING_FLAGS |
|
1852 | 1854 | ): |
|
1853 | 1855 | return False |
|
1854 | 1856 | return True |
|
1855 | 1857 | |
|
1856 | 1858 | def update_caches(self, transaction): |
|
1857 | 1859 | """update on disk cache |
|
1858 | 1860 | |
|
1859 | 1861 | If a transaction is passed, the update may be delayed to transaction |
|
1860 | 1862 | commit.""" |
|
1861 | 1863 | if self._nodemap_file is not None: |
|
1862 | 1864 | if transaction is None: |
|
1863 | 1865 | nodemaputil.update_persistent_nodemap(self) |
|
1864 | 1866 | else: |
|
1865 | 1867 | nodemaputil.setup_persistent_nodemap(transaction, self) |
|
1866 | 1868 | |
|
1867 | 1869 | def clearcaches(self): |
|
1868 | 1870 | """Clear in-memory caches""" |
|
1869 | 1871 | self._chainbasecache.clear() |
|
1870 | 1872 | self._inner.clear_cache() |
|
1871 | 1873 | self._pcache = {} |
|
1872 | 1874 | self._nodemap_docket = None |
|
1873 | 1875 | self.index.clearcaches() |
|
1874 | 1876 | # The python code is the one responsible for validating the docket, we |
|
1875 | 1877 | # end up having to refresh it here. |
|
1876 | 1878 | use_nodemap = ( |
|
1877 | 1879 | not self._inline |
|
1878 | 1880 | and self._nodemap_file is not None |
|
1879 | 1881 | and hasattr(self.index, 'update_nodemap_data') |
|
1880 | 1882 | ) |
|
1881 | 1883 | if use_nodemap: |
|
1882 | 1884 | nodemap_data = nodemaputil.persisted_data(self) |
|
1883 | 1885 | if nodemap_data is not None: |
|
1884 | 1886 | self._nodemap_docket = nodemap_data[0] |
|
1885 | 1887 | self.index.update_nodemap_data(*nodemap_data) |
|
1886 | 1888 | |
|
1887 | 1889 | def rev(self, node): |
|
1888 | 1890 | """return the revision number associated with a <nodeid>""" |
|
1889 | 1891 | try: |
|
1890 | 1892 | return self.index.rev(node) |
|
1891 | 1893 | except TypeError: |
|
1892 | 1894 | raise |
|
1893 | 1895 | except error.RevlogError: |
|
1894 | 1896 | # parsers.c radix tree lookup failed |
|
1895 | 1897 | if ( |
|
1896 | 1898 | node == self.nodeconstants.wdirid |
|
1897 | 1899 | or node in self.nodeconstants.wdirfilenodeids |
|
1898 | 1900 | ): |
|
1899 | 1901 | raise error.WdirUnsupported |
|
1900 | 1902 | raise error.LookupError(node, self.display_id, _(b'no node')) |
|
1901 | 1903 | |
|
1902 | 1904 | # Accessors for index entries. |
|
1903 | 1905 | |
|
1904 | 1906 | # First tuple entry is 8 bytes. First 6 bytes are offset. Last 2 bytes |
|
1905 | 1907 | # are flags. |
|
1906 | 1908 | def start(self, rev): |
|
1907 | 1909 | return int(self.index[rev][0] >> 16) |
|
1908 | 1910 | |
|
1909 | 1911 | def sidedata_cut_off(self, rev): |
|
1910 | 1912 | sd_cut_off = self.index[rev][8] |
|
1911 | 1913 | if sd_cut_off != 0: |
|
1912 | 1914 | return sd_cut_off |
|
1913 | 1915 | # This is some annoying dance, because entries without sidedata |
|
1914 | 1916 | # currently use 0 as their ofsset. (instead of previous-offset + |
|
1915 | 1917 | # previous-size) |
|
1916 | 1918 | # |
|
1917 | 1919 | # We should reconsider this sidedata β 0 sidata_offset policy. |
|
1918 | 1920 | # In the meantime, we need this. |
|
1919 | 1921 | while 0 <= rev: |
|
1920 | 1922 | e = self.index[rev] |
|
1921 | 1923 | if e[9] != 0: |
|
1922 | 1924 | return e[8] + e[9] |
|
1923 | 1925 | rev -= 1 |
|
1924 | 1926 | return 0 |
|
1925 | 1927 | |
|
1926 | 1928 | def flags(self, rev): |
|
1927 | 1929 | return self.index[rev][0] & 0xFFFF |
|
1928 | 1930 | |
|
1929 | 1931 | def length(self, rev): |
|
1930 | 1932 | return self.index[rev][1] |
|
1931 | 1933 | |
|
1932 | 1934 | def sidedata_length(self, rev): |
|
1933 | 1935 | if not self.feature_config.has_side_data: |
|
1934 | 1936 | return 0 |
|
1935 | 1937 | return self.index[rev][9] |
|
1936 | 1938 | |
|
1937 | 1939 | def rawsize(self, rev): |
|
1938 | 1940 | """return the length of the uncompressed text for a given revision""" |
|
1939 | 1941 | l = self.index[rev][2] |
|
1940 | 1942 | if l >= 0: |
|
1941 | 1943 | return l |
|
1942 | 1944 | |
|
1943 | 1945 | t = self.rawdata(rev) |
|
1944 | 1946 | return len(t) |
|
1945 | 1947 | |
|
1946 | 1948 | def size(self, rev): |
|
1947 | 1949 | """length of non-raw text (processed by a "read" flag processor)""" |
|
1948 | 1950 | # fast path: if no "read" flag processor could change the content, |
|
1949 | 1951 | # size is rawsize. note: ELLIPSIS is known to not change the content. |
|
1950 | 1952 | flags = self.flags(rev) |
|
1951 | 1953 | if flags & (flagutil.REVIDX_KNOWN_FLAGS ^ REVIDX_ELLIPSIS) == 0: |
|
1952 | 1954 | return self.rawsize(rev) |
|
1953 | 1955 | |
|
1954 | 1956 | return len(self.revision(rev)) |
|
1955 | 1957 | |
|
1956 | 1958 | def fast_rank(self, rev): |
|
1957 | 1959 | """Return the rank of a revision if already known, or None otherwise. |
|
1958 | 1960 | |
|
1959 | 1961 | The rank of a revision is the size of the sub-graph it defines as a |
|
1960 | 1962 | head. Equivalently, the rank of a revision `r` is the size of the set |
|
1961 | 1963 | `ancestors(r)`, `r` included. |
|
1962 | 1964 | |
|
1963 | 1965 | This method returns the rank retrieved from the revlog in constant |
|
1964 | 1966 | time. It makes no attempt at computing unknown values for versions of |
|
1965 | 1967 | the revlog which do not persist the rank. |
|
1966 | 1968 | """ |
|
1967 | 1969 | rank = self.index[rev][ENTRY_RANK] |
|
1968 | 1970 | if self._format_version != CHANGELOGV2 or rank == RANK_UNKNOWN: |
|
1969 | 1971 | return None |
|
1970 | 1972 | if rev == nullrev: |
|
1971 | 1973 | return 0 # convention |
|
1972 | 1974 | return rank |
|
1973 | 1975 | |
|
1974 | 1976 | def chainbase(self, rev): |
|
1975 | 1977 | base = self._chainbasecache.get(rev) |
|
1976 | 1978 | if base is not None: |
|
1977 | 1979 | return base |
|
1978 | 1980 | |
|
1979 | 1981 | index = self.index |
|
1980 | 1982 | iterrev = rev |
|
1981 | 1983 | base = index[iterrev][3] |
|
1982 | 1984 | while base != iterrev: |
|
1983 | 1985 | iterrev = base |
|
1984 | 1986 | base = index[iterrev][3] |
|
1985 | 1987 | |
|
1986 | 1988 | self._chainbasecache[rev] = base |
|
1987 | 1989 | return base |
|
1988 | 1990 | |
|
1989 | 1991 | def linkrev(self, rev): |
|
1990 | 1992 | return self.index[rev][4] |
|
1991 | 1993 | |
|
1992 | 1994 | def parentrevs(self, rev): |
|
1993 | 1995 | try: |
|
1994 | 1996 | entry = self.index[rev] |
|
1995 | 1997 | except IndexError: |
|
1996 | 1998 | if rev == wdirrev: |
|
1997 | 1999 | raise error.WdirUnsupported |
|
1998 | 2000 | raise |
|
1999 | 2001 | |
|
2000 | 2002 | if self.feature_config.canonical_parent_order and entry[5] == nullrev: |
|
2001 | 2003 | return entry[6], entry[5] |
|
2002 | 2004 | else: |
|
2003 | 2005 | return entry[5], entry[6] |
|
2004 | 2006 | |
|
2005 | 2007 | # fast parentrevs(rev) where rev isn't filtered |
|
2006 | 2008 | _uncheckedparentrevs = parentrevs |
|
2007 | 2009 | |
|
2008 | 2010 | def node(self, rev): |
|
2009 | 2011 | try: |
|
2010 | 2012 | return self.index[rev][7] |
|
2011 | 2013 | except IndexError: |
|
2012 | 2014 | if rev == wdirrev: |
|
2013 | 2015 | raise error.WdirUnsupported |
|
2014 | 2016 | raise |
|
2015 | 2017 | |
|
2016 | 2018 | # Derived from index values. |
|
2017 | 2019 | |
|
2018 | 2020 | def end(self, rev): |
|
2019 | 2021 | return self.start(rev) + self.length(rev) |
|
2020 | 2022 | |
|
2021 | 2023 | def parents(self, node): |
|
2022 | 2024 | i = self.index |
|
2023 | 2025 | d = i[self.rev(node)] |
|
2024 | 2026 | # inline node() to avoid function call overhead |
|
2025 | 2027 | if self.feature_config.canonical_parent_order and d[5] == self.nullid: |
|
2026 | 2028 | return i[d[6]][7], i[d[5]][7] |
|
2027 | 2029 | else: |
|
2028 | 2030 | return i[d[5]][7], i[d[6]][7] |
|
2029 | 2031 | |
|
2030 | 2032 | def chainlen(self, rev): |
|
2031 | 2033 | return self._chaininfo(rev)[0] |
|
2032 | 2034 | |
|
2033 | 2035 | def _chaininfo(self, rev): |
|
2034 | 2036 | chaininfocache = self._chaininfocache |
|
2035 | 2037 | if rev in chaininfocache: |
|
2036 | 2038 | return chaininfocache[rev] |
|
2037 | 2039 | index = self.index |
|
2038 | 2040 | generaldelta = self.delta_config.general_delta |
|
2039 | 2041 | iterrev = rev |
|
2040 | 2042 | e = index[iterrev] |
|
2041 | 2043 | clen = 0 |
|
2042 | 2044 | compresseddeltalen = 0 |
|
2043 | 2045 | while iterrev != e[3]: |
|
2044 | 2046 | clen += 1 |
|
2045 | 2047 | compresseddeltalen += e[1] |
|
2046 | 2048 | if generaldelta: |
|
2047 | 2049 | iterrev = e[3] |
|
2048 | 2050 | else: |
|
2049 | 2051 | iterrev -= 1 |
|
2050 | 2052 | if iterrev in chaininfocache: |
|
2051 | 2053 | t = chaininfocache[iterrev] |
|
2052 | 2054 | clen += t[0] |
|
2053 | 2055 | compresseddeltalen += t[1] |
|
2054 | 2056 | break |
|
2055 | 2057 | e = index[iterrev] |
|
2056 | 2058 | else: |
|
2057 | 2059 | # Add text length of base since decompressing that also takes |
|
2058 | 2060 | # work. For cache hits the length is already included. |
|
2059 | 2061 | compresseddeltalen += e[1] |
|
2060 | 2062 | r = (clen, compresseddeltalen) |
|
2061 | 2063 | chaininfocache[rev] = r |
|
2062 | 2064 | return r |
|
2063 | 2065 | |
|
2064 | 2066 | def _deltachain(self, rev, stoprev=None): |
|
2065 | 2067 | return self._inner._deltachain(rev, stoprev=stoprev) |
|
2066 | 2068 | |
|
2067 | 2069 | def ancestors(self, revs, stoprev=0, inclusive=False): |
|
2068 | 2070 | """Generate the ancestors of 'revs' in reverse revision order. |
|
2069 | 2071 | Does not generate revs lower than stoprev. |
|
2070 | 2072 | |
|
2071 | 2073 | See the documentation for ancestor.lazyancestors for more details.""" |
|
2072 | 2074 | |
|
2073 | 2075 | # first, make sure start revisions aren't filtered |
|
2074 | 2076 | revs = list(revs) |
|
2075 | 2077 | checkrev = self.node |
|
2076 | 2078 | for r in revs: |
|
2077 | 2079 | checkrev(r) |
|
2078 | 2080 | # and we're sure ancestors aren't filtered as well |
|
2079 | 2081 | |
|
2080 | 2082 | if rustancestor is not None and self.index.rust_ext_compat: |
|
2081 | 2083 | lazyancestors = rustancestor.LazyAncestors |
|
2082 | 2084 | arg = self.index |
|
2083 | 2085 | else: |
|
2084 | 2086 | lazyancestors = ancestor.lazyancestors |
|
2085 | 2087 | arg = self._uncheckedparentrevs |
|
2086 | 2088 | return lazyancestors(arg, revs, stoprev=stoprev, inclusive=inclusive) |
|
2087 | 2089 | |
|
2088 | 2090 | def descendants(self, revs): |
|
2089 | 2091 | return dagop.descendantrevs(revs, self.revs, self.parentrevs) |
|
2090 | 2092 | |
|
2091 | 2093 | def findcommonmissing(self, common=None, heads=None): |
|
2092 | 2094 | """Return a tuple of the ancestors of common and the ancestors of heads |
|
2093 | 2095 | that are not ancestors of common. In revset terminology, we return the |
|
2094 | 2096 | tuple: |
|
2095 | 2097 | |
|
2096 | 2098 | ::common, (::heads) - (::common) |
|
2097 | 2099 | |
|
2098 | 2100 | The list is sorted by revision number, meaning it is |
|
2099 | 2101 | topologically sorted. |
|
2100 | 2102 | |
|
2101 | 2103 | 'heads' and 'common' are both lists of node IDs. If heads is |
|
2102 | 2104 | not supplied, uses all of the revlog's heads. If common is not |
|
2103 | 2105 | supplied, uses nullid.""" |
|
2104 | 2106 | if common is None: |
|
2105 | 2107 | common = [self.nullid] |
|
2106 | 2108 | if heads is None: |
|
2107 | 2109 | heads = self.heads() |
|
2108 | 2110 | |
|
2109 | 2111 | common = [self.rev(n) for n in common] |
|
2110 | 2112 | heads = [self.rev(n) for n in heads] |
|
2111 | 2113 | |
|
2112 | 2114 | # we want the ancestors, but inclusive |
|
2113 | 2115 | class lazyset: |
|
2114 | 2116 | def __init__(self, lazyvalues): |
|
2115 | 2117 | self.addedvalues = set() |
|
2116 | 2118 | self.lazyvalues = lazyvalues |
|
2117 | 2119 | |
|
2118 | 2120 | def __contains__(self, value): |
|
2119 | 2121 | return value in self.addedvalues or value in self.lazyvalues |
|
2120 | 2122 | |
|
2121 | 2123 | def __iter__(self): |
|
2122 | 2124 | added = self.addedvalues |
|
2123 | 2125 | for r in added: |
|
2124 | 2126 | yield r |
|
2125 | 2127 | for r in self.lazyvalues: |
|
2126 | 2128 | if not r in added: |
|
2127 | 2129 | yield r |
|
2128 | 2130 | |
|
2129 | 2131 | def add(self, value): |
|
2130 | 2132 | self.addedvalues.add(value) |
|
2131 | 2133 | |
|
2132 | 2134 | def update(self, values): |
|
2133 | 2135 | self.addedvalues.update(values) |
|
2134 | 2136 | |
|
2135 | 2137 | has = lazyset(self.ancestors(common)) |
|
2136 | 2138 | has.add(nullrev) |
|
2137 | 2139 | has.update(common) |
|
2138 | 2140 | |
|
2139 | 2141 | # take all ancestors from heads that aren't in has |
|
2140 | 2142 | missing = set() |
|
2141 | 2143 | visit = collections.deque(r for r in heads if r not in has) |
|
2142 | 2144 | while visit: |
|
2143 | 2145 | r = visit.popleft() |
|
2144 | 2146 | if r in missing: |
|
2145 | 2147 | continue |
|
2146 | 2148 | else: |
|
2147 | 2149 | missing.add(r) |
|
2148 | 2150 | for p in self.parentrevs(r): |
|
2149 | 2151 | if p not in has: |
|
2150 | 2152 | visit.append(p) |
|
2151 | 2153 | missing = list(missing) |
|
2152 | 2154 | missing.sort() |
|
2153 | 2155 | return has, [self.node(miss) for miss in missing] |
|
2154 | 2156 | |
|
2155 | 2157 | def incrementalmissingrevs(self, common=None): |
|
2156 | 2158 | """Return an object that can be used to incrementally compute the |
|
2157 | 2159 | revision numbers of the ancestors of arbitrary sets that are not |
|
2158 | 2160 | ancestors of common. This is an ancestor.incrementalmissingancestors |
|
2159 | 2161 | object. |
|
2160 | 2162 | |
|
2161 | 2163 | 'common' is a list of revision numbers. If common is not supplied, uses |
|
2162 | 2164 | nullrev. |
|
2163 | 2165 | """ |
|
2164 | 2166 | if common is None: |
|
2165 | 2167 | common = [nullrev] |
|
2166 | 2168 | |
|
2167 | 2169 | if rustancestor is not None and self.index.rust_ext_compat: |
|
2168 | 2170 | return rustancestor.MissingAncestors(self.index, common) |
|
2169 | 2171 | return ancestor.incrementalmissingancestors(self.parentrevs, common) |
|
2170 | 2172 | |
|
2171 | 2173 | def findmissingrevs(self, common=None, heads=None): |
|
2172 | 2174 | """Return the revision numbers of the ancestors of heads that |
|
2173 | 2175 | are not ancestors of common. |
|
2174 | 2176 | |
|
2175 | 2177 | More specifically, return a list of revision numbers corresponding to |
|
2176 | 2178 | nodes N such that every N satisfies the following constraints: |
|
2177 | 2179 | |
|
2178 | 2180 | 1. N is an ancestor of some node in 'heads' |
|
2179 | 2181 | 2. N is not an ancestor of any node in 'common' |
|
2180 | 2182 | |
|
2181 | 2183 | The list is sorted by revision number, meaning it is |
|
2182 | 2184 | topologically sorted. |
|
2183 | 2185 | |
|
2184 | 2186 | 'heads' and 'common' are both lists of revision numbers. If heads is |
|
2185 | 2187 | not supplied, uses all of the revlog's heads. If common is not |
|
2186 | 2188 | supplied, uses nullid.""" |
|
2187 | 2189 | if common is None: |
|
2188 | 2190 | common = [nullrev] |
|
2189 | 2191 | if heads is None: |
|
2190 | 2192 | heads = self.headrevs() |
|
2191 | 2193 | |
|
2192 | 2194 | inc = self.incrementalmissingrevs(common=common) |
|
2193 | 2195 | return inc.missingancestors(heads) |
|
2194 | 2196 | |
|
2195 | 2197 | def findmissing(self, common=None, heads=None): |
|
2196 | 2198 | """Return the ancestors of heads that are not ancestors of common. |
|
2197 | 2199 | |
|
2198 | 2200 | More specifically, return a list of nodes N such that every N |
|
2199 | 2201 | satisfies the following constraints: |
|
2200 | 2202 | |
|
2201 | 2203 | 1. N is an ancestor of some node in 'heads' |
|
2202 | 2204 | 2. N is not an ancestor of any node in 'common' |
|
2203 | 2205 | |
|
2204 | 2206 | The list is sorted by revision number, meaning it is |
|
2205 | 2207 | topologically sorted. |
|
2206 | 2208 | |
|
2207 | 2209 | 'heads' and 'common' are both lists of node IDs. If heads is |
|
2208 | 2210 | not supplied, uses all of the revlog's heads. If common is not |
|
2209 | 2211 | supplied, uses nullid.""" |
|
2210 | 2212 | if common is None: |
|
2211 | 2213 | common = [self.nullid] |
|
2212 | 2214 | if heads is None: |
|
2213 | 2215 | heads = self.heads() |
|
2214 | 2216 | |
|
2215 | 2217 | common = [self.rev(n) for n in common] |
|
2216 | 2218 | heads = [self.rev(n) for n in heads] |
|
2217 | 2219 | |
|
2218 | 2220 | inc = self.incrementalmissingrevs(common=common) |
|
2219 | 2221 | return [self.node(r) for r in inc.missingancestors(heads)] |
|
2220 | 2222 | |
|
2221 | 2223 | def nodesbetween(self, roots=None, heads=None): |
|
2222 | 2224 | """Return a topological path from 'roots' to 'heads'. |
|
2223 | 2225 | |
|
2224 | 2226 | Return a tuple (nodes, outroots, outheads) where 'nodes' is a |
|
2225 | 2227 | topologically sorted list of all nodes N that satisfy both of |
|
2226 | 2228 | these constraints: |
|
2227 | 2229 | |
|
2228 | 2230 | 1. N is a descendant of some node in 'roots' |
|
2229 | 2231 | 2. N is an ancestor of some node in 'heads' |
|
2230 | 2232 | |
|
2231 | 2233 | Every node is considered to be both a descendant and an ancestor |
|
2232 | 2234 | of itself, so every reachable node in 'roots' and 'heads' will be |
|
2233 | 2235 | included in 'nodes'. |
|
2234 | 2236 | |
|
2235 | 2237 | 'outroots' is the list of reachable nodes in 'roots', i.e., the |
|
2236 | 2238 | subset of 'roots' that is returned in 'nodes'. Likewise, |
|
2237 | 2239 | 'outheads' is the subset of 'heads' that is also in 'nodes'. |
|
2238 | 2240 | |
|
2239 | 2241 | 'roots' and 'heads' are both lists of node IDs. If 'roots' is |
|
2240 | 2242 | unspecified, uses nullid as the only root. If 'heads' is |
|
2241 | 2243 | unspecified, uses list of all of the revlog's heads.""" |
|
2242 | 2244 | nonodes = ([], [], []) |
|
2243 | 2245 | if roots is not None: |
|
2244 | 2246 | roots = list(roots) |
|
2245 | 2247 | if not roots: |
|
2246 | 2248 | return nonodes |
|
2247 | 2249 | lowestrev = min([self.rev(n) for n in roots]) |
|
2248 | 2250 | else: |
|
2249 | 2251 | roots = [self.nullid] # Everybody's a descendant of nullid |
|
2250 | 2252 | lowestrev = nullrev |
|
2251 | 2253 | if (lowestrev == nullrev) and (heads is None): |
|
2252 | 2254 | # We want _all_ the nodes! |
|
2253 | 2255 | return ( |
|
2254 | 2256 | [self.node(r) for r in self], |
|
2255 | 2257 | [self.nullid], |
|
2256 | 2258 | list(self.heads()), |
|
2257 | 2259 | ) |
|
2258 | 2260 | if heads is None: |
|
2259 | 2261 | # All nodes are ancestors, so the latest ancestor is the last |
|
2260 | 2262 | # node. |
|
2261 | 2263 | highestrev = len(self) - 1 |
|
2262 | 2264 | # Set ancestors to None to signal that every node is an ancestor. |
|
2263 | 2265 | ancestors = None |
|
2264 | 2266 | # Set heads to an empty dictionary for later discovery of heads |
|
2265 | 2267 | heads = {} |
|
2266 | 2268 | else: |
|
2267 | 2269 | heads = list(heads) |
|
2268 | 2270 | if not heads: |
|
2269 | 2271 | return nonodes |
|
2270 | 2272 | ancestors = set() |
|
2271 | 2273 | # Turn heads into a dictionary so we can remove 'fake' heads. |
|
2272 | 2274 | # Also, later we will be using it to filter out the heads we can't |
|
2273 | 2275 | # find from roots. |
|
2274 | 2276 | heads = dict.fromkeys(heads, False) |
|
2275 | 2277 | # Start at the top and keep marking parents until we're done. |
|
2276 | 2278 | nodestotag = set(heads) |
|
2277 | 2279 | # Remember where the top was so we can use it as a limit later. |
|
2278 | 2280 | highestrev = max([self.rev(n) for n in nodestotag]) |
|
2279 | 2281 | while nodestotag: |
|
2280 | 2282 | # grab a node to tag |
|
2281 | 2283 | n = nodestotag.pop() |
|
2282 | 2284 | # Never tag nullid |
|
2283 | 2285 | if n == self.nullid: |
|
2284 | 2286 | continue |
|
2285 | 2287 | # A node's revision number represents its place in a |
|
2286 | 2288 | # topologically sorted list of nodes. |
|
2287 | 2289 | r = self.rev(n) |
|
2288 | 2290 | if r >= lowestrev: |
|
2289 | 2291 | if n not in ancestors: |
|
2290 | 2292 | # If we are possibly a descendant of one of the roots |
|
2291 | 2293 | # and we haven't already been marked as an ancestor |
|
2292 | 2294 | ancestors.add(n) # Mark as ancestor |
|
2293 | 2295 | # Add non-nullid parents to list of nodes to tag. |
|
2294 | 2296 | nodestotag.update( |
|
2295 | 2297 | [p for p in self.parents(n) if p != self.nullid] |
|
2296 | 2298 | ) |
|
2297 | 2299 | elif n in heads: # We've seen it before, is it a fake head? |
|
2298 | 2300 | # So it is, real heads should not be the ancestors of |
|
2299 | 2301 | # any other heads. |
|
2300 | 2302 | heads.pop(n) |
|
2301 | 2303 | if not ancestors: |
|
2302 | 2304 | return nonodes |
|
2303 | 2305 | # Now that we have our set of ancestors, we want to remove any |
|
2304 | 2306 | # roots that are not ancestors. |
|
2305 | 2307 | |
|
2306 | 2308 | # If one of the roots was nullid, everything is included anyway. |
|
2307 | 2309 | if lowestrev > nullrev: |
|
2308 | 2310 | # But, since we weren't, let's recompute the lowest rev to not |
|
2309 | 2311 | # include roots that aren't ancestors. |
|
2310 | 2312 | |
|
2311 | 2313 | # Filter out roots that aren't ancestors of heads |
|
2312 | 2314 | roots = [root for root in roots if root in ancestors] |
|
2313 | 2315 | # Recompute the lowest revision |
|
2314 | 2316 | if roots: |
|
2315 | 2317 | lowestrev = min([self.rev(root) for root in roots]) |
|
2316 | 2318 | else: |
|
2317 | 2319 | # No more roots? Return empty list |
|
2318 | 2320 | return nonodes |
|
2319 | 2321 | else: |
|
2320 | 2322 | # We are descending from nullid, and don't need to care about |
|
2321 | 2323 | # any other roots. |
|
2322 | 2324 | lowestrev = nullrev |
|
2323 | 2325 | roots = [self.nullid] |
|
2324 | 2326 | # Transform our roots list into a set. |
|
2325 | 2327 | descendants = set(roots) |
|
2326 | 2328 | # Also, keep the original roots so we can filter out roots that aren't |
|
2327 | 2329 | # 'real' roots (i.e. are descended from other roots). |
|
2328 | 2330 | roots = descendants.copy() |
|
2329 | 2331 | # Our topologically sorted list of output nodes. |
|
2330 | 2332 | orderedout = [] |
|
2331 | 2333 | # Don't start at nullid since we don't want nullid in our output list, |
