Show More
@@ -1,694 +1,694 b'' | |||
|
1 | 1 | # narrowcommands.py - command modifications for narrowhg extension |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2017 Google, 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 | from __future__ import absolute_import |
|
8 | 8 | |
|
9 | 9 | import itertools |
|
10 | 10 | import os |
|
11 | 11 | |
|
12 | 12 | from mercurial.i18n import _ |
|
13 | 13 | from mercurial.node import ( |
|
14 | 14 | hex, |
|
15 | 15 | short, |
|
16 | 16 | ) |
|
17 | 17 | from mercurial import ( |
|
18 | 18 | bundle2, |
|
19 | 19 | cmdutil, |
|
20 | 20 | commands, |
|
21 | 21 | discovery, |
|
22 | 22 | encoding, |
|
23 | 23 | error, |
|
24 | 24 | exchange, |
|
25 | 25 | extensions, |
|
26 | 26 | hg, |
|
27 | 27 | narrowspec, |
|
28 | 28 | pathutil, |
|
29 | 29 | pycompat, |
|
30 | 30 | registrar, |
|
31 | 31 | repair, |
|
32 | 32 | repoview, |
|
33 | 33 | requirements, |
|
34 | 34 | sparse, |
|
35 | 35 | util, |
|
36 | 36 | wireprototypes, |
|
37 | 37 | ) |
|
38 | 38 | from mercurial.utils import ( |
|
39 | 39 | urlutil, |
|
40 | 40 | ) |
|
41 | 41 | |
|
42 | 42 | table = {} |
|
43 | 43 | command = registrar.command(table) |
|
44 | 44 | |
|
45 | 45 | |
|
46 | 46 | def setup(): |
|
47 | 47 | """Wraps user-facing mercurial commands with narrow-aware versions.""" |
|
48 | 48 | |
|
49 | 49 | entry = extensions.wrapcommand(commands.table, b'clone', clonenarrowcmd) |
|
50 | 50 | entry[1].append( |
|
51 | 51 | (b'', b'narrow', None, _(b"create a narrow clone of select files")) |
|
52 | 52 | ) |
|
53 | 53 | entry[1].append( |
|
54 | 54 | ( |
|
55 | 55 | b'', |
|
56 | 56 | b'depth', |
|
57 | 57 | b'', |
|
58 | 58 | _(b"limit the history fetched by distance from heads"), |
|
59 | 59 | ) |
|
60 | 60 | ) |
|
61 | 61 | entry[1].append((b'', b'narrowspec', b'', _(b"read narrowspecs from file"))) |
|
62 | 62 | # TODO(durin42): unify sparse/narrow --include/--exclude logic a bit |
|
63 | 63 | if b'sparse' not in extensions.enabled(): |
|
64 | 64 | entry[1].append( |
|
65 | 65 | (b'', b'include', [], _(b"specifically fetch this file/directory")) |
|
66 | 66 | ) |
|
67 | 67 | entry[1].append( |
|
68 | 68 | ( |
|
69 | 69 | b'', |
|
70 | 70 | b'exclude', |
|
71 | 71 | [], |
|
72 | 72 | _(b"do not fetch this file/directory, even if included"), |
|
73 | 73 | ) |
|
74 | 74 | ) |
|
75 | 75 | |
|
76 | 76 | entry = extensions.wrapcommand(commands.table, b'pull', pullnarrowcmd) |
|
77 | 77 | entry[1].append( |
|
78 | 78 | ( |
|
79 | 79 | b'', |
|
80 | 80 | b'depth', |
|
81 | 81 | b'', |
|
82 | 82 | _(b"limit the history fetched by distance from heads"), |
|
83 | 83 | ) |
|
84 | 84 | ) |
|
85 | 85 | |
|
86 | 86 | extensions.wrapcommand(commands.table, b'archive', archivenarrowcmd) |
|
87 | 87 | |
|
88 | 88 | |
|
89 | 89 | def clonenarrowcmd(orig, ui, repo, *args, **opts): |
|
90 | 90 | """Wraps clone command, so 'hg clone' first wraps localrepo.clone().""" |
|
91 | 91 | opts = pycompat.byteskwargs(opts) |
|
92 | 92 | wrappedextraprepare = util.nullcontextmanager() |
|
93 | 93 | narrowspecfile = opts[b'narrowspec'] |
|
94 | 94 | |
|
95 | 95 | if narrowspecfile: |
|
96 | 96 | filepath = os.path.join(encoding.getcwd(), narrowspecfile) |
|
97 | 97 | ui.status(_(b"reading narrowspec from '%s'\n") % filepath) |
|
98 | 98 | try: |
|
99 | 99 | fdata = util.readfile(filepath) |
|
100 | 100 | except IOError as inst: |
|
101 | 101 | raise error.Abort( |
|
102 | 102 | _(b"cannot read narrowspecs from '%s': %s") |
|
103 | 103 | % (filepath, encoding.strtolocal(inst.strerror)) |
|
104 | 104 | ) |
|
105 | 105 | |
|
106 | 106 | includes, excludes, profiles = sparse.parseconfig(ui, fdata, b'narrow') |
|
107 | 107 | if profiles: |
|
108 | 108 | raise error.ConfigError( |
|
109 | 109 | _( |
|
110 | 110 | b"cannot specify other files using '%include' in" |
|
111 | 111 | b" narrowspec" |
|
112 | 112 | ) |
|
113 | 113 | ) |
|
114 | 114 | |
|
115 | 115 | narrowspec.validatepatterns(includes) |
|
116 | 116 | narrowspec.validatepatterns(excludes) |
|
117 | 117 | |
|
118 | 118 | # narrowspec is passed so we should assume that user wants narrow clone |
|
119 | 119 | opts[b'narrow'] = True |
|
120 | 120 | opts[b'include'].extend(includes) |
|
121 | 121 | opts[b'exclude'].extend(excludes) |
|
122 | 122 | |
|
123 | 123 | if opts[b'narrow']: |
|
124 | 124 | |
|
125 | 125 | def pullbundle2extraprepare_widen(orig, pullop, kwargs): |
|
126 | 126 | orig(pullop, kwargs) |
|
127 | 127 | |
|
128 | 128 | if opts.get(b'depth'): |
|
129 | 129 | kwargs[b'depth'] = opts[b'depth'] |
|
130 | 130 | |
|
131 | 131 | wrappedextraprepare = extensions.wrappedfunction( |
|
132 | 132 | exchange, b'_pullbundle2extraprepare', pullbundle2extraprepare_widen |
|
133 | 133 | ) |
|
134 | 134 | |
|
135 | 135 | with wrappedextraprepare: |
|
136 | 136 | return orig(ui, repo, *args, **pycompat.strkwargs(opts)) |
|
137 | 137 | |
|
138 | 138 | |
|
139 | 139 | def pullnarrowcmd(orig, ui, repo, *args, **opts): |
|
140 | 140 | """Wraps pull command to allow modifying narrow spec.""" |
|
141 | 141 | wrappedextraprepare = util.nullcontextmanager() |
|
142 | 142 | if requirements.NARROW_REQUIREMENT in repo.requirements: |
|
143 | 143 | |
|
144 | 144 | def pullbundle2extraprepare_widen(orig, pullop, kwargs): |
|
145 | 145 | orig(pullop, kwargs) |
|
146 | 146 | if opts.get('depth'): |
|
147 | 147 | kwargs[b'depth'] = opts['depth'] |
|
148 | 148 | |
|
149 | 149 | wrappedextraprepare = extensions.wrappedfunction( |
|
150 | 150 | exchange, b'_pullbundle2extraprepare', pullbundle2extraprepare_widen |
|
151 | 151 | ) |
|
152 | 152 | |
|
153 | 153 | with wrappedextraprepare: |
|
154 | 154 | return orig(ui, repo, *args, **opts) |
|
155 | 155 | |
|
156 | 156 | |
|
157 | 157 | def archivenarrowcmd(orig, ui, repo, *args, **opts): |