|
2332 | 2334 | # and if nullid shows up in descendants, empty parents will look like |
|
2333 | 2335 | # they're descendants. |
|
2334 | 2336 | for r in self.revs(start=max(lowestrev, 0), stop=highestrev + 1): |
|
2335 | 2337 | n = self.node(r) |
|
2336 | 2338 | isdescendant = False |
|
2337 | 2339 | if lowestrev == nullrev: # Everybody is a descendant of nullid |
|
2338 | 2340 | isdescendant = True |
|
2339 | 2341 | elif n in descendants: |
|
2340 | 2342 | # n is already a descendant |
|
2341 | 2343 | isdescendant = True |
|
2342 | 2344 | # This check only needs to be done here because all the roots |
|
2343 | 2345 | # will start being marked is descendants before the loop. |
|
2344 | 2346 | if n in roots: |
|
2345 | 2347 | # If n was a root, check if it's a 'real' root. |
|
2346 | 2348 | p = tuple(self.parents(n)) |
|
2347 | 2349 | # If any of its parents are descendants, it's not a root. |
|
2348 | 2350 | if (p[0] in descendants) or (p[1] in descendants): |
|
2349 | 2351 | roots.remove(n) |
|
2350 | 2352 | else: |
|
2351 | 2353 | p = tuple(self.parents(n)) |
|
2352 | 2354 | # A node is a descendant if either of its parents are |
|
2353 | 2355 | # descendants. (We seeded the dependents list with the roots |
|
2354 | 2356 | # up there, remember?) |
|
2355 | 2357 | if (p[0] in descendants) or (p[1] in descendants): |
|
2356 | 2358 | descendants.add(n) |
|
2357 | 2359 | isdescendant = True |
|
2358 | 2360 | if isdescendant and ((ancestors is None) or (n in ancestors)): |
|
2359 | 2361 | # Only include nodes that are both descendants and ancestors. |
|
2360 | 2362 | orderedout.append(n) |
|
2361 | 2363 | if (ancestors is not None) and (n in heads): |
|
2362 | 2364 | # We're trying to figure out which heads are reachable |
|
2363 | 2365 | # from roots. |
|
2364 | 2366 | # Mark this head as having been reached |
|
2365 | 2367 | heads[n] = True |
|
2366 | 2368 | elif ancestors is None: |
|
2367 | 2369 | # Otherwise, we're trying to discover the heads. |
|
2368 | 2370 | # Assume this is a head because if it isn't, the next step |
|
2369 | 2371 | # will eventually remove it. |
|
2370 | 2372 | heads[n] = True |
|
2371 | 2373 | # But, obviously its parents aren't. |
|
2372 | 2374 | for p in self.parents(n): |
|
2373 | 2375 | heads.pop(p, None) |
|
2374 | 2376 | heads = [head for head, flag in heads.items() if flag] |
|
2375 | 2377 | roots = list(roots) |
|
2376 | 2378 | assert orderedout |
|
2377 | 2379 | assert roots |
|
2378 | 2380 | assert heads |
|
2379 | 2381 | return (orderedout, roots, heads) |
|
2380 | 2382 | |
|
2381 | 2383 | def headrevs(self, revs=None): |
|
2382 | 2384 | if revs is None: |
|
2383 | 2385 | try: |
|
2384 | 2386 | return self.index.headrevs() |
|
2385 | 2387 | except AttributeError: |
|
2386 | 2388 | return self._headrevs() |
|
2387 | 2389 | if rustdagop is not None and self.index.rust_ext_compat: |
|
2388 | 2390 | return rustdagop.headrevs(self.index, revs) |
|
2389 | 2391 | return dagop.headrevs(revs, self._uncheckedparentrevs) |
|
2390 | 2392 | |
|
2391 | 2393 | def headrevsdiff(self, start, stop): |
|
2392 | 2394 | try: |
|
2393 | 2395 | return self.index.headrevsdiff(start, stop) |
|
2394 | 2396 | except AttributeError: |
|
2395 | 2397 | return dagop.headrevsdiff(self._uncheckedparentrevs, start, stop) |
|
2396 | 2398 | |
|
2397 | 2399 | def computephases(self, roots): |
|
2398 | 2400 | return self.index.computephasesmapsets(roots) |
|
2399 | 2401 | |
|
2400 | 2402 | def _headrevs(self): |
|
2401 | 2403 | count = len(self) |
|
2402 | 2404 | if not count: |
|
2403 | 2405 | return [nullrev] |
|
2404 | 2406 | # we won't iter over filtered rev so nobody is a head at start |
|
2405 | 2407 | ishead = [0] * (count + 1) |
|
2406 | 2408 | index = self.index |
|
2407 | 2409 | for r in self: |
|
2408 | 2410 | ishead[r] = 1 # I may be an head |
|
2409 | 2411 | e = index[r] |
|
2410 | 2412 | ishead[e[5]] = ishead[e[6]] = 0 # my parent are not |
|
2411 | 2413 | return [r for r, val in enumerate(ishead) if val] |
|
2412 | 2414 | |
|
2413 | 2415 | def _head_node_ids(self): |
|
2414 | 2416 | try: |
|
2415 | 2417 | return self.index.head_node_ids() |
|
2416 | 2418 | except AttributeError: |
|
2417 | 2419 | return [self.node(r) for r in self.headrevs()] |
|
2418 | 2420 | |
|
2419 | 2421 | def heads(self, start=None, stop=None): |
|
2420 | 2422 | """return the list of all nodes that have no children |
|
2421 | 2423 | |
|
2422 | 2424 | if start is specified, only heads that are descendants of |
|
2423 | 2425 | start will be returned |
|
2424 | 2426 | if stop is specified, it will consider all the revs from stop |
|
2425 | 2427 | as if they had no children |
|
2426 | 2428 | """ |
|
2427 | 2429 | if start is None and stop is None: |
|
2428 | 2430 | if not len(self): |
|
2429 | 2431 | return [self.nullid] |
|
2430 | 2432 | return self._head_node_ids() |
|
2431 | 2433 | if start is None: |
|
2432 | 2434 | start = nullrev |
|
2433 | 2435 | else: |
|
2434 | 2436 | start = self.rev(start) |
|
2435 | 2437 | |
|
2436 | 2438 | stoprevs = {self.rev(n) for n in stop or []} |
|
2437 | 2439 | |
|
2438 | 2440 | revs = dagop.headrevssubset( |
|
2439 | 2441 | self.revs, self.parentrevs, startrev=start, stoprevs=stoprevs |
|
2440 | 2442 | ) |
|
2441 | 2443 | |
|
2442 | 2444 | return [self.node(rev) for rev in revs] |
|
2443 | 2445 | |
|
2444 | 2446 | def diffheads(self, start, stop): |
|
2445 | 2447 | """return the nodes that make up the difference between |
|
2446 | 2448 | heads of revs before `start` and heads of revs before `stop`""" |
|
2447 | 2449 | removed, added = self.headrevsdiff(start, stop) |
|
2448 | 2450 | return [self.node(r) for r in removed], [self.node(r) for r in added] |
|
2449 | 2451 | |
|
2450 | 2452 | def children(self, node): |
|
2451 | 2453 | """find the children of a given node""" |
|
2452 | 2454 | c = [] |
|
2453 | 2455 | p = self.rev(node) |
|
2454 | 2456 | for r in self.revs(start=p + 1): |
|
2455 | 2457 | prevs = [pr for pr in self.parentrevs(r) if pr != nullrev] |
|
2456 | 2458 | if prevs: |
|
2457 | 2459 | for pr in prevs: |
|
2458 | 2460 | if pr == p: |
|
2459 | 2461 | c.append(self.node(r)) |
|
2460 | 2462 | elif p == nullrev: |
|
2461 | 2463 | c.append(self.node(r)) |
|
2462 | 2464 | return c |
|
2463 | 2465 | |
|
2464 | 2466 | def commonancestorsheads(self, a, b): |
|
2465 | 2467 | """calculate all the heads of the common ancestors of nodes a and b""" |
|
2466 | 2468 | a, b = self.rev(a), self.rev(b) |
|
2467 | 2469 | ancs = self._commonancestorsheads(a, b) |
|
2468 | 2470 | return pycompat.maplist(self.node, ancs) |
|
2469 | 2471 | |
|
2470 | 2472 | def _commonancestorsheads(self, *revs): |
|
2471 | 2473 | """calculate all the heads of the common ancestors of revs""" |
|
2472 | 2474 | try: |
|
2473 | 2475 | ancs = self.index.commonancestorsheads(*revs) |
|
2474 | 2476 | except (AttributeError, OverflowError): # C implementation failed |
|
2475 | 2477 | ancs = ancestor.commonancestorsheads(self.parentrevs, *revs) |
|
2476 | 2478 | return ancs |
|
2477 | 2479 | |
|
2478 | 2480 | def isancestor(self, a, b): |
|
2479 | 2481 | """return True if node a is an ancestor of node b |
|
2480 | 2482 | |
|
2481 | 2483 | A revision is considered an ancestor of itself.""" |
|
2482 | 2484 | a, b = self.rev(a), self.rev(b) |
|
2483 | 2485 | return self.isancestorrev(a, b) |
|
2484 | 2486 | |
|
2485 | 2487 | def isancestorrev(self, a, b): |
|
2486 | 2488 | """return True if revision a is an ancestor of revision b |
|
2487 | 2489 | |
|
2488 | 2490 | A revision is considered an ancestor of itself. |
|
2489 | 2491 | |
|
2490 | 2492 | The implementation of this is trivial but the use of |
|
2491 | 2493 | reachableroots is not.""" |
|
2492 | 2494 | if a == nullrev: |
|
2493 | 2495 | return True |
|
2494 | 2496 | elif a == b: |
|
2495 | 2497 | return True |
|
2496 | 2498 | elif a > b: |
|
2497 | 2499 | return False |
|
2498 | 2500 | return bool(self.reachableroots(a, [b], [a], includepath=False)) |
|
2499 | 2501 | |
|
2500 | 2502 | def reachableroots(self, minroot, heads, roots, includepath=False): |
|
2501 | 2503 | """return (heads(::(<roots> and <roots>::<heads>))) |
|
2502 | 2504 | |
|
2503 | 2505 | If includepath is True, return (<roots>::<heads>).""" |
|
2504 | 2506 | try: |
|
2505 | 2507 | return self.index.reachableroots2( |
|
2506 | 2508 | minroot, heads, roots, includepath |
|
2507 | 2509 | ) |
|
2508 | 2510 | except AttributeError: |
|
2509 | 2511 | return dagop._reachablerootspure( |
|
2510 | 2512 | self.parentrevs, minroot, roots, heads, includepath |
|
2511 | 2513 | ) |
|
2512 | 2514 | |
|
2513 | 2515 | def ancestor(self, a, b): |
|
2514 | 2516 | """calculate the "best" common ancestor of nodes a and b""" |
|
2515 | 2517 | |
|
2516 | 2518 | a, b = self.rev(a), self.rev(b) |
|
2517 | 2519 | try: |
|
2518 | 2520 | ancs = self.index.ancestors(a, b) |
|
2519 | 2521 | except (AttributeError, OverflowError): |
|
2520 | 2522 | ancs = ancestor.ancestors(self.parentrevs, a, b) |
|
2521 | 2523 | if ancs: |
|
2522 | 2524 | # choose a consistent winner when there's a tie |
|
2523 | 2525 | return min(map(self.node, ancs)) |
|
2524 | 2526 | return self.nullid |
|
2525 | 2527 | |
|
2526 | 2528 | def _match(self, id): |
|
2527 | 2529 | if isinstance(id, int): |
|
2528 | 2530 | # rev |
|
2529 | 2531 | return self.node(id) |
|
2530 | 2532 | if len(id) == self.nodeconstants.nodelen: |
|
2531 | 2533 | # possibly a binary node |
|
2532 | 2534 | # odds of a binary node being all hex in ASCII are 1 in 10**25 |
|
2533 | 2535 | try: |
|
2534 | 2536 | node = id |
|
2535 | 2537 | self.rev(node) # quick search the index |
|
2536 | 2538 | return node |
|
2537 | 2539 | except error.LookupError: |
|
2538 | 2540 | pass # may be partial hex id |
|
2539 | 2541 | try: |
|
2540 | 2542 | # str(rev) |
|
2541 | 2543 | rev = int(id) |
|
2542 | 2544 | if b"%d" % rev != id: |
|
2543 | 2545 | raise ValueError |
|
2544 | 2546 | if rev < 0: |
|
2545 | 2547 | rev = len(self) + rev |
|
2546 | 2548 | if rev < 0 or rev >= len(self): |
|
2547 | 2549 | raise ValueError |
|
2548 | 2550 | return self.node(rev) |
|
2549 | 2551 | except (ValueError, OverflowError): |
|
2550 | 2552 | pass |
|
2551 | 2553 | if len(id) == 2 * self.nodeconstants.nodelen: |
|
2552 | 2554 | try: |
|
2553 | 2555 | # a full hex nodeid? |
|
2554 | 2556 | node = bin(id) |
|
2555 | 2557 | self.rev(node) |
|
2556 | 2558 | return node |
|
2557 | 2559 | except (binascii.Error, error.LookupError): |
|
2558 | 2560 | pass |
|
2559 | 2561 | |
|
2560 | 2562 | def _partialmatch(self, id): |
|
2561 | 2563 | # we don't care wdirfilenodeids as they should be always full hash |
|
2562 | 2564 | maybewdir = self.nodeconstants.wdirhex.startswith(id) |
|
2563 | 2565 | ambiguous = False |
|
2564 | 2566 | try: |
|
2565 | 2567 | partial = self.index.partialmatch(id) |
|
2566 | 2568 | if partial and self.hasnode(partial): |
|
2567 | 2569 | if maybewdir: |
|
2568 | 2570 | # single 'ff...' match in radix tree, ambiguous with wdir |
|
2569 | 2571 | ambiguous = True |
|
2570 | 2572 | else: |
|
2571 | 2573 | return partial |
|
2572 | 2574 | elif maybewdir: |
|
2573 | 2575 | # no 'ff...' match in radix tree, wdir identified |
|
2574 | 2576 | raise error.WdirUnsupported |
|
2575 | 2577 | else: |
|
2576 | 2578 | return None |
|
2577 | 2579 | except error.RevlogError: |
|
2578 | 2580 | # parsers.c radix tree lookup gave multiple matches |
|
2579 | 2581 | # fast path: for unfiltered changelog, radix tree is accurate |
|
2580 | 2582 | if not getattr(self, 'filteredrevs', None): |
|
2581 | 2583 | ambiguous = True |
|
2582 | 2584 | # fall through to slow path that filters hidden revisions |
|
2583 | 2585 | except (AttributeError, ValueError): |
|
2584 | 2586 | # we are pure python, or key is not hex |
|
2585 | 2587 | pass |
|
2586 | 2588 | if ambiguous: |
|
2587 | 2589 | raise error.AmbiguousPrefixLookupError( |
|
2588 | 2590 | id, self.display_id, _(b'ambiguous identifier') |
|
2589 | 2591 | ) |
|
2590 | 2592 | |
|
2591 | 2593 | if id in self._pcache: |
|
2592 | 2594 | return self._pcache[id] |
|
2593 | 2595 | |
|
2594 | 2596 | if len(id) <= 40: |
|
2595 | 2597 | # hex(node)[:...] |
|
2596 | 2598 | l = len(id) // 2 * 2 # grab an even number of digits |
|
2597 | 2599 | try: |
|
2598 | 2600 | # we're dropping the last digit, so let's check that it's hex, |
|
2599 | 2601 | # to avoid the expensive computation below if it's not |
|
2600 | 2602 | if len(id) % 2 > 0: |
|
2601 | 2603 | if not (id[-1] in hexdigits): |
|
2602 | 2604 | return None |
|
2603 | 2605 | prefix = bin(id[:l]) |
|
2604 | 2606 | except binascii.Error: |
|
2605 | 2607 | pass |
|
2606 | 2608 | else: |
|
2607 | 2609 | nl = [e[7] for e in self.index if e[7].startswith(prefix)] |
|
2608 | 2610 | nl = [ |
|
2609 | 2611 | n for n in nl if hex(n).startswith(id) and self.hasnode(n) |
|
2610 | 2612 | ] |
|
2611 | 2613 | if self.nodeconstants.nullhex.startswith(id): |
|
2612 | 2614 | nl.append(self.nullid) |
|
2613 | 2615 | if len(nl) > 0: |
|
2614 | 2616 | if len(nl) == 1 and not maybewdir: |
|
2615 | 2617 | self._pcache[id] = nl[0] |
|
2616 | 2618 | return nl[0] |
|
2617 | 2619 | raise error.AmbiguousPrefixLookupError( |
|
2618 | 2620 | id, self.display_id, _(b'ambiguous identifier') |
|
2619 | 2621 | ) |
|
2620 | 2622 | if maybewdir: |
|
2621 | 2623 | raise error.WdirUnsupported |
|
2622 | 2624 | return None |
|
2623 | 2625 | |
|
2624 | 2626 | def lookup(self, id): |
|
2625 | 2627 | """locate a node based on: |
|
2626 | 2628 | - revision number or str(revision number) |
|
2627 | 2629 | - nodeid or subset of hex nodeid |
|
2628 | 2630 | """ |
|
2629 | 2631 | n = self._match(id) |
|
2630 | 2632 | if n is not None: |
|
2631 | 2633 | return n |
|
2632 | 2634 | n = self._partialmatch(id) |
|
2633 | 2635 | if n: |
|
2634 | 2636 | return n |
|
2635 | 2637 | |
|
2636 | 2638 | raise error.LookupError(id, self.display_id, _(b'no match found')) |
|
2637 | 2639 | |
|
2638 | 2640 | def shortest(self, node, minlength=1): |
|
2639 | 2641 | """Find the shortest unambiguous prefix that matches node.""" |
|
2640 | 2642 | |
|
2641 | 2643 | def isvalid(prefix): |
|
2642 | 2644 | try: |
|
2643 | 2645 | matchednode = self._partialmatch(prefix) |
|
2644 | 2646 | except error.AmbiguousPrefixLookupError: |
|
2645 | 2647 | return False |
|
2646 | 2648 | except error.WdirUnsupported: |
|
2647 | 2649 | # single 'ff...' match |
|
2648 | 2650 | return True |
|
2649 | 2651 | if matchednode is None: |
|
2650 | 2652 | raise error.LookupError(node, self.display_id, _(b'no node')) |
|
2651 | 2653 | return True |
|
2652 | 2654 | |
|
2653 | 2655 | def maybewdir(prefix): |
|
2654 | 2656 | return all(c == b'f' for c in pycompat.iterbytestr(prefix)) |
|
2655 | 2657 | |
|
2656 | 2658 | hexnode = hex(node) |
|
2657 | 2659 | |
|
2658 | 2660 | def disambiguate(hexnode, minlength): |
|
2659 | 2661 | """Disambiguate against wdirid.""" |
|
2660 | 2662 | for length in range(minlength, len(hexnode) + 1): |
|
2661 | 2663 | prefix = hexnode[:length] |
|
2662 | 2664 | if not maybewdir(prefix): |
|
2663 | 2665 | return prefix |
|
2664 | 2666 | |
|
2665 | 2667 | if not getattr(self, 'filteredrevs', None): |
|
2666 | 2668 | try: |
|
2667 | 2669 | length = max(self.index.shortest(node), minlength) |
|
2668 | 2670 | return disambiguate(hexnode, length) |
|
2669 | 2671 | except error.RevlogError: |
|
2670 | 2672 | if node != self.nodeconstants.wdirid: |
|
2671 | 2673 | raise error.LookupError( |
|
2672 | 2674 | node, self.display_id, _(b'no node') |
|
2673 | 2675 | ) |
|
2674 | 2676 | except AttributeError: |
|
2675 | 2677 | # Fall through to pure code |
|
2676 | 2678 | pass |
|
2677 | 2679 | |
|
2678 | 2680 | if node == self.nodeconstants.wdirid: |
|
2679 | 2681 | for length in range(minlength, len(hexnode) + 1): |
|
2680 | 2682 | prefix = hexnode[:length] |
|
2681 | 2683 | if isvalid(prefix): |
|
2682 | 2684 | return prefix |
|
2683 | 2685 | |
|
2684 | 2686 | for length in range(minlength, len(hexnode) + 1): |
|
2685 | 2687 | prefix = hexnode[:length] |
|
2686 | 2688 | if isvalid(prefix): |
|
2687 | 2689 | return disambiguate(hexnode, length) |
|
2688 | 2690 | |
|
2689 | 2691 | def cmp(self, node, text): |
|
2690 | 2692 | """compare text with a given file revision |
|
2691 | 2693 | |
|
2692 | 2694 | returns True if text is different than what is stored. |
|
2693 | 2695 | """ |
|
2694 | 2696 | p1, p2 = self.parents(node) |
|
2695 | 2697 | return storageutil.hashrevisionsha1(text, p1, p2) != node |
|
2696 | 2698 | |
|
2697 | 2699 | def deltaparent(self, rev): |
|
2698 | 2700 | """return deltaparent of the given revision""" |
|
2699 | 2701 | base = self.index[rev][3] |
|
2700 | 2702 | if base == rev: |
|
2701 | 2703 | return nullrev |
|
2702 | 2704 | elif self.delta_config.general_delta: |
|
2703 | 2705 | return base |
|
2704 | 2706 | else: |
|
2705 | 2707 | return rev - 1 |
|
2706 | 2708 | |
|
2707 | 2709 | def issnapshot(self, rev): |
|
2708 | 2710 | """tells whether rev is a snapshot""" |
|
2709 | 2711 | ret = self._inner.issnapshot(rev) |
|
2710 | 2712 | self.issnapshot = self._inner.issnapshot |
|
2711 | 2713 | return ret |
|
2712 | 2714 | |
|
2713 | 2715 | def snapshotdepth(self, rev): |
|
2714 | 2716 | """number of snapshot in the chain before this one""" |
|
2715 | 2717 | if not self.issnapshot(rev): |
|
2716 | 2718 | raise error.ProgrammingError(b'revision %d not a snapshot') |
|
2717 | 2719 | return len(self._inner._deltachain(rev)[0]) - 1 |
|
2718 | 2720 | |
|
2719 | 2721 | def revdiff(self, rev1, rev2): |
|
2720 | 2722 | """return or calculate a delta between two revisions |
|
2721 | 2723 | |
|
2722 | 2724 | The delta calculated is in binary form and is intended to be written to |
|
2723 | 2725 | revlog data directly. So this function needs raw revision data. |
|
2724 | 2726 | """ |
|
2725 | 2727 | if rev1 != nullrev and self.deltaparent(rev2) == rev1: |
|
2726 | 2728 | return bytes(self._inner._chunk(rev2)) |
|
2727 | 2729 | |
|
2728 | 2730 | return mdiff.textdiff(self.rawdata(rev1), self.rawdata(rev2)) |
|
2729 | 2731 | |
|
2730 | 2732 | def revision(self, nodeorrev): |
|
2731 | 2733 | """return an uncompressed revision of a given node or revision |
|
2732 | 2734 | number. |
|
2733 | 2735 | """ |
|
2734 | 2736 | return self._revisiondata(nodeorrev) |
|
2735 | 2737 | |
|
2736 | 2738 | def sidedata(self, nodeorrev): |
|
2737 | 2739 | """a map of extra data related to the changeset but not part of the hash |
|
2738 | 2740 | |
|
2739 | 2741 | This function currently return a dictionary. However, more advanced |
|
2740 | 2742 | mapping object will likely be used in the future for a more |
|
2741 | 2743 | efficient/lazy code. |
|
2742 | 2744 | """ |
|
2743 | 2745 | # deal with <nodeorrev> argument type |
|
2744 | 2746 | if isinstance(nodeorrev, int): |
|
2745 | 2747 | rev = nodeorrev |
|
2746 | 2748 | else: |
|
2747 | 2749 | rev = self.rev(nodeorrev) |
|
2748 | 2750 | return self._sidedata(rev) |
|
2749 | 2751 | |
|
2750 | 2752 | def _rawtext(self, node, rev): |
|
2751 | 2753 | """return the possibly unvalidated rawtext for a revision |
|
2752 | 2754 | |
|
2753 | 2755 | returns (rev, rawtext, validated) |
|
2754 | 2756 | """ |
|
2755 | 2757 | # Check if we have the entry in cache |
|
2756 | 2758 | # The cache entry looks like (node, rev, rawtext) |
|
2757 | 2759 | if self._inner._revisioncache: |
|
2758 | 2760 | if self._inner._revisioncache[0] == node: |
|
2759 | 2761 | return (rev, self._inner._revisioncache[2], True) |
|
2760 | 2762 | |
|
2761 | 2763 | if rev is None: |
|
2762 | 2764 | rev = self.rev(node) |
|
2763 | 2765 | |
|
2764 | 2766 | return self._inner.raw_text(node, rev) |
|
2765 | 2767 | |
|
2766 | 2768 | def _revisiondata(self, nodeorrev, raw=False): |
|
2767 | 2769 | # deal with <nodeorrev> argument type |
|
2768 | 2770 | if isinstance(nodeorrev, int): |
|
2769 | 2771 | rev = nodeorrev |
|
2770 | 2772 | node = self.node(rev) |
|
2771 | 2773 | else: |
|
2772 | 2774 | node = nodeorrev |
|
2773 | 2775 | rev = None |
|
2774 | 2776 | |
|
2775 | 2777 | # fast path the special `nullid` rev |
|
2776 | 2778 | if node == self.nullid: |
|
2777 | 2779 | return b"" |
|
2778 | 2780 | |
|
2779 | 2781 | # ``rawtext`` is the text as stored inside the revlog. Might be the |
|
2780 | 2782 | # revision or might need to be processed to retrieve the revision. |
|
2781 | 2783 | rev, rawtext, validated = self._rawtext(node, rev) |
|
2782 | 2784 | |
|
2783 | 2785 | if raw and validated: |
|
2784 | 2786 | # if we don't want to process the raw text and that raw |
|
2785 | 2787 | # text is cached, we can exit early. |
|
2786 | 2788 | return rawtext |
|
2787 | 2789 | if rev is None: |
|
2788 | 2790 | rev = self.rev(node) |
|
2789 | 2791 | # the revlog's flag for this revision |
|
2790 | 2792 | # (usually alter its state or content) |
|
2791 | 2793 | flags = self.flags(rev) |
|
2792 | 2794 | |
|
2793 | 2795 | if validated and flags == REVIDX_DEFAULT_FLAGS: |
|
2794 | 2796 | # no extra flags set, no flag processor runs, text = rawtext |
|
2795 | 2797 | return rawtext |
|
2796 | 2798 | |
|
2797 | 2799 | if raw: |
|
2798 | 2800 | validatehash = flagutil.processflagsraw(self, rawtext, flags) |
|
2799 | 2801 | text = rawtext |
|
2800 | 2802 | else: |
|
2801 | 2803 | r = flagutil.processflagsread(self, rawtext, flags) |
|
2802 | 2804 | text, validatehash = r |
|
2803 | 2805 | if validatehash: |
|
2804 | 2806 | self.checkhash(text, node, rev=rev) |
|
2805 | 2807 | if not validated: |
|
2806 | 2808 | self._inner._revisioncache = (node, rev, rawtext) |
|
2807 | 2809 | |
|
2808 | 2810 | return text |
|
2809 | 2811 | |
|
2810 | 2812 | def _sidedata(self, rev): |
|
2811 | 2813 | """Return the sidedata for a given revision number.""" |
|
2812 | 2814 | sidedata_end = None |
|
2813 | 2815 | if self._docket is not None: |
|
2814 | 2816 | sidedata_end = self._docket.sidedata_end |
|
2815 | 2817 | return self._inner.sidedata(rev, sidedata_end) |
|
2816 | 2818 | |
|
2817 | 2819 | def rawdata(self, nodeorrev): |
|
2818 | 2820 | """return an uncompressed raw data of a given node or revision number.""" |
|
2819 | 2821 | return self._revisiondata(nodeorrev, raw=True) |
|
2820 | 2822 | |
|
2821 | 2823 | def hash(self, text, p1, p2): |
|
2822 | 2824 | """Compute a node hash. |
|
2823 | 2825 | |
|
2824 | 2826 | Available as a function so that subclasses can replace the hash |
|
2825 | 2827 | as needed. |
|
2826 | 2828 | """ |
|
2827 | 2829 | return storageutil.hashrevisionsha1(text, p1, p2) |
|
2828 | 2830 | |
|
2829 | 2831 | def checkhash(self, text, node, p1=None, p2=None, rev=None): |
|
2830 | 2832 | """Check node hash integrity. |
|
2831 | 2833 | |
|
2832 | 2834 | Available as a function so that subclasses can extend hash mismatch |
|
2833 | 2835 | behaviors as needed. |
|
2834 | 2836 | """ |
|
2835 | 2837 | try: |
|
2836 | 2838 | if p1 is None and p2 is None: |
|
2837 | 2839 | p1, p2 = self.parents(node) |
|
2838 | 2840 | if node != self.hash(text, p1, p2): |
|
2839 | 2841 | # Clear the revision cache on hash failure. The revision cache |
|
2840 | 2842 | # only stores the raw revision and clearing the cache does have |
|
2841 | 2843 | # the side-effect that we won't have a cache hit when the raw |
|
2842 | 2844 | # revision data is accessed. But this case should be rare and |
|
2843 | 2845 | # it is extra work to teach the cache about the hash |
|
2844 | 2846 | # verification state. |
|
2845 | 2847 | if ( |
|
2846 | 2848 | self._inner._revisioncache |
|
2847 | 2849 | and self._inner._revisioncache[0] == node |
|
2848 | 2850 | ): |
|
2849 | 2851 | self._inner._revisioncache = None |
|
2850 | 2852 | |
|
2851 | 2853 | revornode = rev |
|
2852 | 2854 | if revornode is None: |
|
2853 | 2855 | revornode = templatefilters.short(hex(node)) |
|
2854 | 2856 | raise error.RevlogError( |
|
2855 | 2857 | _(b"integrity check failed on %s:%s") |
|
2856 | 2858 | % (self.display_id, pycompat.bytestr(revornode)) |
|
2857 | 2859 | ) |
|
2858 | 2860 | except error.RevlogError: |
|
2859 | 2861 | if self.feature_config.censorable and storageutil.iscensoredtext( |
|
2860 | 2862 | text |
|
2861 | 2863 | ): |
|
2862 | 2864 | raise error.CensoredNodeError(self.display_id, node, text) |
|
2863 | 2865 | raise |
|
2864 | 2866 | |
|
2865 | 2867 | @property |
|
2866 | 2868 | def _split_index_file(self): |
|
2867 | 2869 | """the path where to expect the index of an ongoing splitting operation |
|
2868 | 2870 | |
|
2869 | 2871 | The file will only exist if a splitting operation is in progress, but |
|
2870 | 2872 | it is always expected at the same location.""" |
|
2871 | 2873 | parts = self.radix.split(b'/') |
|
2872 | 2874 | if len(parts) > 1: |
|
2873 | 2875 | # adds a '-s' prefix to the ``data/` or `meta/` base |
|
2874 | 2876 | head = parts[0] + b'-s' |
|
2875 | 2877 | mids = parts[1:-1] |
|
2876 | 2878 | tail = parts[-1] + b'.i' |
|
2877 | 2879 | pieces = [head] + mids + [tail] |
|
2878 | 2880 | return b'/'.join(pieces) |
|
2879 | 2881 | else: |
|
2880 | 2882 | # the revlog is stored at the root of the store (changelog or |
|
2881 | 2883 | # manifest), no risk of collision. |
|
2882 | 2884 | return self.radix + b'.i.s' |
|
2883 | 2885 | |
|
2884 | 2886 | def _enforceinlinesize(self, tr): |
|
2885 | 2887 | """Check if the revlog is too big for inline and convert if so. |
|
2886 | 2888 | |
|
2887 | 2889 | This should be called after revisions are added to the revlog. If the |
|
2888 | 2890 | revlog has grown too large to be an inline revlog, it will convert it |
|
2889 | 2891 | to use multiple index and data files. |
|
2890 | 2892 | """ |
|
2891 | 2893 | tiprev = len(self) - 1 |
|
2892 | 2894 | total_size = self.start(tiprev) + self.length(tiprev) |
|
2893 | 2895 | if not self._inline or (self._may_inline and total_size < _maxinline): |
|
2894 | 2896 | return |
|
2895 | 2897 | |
|
2896 | 2898 | if self._docket is not None: |
|
2897 | 2899 | msg = b"inline revlog should not have a docket" |
|
2898 | 2900 | raise error.ProgrammingError(msg) |
|
2899 | 2901 | |
|
2900 | 2902 | # In the common case, we enforce inline size because the revlog has |
|
2901 | 2903 | # been appened too. And in such case, it must have an initial offset |
|
2902 | 2904 | # recorded in the transaction. |
|
2903 | 2905 | troffset = tr.findoffset(self._inner.canonical_index_file) |
|
2904 | 2906 | pre_touched = troffset is not None |
|
2905 | 2907 | if not pre_touched and self.target[0] != KIND_CHANGELOG: |
|
2906 | 2908 | raise error.RevlogError( |
|
2907 | 2909 | _(b"%s not found in the transaction") % self._indexfile |
|
2908 | 2910 | ) |
|
2909 | 2911 | |
|
2910 | 2912 | tr.addbackup(self._inner.canonical_index_file, for_offset=pre_touched) |
|
2911 | 2913 | tr.add(self._datafile, 0) |
|
2912 | 2914 | |
|
2913 | 2915 | new_index_file_path = None |
|
2914 | 2916 | old_index_file_path = self._indexfile |
|
2915 | 2917 | new_index_file_path = self._split_index_file |
|
2916 | 2918 | opener = self.opener |
|
2917 | 2919 | weak_self = weakref.ref(self) |
|
2918 | 2920 | |
|
2919 | 2921 | # the "split" index replace the real index when the transaction is |
|
2920 | 2922 | # finalized |
|
2921 | 2923 | def finalize_callback(tr): |
|
2922 | 2924 | opener.rename( |
|
2923 | 2925 | new_index_file_path, |
|
2924 | 2926 | old_index_file_path, |
|
2925 | 2927 | checkambig=True, |
|
2926 | 2928 | ) |
|
2927 | 2929 | maybe_self = weak_self() |
|
2928 | 2930 | if maybe_self is not None: |
|
2929 | 2931 | maybe_self._indexfile = old_index_file_path |
|
2930 | 2932 | maybe_self._inner.index_file = maybe_self._indexfile |
|
2931 | 2933 | |
|
2932 | 2934 | def abort_callback(tr): |
|
2933 | 2935 | maybe_self = weak_self() |
|
2934 | 2936 | if maybe_self is not None: |
|
2935 | 2937 | maybe_self._indexfile = old_index_file_path |
|
2936 | 2938 | maybe_self._inner.inline = True |
|
2937 | 2939 | maybe_self._inner.index_file = old_index_file_path |
|
2938 | 2940 | |
|
2939 | 2941 | tr.registertmp(new_index_file_path) |
|
2940 | 2942 | # we use 001 here to make this this happens after the finalisation of |
|
2941 | 2943 | # pending changelog write (using 000). Otherwise the two finalizer |
|
2942 | 2944 | # would step over each other and delete the changelog.i file. |
|
2943 | 2945 | if self.target[1] is not None: |
|
2944 | 2946 | callback_id = b'001-revlog-split-%d-%s' % self.target |
|
2945 | 2947 | else: |
|
2946 | 2948 | callback_id = b'001-revlog-split-%d' % self.target[0] |
|
2947 | 2949 | tr.addfinalize(callback_id, finalize_callback) |
|
2948 | 2950 | tr.addabort(callback_id, abort_callback) |
|
2949 | 2951 | |
|
2950 | 2952 | self._format_flags &= ~FLAG_INLINE_DATA |
|
2951 | 2953 | self._inner.split_inline( |
|
2952 | 2954 | tr, |
|
2953 | 2955 | self._format_flags | self._format_version, |
|
2954 | 2956 | new_index_file_path=new_index_file_path, |
|
2955 | 2957 | ) |
|
2956 | 2958 | |
|
2957 | 2959 | self._inline = False |
|
2958 | 2960 | if new_index_file_path is not None: |
|
2959 | 2961 | self._indexfile = new_index_file_path |
|
2960 | 2962 | |
|
2961 | 2963 | nodemaputil.setup_persistent_nodemap(tr, self) |
|
2962 | 2964 | |
|
2963 | 2965 | def _nodeduplicatecallback(self, transaction, node): |
|
2964 | 2966 | """called when trying to add a node already stored.""" |
|
2965 | 2967 | |
|
2966 | 2968 | @contextlib.contextmanager |
|
2967 | 2969 | def reading(self): |
|
2968 | 2970 | with self._inner.reading(): |
|
2969 | 2971 | yield |
|
2970 | 2972 | |
|
2971 | 2973 | @contextlib.contextmanager |
|
2972 | 2974 | def _writing(self, transaction): |
|
2973 | 2975 | if self._trypending: |
|
2974 | 2976 | msg = b'try to write in a `trypending` revlog: %s' |
|
2975 | 2977 | msg %= self.display_id |
|
2976 | 2978 | raise error.ProgrammingError(msg) |
|
2977 | 2979 | if self._inner.is_writing: |
|
2978 | 2980 | yield |
|
2979 | 2981 | else: |
|
2980 | 2982 | data_end = None |
|
2981 | 2983 | sidedata_end = None |
|
2982 | 2984 | if self._docket is not None: |
|
2983 | 2985 | data_end = self._docket.data_end |
|
2984 | 2986 | sidedata_end = self._docket.sidedata_end |
|
2985 | 2987 | with self._inner.writing( |
|
2986 | 2988 | transaction, |
|
2987 | 2989 | data_end=data_end, |
|
2988 | 2990 | sidedata_end=sidedata_end, |
|
2989 | 2991 | ): |
|
2990 | 2992 | yield |
|
2991 | 2993 | if self._docket is not None: |
|
2992 | 2994 | self._write_docket(transaction) |
|
2993 | 2995 | |
|
2994 | 2996 | @property |
|
2995 | 2997 | def is_delaying(self): |
|
2996 | 2998 | return self._inner.is_delaying |
|
2997 | 2999 | |
|
2998 | 3000 | def _write_docket(self, transaction): |
|
2999 | 3001 | """write the current docket on disk |
|
3000 | 3002 | |
|
3001 | 3003 | Exist as a method to help changelog to implement transaction logic |
|
3002 | 3004 | |
|
3003 | 3005 | We could also imagine using the same transaction logic for all revlog |
|
3004 | 3006 | since docket are cheap.""" |
|
3005 | 3007 | self._docket.write(transaction) |
|
3006 | 3008 | |
|
3007 | 3009 | def addrevision( |
|
3008 | 3010 | self, |
|
3009 | 3011 | text, |
|
3010 | 3012 | transaction, |
|
3011 | 3013 | link, |
|
3012 | 3014 | p1, |
|
3013 | 3015 | p2, |
|
3014 | 3016 | cachedelta=None, |
|
3015 | 3017 | node=None, |
|
3016 | 3018 | flags=REVIDX_DEFAULT_FLAGS, |
|
3017 | 3019 | deltacomputer=None, |
|
3018 | 3020 | sidedata=None, |
|
3019 | 3021 | ): |
|
3020 | 3022 | """add a revision to the log |
|
3021 | 3023 | |
|
3022 | 3024 | text - the revision data to add |
|
3023 | 3025 | transaction - the transaction object used for rollback |
|
3024 | 3026 | link - the linkrev data to add |
|
3025 | 3027 | p1, p2 - the parent nodeids of the revision |
|
3026 | 3028 | cachedelta - an optional precomputed delta |
|
3027 | 3029 | node - nodeid of revision; typically node is not specified, and it is |
|
3028 | 3030 | computed by default as hash(text, p1, p2), however subclasses might |
|
3029 | 3031 | use different hashing method (and override checkhash() in such case) |
|
3030 | 3032 | flags - the known flags to set on the revision |
|
3031 | 3033 | deltacomputer - an optional deltacomputer instance shared between |
|
3032 | 3034 | multiple calls |
|
3033 | 3035 | """ |
|
3034 | 3036 | if link == nullrev: |
|
3035 | 3037 | raise error.RevlogError( |
|
3036 | 3038 | _(b"attempted to add linkrev -1 to %s") % self.display_id |
|
3037 | 3039 | ) |
|
3038 | 3040 | |
|
3039 | 3041 | if sidedata is None: |
|
3040 | 3042 | sidedata = {} |
|
3041 | 3043 | elif sidedata and not self.feature_config.has_side_data: |
|
3042 | 3044 | raise error.ProgrammingError( |
|
3043 | 3045 | _(b"trying to add sidedata to a revlog who don't support them") |
|
3044 | 3046 | ) |
|
3045 | 3047 | |
|
3046 | 3048 | if flags: |
|
3047 | 3049 | node = node or self.hash(text, p1, p2) |
|
3048 | 3050 | |
|
3049 | 3051 | rawtext, validatehash = flagutil.processflagswrite(self, text, flags) |
|
3050 | 3052 | |
|
3051 | 3053 | # If the flag processor modifies the revision data, ignore any provided |
|
3052 | 3054 | # cachedelta. |
|
3053 | 3055 | if rawtext != text: |
|
3054 | 3056 | cachedelta = None |
|
3055 | 3057 | |
|
3056 | 3058 | if len(rawtext) > _maxentrysize: |
|
3057 | 3059 | raise error.RevlogError( |
|
3058 | 3060 | _( |
|
3059 | 3061 | b"%s: size of %d bytes exceeds maximum revlog storage of 2GiB" |
|
3060 | 3062 | ) |
|
3061 | 3063 | % (self.display_id, len(rawtext)) |
|
3062 | 3064 | ) |
|
3063 | 3065 | |
|
3064 | 3066 | node = node or self.hash(rawtext, p1, p2) |
|
3065 | 3067 | rev = self.index.get_rev(node) |
|
3066 | 3068 | if rev is not None: |
|
3067 | 3069 | return rev |
|
3068 | 3070 | |
|
3069 | 3071 | if validatehash: |
|
3070 | 3072 | self.checkhash(rawtext, node, p1=p1, p2=p2) |
|
3071 | 3073 | |
|
3072 | 3074 | return self.addrawrevision( |
|
3073 | 3075 | rawtext, |
|
3074 | 3076 | transaction, |
|
3075 | 3077 | link, |
|
3076 | 3078 | p1, |
|
3077 | 3079 | p2, |
|
3078 | 3080 | node, |
|
3079 | 3081 | flags, |
|
3080 | 3082 | cachedelta=cachedelta, |
|
3081 | 3083 | deltacomputer=deltacomputer, |
|
3082 | 3084 | sidedata=sidedata, |
|
3083 | 3085 | ) |
|
3084 | 3086 | |
|
3085 | 3087 | def addrawrevision( |
|
3086 | 3088 | self, |
|
3087 | 3089 | rawtext, |
|
3088 | 3090 | transaction, |
|
3089 | 3091 | link, |
|
3090 | 3092 | p1, |
|
3091 | 3093 | p2, |
|
3092 | 3094 | node, |
|
3093 | 3095 | flags, |
|
3094 | 3096 | cachedelta=None, |
|
3095 | 3097 | deltacomputer=None, |
|
3096 | 3098 | sidedata=None, |
|
3097 | 3099 | ): |
|
3098 | 3100 | """add a raw revision with known flags, node and parents |
|
3099 | 3101 | useful when reusing a revision not stored in this revlog (ex: received |
|
3100 | 3102 | over wire, or read from an external bundle). |
|
3101 | 3103 | """ |
|
3102 | 3104 | with self._writing(transaction): |
|
3103 | 3105 | return self._addrevision( |
|
3104 | 3106 | node, |
|
3105 | 3107 | rawtext, |
|
3106 | 3108 | transaction, |
|
3107 | 3109 | link, |
|
3108 | 3110 | p1, |
|
3109 | 3111 | p2, |
|
3110 | 3112 | flags, |
|
3111 | 3113 | cachedelta, |
|
3112 | 3114 | deltacomputer=deltacomputer, |
|
3113 | 3115 | sidedata=sidedata, |
|
3114 | 3116 | ) |
|
3115 | 3117 | |
|
3116 | 3118 | def compress(self, data: bytes) -> Tuple[bytes, bytes]: |
|
3117 | 3119 | return self._inner.compress(data) |
|
3118 | 3120 | |
|
3119 | 3121 | def decompress(self, data): |
|
3120 | 3122 | return self._inner.decompress(data) |
|
3121 | 3123 | |
|
3122 | 3124 | def _addrevision( |
|
3123 | 3125 | self, |
|
3124 | 3126 | node, |
|
3125 | 3127 | rawtext, |
|
3126 | 3128 | transaction, |
|
3127 | 3129 | link, |
|
3128 | 3130 | p1, |
|
3129 | 3131 | p2, |
|
3130 | 3132 | flags, |
|
3131 | 3133 | cachedelta, |
|
3132 | 3134 | alwayscache=False, |
|
3133 | 3135 | deltacomputer=None, |
|
3134 | 3136 | sidedata=None, |
|
3135 | 3137 | ): |
|
3136 | 3138 | """internal function to add revisions to the log |
|
3137 | 3139 | |
|
3138 | 3140 | see addrevision for argument descriptions. |
|
3139 | 3141 | |
|
3140 | 3142 | note: "addrevision" takes non-raw text, "_addrevision" takes raw text. |
|
3141 | 3143 | |
|
3142 | 3144 | if "deltacomputer" is not provided or None, a defaultdeltacomputer will |
|
3143 | 3145 | be used. |
|
3144 | 3146 | |
|
3145 | 3147 | invariants: |
|
3146 | 3148 | - rawtext is optional (can be None); if not set, cachedelta must be set. |
|
3147 | 3149 | if both are set, they must correspond to each other. |
|
3148 | 3150 | """ |
|
3149 | 3151 | if node == self.nullid: |
|
3150 | 3152 | raise error.RevlogError( |
|
3151 | 3153 | _(b"%s: attempt to add null revision") % self.display_id |
|
3152 | 3154 | ) |
|
3153 | 3155 | if ( |
|
3154 | 3156 | node == self.nodeconstants.wdirid |
|
3155 | 3157 | or node in self.nodeconstants.wdirfilenodeids |
|
3156 | 3158 | ): |
|
3157 | 3159 | raise error.RevlogError( |
|
3158 | 3160 | _(b"%s: attempt to add wdir revision") % self.display_id |
|
3159 | 3161 | ) |
|
3160 | 3162 | if self._inner._writinghandles is None: |
|
3161 | 3163 | msg = b'adding revision outside `revlog._writing` context' |
|
3162 | 3164 | raise error.ProgrammingError(msg) |
|
3163 | 3165 | |
|
3164 | 3166 | btext = [rawtext] |
|
3165 | 3167 | |
|
3166 | 3168 | curr = len(self) |
|
3167 | 3169 | prev = curr - 1 |
|
3168 | 3170 | |
|
3169 | 3171 | offset = self._get_data_offset(prev) |
|
3170 | 3172 | |
|
3171 | 3173 | if self._concurrencychecker: |
|
3172 | 3174 | ifh, dfh, sdfh = self._inner._writinghandles |
|
3173 | 3175 | # XXX no checking for the sidedata file |
|
3174 | 3176 | if self._inline: |
|
3175 | 3177 | # offset is "as if" it were in the .d file, so we need to add on |
|
3176 | 3178 | # the size of the entry metadata. |
|
3177 | 3179 | self._concurrencychecker( |
|
3178 | 3180 | ifh, self._indexfile, offset + curr * self.index.entry_size |
|
3179 | 3181 | ) |
|
3180 | 3182 | else: |
|
3181 | 3183 | # Entries in the .i are a consistent size. |
|
3182 | 3184 | self._concurrencychecker( |
|
3183 | 3185 | ifh, self._indexfile, curr * self.index.entry_size |
|
3184 | 3186 | ) |
|
3185 | 3187 | self._concurrencychecker(dfh, self._datafile, offset) |
|
3186 | 3188 | |
|
3187 | 3189 | p1r, p2r = self.rev(p1), self.rev(p2) |
|
3188 | 3190 | |
|
3189 | 3191 | # full versions are inserted when the needed deltas |
|
3190 | 3192 | # become comparable to the uncompressed text |
|
3191 | 3193 | if rawtext is None: |
|
3192 | 3194 | # need rawtext size, before changed by flag processors, which is |
|
3193 | 3195 | # the non-raw size. use revlog explicitly to avoid filelog's extra |
|
3194 | 3196 | # logic that might remove metadata size. |
|
3195 | 3197 | textlen = mdiff.patchedsize( |
|
3196 | 3198 | revlog.size(self, cachedelta[0]), cachedelta[1] |
|
3197 | 3199 | ) |
|
3198 | 3200 | else: |
|
3199 | 3201 | textlen = len(rawtext) |
|
3200 | 3202 | |
|
3201 | 3203 | if deltacomputer is None: |
|
3202 | 3204 | write_debug = None |
|
3203 | 3205 | if self.delta_config.debug_delta: |
|
3204 | 3206 | write_debug = transaction._report |
|
3205 | 3207 | deltacomputer = deltautil.deltacomputer( |
|
3206 | 3208 | self, write_debug=write_debug |
|
3207 | 3209 | ) |
|
3208 | 3210 | |
|
3209 | 3211 | if cachedelta is not None and len(cachedelta) == 2: |
|
3210 | 3212 | # If the cached delta has no information about how it should be |
|
3211 | 3213 | # reused, add the default reuse instruction according to the |
|
3212 | 3214 | # revlog's configuration. |
|
3213 | 3215 | if ( |
|
3214 | 3216 | self.delta_config.general_delta |
|
3215 | 3217 | and self.delta_config.lazy_delta_base |
|
3216 | 3218 | ): |
|
3217 | 3219 | delta_base_reuse = DELTA_BASE_REUSE_TRY |
|
3218 | 3220 | else: |
|
3219 | 3221 | delta_base_reuse = DELTA_BASE_REUSE_NO |
|
3220 | 3222 | cachedelta = (cachedelta[0], cachedelta[1], delta_base_reuse) |
|
3221 | 3223 | |
|
3222 | 3224 | revinfo = revlogutils.revisioninfo( |
|
3223 | 3225 | node, |
|
3224 | 3226 | p1, |
|
3225 | 3227 | p2, |
|
3226 | 3228 | btext, |
|
3227 | 3229 | textlen, |
|
3228 | 3230 | cachedelta, |
|
3229 | 3231 | flags, |
|
3230 | 3232 | ) |
|
3231 | 3233 | |
|
3232 | 3234 | deltainfo = deltacomputer.finddeltainfo(revinfo) |
|
3233 | 3235 | |
|
3234 | 3236 | compression_mode = COMP_MODE_INLINE |
|
3235 | 3237 | if self._docket is not None: |
|
3236 | 3238 | default_comp = self._docket.default_compression_header |
|
3237 | 3239 | r = deltautil.delta_compression(default_comp, deltainfo) |
|
3238 | 3240 | compression_mode, deltainfo = r |
|
3239 | 3241 | |
|
3240 | 3242 | sidedata_compression_mode = COMP_MODE_INLINE |
|
3241 | 3243 | if sidedata and self.feature_config.has_side_data: |
|
3242 | 3244 | sidedata_compression_mode = COMP_MODE_PLAIN |
|
3243 | 3245 | serialized_sidedata = sidedatautil.serialize_sidedata(sidedata) |
|
3244 | 3246 | sidedata_offset = self._docket.sidedata_end |
|
3245 | 3247 | h, comp_sidedata = self._inner.compress(serialized_sidedata) |
|
3246 | 3248 | if ( |
|
3247 | 3249 | h != b'u' |
|
3248 | 3250 | and comp_sidedata[0:1] != b'\0' |
|
3249 | 3251 | and len(comp_sidedata) < len(serialized_sidedata) |
|
3250 | 3252 | ): |
|
3251 | 3253 | assert not h |
|
3252 | 3254 | if ( |
|
3253 | 3255 | comp_sidedata[0:1] |
|
3254 | 3256 | == self._docket.default_compression_header |
|
3255 | 3257 | ): |
|
3256 | 3258 | sidedata_compression_mode = COMP_MODE_DEFAULT |
|
3257 | 3259 | serialized_sidedata = comp_sidedata |
|
3258 | 3260 | else: |
|
3259 | 3261 | sidedata_compression_mode = COMP_MODE_INLINE |
|
3260 | 3262 | serialized_sidedata = comp_sidedata |
|
3261 | 3263 | else: |
|
3262 | 3264 | serialized_sidedata = b"" |
|
3263 | 3265 | # Don't store the offset if the sidedata is empty, that way |
|
3264 | 3266 | # we can easily detect empty sidedata and they will be no different |
|
3265 | 3267 | # than ones we manually add. |
|
3266 | 3268 | sidedata_offset = 0 |
|
3267 | 3269 | |
|
3268 | 3270 | rank = RANK_UNKNOWN |
|
3269 | 3271 | if self.feature_config.compute_rank: |
|
3270 | 3272 | if (p1r, p2r) == (nullrev, nullrev): |
|
3271 | 3273 | rank = 1 |
|
3272 | 3274 | elif p1r != nullrev and p2r == nullrev: |
|
3273 | 3275 | rank = 1 + self.fast_rank(p1r) |
|
3274 | 3276 | elif p1r == nullrev and p2r != nullrev: |
|
3275 | 3277 | rank = 1 + self.fast_rank(p2r) |
|
3276 | 3278 | else: # merge node |
|
3277 | 3279 | if rustdagop is not None and self.index.rust_ext_compat: |
|
3278 | 3280 | rank = rustdagop.rank(self.index, p1r, p2r) |
|
3279 | 3281 | else: |
|
3280 | 3282 | pmin, pmax = sorted((p1r, p2r)) |
|
3281 | 3283 | rank = 1 + self.fast_rank(pmax) |
|
3282 | 3284 | rank += sum(1 for _ in self.findmissingrevs([pmax], [pmin])) |
|
3283 | 3285 | |
|
3284 | 3286 | e = revlogutils.entry( |
|
3285 | 3287 | flags=flags, |
|
3286 | 3288 | data_offset=offset, |
|
3287 | 3289 | data_compressed_length=deltainfo.deltalen, |
|
3288 | 3290 | data_uncompressed_length=textlen, |
|
3289 | 3291 | data_compression_mode=compression_mode, |
|
3290 | 3292 | data_delta_base=deltainfo.base, |
|
3291 | 3293 | link_rev=link, |
|
3292 | 3294 | parent_rev_1=p1r, |
|
3293 | 3295 | parent_rev_2=p2r, |
|
3294 | 3296 | node_id=node, |
|
3295 | 3297 | sidedata_offset=sidedata_offset, |
|
3296 | 3298 | sidedata_compressed_length=len(serialized_sidedata), |
|
3297 | 3299 | sidedata_compression_mode=sidedata_compression_mode, |
|
3298 | 3300 | rank=rank, |
|
3299 | 3301 | ) |
|
3300 | 3302 | |
|
3301 | 3303 | self.index.append(e) |
|
3302 | 3304 | entry = self.index.entry_binary(curr) |
|
3303 | 3305 | if curr == 0 and self._docket is None: |
|
3304 | 3306 | header = self._format_flags | self._format_version |
|
3305 | 3307 | header = self.index.pack_header(header) |
|
3306 | 3308 | entry = header + entry |
|
3307 | 3309 | self._writeentry( |
|
3308 | 3310 | transaction, |
|
3309 | 3311 | entry, |
|
3310 | 3312 | deltainfo.data, |
|
3311 | 3313 | link, |
|
3312 | 3314 | offset, |
|
3313 | 3315 | serialized_sidedata, |
|
3314 | 3316 | sidedata_offset, |
|
3315 | 3317 | ) |
|
3316 | 3318 | |
|
3317 | 3319 | rawtext = btext[0] |
|
3318 | 3320 | |
|
3319 | 3321 | if alwayscache and rawtext is None: |
|
3320 | 3322 | rawtext = deltacomputer.buildtext(revinfo) |
|
3321 | 3323 | |
|
3322 | 3324 | if type(rawtext) == bytes: # only accept immutable objects |
|
3323 | 3325 | self._inner._revisioncache = (node, curr, rawtext) |
|
3324 | 3326 | self._chainbasecache[curr] = deltainfo.chainbase |
|
3325 | 3327 | return curr |
|
3326 | 3328 | |
|
3327 | 3329 | def _get_data_offset(self, prev): |
|
3328 | 3330 | """Returns the current offset in the (in-transaction) data file. |
|
3329 | 3331 | Versions < 2 of the revlog can get this 0(1), revlog v2 needs a docket |
|
3330 | 3332 | file to store that information: since sidedata can be rewritten to the |
|
3331 | 3333 | end of the data file within a transaction, you can have cases where, for |
|
3332 | 3334 | example, rev `n` does not have sidedata while rev `n - 1` does, leading |
|
3333 | 3335 | to `n - 1`'s sidedata being written after `n`'s data. |
|
3334 | 3336 | |
|
3335 | 3337 | TODO cache this in a docket file before getting out of experimental.""" |
|
3336 | 3338 | if self._docket is None: |
|
3337 | 3339 | return self.end(prev) |
|
3338 | 3340 | else: |
|
3339 | 3341 | return self._docket.data_end |
|
3340 | 3342 | |
|
3341 | 3343 | def _writeentry( |
|
3342 | 3344 | self, |
|
3343 | 3345 | transaction, |
|
3344 | 3346 | entry, |
|
3345 | 3347 | data, |
|
3346 | 3348 | link, |
|
3347 | 3349 | offset, |
|
3348 | 3350 | sidedata, |
|
3349 | 3351 | sidedata_offset, |
|
3350 | 3352 | ): |
|
3351 | 3353 | # Files opened in a+ mode have inconsistent behavior on various |
|
3352 | 3354 | # platforms. Windows requires that a file positioning call be made |
|
3353 | 3355 | # when the file handle transitions between reads and writes. See |
|
3354 | 3356 | # 3686fa2b8eee and the mixedfilemodewrapper in windows.py. On other |
|
3355 | 3357 | # platforms, Python or the platform itself can be buggy. Some versions |
|
3356 | 3358 | # of Solaris have been observed to not append at the end of the file |
|
3357 | 3359 | # if the file was seeked to before the end. See issue4943 for more. |
|
3358 | 3360 | # |
|
3359 | 3361 | # We work around this issue by inserting a seek() before writing. |
|
3360 | 3362 | # Note: This is likely not necessary on Python 3. However, because |
|
3361 | 3363 | # the file handle is reused for reads and may be seeked there, we need |
|
3362 | 3364 | # to be careful before changing this. |
|
3363 | 3365 | index_end = data_end = sidedata_end = None |
|
3364 | 3366 | if self._docket is not None: |
|
3365 | 3367 | index_end = self._docket.index_end |
|
3366 | 3368 | data_end = self._docket.data_end |
|
3367 | 3369 | sidedata_end = self._docket.sidedata_end |
|
3368 | 3370 | |
|
3369 | 3371 | files_end = self._inner.write_entry( |
|
3370 | 3372 | transaction, |
|
3371 | 3373 | entry, |
|
3372 | 3374 | data, |
|
3373 | 3375 | link, |
|
3374 | 3376 | offset, |
|
3375 | 3377 | sidedata, |
|
3376 | 3378 | sidedata_offset, |
|
3377 | 3379 | index_end, |
|
3378 | 3380 | data_end, |
|
3379 | 3381 | sidedata_end, |
|
3380 | 3382 | ) |
|
3381 | 3383 | self._enforceinlinesize(transaction) |
|
3382 | 3384 | if self._docket is not None: |
|
3383 | 3385 | self._docket.index_end = files_end[0] |
|
3384 | 3386 | self._docket.data_end = files_end[1] |
|
3385 | 3387 | self._docket.sidedata_end = files_end[2] |
|
3386 | 3388 | |
|
3387 | 3389 | nodemaputil.setup_persistent_nodemap(transaction, self) |
|
3388 | 3390 | |
|
3389 | 3391 | def addgroup( |
|
3390 | 3392 | self, |
|
3391 | 3393 | deltas, |
|
3392 | 3394 | linkmapper, |
|
3393 | 3395 | transaction, |
|
3394 | 3396 | alwayscache=False, |
|
3395 | 3397 | addrevisioncb=None, |
|
3396 | 3398 | duplicaterevisioncb=None, |
|
3397 | 3399 | debug_info=None, |
|
3398 | 3400 | delta_base_reuse_policy=None, |
|
3399 | 3401 | ): |
|
3400 | 3402 | """ |
|
3401 | 3403 | add a delta group |
|
3402 | 3404 | |
|
3403 | 3405 | given a set of deltas, add them to the revision log. the |
|
3404 | 3406 | first delta is against its parent, which should be in our |
|
3405 | 3407 | log, the rest are against the previous delta. |
|
3406 | 3408 | |
|
3407 | 3409 | If ``addrevisioncb`` is defined, it will be called with arguments of |
|
3408 | 3410 | this revlog and the node that was added. |
|
3409 | 3411 | """ |
|
3410 | 3412 | |
|
3411 | 3413 | if self._adding_group: |
|
3412 | 3414 | raise error.ProgrammingError(b'cannot nest addgroup() calls') |
|
3413 | 3415 | |
|
3414 | 3416 | # read the default delta-base reuse policy from revlog config if the |
|
3415 | 3417 | # group did not specify one. |
|
3416 | 3418 | if delta_base_reuse_policy is None: |
|
3417 | 3419 | if ( |
|
3418 | 3420 | self.delta_config.general_delta |
|
3419 | 3421 | and self.delta_config.lazy_delta_base |
|
3420 | 3422 | ): |
|
3421 | 3423 | delta_base_reuse_policy = DELTA_BASE_REUSE_TRY |
|
3422 | 3424 | else: |
|
3423 | 3425 | delta_base_reuse_policy = DELTA_BASE_REUSE_NO |
|
3424 | 3426 | |
|
3425 | 3427 | self._adding_group = True |
|
3426 | 3428 | empty = True |
|
3427 | 3429 | try: |
|
3428 | 3430 | with self._writing(transaction): |
|
3429 | 3431 | write_debug = None |
|
3430 | 3432 | if self.delta_config.debug_delta: |
|
3431 | 3433 | write_debug = transaction._report |
|
3432 | 3434 | deltacomputer = deltautil.deltacomputer( |
|
3433 | 3435 | self, |
|
3434 | 3436 | write_debug=write_debug, |
|
3435 | 3437 | debug_info=debug_info, |
|
3436 | 3438 | ) |
|
3437 | 3439 | # loop through our set of deltas |
|
3438 | 3440 | for data in deltas: |
|
3439 | 3441 | ( |
|
3440 | 3442 | node, |
|
3441 | 3443 | p1, |
|
3442 | 3444 | p2, |
|
3443 | 3445 | linknode, |
|
3444 | 3446 | deltabase, |
|
3445 | 3447 | delta, |
|
3446 | 3448 | flags, |
|
3447 | 3449 | sidedata, |
|
3448 | 3450 | ) = data |
|
3449 | 3451 | link = linkmapper(linknode) |
|
3450 | 3452 | flags = flags or REVIDX_DEFAULT_FLAGS |
|
3451 | 3453 | |
|
3452 | 3454 | rev = self.index.get_rev(node) |
|
3453 | 3455 | if rev is not None: |
|
3454 | 3456 | # this can happen if two branches make the same change |
|
3455 | 3457 | self._nodeduplicatecallback(transaction, rev) |
|
3456 | 3458 | if duplicaterevisioncb: |
|
3457 | 3459 | duplicaterevisioncb(self, rev) |
|
3458 | 3460 | empty = False |
|
3459 | 3461 | continue |
|
3460 | 3462 | |
|
3461 | 3463 | for p in (p1, p2): |
|
3462 | 3464 | if not self.index.has_node(p): |
|
3463 | 3465 | raise error.LookupError( |
|
3464 | 3466 | p, self.radix, _(b'unknown parent') |
|
3465 | 3467 | ) |
|
3466 | 3468 | |
|
3467 | 3469 | if not self.index.has_node(deltabase): |
|
3468 | 3470 | raise error.LookupError( |
|
3469 | 3471 | deltabase, self.display_id, _(b'unknown delta base') |
|
3470 | 3472 | ) |
|
3471 | 3473 | |
|
3472 | 3474 | baserev = self.rev(deltabase) |
|
3473 | 3475 | |
|
3474 | 3476 | if baserev != nullrev and self.iscensored(baserev): |
|
3475 | 3477 | # if base is censored, delta must be full replacement in a |
|
3476 | 3478 | # single patch operation |
|
3477 | 3479 | hlen = struct.calcsize(b">lll") |
|
3478 | 3480 | oldlen = self.rawsize(baserev) |
|
3479 | 3481 | newlen = len(delta) - hlen |
|
3480 | 3482 | if delta[:hlen] != mdiff.replacediffheader( |
|
3481 | 3483 | oldlen, newlen |
|
3482 | 3484 | ): |
|
3483 | 3485 | raise error.CensoredBaseError( |
|
3484 | 3486 | self.display_id, self.node(baserev) |
|
3485 | 3487 | ) |
|
3486 | 3488 | |
|
3487 | 3489 | if not flags and self._peek_iscensored(baserev, delta): |
|
3488 | 3490 | flags |= REVIDX_ISCENSORED |
|
3489 | 3491 | |
|
3490 | 3492 | # We assume consumers of addrevisioncb will want to retrieve |
|
3491 | 3493 | # the added revision, which will require a call to |
|
3492 | 3494 | # revision(). revision() will fast path if there is a cache |
|
3493 | 3495 | # hit. So, we tell _addrevision() to always cache in this case. |
|
3494 | 3496 | # We're only using addgroup() in the context of changegroup |
|
3495 | 3497 | # generation so the revision data can always be handled as raw |
|
3496 | 3498 | # by the flagprocessor. |
|
3497 | 3499 | rev = self._addrevision( |
|
3498 | 3500 | node, |
|
3499 | 3501 | None, |
|
3500 | 3502 | transaction, |
|
3501 | 3503 | link, |
|
3502 | 3504 | p1, |
|
3503 | 3505 | p2, |
|
3504 | 3506 | flags, |
|
3505 | 3507 | (baserev, delta, delta_base_reuse_policy), |
|
3506 | 3508 | alwayscache=alwayscache, |
|
3507 | 3509 | deltacomputer=deltacomputer, |
|
3508 | 3510 | sidedata=sidedata, |
|
3509 | 3511 | ) |
|
3510 | 3512 | |
|
3511 | 3513 | if addrevisioncb: |
|
3512 | 3514 | addrevisioncb(self, rev) |
|
3513 | 3515 | empty = False |
|
3514 | 3516 | finally: |
|
3515 | 3517 | self._adding_group = False |
|
3516 | 3518 | return not empty |
|
3517 | 3519 | |
|
3518 | 3520 | def iscensored(self, rev): |
|
3519 | 3521 | """Check if a file revision is censored.""" |
|
3520 | 3522 | if not self.feature_config.censorable: |
|
3521 | 3523 | return False |
|
3522 | 3524 | |
|
3523 | 3525 | return self.flags(rev) & REVIDX_ISCENSORED |
|
3524 | 3526 | |
|
3525 | 3527 | def _peek_iscensored(self, baserev, delta): |
|
3526 | 3528 | """Quickly check if a delta produces a censored revision.""" |
|
3527 | 3529 | if not self.feature_config.censorable: |
|
3528 | 3530 | return False |
|
3529 | 3531 | |
|
3530 | 3532 | return storageutil.deltaiscensored(delta, baserev, self.rawsize) |
|
3531 | 3533 | |
|
3532 | 3534 | def getstrippoint(self, minlink): |
|
3533 | 3535 | """find the minimum rev that must be stripped to strip the linkrev |
|
3534 | 3536 | |
|
3535 | 3537 | Returns a tuple containing the minimum rev and a set of all revs that |
|
3536 | 3538 | have linkrevs that will be broken by this strip. |
|
3537 | 3539 | """ |
|
3538 | 3540 | return storageutil.resolvestripinfo( |
|
3539 | 3541 | minlink, |
|
3540 | 3542 | len(self) - 1, |
|
3541 | 3543 | self.headrevs(), |
|
3542 | 3544 | self.linkrev, |
|
3543 | 3545 | self.parentrevs, |
|
3544 | 3546 | ) |
|
3545 | 3547 | |
|
3546 | 3548 | def strip(self, minlink, transaction): |
|
3547 | 3549 | """truncate the revlog on the first revision with a linkrev >= minlink |
|
3548 | 3550 | |
|
3549 | 3551 | This function is called when we're stripping revision minlink and |
|
3550 | 3552 | its descendants from the repository. |
|
3551 | 3553 | |
|
3552 | 3554 | We have to remove all revisions with linkrev >= minlink, because |
|
3553 | 3555 | the equivalent changelog revisions will be renumbered after the |
|
3554 | 3556 | strip. |
|
3555 | 3557 | |
|
3556 | 3558 | So we truncate the revlog on the first of these revisions, and |
|
3557 | 3559 | trust that the caller has saved the revisions that shouldn't be |
|
3558 | 3560 | removed and that it'll re-add them after this truncation. |
|
3559 | 3561 | """ |
|
3560 | 3562 | if len(self) == 0: |
|
3561 | 3563 | return |
|
3562 | 3564 | |
|
3563 | 3565 | rev, _ = self.getstrippoint(minlink) |
|
3564 | 3566 | if rev == len(self): |
|
3565 | 3567 | return |
|
3566 | 3568 | |
|
3567 | 3569 | # first truncate the files on disk |
|
3568 | 3570 | data_end = self.start(rev) |
|
3569 | 3571 | if not self._inline: |
|
3570 | 3572 | transaction.add(self._datafile, data_end) |
|
3571 | 3573 | end = rev * self.index.entry_size |
|
3572 | 3574 | else: |
|
3573 | 3575 | end = data_end + (rev * self.index.entry_size) |
|
3574 | 3576 | |
|
3575 | 3577 | if self._sidedatafile: |
|
3576 | 3578 | sidedata_end = self.sidedata_cut_off(rev) |
|
3577 | 3579 | transaction.add(self._sidedatafile, sidedata_end) |
|
3578 | 3580 | |
|
3579 | 3581 | transaction.add(self._indexfile, end) |
|
3580 | 3582 | if self._docket is not None: |
|
3581 | 3583 | # XXX we could, leverage the docket while stripping. However it is |
|
3582 | 3584 | # not powerfull enough at the time of this comment |
|
3583 | 3585 | self._docket.index_end = end |
|
3584 | 3586 | self._docket.data_end = data_end |
|
3585 | 3587 | self._docket.sidedata_end = sidedata_end |
|
3586 | 3588 | self._docket.write(transaction, stripping=True) |
|
3587 | 3589 | |
|
3588 | 3590 | # then reset internal state in memory to forget those revisions |
|
3589 | 3591 | self._chaininfocache = util.lrucachedict(500) |
|
3590 | 3592 | self._inner.clear_cache() |
|
3591 | 3593 | |
|
3592 | 3594 | del self.index[rev:-1] |
|
3593 | 3595 | |
|
3594 | 3596 | def checksize(self): |
|
3595 | 3597 | """Check size of index and data files |
|
3596 | 3598 | |
|
3597 | 3599 | return a (dd, di) tuple. |
|
3598 | 3600 | - dd: extra bytes for the "data" file |
|
3599 | 3601 | - di: extra bytes for the "index" file |
|
3600 | 3602 | |
|
3601 | 3603 | A healthy revlog will return (0, 0). |
|
3602 | 3604 | """ |
|
3603 | 3605 | expected = 0 |
|
3604 | 3606 | if len(self): |
|
3605 | 3607 | expected = max(0, self.end(len(self) - 1)) |
|
3606 | 3608 | |
|
3607 | 3609 | try: |
|
3608 | 3610 | with self._datafp() as f: |
|
3609 | 3611 | f.seek(0, io.SEEK_END) |
|
3610 | 3612 | actual = f.tell() |
|
3611 | 3613 | dd = actual - expected |
|
3612 | 3614 | except FileNotFoundError: |
|
3613 | 3615 | dd = 0 |
|
3614 | 3616 | |
|
3615 | 3617 | try: |
|
3616 | 3618 | f = self.opener(self._indexfile) |
|
3617 | 3619 | f.seek(0, io.SEEK_END) |
|
3618 | 3620 | actual = f.tell() |
|
3619 | 3621 | f.close() |
|
3620 | 3622 | s = self.index.entry_size |
|
3621 | 3623 | i = max(0, actual // s) |
|
3622 | 3624 | di = actual - (i * s) |
|
3623 | 3625 | if self._inline: |
|
3624 | 3626 | databytes = 0 |
|
3625 | 3627 | for r in self: |
|
3626 | 3628 | databytes += max(0, self.length(r)) |
|
3627 | 3629 | dd = 0 |
|
3628 | 3630 | di = actual - len(self) * s - databytes |
|
3629 | 3631 | except FileNotFoundError: |
|
3630 | 3632 | di = 0 |
|
3631 | 3633 | |
|
3632 | 3634 | return (dd, di) |
|
3633 | 3635 | |
|
3634 | 3636 | def files(self): |
|
3635 | 3637 | """return list of files that compose this revlog""" |
|
3636 | 3638 | res = [self._indexfile] |
|
3637 | 3639 | if self._docket_file is None: |
|
3638 | 3640 | if not self._inline: |
|
3639 | 3641 | res.append(self._datafile) |
|
3640 | 3642 | else: |
|
3641 | 3643 | res.append(self._docket_file) |
|
3642 | 3644 | res.extend(self._docket.old_index_filepaths(include_empty=False)) |
|
3643 | 3645 | if self._docket.data_end: |
|
3644 | 3646 | res.append(self._datafile) |
|
3645 | 3647 | res.extend(self._docket.old_data_filepaths(include_empty=False)) |
|
3646 | 3648 | if self._docket.sidedata_end: |
|
3647 | 3649 | res.append(self._sidedatafile) |
|
3648 | 3650 | res.extend(self._docket.old_sidedata_filepaths(include_empty=False)) |
|
3649 | 3651 | return res |
|
3650 | 3652 | |
|
3651 | 3653 | def emitrevisions( |
|
3652 | 3654 | self, |
|
3653 | 3655 | nodes, |
|
3654 | 3656 | nodesorder=None, |
|
3655 | 3657 | revisiondata=False, |
|
3656 | 3658 | assumehaveparentrevisions=False, |
|
3657 | 3659 | deltamode=repository.CG_DELTAMODE_STD, |
|
3658 | 3660 | sidedata_helpers=None, |
|
3659 | 3661 | debug_info=None, |
|
3660 | 3662 | ): |
|
3661 | 3663 | if nodesorder not in (b'nodes', b'storage', b'linear', None): |
|
3662 | 3664 | raise error.ProgrammingError( |
|
3663 | 3665 | b'unhandled value for nodesorder: %s' % nodesorder |
|
3664 | 3666 | ) |
|
3665 | 3667 | |
|
3666 | 3668 | if nodesorder is None and not self.delta_config.general_delta: |
|
3667 | 3669 | nodesorder = b'storage' |
|
3668 | 3670 | |
|
3669 | 3671 | if ( |
|
3670 | 3672 | not self._storedeltachains |
|
3671 | 3673 | and deltamode != repository.CG_DELTAMODE_PREV |
|
3672 | 3674 | ): |
|
3673 | 3675 | deltamode = repository.CG_DELTAMODE_FULL |
|
3674 | 3676 | |
|
3675 | 3677 | return storageutil.emitrevisions( |
|
3676 | 3678 | self, |
|
3677 | 3679 | nodes, |
|
3678 | 3680 | nodesorder, |
|
3679 | 3681 | revlogrevisiondelta, |
|
3680 | 3682 | deltaparentfn=self.deltaparent, |
|
3681 | 3683 | candeltafn=self._candelta, |
|
3682 | 3684 | rawsizefn=self.rawsize, |
|
3683 | 3685 | revdifffn=self.revdiff, |
|
3684 | 3686 | flagsfn=self.flags, |
|
3685 | 3687 | deltamode=deltamode, |
|
3686 | 3688 | revisiondata=revisiondata, |
|
3687 | 3689 | assumehaveparentrevisions=assumehaveparentrevisions, |
|
3688 | 3690 | sidedata_helpers=sidedata_helpers, |
|
3689 | 3691 | debug_info=debug_info, |
|
3690 | 3692 | ) |
|
3691 | 3693 | |
|
3692 | 3694 | DELTAREUSEALWAYS = b'always' |
|
3693 | 3695 | DELTAREUSESAMEREVS = b'samerevs' |
|
3694 | 3696 | DELTAREUSENEVER = b'never' |
|
3695 | 3697 | |
|
3696 | 3698 | DELTAREUSEFULLADD = b'fulladd' |
|
3697 | 3699 | |
|
3698 | 3700 | DELTAREUSEALL = {b'always', b'samerevs', b'never', b'fulladd'} |
|
3699 | 3701 | |
|
3700 | 3702 | def clone( |
|
3701 | 3703 | self, |
|
3702 | 3704 | tr, |
|
3703 | 3705 | destrevlog, |
|
3704 | 3706 | addrevisioncb=None, |
|
3705 | 3707 | deltareuse=DELTAREUSESAMEREVS, |
|
3706 | 3708 | forcedeltabothparents=None, |
|
3707 | 3709 | sidedata_helpers=None, |
|
3708 | 3710 | ): |
|
3709 | 3711 | """Copy this revlog to another, possibly with format changes. |
|
3710 | 3712 | |
|
3711 | 3713 | The destination revlog will contain the same revisions and nodes. |
|
3712 | 3714 | However, it may not be bit-for-bit identical due to e.g. delta encoding |
|
3713 | 3715 | differences. |
|
3714 | 3716 | |
|
3715 | 3717 | The ``deltareuse`` argument control how deltas from the existing revlog |
|
3716 | 3718 | are preserved in the destination revlog. The argument can have the |
|
3717 | 3719 | following values: |
|
3718 | 3720 | |
|
3719 | 3721 | DELTAREUSEALWAYS |
|
3720 | 3722 | Deltas will always be reused (if possible), even if the destination |
|
3721 | 3723 | revlog would not select the same revisions for the delta. This is the |
|
3722 | 3724 | fastest mode of operation. |
|
3723 | 3725 | DELTAREUSESAMEREVS |
|
3724 | 3726 | Deltas will be reused if the destination revlog would pick the same |
|
3725 | 3727 | revisions for the delta. This mode strikes a balance between speed |
|
3726 | 3728 | and optimization. |
|
3727 | 3729 | DELTAREUSENEVER |
|
3728 | 3730 | Deltas will never be reused. This is the slowest mode of execution. |
|
3729 | 3731 | This mode can be used to recompute deltas (e.g. if the diff/delta |
|
3730 | 3732 | algorithm changes). |
|
3731 | 3733 | DELTAREUSEFULLADD |
|
3732 | 3734 | Revision will be re-added as if their were new content. This is |
|
3733 | 3735 | slower than DELTAREUSEALWAYS but allow more mechanism to kicks in. |
|
3734 | 3736 | eg: large file detection and handling. |
|
3735 | 3737 | |
|
3736 | 3738 | Delta computation can be slow, so the choice of delta reuse policy can |
|
3737 | 3739 | significantly affect run time. |
|
3738 | 3740 | |
|
3739 | 3741 | The default policy (``DELTAREUSESAMEREVS``) strikes a balance between |
|
3740 | 3742 | two extremes. Deltas will be reused if they are appropriate. But if the |
|
3741 | 3743 | delta could choose a better revision, it will do so. This means if you |
|
3742 | 3744 | are converting a non-generaldelta revlog to a generaldelta revlog, |
|
3743 | 3745 | deltas will be recomputed if the delta's parent isn't a parent of the |
|
3744 | 3746 | revision. |
|
3745 | 3747 | |
|
3746 | 3748 | In addition to the delta policy, the ``forcedeltabothparents`` |
|
3747 | 3749 | argument controls whether to force compute deltas against both parents |
|
3748 | 3750 | for merges. By default, the current default is used. |
|
3749 | 3751 | |
|
3750 | 3752 | See `revlogutil.sidedata.get_sidedata_helpers` for the doc on |
|
3751 | 3753 | `sidedata_helpers`. |
|
3752 | 3754 | """ |
|
3753 | 3755 | if deltareuse not in self.DELTAREUSEALL: |
|
3754 | 3756 | raise ValueError( |
|
3755 | 3757 | _(b'value for deltareuse invalid: %s') % deltareuse |
|
3756 | 3758 | ) |
|
3757 | 3759 | |
|
3758 | 3760 | if len(destrevlog): |
|
3759 | 3761 | raise ValueError(_(b'destination revlog is not empty')) |
|
3760 | 3762 | |
|
3761 | 3763 | if getattr(self, 'filteredrevs', None): |
|
3762 | 3764 | raise ValueError(_(b'source revlog has filtered revisions')) |
|
3763 | 3765 | if getattr(destrevlog, 'filteredrevs', None): |
|
3764 | 3766 | raise ValueError(_(b'destination revlog has filtered revisions')) |
|
3765 | 3767 | |
|
3766 | 3768 | # lazydelta and lazydeltabase controls whether to reuse a cached delta, |
|
3767 | 3769 | # if possible. |
|
3768 | 3770 | old_delta_config = destrevlog.delta_config |
|
3769 | 3771 | destrevlog.delta_config = destrevlog.delta_config.copy() |
|
3770 | 3772 | |
|
3771 | 3773 | try: |
|
3772 | 3774 | if deltareuse == self.DELTAREUSEALWAYS: |
|
3773 | 3775 | destrevlog.delta_config.lazy_delta_base = True |
|
3774 | 3776 | destrevlog.delta_config.lazy_delta = True |
|
3775 | 3777 | elif deltareuse == self.DELTAREUSESAMEREVS: |
|
3776 | 3778 | destrevlog.delta_config.lazy_delta_base = False |
|
3777 | 3779 | destrevlog.delta_config.lazy_delta = True |
|
3778 | 3780 | elif deltareuse == self.DELTAREUSENEVER: |
|
3779 | 3781 | destrevlog.delta_config.lazy_delta_base = False |
|
3780 | 3782 | destrevlog.delta_config.lazy_delta = False |
|
3781 | 3783 | |
|
3782 | 3784 | delta_both_parents = ( |
|
3783 | 3785 | forcedeltabothparents or old_delta_config.delta_both_parents |
|
3784 | 3786 | ) |
|
3785 | 3787 | destrevlog.delta_config.delta_both_parents = delta_both_parents |
|
3786 | 3788 | |
|
3787 | 3789 | with self.reading(), destrevlog._writing(tr): |
|
3788 | 3790 | self._clone( |
|
3789 | 3791 | tr, |
|
3790 | 3792 | destrevlog, |
|
3791 | 3793 | addrevisioncb, |
|
3792 | 3794 | deltareuse, |
|
3793 | 3795 | forcedeltabothparents, |
|
3794 | 3796 | sidedata_helpers, |
|
3795 | 3797 | ) |
|
3796 | 3798 | |
|
3797 | 3799 | finally: |
|
3798 | 3800 | destrevlog.delta_config = old_delta_config |
|
3799 | 3801 | |
|
3800 | 3802 | def _clone( |
|
3801 | 3803 | self, |
|
3802 | 3804 | tr, |
|
3803 | 3805 | destrevlog, |
|
3804 | 3806 | addrevisioncb, |
|
3805 | 3807 | deltareuse, |
|
3806 | 3808 | forcedeltabothparents, |
|
3807 | 3809 | sidedata_helpers, |
|
3808 | 3810 | ): |
|
3809 | 3811 | """perform the core duty of `revlog.clone` after parameter processing""" |
|
3810 | 3812 | write_debug = None |
|
3811 | 3813 | if self.delta_config.debug_delta: |
|
3812 | 3814 | write_debug = tr._report |
|
3813 | 3815 | deltacomputer = deltautil.deltacomputer( |
|
3814 | 3816 | destrevlog, |
|
3815 | 3817 | write_debug=write_debug, |
|
3816 | 3818 | ) |
|
3817 | 3819 | index = self.index |
|
3818 | 3820 | for rev in self: |
|
3819 | 3821 | entry = index[rev] |
|
3820 | 3822 | |
|
3821 | 3823 | # Some classes override linkrev to take filtered revs into |
|
3822 | 3824 | # account. Use raw entry from index. |
|
3823 | 3825 | flags = entry[0] & 0xFFFF |
|
3824 | 3826 | linkrev = entry[4] |
|
3825 | 3827 | p1 = index[entry[5]][7] |
|
3826 | 3828 | p2 = index[entry[6]][7] |
|
3827 | 3829 | node = entry[7] |
|
3828 | 3830 | |
|
3829 | 3831 | # (Possibly) reuse the delta from the revlog if allowed and |
|
3830 | 3832 | # the revlog chunk is a delta. |
|
3831 | 3833 | cachedelta = None |
|
3832 | 3834 | rawtext = None |
|
3833 | 3835 | if deltareuse == self.DELTAREUSEFULLADD: |
|
3834 | 3836 | text = self._revisiondata(rev) |
|
3835 | 3837 | sidedata = self.sidedata(rev) |
|
3836 | 3838 | |
|
3837 | 3839 | if sidedata_helpers is not None: |
|
3838 | 3840 | (sidedata, new_flags) = sidedatautil.run_sidedata_helpers( |
|
3839 | 3841 | self, sidedata_helpers, sidedata, rev |
|
3840 | 3842 | ) |
|
3841 | 3843 | flags = flags | new_flags[0] & ~new_flags[1] |
|
3842 | 3844 | |
|
3843 | 3845 | destrevlog.addrevision( |
|
3844 | 3846 | text, |
|
3845 | 3847 | tr, |
|
3846 | 3848 | linkrev, |
|
3847 | 3849 | p1, |
|
3848 | 3850 | p2, |
|
3849 | 3851 | cachedelta=cachedelta, |
|
3850 | 3852 | node=node, |
|
3851 | 3853 | flags=flags, |
|
3852 | 3854 | deltacomputer=deltacomputer, |
|
3853 | 3855 | sidedata=sidedata, |
|
3854 | 3856 | ) |
|
3855 | 3857 | else: |
|
3856 | 3858 | if destrevlog.delta_config.lazy_delta: |
|
3857 | 3859 | dp = self.deltaparent(rev) |
|
3858 | 3860 | if dp != nullrev: |
|
3859 | 3861 | cachedelta = (dp, bytes(self._inner._chunk(rev))) |
|
3860 | 3862 | |
|
3861 | 3863 | sidedata = None |
|
3862 | 3864 | if not cachedelta: |
|
3863 | 3865 | try: |
|
3864 | 3866 | rawtext = self._revisiondata(rev) |
|
3865 | 3867 | except error.CensoredNodeError as censored: |
|
3866 | 3868 | assert flags & REVIDX_ISCENSORED |
|
3867 | 3869 | rawtext = censored.tombstone |
|
3868 | 3870 | sidedata = self.sidedata(rev) |
|
3869 | 3871 | if sidedata is None: |
|
3870 | 3872 | sidedata = self.sidedata(rev) |
|
3871 | 3873 | |
|
3872 | 3874 | if sidedata_helpers is not None: |
|
3873 | 3875 | (sidedata, new_flags) = sidedatautil.run_sidedata_helpers( |
|
3874 | 3876 | self, sidedata_helpers, sidedata, rev |
|
3875 | 3877 | ) |
|
3876 | 3878 | flags = flags | new_flags[0] & ~new_flags[1] |
|
3877 | 3879 | |
|
3878 | 3880 | destrevlog._addrevision( |
|
3879 | 3881 | node, |
|
3880 | 3882 | rawtext, |
|
3881 | 3883 | tr, |
|
3882 | 3884 | linkrev, |
|
3883 | 3885 | p1, |
|
3884 | 3886 | p2, |
|
3885 | 3887 | flags, |
|
3886 | 3888 | cachedelta, |
|
3887 | 3889 | deltacomputer=deltacomputer, |
|
3888 | 3890 | sidedata=sidedata, |
|
3889 | 3891 | ) |
|
3890 | 3892 | |
|
3891 | 3893 | if addrevisioncb: |
|
3892 | 3894 | addrevisioncb(self, rev, node) |
|
3893 | 3895 | |
|
3894 | 3896 | def censorrevision(self, tr, censor_nodes, tombstone=b''): |
|
3895 | 3897 | if self._format_version == REVLOGV0: |
|
3896 | 3898 | raise error.RevlogError( |
|
3897 | 3899 | _(b'cannot censor with version %d revlogs') |
|
3898 | 3900 | % self._format_version |
|
3899 | 3901 | ) |
|
3900 | 3902 | elif self._format_version == REVLOGV1: |
|
3901 | 3903 | rewrite.v1_censor(self, tr, censor_nodes, tombstone) |
|
3902 | 3904 | else: |
|
3903 | 3905 | rewrite.v2_censor(self, tr, censor_nodes, tombstone) |
|
3904 | 3906 | |
|
3905 | def verifyintegrity(self, state): | |
|
3907 | def verifyintegrity(self, state) -> Iterable[RevLogProblem]: | |
|
3906 | 3908 | """Verifies the integrity of the revlog. |
|
3907 | 3909 | |
|
3908 | 3910 | Yields ``revlogproblem`` instances describing problems that are |
|
3909 | 3911 | found. |
|
3910 | 3912 | """ |
|
3911 | 3913 | dd, di = self.checksize() |
|
3912 | 3914 | if dd: |
|
3913 | 3915 | yield revlogproblem(error=_(b'data length off by %d bytes') % dd) |
|
3914 | 3916 | if di: |
|
3915 | 3917 | yield revlogproblem(error=_(b'index contains %d extra bytes') % di) |
|
3916 | 3918 | |
|
3917 | 3919 | version = self._format_version |
|
3918 | 3920 | |
|
3919 | 3921 | # The verifier tells us what version revlog we should be. |
|
3920 | 3922 | if version != state[b'expectedversion']: |
|
3921 | 3923 | yield revlogproblem( |
|
3922 | 3924 | warning=_(b"warning: '%s' uses revlog format %d; expected %d") |
|
3923 | 3925 | % (self.display_id, version, state[b'expectedversion']) |
|
3924 | 3926 | ) |
|
3925 | 3927 | |
|
3926 | 3928 | state[b'skipread'] = set() |
|
3927 | 3929 | state[b'safe_renamed'] = set() |
|
3928 | 3930 | |
|
3929 | 3931 | for rev in self: |
|
3930 | 3932 | node = self.node(rev) |
|
3931 | 3933 | |
|
3932 | 3934 | # Verify contents. 4 cases to care about: |
|
3933 | 3935 | # |
|
3934 | 3936 | # common: the most common case |
|
3935 | 3937 | # rename: with a rename |
|
3936 | 3938 | # meta: file content starts with b'\1\n', the metadata |
|
3937 | 3939 | # header defined in filelog.py, but without a rename |
|
3938 | 3940 | # ext: content stored externally |
|
3939 | 3941 | # |
|
3940 | 3942 | # More formally, their differences are shown below: |
|
3941 | 3943 | # |
|
3942 | 3944 | # | common | rename | meta | ext |
|
3943 | 3945 | # ------------------------------------------------------- |
|
3944 | 3946 | # flags() | 0 | 0 | 0 | not 0 |
|
3945 | 3947 | # renamed() | False | True | False | ? |
|
3946 | 3948 | # rawtext[0:2]=='\1\n'| False | True | True | ? |
|
3947 | 3949 | # |
|
3948 | 3950 | # "rawtext" means the raw text stored in revlog data, which |
|
3949 | 3951 | # could be retrieved by "rawdata(rev)". "text" |
|
3950 | 3952 | # mentioned below is "revision(rev)". |
|
3951 | 3953 | # |
|
3952 | 3954 | # There are 3 different lengths stored physically: |
|
3953 | 3955 | # 1. L1: rawsize, stored in revlog index |
|
3954 | 3956 | # 2. L2: len(rawtext), stored in revlog data |
|
3955 | 3957 | # 3. L3: len(text), stored in revlog data if flags==0, or |
|
3956 | 3958 | # possibly somewhere else if flags!=0 |
|
3957 | 3959 | # |
|
3958 | 3960 | # L1 should be equal to L2. L3 could be different from them. |
|
3959 | 3961 | # "text" may or may not affect commit hash depending on flag |
|
3960 | 3962 | # processors (see flagutil.addflagprocessor). |
|
3961 | 3963 | # |
|
3962 | 3964 | # | common | rename | meta | ext |
|
3963 | 3965 | # ------------------------------------------------- |
|
3964 | 3966 | # rawsize() | L1 | L1 | L1 | L1 |
|
3965 | 3967 | # size() | L1 | L2-LM | L1(*) | L1 (?) |
|
3966 | 3968 | # len(rawtext) | L2 | L2 | L2 | L2 |
|
3967 | 3969 | # len(text) | L2 | L2 | L2 | L3 |
|
3968 | 3970 | # len(read()) | L2 | L2-LM | L2-LM | L3 (?) |
|
3969 | 3971 | # |
|
3970 | 3972 | # LM: length of metadata, depending on rawtext |
|
3971 | 3973 | # (*): not ideal, see comment in filelog.size |
|
3972 | 3974 | # (?): could be "- len(meta)" if the resolved content has |
|
3973 | 3975 | # rename metadata |
|
3974 | 3976 | # |
|
3975 | 3977 | # Checks needed to be done: |
|
3976 | 3978 | # 1. length check: L1 == L2, in all cases. |
|
3977 | 3979 | # 2. hash check: depending on flag processor, we may need to |
|
3978 | 3980 | # use either "text" (external), or "rawtext" (in revlog). |
|
3979 | 3981 | |
|
3980 | 3982 | try: |
|
3981 | 3983 | skipflags = state.get(b'skipflags', 0) |
|
3982 | 3984 | if skipflags: |
|
3983 | 3985 | skipflags &= self.flags(rev) |
|
3984 | 3986 | |
|
3985 | 3987 | _verify_revision(self, skipflags, state, node) |
|
3986 | 3988 | |
|
3987 | 3989 | l1 = self.rawsize(rev) |
|
3988 | 3990 | l2 = len(self.rawdata(node)) |
|
3989 | 3991 | |
|
3990 | 3992 | if l1 != l2: |
|
3991 | 3993 | yield revlogproblem( |
|
3992 | 3994 | error=_(b'unpacked size is %d, %d expected') % (l2, l1), |
|
3993 | 3995 | node=node, |
|
3994 | 3996 | ) |
|
3995 | 3997 | |
|
3996 | 3998 | except error.CensoredNodeError: |
|
3997 | 3999 | if state[b'erroroncensored']: |
|
3998 | 4000 | yield revlogproblem( |
|
3999 | 4001 | error=_(b'censored file data'), node=node |
|
4000 | 4002 | ) |
|
4001 | 4003 | state[b'skipread'].add(node) |
|
4002 | 4004 | except Exception as e: |
|
4003 | 4005 | yield revlogproblem( |
|
4004 | 4006 | error=_(b'unpacking %s: %s') |
|
4005 | 4007 | % (short(node), stringutil.forcebytestr(e)), |
|
4006 | 4008 | node=node, |
|
4007 | 4009 | ) |
|
4008 | 4010 | state[b'skipread'].add(node) |
|
4009 | 4011 | |
|
4010 | 4012 | def storageinfo( |
|
4011 | 4013 | self, |
|
4012 | 4014 | exclusivefiles=False, |
|
4013 | 4015 | sharedfiles=False, |
|
4014 | 4016 | revisionscount=False, |
|
4015 | 4017 | trackedsize=False, |
|
4016 | 4018 | storedsize=False, |
|
4017 | 4019 | ): |
|
4018 | 4020 | d = {} |
|
4019 | 4021 | |
|
4020 | 4022 | if exclusivefiles: |
|
4021 | 4023 | d[b'exclusivefiles'] = [(self.opener, self._indexfile)] |
|
4022 | 4024 | if not self._inline: |
|
4023 | 4025 | d[b'exclusivefiles'].append((self.opener, self._datafile)) |
|
4024 | 4026 | |
|
4025 | 4027 | if sharedfiles: |
|
4026 | 4028 | d[b'sharedfiles'] = [] |
|
4027 | 4029 | |
|
4028 | 4030 | if revisionscount: |
|
4029 | 4031 | d[b'revisionscount'] = len(self) |
|
4030 | 4032 | |
|
4031 | 4033 | if trackedsize: |
|
4032 | 4034 | d[b'trackedsize'] = sum(map(self.rawsize, iter(self))) |
|
4033 | 4035 | |
|
4034 | 4036 | if storedsize: |
|
4035 | 4037 | d[b'storedsize'] = sum( |
|
4036 | 4038 | self.opener.stat(path).st_size for path in self.files() |
|
4037 | 4039 | ) |
|
4038 | 4040 | |
|
4039 | 4041 | return d |
|
4040 | 4042 | |
|
4041 | 4043 | def rewrite_sidedata(self, transaction, helpers, startrev, endrev): |
|
4042 | 4044 | if not self.feature_config.has_side_data: |
|
4043 | 4045 | return |
|
4044 | 4046 | # revlog formats with sidedata support does not support inline |
|
4045 | 4047 | assert not self._inline |
|
4046 | 4048 | if not helpers[1] and not helpers[2]: |
|
4047 | 4049 | # Nothing to generate or remove |
|
4048 | 4050 | return |
|
4049 | 4051 | |
|
4050 | 4052 | new_entries = [] |
|
4051 | 4053 | # append the new sidedata |
|
4052 | 4054 | with self._writing(transaction): |
|
4053 | 4055 | ifh, dfh, sdfh = self._inner._writinghandles |
|
4054 | 4056 | dfh.seek(self._docket.sidedata_end, os.SEEK_SET) |
|
4055 | 4057 | |
|
4056 | 4058 | current_offset = sdfh.tell() |
|
4057 | 4059 | for rev in range(startrev, endrev + 1): |
|
4058 | 4060 | entry = self.index[rev] |
|
4059 | 4061 | new_sidedata, flags = sidedatautil.run_sidedata_helpers( |
|
4060 | 4062 | store=self, |
|
4061 | 4063 | sidedata_helpers=helpers, |
|
4062 | 4064 | sidedata={}, |
|
4063 | 4065 | rev=rev, |
|
4064 | 4066 | ) |
|
4065 | 4067 | |
|
4066 | 4068 | serialized_sidedata = sidedatautil.serialize_sidedata( |
|
4067 | 4069 | new_sidedata |
|
4068 | 4070 | ) |
|
4069 | 4071 | |
|
4070 | 4072 | sidedata_compression_mode = COMP_MODE_INLINE |
|
4071 | 4073 | if serialized_sidedata and self.feature_config.has_side_data: |
|
4072 | 4074 | sidedata_compression_mode = COMP_MODE_PLAIN |
|
4073 | 4075 | h, comp_sidedata = self._inner.compress(serialized_sidedata) |
|
4074 | 4076 | if ( |
|
4075 | 4077 | h != b'u' |
|
4076 | 4078 | and comp_sidedata[0] != b'\0' |
|
4077 | 4079 | and len(comp_sidedata) < len(serialized_sidedata) |
|
4078 | 4080 | ): |
|
4079 | 4081 | assert not h |
|
4080 | 4082 | if ( |
|
4081 | 4083 | comp_sidedata[0] |
|
4082 | 4084 | == self._docket.default_compression_header |
|
4083 | 4085 | ): |
|
4084 | 4086 | sidedata_compression_mode = COMP_MODE_DEFAULT |
|
4085 | 4087 | serialized_sidedata = comp_sidedata |
|
4086 | 4088 | else: |
|
4087 | 4089 | sidedata_compression_mode = COMP_MODE_INLINE |
|
4088 | 4090 | serialized_sidedata = comp_sidedata |
|
4089 | 4091 | if entry[8] != 0 or entry[9] != 0: |
|
4090 | 4092 | # rewriting entries that already have sidedata is not |
|
4091 | 4093 | # supported yet, because it introduces garbage data in the |
|
4092 | 4094 | # revlog. |
|
4093 | 4095 | msg = b"rewriting existing sidedata is not supported yet" |
|
4094 | 4096 | raise error.Abort(msg) |
|
4095 | 4097 | |
|
4096 | 4098 | # Apply (potential) flags to add and to remove after running |
|
4097 | 4099 | # the sidedata helpers |
|
4098 | 4100 | new_offset_flags = entry[0] | flags[0] & ~flags[1] |
|
4099 | 4101 | entry_update = ( |
|
4100 | 4102 | current_offset, |
|
4101 | 4103 | len(serialized_sidedata), |
|
4102 | 4104 | new_offset_flags, |
|
4103 | 4105 | sidedata_compression_mode, |
|
4104 | 4106 | ) |
|
4105 | 4107 | |
|
4106 | 4108 | # the sidedata computation might have move the file cursors around |
|
4107 | 4109 | sdfh.seek(current_offset, os.SEEK_SET) |
|
4108 | 4110 | sdfh.write(serialized_sidedata) |
|
4109 | 4111 | new_entries.append(entry_update) |
|
4110 | 4112 | current_offset += len(serialized_sidedata) |
|
4111 | 4113 | self._docket.sidedata_end = sdfh.tell() |
|
4112 | 4114 | |
|
4113 | 4115 | # rewrite the new index entries |
|
4114 | 4116 | ifh.seek(startrev * self.index.entry_size) |
|
4115 | 4117 | for i, e in enumerate(new_entries): |
|
4116 | 4118 | rev = startrev + i |
|
4117 | 4119 | self.index.replace_sidedata_info(rev, *e) |
|
4118 | 4120 | packed = self.index.entry_binary(rev) |
|
4119 | 4121 | if rev == 0 and self._docket is None: |
|
4120 | 4122 | header = self._format_flags | self._format_version |
|
4121 | 4123 | header = self.index.pack_header(header) |
|
4122 | 4124 | packed = header + packed |
|
4123 | 4125 | ifh.write(packed) |
@@ -1,271 +1,273 b'' | |||
|
1 | 1 | # statichttprepo.py - simple http repository class for mercurial |
|
2 | 2 | # |
|
3 | 3 | # This provides read-only repo access to repositories exported via static http |
|
4 | 4 | # |
|
5 | 5 | # Copyright 2005-2007 Olivia Mackall <olivia@selenic.com> |
|
6 | 6 | # |
|
7 | 7 | # This software may be used and distributed according to the terms of the |
|
8 | 8 | # GNU General Public License version 2 or any later version. |
|
9 | 9 | |
|
10 | 10 | |
|
11 | 11 | import errno |
|
12 | 12 | |
|
13 | 13 | from .i18n import _ |
|
14 | 14 | from .node import sha1nodeconstants |
|
15 | 15 | from . import ( |
|
16 | 16 | branchmap, |
|
17 | 17 | changelog, |
|
18 | 18 | error, |
|
19 | 19 | localrepo, |
|
20 | 20 | manifest, |
|
21 | 21 | namespaces, |
|
22 | 22 | pathutil, |
|
23 | 23 | pycompat, |
|
24 | 24 | requirements as requirementsmod, |
|
25 | 25 | url, |
|
26 | 26 | util, |
|
27 | 27 | vfs as vfsmod, |
|
28 | 28 | ) |
|
29 | 29 | from .utils import ( |
|
30 | 30 | urlutil, |
|
31 | 31 | ) |
|
32 | 32 | |
|
33 | 33 | urlerr = util.urlerr |
|
34 | 34 | urlreq = util.urlreq |
|
35 | 35 | |
|
36 | 36 | |
|
37 | 37 | class httprangereader: |
|
38 | 38 | def __init__(self, url, opener): |
|
39 | 39 | # we assume opener has HTTPRangeHandler |
|
40 | 40 | self.url = url |
|
41 | 41 | self.pos = 0 |
|
42 | 42 | self.opener = opener |
|
43 | 43 | self.name = url |
|
44 | 44 | |
|
45 | 45 | def __enter__(self): |
|
46 | 46 | return self |
|
47 | 47 | |
|
48 | 48 | def __exit__(self, exc_type, exc_value, traceback): |
|
49 | 49 | self.close() |
|
50 | 50 | |
|
51 | 51 | def seek(self, pos): |
|
52 | 52 | self.pos = pos |
|
53 | 53 | |
|
54 | 54 | def read(self, bytes=None): |
|
55 | 55 | req = urlreq.request(pycompat.strurl(self.url)) |
|
56 | 56 | end = b'' |
|
57 | 57 | if bytes: |
|
58 | 58 | end = self.pos + bytes - 1 |
|
59 | 59 | if self.pos or end: |
|
60 | 60 | req.add_header('Range', 'bytes=%d-%s' % (self.pos, end)) |
|
61 | 61 | |
|
62 | 62 | try: |
|
63 | 63 | f = self.opener.open(req) |
|
64 | 64 | data = f.read() |
|
65 | 65 | code = f.code |
|
66 | 66 | except urlerr.httperror as inst: |
|
67 | 67 | num = inst.code == 404 and errno.ENOENT or None |
|
68 | 68 | # Explicitly convert the exception to str as Py3 will try |
|
69 | 69 | # convert it to local encoding and with as the HTTPResponse |
|
70 | 70 | # instance doesn't support encode. |
|
71 | 71 | raise IOError(num, str(inst)) |
|
72 | 72 | except urlerr.urlerror as inst: |
|
73 | 73 | raise IOError(None, inst.reason) |
|
74 | 74 | |
|
75 | 75 | if code == 200: |
|
76 | 76 | # HTTPRangeHandler does nothing if remote does not support |
|
77 | 77 | # Range headers and returns the full entity. Let's slice it. |
|
78 | 78 | if bytes: |
|
79 | 79 | data = data[self.pos : self.pos + bytes] |
|
80 | 80 | else: |
|
81 | 81 | data = data[self.pos :] |
|
82 | 82 | elif bytes: |
|
83 | 83 | data = data[:bytes] |
|
84 | 84 | self.pos += len(data) |
|
85 | 85 | return data |
|
86 | 86 | |
|
87 | 87 | def readlines(self): |
|
88 | 88 | return self.read().splitlines(True) |
|
89 | 89 | |
|
90 | 90 | def __iter__(self): |
|
91 | 91 | return iter(self.readlines()) |
|
92 | 92 | |
|
93 | 93 | def close(self): |
|
94 | 94 | pass |
|
95 | 95 | |
|
96 | 96 | |
|
97 | 97 | # _RangeError and _HTTPRangeHandler were originally in byterange.py, |
|
98 | 98 | # which was itself extracted from urlgrabber. See the last version of |
|
99 | 99 | # byterange.py from history if you need more information. |
|
100 | 100 | class _RangeError(IOError): |
|
101 | 101 | """Error raised when an unsatisfiable range is requested.""" |
|
102 | 102 | |
|
103 | 103 | |
|
104 | 104 | class _HTTPRangeHandler(urlreq.basehandler): |
|
105 | 105 | """Handler that enables HTTP Range headers. |
|
106 | 106 | |
|
107 | 107 | This was extremely simple. The Range header is a HTTP feature to |
|
108 | 108 | begin with so all this class does is tell urllib2 that the |
|
109 | 109 | "206 Partial Content" response from the HTTP server is what we |
|
110 | 110 | expected. |
|
111 | 111 | """ |
|
112 | 112 | |
|
113 | 113 | def http_error_206(self, req, fp, code, msg, hdrs): |
|
114 | 114 | # 206 Partial Content Response |
|
115 | 115 | r = urlreq.addinfourl(fp, hdrs, req.get_full_url()) |
|
116 | 116 | r.code = code |
|
117 | 117 | r.msg = msg |
|
118 | 118 | return r |
|
119 | 119 | |
|
120 | 120 | def http_error_416(self, req, fp, code, msg, hdrs): |
|
121 | 121 | # HTTP's Range Not Satisfiable error |
|
122 | 122 | raise _RangeError('Requested Range Not Satisfiable') |
|
123 | 123 | |
|
124 | 124 | |
|
125 | 125 | def build_opener(ui, authinfo): |
|
126 | 126 | # urllib cannot handle URLs with embedded user or passwd |
|
127 | 127 | urlopener = url.opener(ui, authinfo) |
|
128 | 128 | urlopener.add_handler(_HTTPRangeHandler()) |
|
129 | 129 | |
|
130 | 130 | class statichttpvfs(vfsmod.abstractvfs): |
|
131 | 131 | def __init__(self, base): |
|
132 | 132 | self.base = base |
|
133 | 133 | self.options = {} |
|
134 | 134 | |
|
135 | 135 | def __call__(self, path, mode=b'r', *args, **kw): |
|
136 | 136 | if mode not in (b'r', b'rb'): |
|
137 | 137 | raise IOError('Permission denied') |
|
138 | 138 | f = b"/".join((self.base, urlreq.quote(path))) |
|
139 | 139 | return httprangereader(f, urlopener) |
|
140 | 140 | |
|
141 | 141 | def join(self, path, *insidef): |
|
142 | 142 | if path: |
|
143 | 143 | return pathutil.join(self.base, path, *insidef) |
|
144 | 144 | else: |
|
145 | 145 | return self.base |
|
146 | 146 | |
|
147 | 147 | return statichttpvfs |
|
148 | 148 | |
|
149 | 149 | |
|
150 | 150 | class statichttppeer(localrepo.localpeer): |
|
151 | 151 | def local(self): |
|
152 | 152 | return None |
|
153 | 153 | |
|
154 | 154 | def canpush(self): |
|
155 | 155 | return False |
|
156 | 156 | |
|
157 | 157 | |
|
158 | 158 | class statichttprepository( |
|
159 | 159 | localrepo.localrepository, localrepo.revlogfilestorage |
|
160 | 160 | ): |
|
161 | 161 | supported = localrepo.localrepository._basesupported |
|
162 | 162 | |
|
163 | manifestlog: manifest.ManifestLog | |
|
164 | ||
|
163 | 165 | def __init__(self, ui, path): |
|
164 | 166 | self._url = path |
|
165 | 167 | self.ui = ui |
|
166 | 168 | |
|
167 | 169 | self.root = path |
|
168 | 170 | u = urlutil.url(path.rstrip(b'/') + b"/.hg") |
|
169 | 171 | self.path, authinfo = u.authinfo() |
|
170 | 172 | |
|
171 | 173 | vfsclass = build_opener(ui, authinfo) |
|
172 | 174 | self.vfs = vfsclass(self.path) |
|
173 | 175 | self.cachevfs = vfsclass(self.vfs.join(b'cache')) |
|
174 | 176 | self._phasedefaults = [] |
|
175 | 177 | |
|
176 | 178 | self.names = namespaces.namespaces() |
|
177 | 179 | self.filtername = None |
|
178 | 180 | self._extrafilterid = None |
|
179 | 181 | self._wanted_sidedata = set() |
|
180 | 182 | self.features = set() |
|
181 | 183 | |
|
182 | 184 | try: |
|
183 | 185 | requirements = set(self.vfs.read(b'requires').splitlines()) |
|
184 | 186 | except FileNotFoundError: |
|
185 | 187 | requirements = set() |
|
186 | 188 | |
|
187 | 189 | # check if it is a non-empty old-style repository |
|
188 | 190 | try: |
|
189 | 191 | fp = self.vfs(b"00changelog.i") |
|
190 | 192 | fp.read(1) |
|
191 | 193 | fp.close() |
|
192 | 194 | except FileNotFoundError: |
|
193 | 195 | # we do not care about empty old-style repositories here |
|
194 | 196 | msg = _(b"'%s' does not appear to be an hg repository") % path |
|
195 | 197 | raise error.RepoError(msg) |
|
196 | 198 | if requirementsmod.SHARESAFE_REQUIREMENT in requirements: |
|
197 | 199 | storevfs = vfsclass(self.vfs.join(b'store')) |
|
198 | 200 | requirements |= set(storevfs.read(b'requires').splitlines()) |
|
199 | 201 | |
|
200 | 202 | supportedrequirements = localrepo.gathersupportedrequirements(ui) |
|
201 | 203 | localrepo.ensurerequirementsrecognized( |
|
202 | 204 | requirements, supportedrequirements |
|
203 | 205 | ) |
|
204 | 206 | localrepo.ensurerequirementscompatible(ui, requirements) |
|
205 | 207 | self.nodeconstants = sha1nodeconstants |
|
206 | 208 | self.nullid = self.nodeconstants.nullid |
|
207 | 209 | |
|
208 | 210 | # setup store |
|
209 | 211 | self.store = localrepo.makestore(requirements, self.path, vfsclass) |
|
210 | 212 | self.spath = self.store.path |
|
211 | 213 | self.svfs = self.store.opener |
|
212 | 214 | self.sjoin = self.store.join |
|
213 | 215 | self._filecache = {} |
|
214 | 216 | self.requirements = requirements |
|
215 | 217 | |
|
216 | 218 | rootmanifest = manifest.manifestrevlog(self.nodeconstants, self.svfs) |
|
217 | 219 | self.manifestlog = manifest.manifestlog( |
|
218 | 220 | self.svfs, self, rootmanifest, self.narrowmatch() |
|
219 | 221 | ) |
|
220 | 222 | self.changelog = changelog.changelog(self.svfs) |
|
221 | 223 | self._tags = None |
|
222 | 224 | self.nodetagscache = None |
|
223 | 225 | self._branchcaches = branchmap.BranchMapCache() |
|
224 | 226 | self._revbranchcache = None |
|
225 | 227 | self.encodepats = None |
|
226 | 228 | self.decodepats = None |
|
227 | 229 | self._transref = None |
|
228 | 230 | self._dirstate = None |
|
229 | 231 | |
|
230 | 232 | def _restrictcapabilities(self, caps): |
|
231 | 233 | caps = super(statichttprepository, self)._restrictcapabilities(caps) |
|
232 | 234 | return caps.difference([b"pushkey"]) |
|
233 | 235 | |
|
234 | 236 | def url(self): |
|
235 | 237 | return self._url |
|
236 | 238 | |
|
237 | 239 | def local(self): |
|
238 | 240 | return False |
|
239 | 241 | |
|
240 | 242 | def peer(self, path=None, remotehidden=False): |
|
241 | 243 | return statichttppeer(self, path=path, remotehidden=remotehidden) |
|
242 | 244 | |
|
243 | 245 | def wlock(self, wait=True): |
|
244 | 246 | raise error.LockUnavailable( |
|
245 | 247 | 0, |
|
246 | 248 | pycompat.sysstr(_(b'lock not available')), |
|
247 | 249 | b'lock', |
|
248 | 250 | _(b'cannot lock static-http repository'), |
|
249 | 251 | ) |
|
250 | 252 | |
|
251 | 253 | def lock(self, wait=True): |
|
252 | 254 | raise error.LockUnavailable( |
|
253 | 255 | 0, |
|
254 | 256 | pycompat.sysstr(_(b'lock not available')), |
|
255 | 257 | b'lock', |
|
256 | 258 | _(b'cannot lock static-http repository'), |
|
257 | 259 | ) |
|
258 | 260 | |
|
259 | 261 | def _writecaches(self): |
|
260 | 262 | pass # statichttprepository are read only |
|
261 | 263 | |
|
262 | 264 | |
|
263 | 265 | def make_peer( |
|
264 | 266 | ui, path, create, intents=None, createopts=None, remotehidden=False |
|
265 | 267 | ): |
|
266 | 268 | if create: |
|
267 | 269 | raise error.Abort(_(b'cannot create new static-http repository')) |
|
268 | 270 | url = path.loc[7:] |
|
269 | 271 | return statichttprepository(ui, url).peer( |
|
270 | 272 | path=path, remotehidden=remotehidden |
|
271 | 273 | ) |
@@ -1,1250 +1,1250 b'' | |||
|
1 | 1 | # store.py - repository store handling for Mercurial) |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2008 Olivia Mackall <olivia@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | import collections |
|
9 | 9 | import functools |
|
10 | 10 | import os |
|
11 | 11 | import re |
|
12 | 12 | import stat |
|
13 | 13 | import typing |
|
14 | 14 | |
|
15 | 15 | from typing import Generator, List |
|
16 | 16 | |
|
17 | 17 | from .i18n import _ |
|
18 | 18 | from .thirdparty import attr |
|
19 | 19 | |
|
20 | 20 | # Force pytype to use the non-vendored package |
|
21 | 21 | if typing.TYPE_CHECKING: |
|
22 | 22 | # noinspection PyPackageRequirements |
|
23 | 23 | import attr |
|
24 | 24 | |
|
25 | 25 | from .node import hex |
|
26 | 26 | from .revlogutils.constants import ( |
|
27 | 27 | INDEX_HEADER, |
|
28 | 28 | KIND_CHANGELOG, |
|
29 | 29 | KIND_FILELOG, |
|
30 | 30 | KIND_MANIFESTLOG, |
|
31 | 31 | ) |
|
32 | 32 | from . import ( |
|
33 | 33 | changelog, |
|
34 | 34 | error, |
|
35 | 35 | filelog, |
|
36 | 36 | manifest, |
|
37 | 37 | policy, |
|
38 | 38 | pycompat, |
|
39 | 39 | revlog as revlogmod, |
|
40 | 40 | util, |
|
41 | 41 | vfs as vfsmod, |
|
42 | 42 | ) |
|
43 | 43 | from .utils import hashutil |
|
44 | 44 | |
|
45 | 45 | parsers = policy.importmod('parsers') |
|
46 | 46 | # how much bytes should be read from fncache in one read |
|
47 | 47 | # It is done to prevent loading large fncache files into memory |
|
48 | 48 | fncache_chunksize = 10**6 |
|
49 | 49 | |
|
50 | 50 | |
|
51 | 51 | def _match_tracked_entry(entry: "BaseStoreEntry", matcher): |
|
52 | 52 | """parses a fncache entry and returns whether the entry is tracking a path |
|
53 | 53 | matched by matcher or not. |
|
54 | 54 | |
|
55 | 55 | If matcher is None, returns True""" |
|
56 | 56 | |
|
57 | 57 | if matcher is None: |
|
58 | 58 | return True |
|
59 | 59 | |
|
60 | 60 | # TODO: make this safe for other entry types. Currently, the various |
|
61 | 61 | # store.data_entry generators only yield RevlogStoreEntry, so the |
|
62 | 62 | # attributes do exist on `entry`. |
|
63 | 63 | # pytype: disable=attribute-error |
|
64 | 64 | if entry.is_filelog: |
|
65 | 65 | return matcher(entry.target_id) |
|
66 | 66 | elif entry.is_manifestlog: |
|
67 | 67 | return matcher.visitdir(entry.target_id.rstrip(b'/')) |
|
68 | 68 | # pytype: enable=attribute-error |
|
69 | 69 | raise error.ProgrammingError(b"cannot process entry %r" % entry) |
|
70 | 70 | |
|
71 | 71 | |
|
72 | 72 | # This avoids a collision between a file named foo and a dir named |
|
73 | 73 | # foo.i or foo.d |
|
74 | 74 | def _encodedir(path): |
|
75 | 75 | """ |
|
76 | 76 | >>> _encodedir(b'data/foo.i') |
|
77 | 77 | 'data/foo.i' |
|
78 | 78 | >>> _encodedir(b'data/foo.i/bla.i') |
|
79 | 79 | 'data/foo.i.hg/bla.i' |
|
80 | 80 | >>> _encodedir(b'data/foo.i.hg/bla.i') |
|
81 | 81 | 'data/foo.i.hg.hg/bla.i' |
|
82 | 82 | >>> _encodedir(b'data/foo.i\\ndata/foo.i/bla.i\\ndata/foo.i.hg/bla.i\\n') |
|
83 | 83 | 'data/foo.i\\ndata/foo.i.hg/bla.i\\ndata/foo.i.hg.hg/bla.i\\n' |
|
84 | 84 | """ |
|
85 | 85 | return ( |
|
86 | 86 | path.replace(b".hg/", b".hg.hg/") |
|
87 | 87 | .replace(b".i/", b".i.hg/") |
|
88 | 88 | .replace(b".d/", b".d.hg/") |
|
89 | 89 | ) |
|
90 | 90 | |
|
91 | 91 | |
|
92 | 92 | encodedir = getattr(parsers, 'encodedir', _encodedir) |
|
93 | 93 | |
|
94 | 94 | |
|
95 | 95 | def decodedir(path): |
|
96 | 96 | """ |
|
97 | 97 | >>> decodedir(b'data/foo.i') |
|
98 | 98 | 'data/foo.i' |
|
99 | 99 | >>> decodedir(b'data/foo.i.hg/bla.i') |
|
100 | 100 | 'data/foo.i/bla.i' |
|
101 | 101 | >>> decodedir(b'data/foo.i.hg.hg/bla.i') |
|
102 | 102 | 'data/foo.i.hg/bla.i' |
|
103 | 103 | """ |
|
104 | 104 | if b".hg/" not in path: |
|
105 | 105 | return path |
|
106 | 106 | return ( |
|
107 | 107 | path.replace(b".d.hg/", b".d/") |
|
108 | 108 | .replace(b".i.hg/", b".i/") |
|
109 | 109 | .replace(b".hg.hg/", b".hg/") |
|
110 | 110 | ) |
|
111 | 111 | |
|
112 | 112 | |
|
113 | 113 | def _reserved(): |
|
114 | 114 | """characters that are problematic for filesystems |
|
115 | 115 | |
|
116 | 116 | * ascii escapes (0..31) |
|
117 | 117 | * ascii hi (126..255) |
|
118 | 118 | * windows specials |
|
119 | 119 | |
|
120 | 120 | these characters will be escaped by encodefunctions |
|
121 | 121 | """ |
|
122 | 122 | winreserved = [ord(x) for x in u'\\:*?"<>|'] |
|
123 | 123 | for x in range(32): |
|
124 | 124 | yield x |
|
125 | 125 | for x in range(126, 256): |
|
126 | 126 | yield x |
|
127 | 127 | for x in winreserved: |
|
128 | 128 | yield x |
|
129 | 129 | |
|
130 | 130 | |
|
131 | 131 | def _buildencodefun(): |
|
132 | 132 | """ |
|
133 | 133 | >>> enc, dec = _buildencodefun() |
|
134 | 134 | |
|
135 | 135 | >>> enc(b'nothing/special.txt') |
|
136 | 136 | 'nothing/special.txt' |
|
137 | 137 | >>> dec(b'nothing/special.txt') |
|
138 | 138 | 'nothing/special.txt' |
|
139 | 139 | |
|
140 | 140 | >>> enc(b'HELLO') |
|
141 | 141 | '_h_e_l_l_o' |
|
142 | 142 | >>> dec(b'_h_e_l_l_o') |
|
143 | 143 | 'HELLO' |
|
144 | 144 | |
|
145 | 145 | >>> enc(b'hello:world?') |
|
146 | 146 | 'hello~3aworld~3f' |
|
147 | 147 | >>> dec(b'hello~3aworld~3f') |
|
148 | 148 | 'hello:world?' |
|
149 | 149 | |
|
150 | 150 | >>> enc(b'the\\x07quick\\xADshot') |
|
151 | 151 | 'the~07quick~adshot' |
|
152 | 152 | >>> dec(b'the~07quick~adshot') |
|
153 | 153 | 'the\\x07quick\\xadshot' |
|
154 | 154 | """ |
|
155 | 155 | e = b'_' |
|
156 | 156 | xchr = pycompat.bytechr |
|
157 | 157 | asciistr = list(map(xchr, range(127))) |
|
158 | 158 | capitals = list(range(ord(b"A"), ord(b"Z") + 1)) |
|
159 | 159 | |
|
160 | 160 | cmap = {x: x for x in asciistr} |
|
161 | 161 | for x in _reserved(): |
|
162 | 162 | cmap[xchr(x)] = b"~%02x" % x |
|
163 | 163 | for x in capitals + [ord(e)]: |
|
164 | 164 | cmap[xchr(x)] = e + xchr(x).lower() |
|
165 | 165 | |
|
166 | 166 | dmap = {} |
|
167 | 167 | for k, v in cmap.items(): |
|
168 | 168 | dmap[v] = k |
|
169 | 169 | |
|
170 | 170 | def decode(s): |
|
171 | 171 | i = 0 |
|
172 | 172 | while i < len(s): |
|
173 | 173 | for l in range(1, 4): |
|
174 | 174 | try: |
|
175 | 175 | yield dmap[s[i : i + l]] |
|
176 | 176 | i += l |
|
177 | 177 | break |
|
178 | 178 | except KeyError: |
|
179 | 179 | pass |
|
180 | 180 | else: |
|
181 | 181 | raise KeyError |
|
182 | 182 | |
|
183 | 183 | return ( |
|
184 | 184 | lambda s: b''.join([cmap[s[c : c + 1]] for c in range(len(s))]), |
|
185 | 185 | lambda s: b''.join(list(decode(s))), |
|
186 | 186 | ) |
|
187 | 187 | |
|
188 | 188 | |
|
189 | 189 | _encodefname, _decodefname = _buildencodefun() |
|
190 | 190 | |
|
191 | 191 | |
|
192 | 192 | def encodefilename(s): |
|
193 | 193 | """ |
|
194 | 194 | >>> encodefilename(b'foo.i/bar.d/bla.hg/hi:world?/HELLO') |
|
195 | 195 | 'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o' |
|
196 | 196 | """ |
|
197 | 197 | return _encodefname(encodedir(s)) |
|
198 | 198 | |
|
199 | 199 | |
|
200 | 200 | def decodefilename(s): |
|
201 | 201 | """ |
|
202 | 202 | >>> decodefilename(b'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o') |
|
203 | 203 | 'foo.i/bar.d/bla.hg/hi:world?/HELLO' |
|
204 | 204 | """ |
|
205 | 205 | return decodedir(_decodefname(s)) |
|
206 | 206 | |
|
207 | 207 | |
|
208 | 208 | def _buildlowerencodefun(): |
|
209 | 209 | """ |
|
210 | 210 | >>> f = _buildlowerencodefun() |
|
211 | 211 | >>> f(b'nothing/special.txt') |
|
212 | 212 | 'nothing/special.txt' |
|
213 | 213 | >>> f(b'HELLO') |
|
214 | 214 | 'hello' |
|
215 | 215 | >>> f(b'hello:world?') |
|
216 | 216 | 'hello~3aworld~3f' |
|
217 | 217 | >>> f(b'the\\x07quick\\xADshot') |
|
218 | 218 | 'the~07quick~adshot' |
|
219 | 219 | """ |
|
220 | 220 | xchr = pycompat.bytechr |
|
221 | 221 | cmap = {xchr(x): xchr(x) for x in range(127)} |
|
222 | 222 | for x in _reserved(): |
|
223 | 223 | cmap[xchr(x)] = b"~%02x" % x |
|
224 | 224 | for x in range(ord(b"A"), ord(b"Z") + 1): |
|
225 | 225 | cmap[xchr(x)] = xchr(x).lower() |
|
226 | 226 | |
|
227 | 227 | def lowerencode(s): |
|
228 | 228 | return b"".join([cmap[c] for c in pycompat.iterbytestr(s)]) |
|
229 | 229 | |
|
230 | 230 | return lowerencode |
|
231 | 231 | |
|
232 | 232 | |
|
233 | 233 | lowerencode = getattr(parsers, 'lowerencode', None) or _buildlowerencodefun() |
|
234 | 234 | |
|
235 | 235 | # Windows reserved names: con, prn, aux, nul, com1..com9, lpt1..lpt9 |
|
236 | 236 | _winres3 = (b'aux', b'con', b'prn', b'nul') # length 3 |
|
237 | 237 | _winres4 = (b'com', b'lpt') # length 4 (with trailing 1..9) |
|
238 | 238 | |
|
239 | 239 | |
|
240 | 240 | def _auxencode(path, dotencode): |
|
241 | 241 | """ |
|
242 | 242 | Encodes filenames containing names reserved by Windows or which end in |
|
243 | 243 | period or space. Does not touch other single reserved characters c. |
|
244 | 244 | Specifically, c in '\\:*?"<>|' or ord(c) <= 31 are *not* encoded here. |
|
245 | 245 | Additionally encodes space or period at the beginning, if dotencode is |
|
246 | 246 | True. Parameter path is assumed to be all lowercase. |
|
247 | 247 | A segment only needs encoding if a reserved name appears as a |
|
248 | 248 | basename (e.g. "aux", "aux.foo"). A directory or file named "foo.aux" |
|
249 | 249 | doesn't need encoding. |
|
250 | 250 | |
|
251 | 251 | >>> s = b'.foo/aux.txt/txt.aux/con/prn/nul/foo.' |
|
252 | 252 | >>> _auxencode(s.split(b'/'), True) |
|
253 | 253 | ['~2efoo', 'au~78.txt', 'txt.aux', 'co~6e', 'pr~6e', 'nu~6c', 'foo~2e'] |
|
254 | 254 | >>> s = b'.com1com2/lpt9.lpt4.lpt1/conprn/com0/lpt0/foo.' |
|
255 | 255 | >>> _auxencode(s.split(b'/'), False) |
|
256 | 256 | ['.com1com2', 'lp~749.lpt4.lpt1', 'conprn', 'com0', 'lpt0', 'foo~2e'] |
|
257 | 257 | >>> _auxencode([b'foo. '], True) |
|
258 | 258 | ['foo.~20'] |
|
259 | 259 | >>> _auxencode([b' .foo'], True) |
|
260 | 260 | ['~20.foo'] |
|
261 | 261 | """ |
|
262 | 262 | for i, n in enumerate(path): |
|
263 | 263 | if not n: |
|
264 | 264 | continue |
|
265 | 265 | if dotencode and n[0] in b'. ': |
|
266 | 266 | n = b"~%02x" % ord(n[0:1]) + n[1:] |
|
267 | 267 | path[i] = n |
|
268 | 268 | else: |
|
269 | 269 | l = n.find(b'.') |
|
270 | 270 | if l == -1: |
|
271 | 271 | l = len(n) |
|
272 | 272 | if (l == 3 and n[:3] in _winres3) or ( |
|
273 | 273 | l == 4 |
|
274 | 274 | and n[3:4] <= b'9' |
|
275 | 275 | and n[3:4] >= b'1' |
|
276 | 276 | and n[:3] in _winres4 |
|
277 | 277 | ): |
|
278 | 278 | # encode third letter ('aux' -> 'au~78') |
|
279 | 279 | ec = b"~%02x" % ord(n[2:3]) |
|
280 | 280 | n = n[0:2] + ec + n[3:] |
|
281 | 281 | path[i] = n |
|
282 | 282 | if n[-1] in b'. ': |
|
283 | 283 | # encode last period or space ('foo...' -> 'foo..~2e') |
|
284 | 284 | path[i] = n[:-1] + b"~%02x" % ord(n[-1:]) |
|
285 | 285 | return path |
|
286 | 286 | |
|
287 | 287 | |
|
288 | 288 | _maxstorepathlen = 120 |
|
289 | 289 | _dirprefixlen = 8 |
|
290 | 290 | _maxshortdirslen = 8 * (_dirprefixlen + 1) - 4 |
|
291 | 291 | |
|
292 | 292 | |
|
293 | 293 | def _hashencode(path, dotencode): |
|
294 | 294 | digest = hex(hashutil.sha1(path).digest()) |
|
295 | 295 | le = lowerencode(path[5:]).split(b'/') # skips prefix 'data/' or 'meta/' |
|
296 | 296 | parts = _auxencode(le, dotencode) |
|
297 | 297 | basename = parts[-1] |
|
298 | 298 | _root, ext = os.path.splitext(basename) |
|
299 | 299 | sdirs = [] |
|
300 | 300 | sdirslen = 0 |
|
301 | 301 | for p in parts[:-1]: |
|
302 | 302 | d = p[:_dirprefixlen] |
|
303 | 303 | if d[-1] in b'. ': |
|
304 | 304 | # Windows can't access dirs ending in period or space |
|
305 | 305 | d = d[:-1] + b'_' |
|
306 | 306 | if sdirslen == 0: |
|
307 | 307 | t = len(d) |
|
308 | 308 | else: |
|
309 | 309 | t = sdirslen + 1 + len(d) |
|
310 | 310 | if t > _maxshortdirslen: |
|
311 | 311 | break |
|
312 | 312 | sdirs.append(d) |
|
313 | 313 | sdirslen = t |
|
314 | 314 | dirs = b'/'.join(sdirs) |
|
315 | 315 | if len(dirs) > 0: |
|
316 | 316 | dirs += b'/' |
|
317 | 317 | res = b'dh/' + dirs + digest + ext |
|
318 | 318 | spaceleft = _maxstorepathlen - len(res) |
|
319 | 319 | if spaceleft > 0: |
|
320 | 320 | filler = basename[:spaceleft] |
|
321 | 321 | res = b'dh/' + dirs + filler + digest + ext |
|
322 | 322 | return res |
|
323 | 323 | |
|
324 | 324 | |
|
325 | 325 | def _hybridencode(path, dotencode): |
|
326 | 326 | """encodes path with a length limit |
|
327 | 327 | |
|
328 | 328 | Encodes all paths that begin with 'data/', according to the following. |
|
329 | 329 | |
|
330 | 330 | Default encoding (reversible): |
|
331 | 331 | |
|
332 | 332 | Encodes all uppercase letters 'X' as '_x'. All reserved or illegal |
|
333 | 333 | characters are encoded as '~xx', where xx is the two digit hex code |
|
334 | 334 | of the character (see encodefilename). |
|
335 | 335 | Relevant path components consisting of Windows reserved filenames are |
|
336 | 336 | masked by encoding the third character ('aux' -> 'au~78', see _auxencode). |
|
337 | 337 | |
|
338 | 338 | Hashed encoding (not reversible): |
|
339 | 339 | |
|
340 | 340 | If the default-encoded path is longer than _maxstorepathlen, a |
|
341 | 341 | non-reversible hybrid hashing of the path is done instead. |
|
342 | 342 | This encoding uses up to _dirprefixlen characters of all directory |
|
343 | 343 | levels of the lowerencoded path, but not more levels than can fit into |
|
344 | 344 | _maxshortdirslen. |
|
345 | 345 | Then follows the filler followed by the sha digest of the full path. |
|
346 | 346 | The filler is the beginning of the basename of the lowerencoded path |
|
347 | 347 | (the basename is everything after the last path separator). The filler |
|
348 | 348 | is as long as possible, filling in characters from the basename until |
|
349 | 349 | the encoded path has _maxstorepathlen characters (or all chars of the |
|
350 | 350 | basename have been taken). |
|
351 | 351 | The extension (e.g. '.i' or '.d') is preserved. |
|
352 | 352 | |
|
353 | 353 | The string 'data/' at the beginning is replaced with 'dh/', if the hashed |
|
354 | 354 | encoding was used. |
|
355 | 355 | """ |
|
356 | 356 | path = encodedir(path) |
|
357 | 357 | ef = _encodefname(path).split(b'/') |
|
358 | 358 | res = b'/'.join(_auxencode(ef, dotencode)) |
|
359 | 359 | if len(res) > _maxstorepathlen: |
|
360 | 360 | res = _hashencode(path, dotencode) |
|
361 | 361 | return res |
|
362 | 362 | |
|
363 | 363 | |
|
364 | 364 | def _pathencode(path): |
|
365 | 365 | de = encodedir(path) |
|
366 | 366 | if len(path) > _maxstorepathlen: |
|
367 | 367 | return _hashencode(de, True) |
|
368 | 368 | ef = _encodefname(de).split(b'/') |
|
369 | 369 | res = b'/'.join(_auxencode(ef, True)) |
|
370 | 370 | if len(res) > _maxstorepathlen: |
|
371 | 371 | return _hashencode(de, True) |
|
372 | 372 | return res |
|
373 | 373 | |
|
374 | 374 | |
|
375 | 375 | _pathencode = getattr(parsers, 'pathencode', _pathencode) |
|
376 | 376 | |
|
377 | 377 | |
|
378 | 378 | def _plainhybridencode(f): |
|
379 | 379 | return _hybridencode(f, False) |
|
380 | 380 | |
|
381 | 381 | |
|
382 | 382 | def _calcmode(vfs): |
|
383 | 383 | try: |
|
384 | 384 | # files in .hg/ will be created using this mode |
|
385 | 385 | mode = vfs.stat().st_mode |
|
386 | 386 | # avoid some useless chmods |
|
387 | 387 | if (0o777 & ~util.umask) == (0o777 & mode): |
|
388 | 388 | mode = None |
|
389 | 389 | except OSError: |
|
390 | 390 | mode = None |
|
391 | 391 | return mode |
|
392 | 392 | |
|
393 | 393 | |
|
394 | 394 | _data = [ |
|
395 | 395 | b'bookmarks', |
|
396 | 396 | b'narrowspec', |
|
397 | 397 | b'data', |
|
398 | 398 | b'meta', |
|
399 | 399 | b'00manifest.d', |
|
400 | 400 | b'00manifest.i', |
|
401 | 401 | b'00changelog.d', |
|
402 | 402 | b'00changelog.i', |
|
403 | 403 | b'phaseroots', |
|
404 | 404 | b'obsstore', |
|
405 | 405 | b'requires', |
|
406 | 406 | ] |
|
407 | 407 | |
|
408 | 408 | REVLOG_FILES_EXT = ( |
|
409 | 409 | b'.i', |
|
410 | 410 | b'.idx', |
|
411 | 411 | b'.d', |
|
412 | 412 | b'.dat', |
|
413 | 413 | b'.n', |
|
414 | 414 | b'.nd', |
|
415 | 415 | b'.sda', |
|
416 | 416 | ) |
|
417 | 417 | # file extension that also use a `-SOMELONGIDHASH.ext` form |
|
418 | 418 | REVLOG_FILES_LONG_EXT = ( |
|
419 | 419 | b'.nd', |
|
420 | 420 | b'.idx', |
|
421 | 421 | b'.dat', |
|
422 | 422 | b'.sda', |
|
423 | 423 | ) |
|
424 | 424 | # files that are "volatile" and might change between listing and streaming |
|
425 | 425 | # |
|
426 | 426 | # note: the ".nd" file are nodemap data and won't "change" but they might be |
|
427 | 427 | # deleted. |
|
428 | 428 | REVLOG_FILES_VOLATILE_EXT = (b'.n', b'.nd') |
|
429 | 429 | |
|
430 | 430 | # some exception to the above matching |
|
431 | 431 | # |
|
432 | 432 | # XXX This is currently not in use because of issue6542 |
|
433 | 433 | EXCLUDED = re.compile(br'.*undo\.[^/]+\.(nd?|i)$') |
|
434 | 434 | |
|
435 | 435 | |
|
436 | 436 | def is_revlog(f, kind, st): |
|
437 | 437 | if kind != stat.S_IFREG: |
|
438 | 438 | return False |
|
439 | 439 | if f.endswith(REVLOG_FILES_EXT): |
|
440 | 440 | return True |
|
441 | 441 | return False |
|
442 | 442 | |
|
443 | 443 | |
|
444 | 444 | def is_revlog_file(f): |
|
445 | 445 | if f.endswith(REVLOG_FILES_EXT): |
|
446 | 446 | return True |
|
447 | 447 | return False |
|
448 | 448 | |
|
449 | 449 | |
|
450 | 450 | @attr.s(slots=True) |
|
451 | 451 | class StoreFile: |
|
452 | 452 | """a file matching a store entry""" |
|
453 | 453 | |
|
454 | 454 | unencoded_path = attr.ib() |
|
455 | 455 | _file_size = attr.ib(default=None) |
|
456 | 456 | is_volatile = attr.ib(default=False) |
|
457 | 457 | |
|
458 | 458 | def file_size(self, vfs): |
|
459 | 459 | if self._file_size is None: |
|
460 | 460 | if vfs is None: |
|
461 | 461 | msg = b"calling vfs-less file_size without prior call: %s" |
|
462 | 462 | msg %= self.unencoded_path |
|
463 | 463 | raise error.ProgrammingError(msg) |
|
464 | 464 | try: |
|
465 | 465 | self._file_size = vfs.stat(self.unencoded_path).st_size |
|
466 | 466 | except FileNotFoundError: |
|
467 | 467 | self._file_size = 0 |
|
468 | 468 | return self._file_size |
|
469 | 469 | |
|
470 | 470 | @property |
|
471 | 471 | def has_size(self): |
|
472 | 472 | return self._file_size is not None |
|
473 | 473 | |
|
474 | 474 | def get_stream(self, vfs, copies): |
|
475 | 475 | """return data "stream" information for this file |
|
476 | 476 | |
|
477 | 477 | (unencoded_file_path, content_iterator, content_size) |
|
478 | 478 | """ |
|
479 | 479 | size = self.file_size(None) |
|
480 | 480 | |
|
481 | 481 | def get_stream(): |
|
482 | 482 | actual_path = copies[vfs.join(self.unencoded_path)] |
|
483 | 483 | with open(actual_path, 'rb') as fp: |
|
484 | 484 | yield None # ready to stream |
|
485 | 485 | if size <= 65536: |
|
486 | 486 | yield fp.read(size) |
|
487 | 487 | else: |
|
488 | 488 | yield from util.filechunkiter(fp, limit=size) |
|
489 | 489 | |
|
490 | 490 | s = get_stream() |
|
491 | 491 | next(s) |
|
492 | 492 | return (self.unencoded_path, s, size) |
|
493 | 493 | |
|
494 | 494 | |
|
495 | 495 | @attr.s(slots=True, init=False) |
|
496 | 496 | class BaseStoreEntry: |
|
497 | 497 | """An entry in the store |
|
498 | 498 | |
|
499 | 499 | This is returned by `store.walk` and represent some data in the store.""" |
|
500 | 500 | |
|
501 | 501 | maybe_volatile = True |
|
502 | 502 | |
|
503 | 503 | def files(self) -> List[StoreFile]: |
|
504 | 504 | raise NotImplementedError |
|
505 | 505 | |
|
506 | 506 | def get_streams( |
|
507 | 507 | self, |
|
508 | 508 | repo=None, |
|
509 | 509 | vfs=None, |
|
510 | 510 | copies=None, |
|
511 | 511 | max_changeset=None, |
|
512 | 512 | preserve_file_count=False, |
|
513 | 513 | ): |
|
514 | 514 | """return a list of data stream associated to files for this entry |
|
515 | 515 | |
|
516 | 516 | return [(unencoded_file_path, content_iterator, content_size), β¦] |
|
517 | 517 | """ |
|
518 | 518 | assert vfs is not None |
|
519 | 519 | return [f.get_stream(vfs, copies) for f in self.files()] |
|
520 | 520 | |
|
521 | 521 | |
|
522 | 522 | @attr.s(slots=True, init=False) |
|
523 | 523 | class SimpleStoreEntry(BaseStoreEntry): |
|
524 | 524 | """A generic entry in the store""" |
|
525 | 525 | |
|
526 | 526 | is_revlog = False |
|
527 | 527 | |
|
528 | 528 | maybe_volatile = attr.ib() |
|
529 | 529 | _entry_path = attr.ib() |
|
530 | 530 | _is_volatile = attr.ib(default=False) |
|
531 | 531 | _file_size = attr.ib(default=None) |
|
532 | 532 | _files = attr.ib(default=None) |
|
533 | 533 | |
|
534 | 534 | def __init__( |
|
535 | 535 | self, |
|
536 | 536 | entry_path, |
|
537 | 537 | is_volatile=False, |
|
538 | 538 | file_size=None, |
|
539 | 539 | ): |
|
540 | 540 | super().__init__() |
|
541 | 541 | self._entry_path = entry_path |
|
542 | 542 | self._is_volatile = is_volatile |
|
543 | 543 | self._file_size = file_size |
|
544 | 544 | self._files = None |
|
545 | 545 | self.maybe_volatile = is_volatile |
|
546 | 546 | |
|
547 | 547 | def files(self) -> List[StoreFile]: |
|
548 | 548 | if self._files is None: |
|
549 | 549 | self._files = [ |
|
550 | 550 | StoreFile( |
|
551 | 551 | unencoded_path=self._entry_path, |
|
552 | 552 | file_size=self._file_size, |
|
553 | 553 | is_volatile=self._is_volatile, |
|
554 | 554 | ) |
|
555 | 555 | ] |
|
556 | 556 | return self._files |
|
557 | 557 | |
|
558 | 558 | |
|
559 | 559 | @attr.s(slots=True, init=False) |
|
560 | 560 | class RevlogStoreEntry(BaseStoreEntry): |
|
561 | 561 | """A revlog entry in the store""" |
|
562 | 562 | |
|
563 | 563 | is_revlog = True |
|
564 | 564 | |
|
565 | 565 | revlog_type = attr.ib(default=None) |
|
566 | 566 | target_id = attr.ib(default=None) |
|
567 | 567 | maybe_volatile = attr.ib(default=True) |
|
568 | 568 | _path_prefix = attr.ib(default=None) |
|
569 | 569 | _details = attr.ib(default=None) |
|
570 | 570 | _files = attr.ib(default=None) |
|
571 | 571 | |
|
572 | 572 | def __init__( |
|
573 | 573 | self, |
|
574 | 574 | revlog_type, |
|
575 | 575 | path_prefix, |
|
576 | 576 | target_id, |
|
577 | 577 | details, |
|
578 | 578 | ): |
|
579 | 579 | super().__init__() |
|
580 | 580 | self.revlog_type = revlog_type |
|
581 | 581 | self.target_id = target_id |
|
582 | 582 | self._path_prefix = path_prefix |
|
583 | 583 | assert b'.i' in details, (path_prefix, details) |
|
584 | 584 | for ext in details: |
|
585 | 585 | if ext.endswith(REVLOG_FILES_VOLATILE_EXT): |
|
586 | 586 | self.maybe_volatile = True |
|
587 | 587 | break |
|
588 | 588 | else: |
|
589 | 589 | self.maybe_volatile = False |
|
590 | 590 | self._details = details |
|
591 | 591 | self._files = None |
|
592 | 592 | |
|
593 | 593 | @property |
|
594 | 594 | def is_changelog(self): |
|
595 | 595 | return self.revlog_type == KIND_CHANGELOG |
|
596 | 596 | |
|
597 | 597 | @property |
|
598 | 598 | def is_manifestlog(self): |
|
599 | 599 | return self.revlog_type == KIND_MANIFESTLOG |
|
600 | 600 | |
|
601 | 601 | @property |
|
602 | 602 | def is_filelog(self): |
|
603 | 603 | return self.revlog_type == KIND_FILELOG |
|
604 | 604 | |
|
605 | 605 | def main_file_path(self): |
|
606 | 606 | """unencoded path of the main revlog file""" |
|
607 | 607 | return self._path_prefix + b'.i' |
|
608 | 608 | |
|
609 | 609 | def files(self) -> List[StoreFile]: |
|
610 | 610 | if self._files is None: |
|
611 | 611 | self._files = [] |
|
612 | 612 | for ext in sorted(self._details, key=_ext_key): |
|
613 | 613 | path = self._path_prefix + ext |
|
614 | 614 | file_size = self._details[ext] |
|
615 | 615 | # files that are "volatile" and might change between |
|
616 | 616 | # listing and streaming |
|
617 | 617 | # |
|
618 | 618 | # note: the ".nd" file are nodemap data and won't "change" |
|
619 | 619 | # but they might be deleted. |
|
620 | 620 | volatile = ext.endswith(REVLOG_FILES_VOLATILE_EXT) |
|
621 | 621 | f = StoreFile(path, file_size, volatile) |
|
622 | 622 | self._files.append(f) |
|
623 | 623 | return self._files |
|
624 | 624 | |
|
625 | 625 | def get_streams( |
|
626 | 626 | self, |
|
627 | 627 | repo=None, |
|
628 | 628 | vfs=None, |
|
629 | 629 | copies=None, |
|
630 | 630 | max_changeset=None, |
|
631 | 631 | preserve_file_count=False, |
|
632 | 632 | ): |
|
633 | 633 | pre_sized = all(f.has_size for f in self.files()) |
|
634 | 634 | if pre_sized and ( |
|
635 | 635 | repo is None |
|
636 | 636 | or max_changeset is None |
|
637 | 637 | # This use revlog-v2, ignore for now |
|
638 | 638 | or any(k.endswith(b'.idx') for k in self._details.keys()) |
|
639 | 639 | # This is not inline, no race expected |
|
640 | 640 | or b'.d' in self._details |
|
641 | 641 | ): |
|
642 | 642 | return super().get_streams( |
|
643 | 643 | repo=repo, |
|
644 | 644 | vfs=vfs, |
|
645 | 645 | copies=copies, |
|
646 | 646 | max_changeset=max_changeset, |
|
647 | 647 | preserve_file_count=preserve_file_count, |
|
648 | 648 | ) |
|
649 | 649 | elif not preserve_file_count: |
|
650 | 650 | stream = [ |
|
651 | 651 | f.get_stream(vfs, copies) |
|
652 | 652 | for f in self.files() |
|
653 | 653 | if not f.unencoded_path.endswith((b'.i', b'.d')) |
|
654 | 654 | ] |
|
655 | 655 | rl = self.get_revlog_instance(repo).get_revlog() |
|
656 | 656 | rl_stream = rl.get_streams(max_changeset) |
|
657 | 657 | stream.extend(rl_stream) |
|
658 | 658 | return stream |
|
659 | 659 | |
|
660 | 660 | name_to_size = {} |
|
661 | 661 | for f in self.files(): |
|
662 | 662 | name_to_size[f.unencoded_path] = f.file_size(None) |
|
663 | 663 | |
|
664 | 664 | stream = [ |
|
665 | 665 | f.get_stream(vfs, copies) |
|
666 | 666 | for f in self.files() |
|
667 | 667 | if not f.unencoded_path.endswith(b'.i') |
|
668 | 668 | ] |
|
669 | 669 | |
|
670 | 670 | index_path = self._path_prefix + b'.i' |
|
671 | 671 | |
|
672 | 672 | index_file = None |
|
673 | 673 | try: |
|
674 | 674 | index_file = vfs(index_path) |
|
675 | 675 | header = index_file.read(INDEX_HEADER.size) |
|
676 | 676 | if revlogmod.revlog.is_inline_index(header): |
|
677 | 677 | size = name_to_size[index_path] |
|
678 | 678 | |
|
679 | 679 | # no split underneath, just return the stream |
|
680 | 680 | def get_stream(): |
|
681 | 681 | fp = index_file |
|
682 | 682 | try: |
|
683 | 683 | fp.seek(0) |
|
684 | 684 | yield None |
|
685 | 685 | if size <= 65536: |
|
686 | 686 | yield fp.read(size) |
|
687 | 687 | else: |
|
688 | 688 | yield from util.filechunkiter(fp, limit=size) |
|
689 | 689 | finally: |
|
690 | 690 | fp.close() |
|
691 | 691 | |
|
692 | 692 | s = get_stream() |
|
693 | 693 | next(s) |
|
694 | 694 | index_file = None |
|
695 | 695 | stream.append((index_path, s, size)) |
|
696 | 696 | else: |
|
697 | 697 | rl = self.get_revlog_instance(repo).get_revlog() |
|
698 | 698 | rl_stream = rl.get_streams(max_changeset, force_inline=True) |
|
699 | 699 | for name, s, size in rl_stream: |
|
700 | 700 | if name_to_size.get(name, 0) != size: |
|
701 | 701 | msg = _(b"expected %d bytes but %d provided for %s") |
|
702 | 702 | msg %= name_to_size.get(name, 0), size, name |
|
703 | 703 | raise error.Abort(msg) |
|
704 | 704 | stream.extend(rl_stream) |
|
705 | 705 | finally: |
|
706 | 706 | if index_file is not None: |
|
707 | 707 | index_file.close() |
|
708 | 708 | |
|
709 | 709 | files = self.files() |
|
710 | 710 | assert len(stream) == len(files), ( |
|
711 | 711 | stream, |
|
712 | 712 | files, |
|
713 | 713 | self._path_prefix, |
|
714 | 714 | self.target_id, |
|
715 | 715 | ) |
|
716 | 716 | return stream |
|
717 | 717 | |
|
718 | 718 | def get_revlog_instance(self, repo): |
|
719 | 719 | """Obtain a revlog instance from this store entry |
|
720 | 720 | |
|
721 | 721 | An instance of the appropriate class is returned. |
|
722 | 722 | """ |
|
723 | 723 | if self.is_changelog: |
|
724 | 724 | return changelog.changelog(repo.svfs) |
|
725 | 725 | elif self.is_manifestlog: |
|
726 | 726 | mandir = self.target_id |
|
727 | 727 | return manifest.manifestrevlog( |
|
728 | 728 | repo.nodeconstants, repo.svfs, tree=mandir |
|
729 | 729 | ) |
|
730 | 730 | else: |
|
731 | 731 | return filelog.filelog(repo.svfs, self.target_id) |
|
732 | 732 | |
|
733 | 733 | |
|
734 | 734 | def _gather_revlog(files_data): |
|
735 | 735 | """group files per revlog prefix |
|
736 | 736 | |
|
737 | 737 | The returns a two level nested dict. The top level key is the revlog prefix |
|
738 | 738 | without extension, the second level is all the file "suffix" that were |
|
739 | 739 | seen for this revlog and arbitrary file data as value. |
|
740 | 740 | """ |
|
741 | 741 | revlogs = collections.defaultdict(dict) |
|
742 | 742 | for u, value in files_data: |
|
743 | 743 | name, ext = _split_revlog_ext(u) |
|
744 | 744 | revlogs[name][ext] = value |
|
745 | 745 | return sorted(revlogs.items()) |
|
746 | 746 | |
|
747 | 747 | |
|
748 | 748 | def _split_revlog_ext(filename): |
|
749 | 749 | """split the revlog file prefix from the variable extension""" |
|
750 | 750 | if filename.endswith(REVLOG_FILES_LONG_EXT): |
|
751 | 751 | char = b'-' |
|
752 | 752 | else: |
|
753 | 753 | char = b'.' |
|
754 | 754 | idx = filename.rfind(char) |
|
755 | 755 | return filename[:idx], filename[idx:] |
|
756 | 756 | |
|
757 | 757 | |
|
758 | 758 | def _ext_key(ext): |
|
759 | 759 | """a key to order revlog suffix |
|
760 | 760 | |
|
761 | 761 | important to issue .i after other entry.""" |
|
762 | 762 | # the only important part of this order is to keep the `.i` last. |
|
763 | 763 | if ext.endswith(b'.n'): |
|
764 | 764 | return (0, ext) |
|
765 | 765 | elif ext.endswith(b'.nd'): |
|
766 | 766 | return (10, ext) |
|
767 | 767 | elif ext.endswith(b'.d'): |
|
768 | 768 | return (20, ext) |
|
769 | 769 | elif ext.endswith(b'.i'): |
|
770 | 770 | return (50, ext) |
|
771 | 771 | else: |
|
772 | 772 | return (40, ext) |
|
773 | 773 | |
|
774 | 774 | |
|
775 | 775 | class basicstore: |
|
776 | 776 | '''base class for local repository stores''' |
|
777 | 777 | |
|
778 | 778 | def __init__(self, path, vfstype): |
|
779 | 779 | vfs = vfstype(path) |
|
780 | 780 | self.path = vfs.base |
|
781 | 781 | self.createmode = _calcmode(vfs) |
|
782 | 782 | vfs.createmode = self.createmode |
|
783 | 783 | self.rawvfs = vfs |
|
784 | 784 | self.vfs = vfsmod.filtervfs(vfs, encodedir) |
|
785 | 785 | self.opener = self.vfs |
|
786 | 786 | |
|
787 | 787 | def join(self, f): |
|
788 | 788 | return self.path + b'/' + encodedir(f) |
|
789 | 789 | |
|
790 | 790 | def _walk(self, relpath, recurse, undecodable=None): |
|
791 | 791 | '''yields (revlog_type, unencoded, size)''' |
|
792 | 792 | path = self.path |
|
793 | 793 | if relpath: |
|
794 | 794 | path += b'/' + relpath |
|
795 | 795 | striplen = len(self.path) + 1 |
|
796 | 796 | l = [] |
|
797 | 797 | if self.rawvfs.isdir(path): |
|
798 | 798 | visit = [path] |
|
799 | 799 | readdir = self.rawvfs.readdir |
|
800 | 800 | while visit: |
|
801 | 801 | p = visit.pop() |
|
802 | 802 | for f, kind, st in readdir(p, stat=True): |
|
803 | 803 | fp = p + b'/' + f |
|
804 | 804 | if is_revlog(f, kind, st): |
|
805 | 805 | n = util.pconvert(fp[striplen:]) |
|
806 | 806 | l.append((decodedir(n), st.st_size)) |
|
807 | 807 | elif kind == stat.S_IFDIR and recurse: |
|
808 | 808 | visit.append(fp) |
|
809 | 809 | |
|
810 | 810 | l.sort() |
|
811 | 811 | return l |
|
812 | 812 | |
|
813 | 813 | def changelog(self, trypending, concurrencychecker=None): |
|
814 | 814 | return changelog.changelog( |
|
815 | 815 | self.vfs, |
|
816 | 816 | trypending=trypending, |
|
817 | 817 | concurrencychecker=concurrencychecker, |
|
818 | 818 | ) |
|
819 | 819 | |
|
820 | def manifestlog(self, repo, storenarrowmatch): | |
|
820 | def manifestlog(self, repo, storenarrowmatch) -> manifest.ManifestLog: | |
|
821 | 821 | rootstore = manifest.manifestrevlog(repo.nodeconstants, self.vfs) |
|
822 | 822 | return manifest.manifestlog(self.vfs, repo, rootstore, storenarrowmatch) |
|
823 | 823 | |
|
824 | 824 | def data_entries( |
|
825 | 825 | self, matcher=None, undecodable=None |
|
826 | 826 | ) -> Generator[BaseStoreEntry, None, None]: |
|
827 | 827 | """Like walk, but excluding the changelog and root manifest. |
|
828 | 828 | |
|
829 | 829 | When [undecodable] is None, revlogs names that can't be |
|
830 | 830 | decoded cause an exception. When it is provided, it should |
|
831 | 831 | be a list and the filenames that can't be decoded are added |
|
832 | 832 | to it instead. This is very rarely needed.""" |
|
833 | 833 | dirs = [ |
|
834 | 834 | (b'data', KIND_FILELOG, False), |
|
835 | 835 | (b'meta', KIND_MANIFESTLOG, True), |
|
836 | 836 | ] |
|
837 | 837 | for base_dir, rl_type, strip_filename in dirs: |
|
838 | 838 | files = self._walk(base_dir, True, undecodable=undecodable) |
|
839 | 839 | for revlog, details in _gather_revlog(files): |
|
840 | 840 | revlog_target_id = revlog.split(b'/', 1)[1] |
|
841 | 841 | if strip_filename and b'/' in revlog: |
|
842 | 842 | revlog_target_id = revlog_target_id.rsplit(b'/', 1)[0] |
|
843 | 843 | revlog_target_id += b'/' |
|
844 | 844 | yield RevlogStoreEntry( |
|
845 | 845 | path_prefix=revlog, |
|
846 | 846 | revlog_type=rl_type, |
|
847 | 847 | target_id=revlog_target_id, |
|
848 | 848 | details=details, |
|
849 | 849 | ) |
|
850 | 850 | |
|
851 | 851 | def top_entries( |
|
852 | 852 | self, phase=False, obsolescence=False |
|
853 | 853 | ) -> Generator[BaseStoreEntry, None, None]: |
|
854 | 854 | if phase and self.vfs.exists(b'phaseroots'): |
|
855 | 855 | yield SimpleStoreEntry( |
|
856 | 856 | entry_path=b'phaseroots', |
|
857 | 857 | is_volatile=True, |
|
858 | 858 | ) |
|
859 | 859 | |
|
860 | 860 | if obsolescence and self.vfs.exists(b'obsstore'): |
|
861 | 861 | # XXX if we had the file size it could be non-volatile |
|
862 | 862 | yield SimpleStoreEntry( |
|
863 | 863 | entry_path=b'obsstore', |
|
864 | 864 | is_volatile=True, |
|
865 | 865 | ) |
|
866 | 866 | |
|
867 | 867 | files = reversed(self._walk(b'', False)) |
|
868 | 868 | |
|
869 | 869 | changelogs = collections.defaultdict(dict) |
|
870 | 870 | manifestlogs = collections.defaultdict(dict) |
|
871 | 871 | |
|
872 | 872 | for u, s in files: |
|
873 | 873 | if u.startswith(b'00changelog'): |
|
874 | 874 | name, ext = _split_revlog_ext(u) |
|
875 | 875 | changelogs[name][ext] = s |
|
876 | 876 | elif u.startswith(b'00manifest'): |
|
877 | 877 | name, ext = _split_revlog_ext(u) |
|
878 | 878 | manifestlogs[name][ext] = s |
|
879 | 879 | else: |
|
880 | 880 | yield SimpleStoreEntry( |
|
881 | 881 | entry_path=u, |
|
882 | 882 | is_volatile=False, |
|
883 | 883 | file_size=s, |
|
884 | 884 | ) |
|
885 | 885 | # yield manifest before changelog |
|
886 | 886 | top_rl = [ |
|
887 | 887 | (manifestlogs, KIND_MANIFESTLOG), |
|
888 | 888 | (changelogs, KIND_CHANGELOG), |
|
889 | 889 | ] |
|
890 | 890 | assert len(manifestlogs) <= 1 |
|
891 | 891 | assert len(changelogs) <= 1 |
|
892 | 892 | for data, revlog_type in top_rl: |
|
893 | 893 | for revlog, details in sorted(data.items()): |
|
894 | 894 | yield RevlogStoreEntry( |
|
895 | 895 | path_prefix=revlog, |
|
896 | 896 | revlog_type=revlog_type, |
|
897 | 897 | target_id=b'', |
|
898 | 898 | details=details, |
|
899 | 899 | ) |
|
900 | 900 | |
|
901 | 901 | def walk( |
|
902 | 902 | self, matcher=None, phase=False, obsolescence=False |
|
903 | 903 | ) -> Generator[BaseStoreEntry, None, None]: |
|
904 | 904 | """return files related to data storage (ie: revlogs) |
|
905 | 905 | |
|
906 | 906 | yields instance from BaseStoreEntry subclasses |
|
907 | 907 | |
|
908 | 908 | if a matcher is passed, storage files of only those tracked paths |
|
909 | 909 | are passed with matches the matcher |
|
910 | 910 | """ |
|
911 | 911 | # yield data files first |
|
912 | 912 | for x in self.data_entries(matcher): |
|
913 | 913 | yield x |
|
914 | 914 | for x in self.top_entries(phase=phase, obsolescence=obsolescence): |
|
915 | 915 | yield x |
|
916 | 916 | |
|
917 | 917 | def copylist(self): |
|
918 | 918 | return _data |
|
919 | 919 | |
|
920 | 920 | def write(self, tr): |
|
921 | 921 | pass |
|
922 | 922 | |
|
923 | 923 | def invalidatecaches(self): |
|
924 | 924 | pass |
|
925 | 925 | |
|
926 | 926 | def markremoved(self, fn): |
|
927 | 927 | pass |
|
928 | 928 | |
|
929 | 929 | def __contains__(self, path): |
|
930 | 930 | '''Checks if the store contains path''' |
|
931 | 931 | path = b"/".join((b"data", path)) |
|
932 | 932 | # file? |
|
933 | 933 | if self.vfs.exists(path + b".i"): |
|
934 | 934 | return True |
|
935 | 935 | # dir? |
|
936 | 936 | if not path.endswith(b"/"): |
|
937 | 937 | path = path + b"/" |
|
938 | 938 | return self.vfs.exists(path) |
|
939 | 939 | |
|
940 | 940 | |
|
941 | 941 | class encodedstore(basicstore): |
|
942 | 942 | def __init__(self, path, vfstype): |
|
943 | 943 | vfs = vfstype(path + b'/store') |
|
944 | 944 | self.path = vfs.base |
|
945 | 945 | self.createmode = _calcmode(vfs) |
|
946 | 946 | vfs.createmode = self.createmode |
|
947 | 947 | self.rawvfs = vfs |
|
948 | 948 | self.vfs = vfsmod.filtervfs(vfs, encodefilename) |
|
949 | 949 | self.opener = self.vfs |
|
950 | 950 | |
|
951 | 951 | def _walk(self, relpath, recurse, undecodable=None): |
|
952 | 952 | old = super()._walk(relpath, recurse) |
|
953 | 953 | new = [] |
|
954 | 954 | for f1, value in old: |
|
955 | 955 | try: |
|
956 | 956 | f2 = decodefilename(f1) |
|
957 | 957 | except KeyError: |
|
958 | 958 | if undecodable is None: |
|
959 | 959 | msg = _(b'undecodable revlog name %s') % f1 |
|
960 | 960 | raise error.StorageError(msg) |
|
961 | 961 | else: |
|
962 | 962 | undecodable.append(f1) |
|
963 | 963 | continue |
|
964 | 964 | new.append((f2, value)) |
|
965 | 965 | return new |
|
966 | 966 | |
|
967 | 967 | def data_entries( |
|
968 | 968 | self, matcher=None, undecodable=None |
|
969 | 969 | ) -> Generator[BaseStoreEntry, None, None]: |
|
970 | 970 | entries = super(encodedstore, self).data_entries( |
|
971 | 971 | undecodable=undecodable |
|
972 | 972 | ) |
|
973 | 973 | for entry in entries: |
|
974 | 974 | if _match_tracked_entry(entry, matcher): |
|
975 | 975 | yield entry |
|
976 | 976 | |
|
977 | 977 | def join(self, f): |
|
978 | 978 | return self.path + b'/' + encodefilename(f) |
|
979 | 979 | |
|
980 | 980 | def copylist(self): |
|
981 | 981 | return [b'requires', b'00changelog.i'] + [b'store/' + f for f in _data] |
|
982 | 982 | |
|
983 | 983 | |
|
984 | 984 | class fncache: |
|
985 | 985 | # the filename used to be partially encoded |
|
986 | 986 | # hence the encodedir/decodedir dance |
|
987 | 987 | def __init__(self, vfs): |
|
988 | 988 | self.vfs = vfs |
|
989 | 989 | self._ignores = set() |
|
990 | 990 | self.entries = None |
|
991 | 991 | self._dirty = False |
|
992 | 992 | # set of new additions to fncache |
|
993 | 993 | self.addls = set() |
|
994 | 994 | |
|
995 | 995 | def ensureloaded(self, warn=None): |
|
996 | 996 | """read the fncache file if not already read. |
|
997 | 997 | |
|
998 | 998 | If the file on disk is corrupted, raise. If warn is provided, |
|
999 | 999 | warn and keep going instead.""" |
|
1000 | 1000 | if self.entries is None: |
|
1001 | 1001 | self._load(warn) |
|
1002 | 1002 | |
|
1003 | 1003 | def _load(self, warn=None): |
|
1004 | 1004 | '''fill the entries from the fncache file''' |
|
1005 | 1005 | self._dirty = False |
|
1006 | 1006 | try: |
|
1007 | 1007 | fp = self.vfs(b'fncache', mode=b'rb') |
|
1008 | 1008 | except IOError: |
|
1009 | 1009 | # skip nonexistent file |
|
1010 | 1010 | self.entries = set() |
|
1011 | 1011 | return |
|
1012 | 1012 | |
|
1013 | 1013 | self.entries = set() |
|
1014 | 1014 | chunk = b'' |
|
1015 | 1015 | for c in iter(functools.partial(fp.read, fncache_chunksize), b''): |
|
1016 | 1016 | chunk += c |
|
1017 | 1017 | try: |
|
1018 | 1018 | p = chunk.rindex(b'\n') |
|
1019 | 1019 | self.entries.update(decodedir(chunk[: p + 1]).splitlines()) |
|
1020 | 1020 | chunk = chunk[p + 1 :] |
|
1021 | 1021 | except ValueError: |
|
1022 | 1022 | # substring '\n' not found, maybe the entry is bigger than the |
|
1023 | 1023 | # chunksize, so let's keep iterating |
|
1024 | 1024 | pass |
|
1025 | 1025 | |
|
1026 | 1026 | if chunk: |
|
1027 | 1027 | msg = _(b"fncache does not ends with a newline") |
|
1028 | 1028 | if warn: |
|
1029 | 1029 | warn(msg + b'\n') |
|
1030 | 1030 | else: |
|
1031 | 1031 | raise error.Abort( |
|
1032 | 1032 | msg, |
|
1033 | 1033 | hint=_( |
|
1034 | 1034 | b"use 'hg debugrebuildfncache' to " |
|
1035 | 1035 | b"rebuild the fncache" |
|
1036 | 1036 | ), |
|
1037 | 1037 | ) |
|
1038 | 1038 | self._checkentries(fp, warn) |
|
1039 | 1039 | fp.close() |
|
1040 | 1040 | |
|
1041 | 1041 | def _checkentries(self, fp, warn): |
|
1042 | 1042 | """make sure there is no empty string in entries""" |
|
1043 | 1043 | if b'' in self.entries: |
|
1044 | 1044 | fp.seek(0) |
|
1045 | 1045 | for n, line in enumerate(fp): |
|
1046 | 1046 | if not line.rstrip(b'\n'): |
|
1047 | 1047 | t = _(b'invalid entry in fncache, line %d') % (n + 1) |
|
1048 | 1048 | if warn: |
|
1049 | 1049 | warn(t + b'\n') |
|
1050 | 1050 | else: |
|
1051 | 1051 | raise error.Abort(t) |
|
1052 | 1052 | |
|
1053 | 1053 | def write(self, tr): |
|
1054 | 1054 | if self._dirty: |
|
1055 | 1055 | assert self.entries is not None |
|
1056 | 1056 | self.entries = self.entries | self.addls |
|
1057 | 1057 | self.addls = set() |
|
1058 | 1058 | tr.addbackup(b'fncache') |
|
1059 | 1059 | fp = self.vfs(b'fncache', mode=b'wb', atomictemp=True) |
|
1060 | 1060 | if self.entries: |
|
1061 | 1061 | fp.write(encodedir(b'\n'.join(self.entries) + b'\n')) |
|
1062 | 1062 | fp.close() |
|
1063 | 1063 | self._dirty = False |
|
1064 | 1064 | if self.addls: |
|
1065 | 1065 | # if we have just new entries, let's append them to the fncache |
|
1066 | 1066 | tr.addbackup(b'fncache') |
|
1067 | 1067 | fp = self.vfs(b'fncache', mode=b'ab', atomictemp=True) |
|
1068 | 1068 | if self.addls: |
|
1069 | 1069 | fp.write(encodedir(b'\n'.join(self.addls) + b'\n')) |
|
1070 | 1070 | fp.close() |
|
1071 | 1071 | self.entries = None |
|
1072 | 1072 | self.addls = set() |
|
1073 | 1073 | |
|
1074 | 1074 | def addignore(self, fn): |
|
1075 | 1075 | self._ignores.add(fn) |
|
1076 | 1076 | |
|
1077 | 1077 | def add(self, fn): |
|
1078 | 1078 | if fn in self._ignores: |
|
1079 | 1079 | return |
|
1080 | 1080 | if self.entries is None: |
|
1081 | 1081 | self._load() |
|
1082 | 1082 | if fn not in self.entries: |
|
1083 | 1083 | self.addls.add(fn) |
|
1084 | 1084 | |
|
1085 | 1085 | def remove(self, fn): |
|
1086 | 1086 | if self.entries is None: |
|
1087 | 1087 | self._load() |
|
1088 | 1088 | if fn in self.addls: |
|
1089 | 1089 | self.addls.remove(fn) |
|
1090 | 1090 | return |
|
1091 | 1091 | try: |
|
1092 | 1092 | self.entries.remove(fn) |
|
1093 | 1093 | self._dirty = True |
|
1094 | 1094 | except KeyError: |
|
1095 | 1095 | pass |
|
1096 | 1096 | |
|
1097 | 1097 | def __contains__(self, fn): |
|
1098 | 1098 | if fn in self.addls: |
|
1099 | 1099 | return True |
|
1100 | 1100 | if self.entries is None: |
|
1101 | 1101 | self._load() |
|
1102 | 1102 | return fn in self.entries |
|
1103 | 1103 | |
|
1104 | 1104 | def __iter__(self): |
|
1105 | 1105 | if self.entries is None: |
|
1106 | 1106 | self._load() |
|
1107 | 1107 | return iter(self.entries | self.addls) |
|
1108 | 1108 | |
|
1109 | 1109 | |
|
1110 | 1110 | class _fncachevfs(vfsmod.proxyvfs): |
|
1111 | 1111 | def __init__(self, vfs, fnc, encode): |
|
1112 | 1112 | vfsmod.proxyvfs.__init__(self, vfs) |
|
1113 | 1113 | self.fncache = fnc |
|
1114 | 1114 | self.encode = encode |
|
1115 | 1115 | |
|
1116 | 1116 | def __call__(self, path, mode=b'r', *args, **kw): |
|
1117 | 1117 | encoded = self.encode(path) |
|
1118 | 1118 | if ( |
|
1119 | 1119 | mode not in (b'r', b'rb') |
|
1120 | 1120 | and (path.startswith(b'data/') or path.startswith(b'meta/')) |
|
1121 | 1121 | and is_revlog_file(path) |
|
1122 | 1122 | ): |
|
1123 | 1123 | # do not trigger a fncache load when adding a file that already is |
|
1124 | 1124 | # known to exist. |
|
1125 | 1125 | notload = self.fncache.entries is None and ( |
|
1126 | 1126 | # if the file has size zero, it should be considered as missing. |
|
1127 | 1127 | # Such zero-size files are the result of truncation when a |
|
1128 | 1128 | # transaction is aborted. |
|
1129 | 1129 | self.vfs.exists(encoded) |
|
1130 | 1130 | and self.vfs.stat(encoded).st_size |
|
1131 | 1131 | ) |
|
1132 | 1132 | if not notload: |
|
1133 | 1133 | self.fncache.add(path) |
|
1134 | 1134 | return self.vfs(encoded, mode, *args, **kw) |
|
1135 | 1135 | |
|
1136 | 1136 | def join(self, path): |
|
1137 | 1137 | if path: |
|
1138 | 1138 | return self.vfs.join(self.encode(path)) |
|
1139 | 1139 | else: |
|
1140 | 1140 | return self.vfs.join(path) |
|
1141 | 1141 | |
|
1142 | 1142 | def register_file(self, path): |
|
1143 | 1143 | """generic hook point to lets fncache steer its stew""" |
|
1144 | 1144 | if path.startswith(b'data/') or path.startswith(b'meta/'): |
|
1145 | 1145 | self.fncache.add(path) |
|
1146 | 1146 | |
|
1147 | 1147 | |
|
1148 | 1148 | class fncachestore(basicstore): |
|
1149 | 1149 | def __init__(self, path, vfstype, dotencode): |
|
1150 | 1150 | if dotencode: |
|
1151 | 1151 | encode = _pathencode |
|
1152 | 1152 | else: |
|
1153 | 1153 | encode = _plainhybridencode |
|
1154 | 1154 | self.encode = encode |
|
1155 | 1155 | vfs = vfstype(path + b'/store') |
|
1156 | 1156 | self.path = vfs.base |
|
1157 | 1157 | self.pathsep = self.path + b'/' |
|
1158 | 1158 | self.createmode = _calcmode(vfs) |
|
1159 | 1159 | vfs.createmode = self.createmode |
|
1160 | 1160 | self.rawvfs = vfs |
|
1161 | 1161 | fnc = fncache(vfs) |
|
1162 | 1162 | self.fncache = fnc |
|
1163 | 1163 | self.vfs = _fncachevfs(vfs, fnc, encode) |
|
1164 | 1164 | self.opener = self.vfs |
|
1165 | 1165 | |
|
1166 | 1166 | def join(self, f): |
|
1167 | 1167 | return self.pathsep + self.encode(f) |
|
1168 | 1168 | |
|
1169 | 1169 | def getsize(self, path): |
|
1170 | 1170 | return self.rawvfs.stat(path).st_size |
|
1171 | 1171 | |
|
1172 | 1172 | def data_entries( |
|
1173 | 1173 | self, matcher=None, undecodable=None |
|
1174 | 1174 | ) -> Generator[BaseStoreEntry, None, None]: |
|
1175 | 1175 | # Note: all files in fncache should be revlog related, However the |
|
1176 | 1176 | # fncache might contains such file added by previous version of |
|
1177 | 1177 | # Mercurial. |
|
1178 | 1178 | files = ((f, None) for f in self.fncache if is_revlog_file(f)) |
|
1179 | 1179 | by_revlog = _gather_revlog(files) |
|
1180 | 1180 | for revlog, details in by_revlog: |
|
1181 | 1181 | if revlog.startswith(b'data/'): |
|
1182 | 1182 | rl_type = KIND_FILELOG |
|
1183 | 1183 | revlog_target_id = revlog.split(b'/', 1)[1] |
|
1184 | 1184 | elif revlog.startswith(b'meta/'): |
|
1185 | 1185 | rl_type = KIND_MANIFESTLOG |
|
1186 | 1186 | # drop the initial directory and the `00manifest` file part |
|
1187 | 1187 | tmp = revlog.split(b'/', 1)[1] |
|
1188 | 1188 | revlog_target_id = tmp.rsplit(b'/', 1)[0] + b'/' |
|
1189 | 1189 | else: |
|
1190 | 1190 | # unreachable |
|
1191 | 1191 | assert False, revlog |
|
1192 | 1192 | entry = RevlogStoreEntry( |
|
1193 | 1193 | path_prefix=revlog, |
|
1194 | 1194 | revlog_type=rl_type, |
|
1195 | 1195 | target_id=revlog_target_id, |
|
1196 | 1196 | details=details, |
|
1197 | 1197 | ) |
|
1198 | 1198 | if _match_tracked_entry(entry, matcher): |
|
1199 | 1199 | yield entry |
|
1200 | 1200 | |
|
1201 | 1201 | def copylist(self): |
|
1202 | 1202 | d = ( |
|
1203 | 1203 | b'bookmarks', |
|
1204 | 1204 | b'narrowspec', |
|
1205 | 1205 | b'data', |
|
1206 | 1206 | b'meta', |
|
1207 | 1207 | b'dh', |
|
1208 | 1208 | b'fncache', |
|
1209 | 1209 | b'phaseroots', |
|
1210 | 1210 | b'obsstore', |
|
1211 | 1211 | b'00manifest.d', |
|
1212 | 1212 | b'00manifest.i', |
|
1213 | 1213 | b'00changelog.d', |
|
1214 | 1214 | b'00changelog.i', |
|
1215 | 1215 | b'requires', |
|
1216 | 1216 | ) |
|
1217 | 1217 | return [b'requires', b'00changelog.i'] + [b'store/' + f for f in d] |
|
1218 | 1218 | |
|
1219 | 1219 | def write(self, tr): |
|
1220 | 1220 | self.fncache.write(tr) |
|
1221 | 1221 | |
|
1222 | 1222 | def invalidatecaches(self): |
|
1223 | 1223 | self.fncache.entries = None |
|
1224 | 1224 | self.fncache.addls = set() |
|
1225 | 1225 | |
|
1226 | 1226 | def markremoved(self, fn): |
|
1227 | 1227 | self.fncache.remove(fn) |
|
1228 | 1228 | |
|
1229 | 1229 | def _exists(self, f): |
|
1230 | 1230 | ef = self.encode(f) |
|
1231 | 1231 | try: |
|
1232 | 1232 | self.getsize(ef) |
|
1233 | 1233 | return True |
|
1234 | 1234 | except FileNotFoundError: |
|
1235 | 1235 | return False |
|
1236 | 1236 | |
|
1237 | 1237 | def __contains__(self, path): |
|
1238 | 1238 | '''Checks if the store contains path''' |
|
1239 | 1239 | path = b"/".join((b"data", path)) |
|
1240 | 1240 | # check for files (exact match) |
|
1241 | 1241 | e = path + b'.i' |
|
1242 | 1242 | if e in self.fncache and self._exists(e): |
|
1243 | 1243 | return True |
|
1244 | 1244 | # now check for directories (prefix match) |
|
1245 | 1245 | if not path.endswith(b'/'): |
|
1246 | 1246 | path += b'/' |
|
1247 | 1247 | for e in self.fncache: |
|
1248 | 1248 | if e.startswith(path) and self._exists(e): |
|
1249 | 1249 | return True |
|
1250 | 1250 | return False |
@@ -1,344 +1,351 b'' | |||
|
1 | 1 | # unionrepo.py - repository class for viewing union of repository changesets |
|
2 | 2 | # |
|
3 | 3 | # Derived from bundlerepo.py |
|
4 | 4 | # Copyright 2006, 2007 Benoit Boissinot <bboissin@gmail.com> |
|
5 | 5 | # Copyright 2013 Unity Technologies, Mads Kiilerich <madski@unity3d.com> |
|
6 | 6 | # |
|
7 | 7 | # This software may be used and distributed according to the terms of the |
|
8 | 8 | # GNU General Public License version 2 or any later version. |
|
9 | 9 | |
|
10 | 10 | """Repository class for "in-memory pull" of one local repository to another, |
|
11 | 11 | allowing operations like diff and log with revsets. |
|
12 | 12 | """ |
|
13 | 13 | |
|
14 | 14 | import contextlib |
|
15 | 15 | |
|
16 | 16 | |
|
17 | 17 | from .i18n import _ |
|
18 | 18 | |
|
19 | 19 | from . import ( |
|
20 | 20 | changelog, |
|
21 | 21 | cmdutil, |
|
22 | 22 | encoding, |
|
23 | 23 | error, |
|
24 | 24 | filelog, |
|
25 | 25 | localrepo, |
|
26 | 26 | manifest, |
|
27 | 27 | mdiff, |
|
28 | 28 | pathutil, |
|
29 | 29 | revlog, |
|
30 | 30 | util, |
|
31 | 31 | vfs as vfsmod, |
|
32 | 32 | ) |
|
33 | 33 | |
|
34 | 34 | from .revlogutils import ( |
|
35 | 35 | constants as revlog_constants, |
|
36 | 36 | ) |
|
37 | 37 | |
|
38 | 38 | |
|
39 | 39 | class unionrevlog(revlog.revlog): |
|
40 | 40 | def __init__(self, opener, radix, revlog2, linkmapper): |
|
41 | 41 | # How it works: |
|
42 | 42 | # To retrieve a revision, we just need to know the node id so we can |
|
43 | 43 | # look it up in revlog2. |
|
44 | 44 | # |
|
45 | 45 | # To differentiate a rev in the second revlog from a rev in the revlog, |
|
46 | 46 | # we check revision against repotiprev. |
|
47 | 47 | opener = vfsmod.readonlyvfs(opener) |
|
48 | 48 | target = getattr(revlog2, 'target', None) |
|
49 | 49 | if target is None: |
|
50 | 50 | # a revlog wrapper, eg: the manifestlog that is not an actual revlog |
|
51 | 51 | target = revlog2._revlog.target |
|
52 | 52 | revlog.revlog.__init__(self, opener, target=target, radix=radix) |
|
53 | 53 | self.revlog2 = revlog2 |
|
54 | 54 | |
|
55 | 55 | n = len(self) |
|
56 | 56 | self.repotiprev = n - 1 |
|
57 | 57 | self.bundlerevs = set() # used by 'bundle()' revset expression |
|
58 | 58 | for rev2 in self.revlog2: |
|
59 | 59 | rev = self.revlog2.index[rev2] |
|
60 | 60 | # rev numbers - in revlog2, very different from self.rev |
|
61 | 61 | ( |
|
62 | 62 | _start, |
|
63 | 63 | _csize, |
|
64 | 64 | rsize, |
|
65 | 65 | base, |
|
66 | 66 | linkrev, |
|
67 | 67 | p1rev, |
|
68 | 68 | p2rev, |
|
69 | 69 | node, |
|
70 | 70 | _sdo, |
|
71 | 71 | _sds, |
|
72 | 72 | _dcm, |
|
73 | 73 | _sdcm, |
|
74 | 74 | rank, |
|
75 | 75 | ) = rev |
|
76 | 76 | flags = _start & 0xFFFF |
|
77 | 77 | |
|
78 | 78 | if linkmapper is None: # link is to same revlog |
|
79 | 79 | assert linkrev == rev2 # we never link back |
|
80 | 80 | link = n |
|
81 | 81 | else: # rev must be mapped from repo2 cl to unified cl by linkmapper |
|
82 | 82 | link = linkmapper(linkrev) |
|
83 | 83 | |
|
84 | 84 | if linkmapper is not None: # link is to same revlog |
|
85 | 85 | base = linkmapper(base) |
|
86 | 86 | |
|
87 | 87 | this_rev = self.index.get_rev(node) |
|
88 | 88 | if this_rev is not None: |
|
89 | 89 | # this happens for the common revlog revisions |
|
90 | 90 | self.bundlerevs.add(this_rev) |
|
91 | 91 | continue |
|
92 | 92 | |
|
93 | 93 | p1node = self.revlog2.node(p1rev) |
|
94 | 94 | p2node = self.revlog2.node(p2rev) |
|
95 | 95 | |
|
96 | 96 | # TODO: it's probably wrong to set compressed length to -1, but |
|
97 | 97 | # I have no idea if csize is valid in the base revlog context. |
|
98 | 98 | e = ( |
|
99 | 99 | flags, |
|
100 | 100 | -1, |
|
101 | 101 | rsize, |
|
102 | 102 | base, |
|
103 | 103 | link, |
|
104 | 104 | self.rev(p1node), |
|
105 | 105 | self.rev(p2node), |
|
106 | 106 | node, |
|
107 | 107 | 0, # sidedata offset |
|
108 | 108 | 0, # sidedata size |
|
109 | 109 | revlog_constants.COMP_MODE_INLINE, |
|
110 | 110 | revlog_constants.COMP_MODE_INLINE, |
|
111 | 111 | rank, |
|
112 | 112 | ) |
|
113 | 113 | self.index.append(e) |
|
114 | 114 | self.bundlerevs.add(n) |
|
115 | 115 | n += 1 |
|
116 | 116 | |
|
117 | 117 | @contextlib.contextmanager |
|
118 | 118 | def reading(self): |
|
119 | 119 | if 0 <= len(self.bundlerevs) < len(self.index): |
|
120 | 120 | read_1 = super().reading |
|
121 | 121 | else: |
|
122 | 122 | read_1 = util.nullcontextmanager |
|
123 | 123 | if 0 < len(self.bundlerevs): |
|
124 | 124 | read_2 = self.revlog2.reading |
|
125 | 125 | else: |
|
126 | 126 | read_2 = util.nullcontextmanager |
|
127 | 127 | with read_1(), read_2(): |
|
128 | 128 | yield |
|
129 | 129 | |
|
130 | 130 | def _chunk(self, rev): |
|
131 | 131 | if rev <= self.repotiprev: |
|
132 | 132 | return revlog.revlog._chunk(self, rev) |
|
133 | 133 | return self.revlog2._chunk(self.node(rev)) |
|
134 | 134 | |
|
135 | 135 | def revdiff(self, rev1, rev2): |
|
136 | 136 | """return or calculate a delta between two revisions""" |
|
137 | 137 | if rev1 > self.repotiprev and rev2 > self.repotiprev: |
|
138 | 138 | return self.revlog2.revdiff( |
|
139 | 139 | self.revlog2.rev(self.node(rev1)), |
|
140 | 140 | self.revlog2.rev(self.node(rev2)), |
|
141 | 141 | ) |
|
142 | 142 | elif rev1 <= self.repotiprev and rev2 <= self.repotiprev: |
|
143 | 143 | return super(unionrevlog, self).revdiff(rev1, rev2) |
|
144 | 144 | |
|
145 | 145 | return mdiff.textdiff(self.rawdata(rev1), self.rawdata(rev2)) |
|
146 | 146 | |
|
147 | 147 | def _revisiondata(self, nodeorrev, raw=False): |
|
148 | 148 | if isinstance(nodeorrev, int): |
|
149 | 149 | rev = nodeorrev |
|
150 | 150 | node = self.node(rev) |
|
151 | 151 | else: |
|
152 | 152 | node = nodeorrev |
|
153 | 153 | rev = self.rev(node) |
|
154 | 154 | |
|
155 | 155 | if rev > self.repotiprev: |
|
156 | 156 | # work around manifestrevlog NOT being a revlog |
|
157 | 157 | revlog2 = getattr(self.revlog2, '_revlog', self.revlog2) |
|
158 | 158 | func = revlog2._revisiondata |
|
159 | 159 | else: |
|
160 | 160 | func = super(unionrevlog, self)._revisiondata |
|
161 | 161 | return func(node, raw=raw) |
|
162 | 162 | |
|
163 | 163 | def addrevision( |
|
164 | 164 | self, |
|
165 | 165 | text, |
|
166 | 166 | transaction, |
|
167 | 167 | link, |
|
168 | 168 | p1, |
|
169 | 169 | p2, |
|
170 | 170 | cachedelta=None, |
|
171 | 171 | node=None, |
|
172 | 172 | flags=revlog.REVIDX_DEFAULT_FLAGS, |
|
173 | 173 | deltacomputer=None, |
|
174 | 174 | sidedata=None, |
|
175 | 175 | ): |
|
176 | 176 | raise NotImplementedError |
|
177 | 177 | |
|
178 | 178 | def addgroup( |
|
179 | 179 | self, |
|
180 | 180 | deltas, |
|
181 | 181 | linkmapper, |
|
182 | 182 | transaction, |
|
183 | 183 | alwayscache=False, |
|
184 | 184 | addrevisioncb=None, |
|
185 | 185 | duplicaterevisioncb=None, |
|
186 | 186 | debug_info=None, |
|
187 | 187 | delta_base_reuse_policy=None, |
|
188 | 188 | ): |
|
189 | 189 | raise NotImplementedError |
|
190 | 190 | |
|
191 | 191 | def strip(self, minlink, transaction): |
|
192 | 192 | raise NotImplementedError |
|
193 | 193 | |
|
194 | 194 | def checksize(self): |
|
195 | 195 | raise NotImplementedError |
|
196 | 196 | |
|
197 | 197 | |
|
198 | 198 | class unionchangelog(unionrevlog, changelog.changelog): |
|
199 | 199 | def __init__(self, opener, opener2): |
|
200 | 200 | changelog.changelog.__init__(self, opener) |
|
201 | 201 | linkmapper = None |
|
202 | 202 | changelog2 = changelog.changelog(opener2) |
|
203 | 203 | unionrevlog.__init__(self, opener, self.radix, changelog2, linkmapper) |
|
204 | 204 | |
|
205 | 205 | |
|
206 | 206 | class unionmanifest(unionrevlog, manifest.manifestrevlog): |
|
207 | repotiprev: int | |
|
208 | revlog2: manifest.ManifestRevlog | |
|
209 | ||
|
207 | 210 | def __init__(self, nodeconstants, opener, opener2, linkmapper): |
|
208 | 211 | # XXX manifestrevlog is not actually a revlog , so mixing it with |
|
209 | 212 | # bundlerevlog is not a good idea. |
|
210 | 213 | manifest.manifestrevlog.__init__(self, nodeconstants, opener) |
|
211 | 214 | manifest2 = manifest.manifestrevlog(nodeconstants, opener2) |
|
212 | 215 | unionrevlog.__init__( |
|
213 | 216 | self, opener, self._revlog.radix, manifest2, linkmapper |
|
214 | 217 | ) |
|
215 | 218 | |
|
216 | 219 | |
|
217 | 220 | class unionfilelog(filelog.filelog): |
|
221 | _revlog: unionrevlog | |
|
222 | repotiprev: int | |
|
223 | revlog2: revlog.revlog | |
|
224 | ||
|
218 | 225 | def __init__(self, opener, path, opener2, linkmapper, repo): |
|
219 | 226 | filelog.filelog.__init__(self, opener, path) |
|
220 | 227 | filelog2 = filelog.filelog(opener2, path) |
|
221 | 228 | self._revlog = unionrevlog( |
|
222 | 229 | opener, self._revlog.radix, filelog2._revlog, linkmapper |
|
223 | 230 | ) |
|
224 | 231 | self._repo = repo |
|
225 | 232 | self.repotiprev = self._revlog.repotiprev |
|
226 | 233 | self.revlog2 = self._revlog.revlog2 |
|
227 | 234 | |
|
228 | 235 | def iscensored(self, rev): |
|
229 | 236 | """Check if a revision is censored.""" |
|
230 | 237 | if rev <= self.repotiprev: |
|
231 | 238 | return filelog.filelog.iscensored(self, rev) |
|
232 | 239 | node = self.node(rev) |
|
233 | 240 | return self.revlog2.iscensored(self.revlog2.rev(node)) |
|
234 | 241 | |
|
235 | 242 | |
|
236 | 243 | class unionpeer(localrepo.localpeer): |
|
237 | 244 | def canpush(self): |
|
238 | 245 | return False |
|
239 | 246 | |
|
240 | 247 | |
|
241 | 248 | class unionrepository: |
|
242 | 249 | """Represents the union of data in 2 repositories. |
|
243 | 250 | |
|
244 | 251 | Instances are not usable if constructed directly. Use ``instance()`` |
|
245 | 252 | or ``makeunionrepository()`` to create a usable instance. |
|
246 | 253 | """ |
|
247 | 254 | |
|
248 | 255 | def __init__(self, repo2, url): |
|
249 | 256 | self.repo2 = repo2 |
|
250 | 257 | self._url = url |
|
251 | 258 | |
|
252 | 259 | self.ui.setconfig(b'phases', b'publish', False, b'unionrepo') |
|
253 | 260 | |
|
254 | 261 | @localrepo.unfilteredpropertycache |
|
255 | 262 | def changelog(self): |
|
256 | 263 | return unionchangelog(self.svfs, self.repo2.svfs) |
|
257 | 264 | |
|
258 | 265 | @localrepo.unfilteredpropertycache |
|
259 | 266 | def manifestlog(self): |
|
260 | 267 | rootstore = unionmanifest( |
|
261 | 268 | self.nodeconstants, |
|
262 | 269 | self.svfs, |
|
263 | 270 | self.repo2.svfs, |
|
264 | 271 | self.unfiltered()._clrev, |
|
265 | 272 | ) |
|
266 | 273 | return manifest.manifestlog( |
|
267 | 274 | self.svfs, self, rootstore, self.narrowmatch() |
|
268 | 275 | ) |
|
269 | 276 | |
|
270 | 277 | def _clrev(self, rev2): |
|
271 | 278 | """map from repo2 changelog rev to temporary rev in self.changelog""" |
|
272 | 279 | node = self.repo2.changelog.node(rev2) |
|
273 | 280 | return self.changelog.rev(node) |
|
274 | 281 | |
|
275 | 282 | def url(self): |
|
276 | 283 | return self._url |
|
277 | 284 | |
|
278 | 285 | def file(self, f): |
|
279 | 286 | return unionfilelog( |
|
280 | 287 | self.svfs, f, self.repo2.svfs, self.unfiltered()._clrev, self |
|
281 | 288 | ) |
|
282 | 289 | |
|
283 | 290 | def close(self): |
|
284 | 291 | self.repo2.close() |
|
285 | 292 | |
|
286 | 293 | def cancopy(self): |
|
287 | 294 | return False |
|
288 | 295 | |
|
289 | 296 | def peer(self, path=None, remotehidden=False): |
|
290 | 297 | return unionpeer(self, path=None, remotehidden=remotehidden) |
|
291 | 298 | |
|
292 | 299 | def getcwd(self): |
|
293 | 300 | return encoding.getcwd() # always outside the repo |
|
294 | 301 | |
|
295 | 302 | |
|
296 | 303 | def instance(ui, path, create, intents=None, createopts=None): |
|
297 | 304 | if create: |
|
298 | 305 | raise error.Abort(_(b'cannot create new union repository')) |
|
299 | 306 | parentpath = ui.config(b"bundle", b"mainreporoot") |
|
300 | 307 | if not parentpath: |
|
301 | 308 | # try to find the correct path to the working directory repo |
|
302 | 309 | parentpath = cmdutil.findrepo(encoding.getcwd()) |
|
303 | 310 | if parentpath is None: |
|
304 | 311 | parentpath = b'' |
|
305 | 312 | if parentpath: |
|
306 | 313 | # Try to make the full path relative so we get a nice, short URL. |
|
307 | 314 | # In particular, we don't want temp dir names in test outputs. |
|
308 | 315 | cwd = encoding.getcwd() |
|
309 | 316 | if parentpath == cwd: |
|
310 | 317 | parentpath = b'' |
|
311 | 318 | else: |
|
312 | 319 | cwd = pathutil.normasprefix(cwd) |
|
313 | 320 | if parentpath.startswith(cwd): |
|
314 | 321 | parentpath = parentpath[len(cwd) :] |
|
315 | 322 | if path.startswith(b'union:'): |
|
316 | 323 | s = path.split(b":", 1)[1].split(b"+", 1) |
|
317 | 324 | if len(s) == 1: |
|
318 | 325 | repopath, repopath2 = parentpath, s[0] |
|
319 | 326 | else: |
|
320 | 327 | repopath, repopath2 = s |
|
321 | 328 | else: |
|
322 | 329 | repopath, repopath2 = parentpath, path |
|
323 | 330 | |
|
324 | 331 | return makeunionrepository(ui, repopath, repopath2) |
|
325 | 332 | |
|
326 | 333 | |
|
327 | 334 | def makeunionrepository(ui, repopath1, repopath2): |
|
328 | 335 | """Make a union repository object from 2 local repo paths.""" |
|
329 | 336 | repo1 = localrepo.instance(ui, repopath1, create=False) |
|
330 | 337 | repo2 = localrepo.instance(ui, repopath2, create=False) |
|
331 | 338 | |
|
332 | 339 | url = b'union:%s+%s' % ( |
|
333 | 340 | util.expandpath(repopath1), |
|
334 | 341 | util.expandpath(repopath2), |
|
335 | 342 | ) |
|
336 | 343 | |
|
337 | 344 | class derivedunionrepository(unionrepository, repo1.__class__): |
|
338 | 345 | pass |
|
339 | 346 | |
|
340 | 347 | repo = repo1 |
|
341 | 348 | repo.__class__ = derivedunionrepository |
|
342 | 349 | unionrepository.__init__(repo1, repo2, url) |
|
343 | 350 | |
|
344 | 351 | return repo |
General Comments 0
You need to be logged in to leave comments.
Login now