|
158 | 158 | """Wraps archive command to narrow the default includes.""" |
|
159 | 159 | if requirements.NARROW_REQUIREMENT in repo.requirements: |
|
160 | 160 | repo_includes, repo_excludes = repo.narrowpats |
|
161 | 161 | includes = set(opts.get('include', [])) |
|
162 | 162 | excludes = set(opts.get('exclude', [])) |
|
163 | 163 | includes, excludes, unused_invalid = narrowspec.restrictpatterns( |
|
164 | 164 | includes, excludes, repo_includes, repo_excludes |
|
165 | 165 | ) |
|
166 | 166 | if includes: |
|
167 | 167 | opts['include'] = includes |
|
168 | 168 | if excludes: |
|
169 | 169 | opts['exclude'] = excludes |
|
170 | 170 | return orig(ui, repo, *args, **opts) |
|
171 | 171 | |
|
172 | 172 | |
|
173 | 173 | def pullbundle2extraprepare(orig, pullop, kwargs): |
|
174 | 174 | repo = pullop.repo |
|
175 | 175 | if requirements.NARROW_REQUIREMENT not in repo.requirements: |
|
176 | 176 | return orig(pullop, kwargs) |
|
177 | 177 | |
|
178 | 178 | if wireprototypes.NARROWCAP not in pullop.remote.capabilities(): |
|
179 | 179 | raise error.Abort(_(b"server does not support narrow clones")) |
|
180 | 180 | orig(pullop, kwargs) |
|
181 | 181 | kwargs[b'narrow'] = True |
|
182 | 182 | include, exclude = repo.narrowpats |
|
183 | 183 | kwargs[b'oldincludepats'] = include |
|
184 | 184 | kwargs[b'oldexcludepats'] = exclude |
|
185 | 185 | if include: |
|
186 | 186 | kwargs[b'includepats'] = include |
|
187 | 187 | if exclude: |
|
188 | 188 | kwargs[b'excludepats'] = exclude |
|
189 | 189 | # calculate known nodes only in ellipses cases because in non-ellipses cases |
|
190 | 190 | # we have all the nodes |
|
191 | 191 | if wireprototypes.ELLIPSESCAP1 in pullop.remote.capabilities(): |
|
192 | 192 | kwargs[b'known'] = [ |
|
193 | 193 | hex(ctx.node()) |
|
194 | 194 | for ctx in repo.set(b'::%ln', pullop.common) |
|
195 | 195 | if ctx.node() != repo.nullid |
|
196 | 196 | ] |
|
197 | 197 | if not kwargs[b'known']: |
|
198 | 198 | # Mercurial serializes an empty list as '' and deserializes it as |
|
199 | 199 | # [''], so delete it instead to avoid handling the empty string on |
|
200 | 200 | # the server. |
|
201 | 201 | del kwargs[b'known'] |
|
202 | 202 | |
|
203 | 203 | |
|
204 | 204 | extensions.wrapfunction( |
|
205 | 205 | exchange, b'_pullbundle2extraprepare', pullbundle2extraprepare |
|
206 | 206 | ) |
|
207 | 207 | |
|
208 | 208 | |
|
209 | 209 | def _narrow( |
|
210 | 210 | ui, |
|
211 | 211 | repo, |
|
212 | 212 | remote, |
|
213 | 213 | commoninc, |
|
214 | 214 | oldincludes, |
|
215 | 215 | oldexcludes, |
|
216 | 216 | newincludes, |
|
217 | 217 | newexcludes, |
|
218 | 218 | force, |
|
219 | 219 | backup, |
|
220 | 220 | ): |
|
221 | 221 | oldmatch = narrowspec.match(repo.root, oldincludes, oldexcludes) |
|
222 | 222 | newmatch = narrowspec.match(repo.root, newincludes, newexcludes) |
|
223 | 223 | |
|
224 | 224 | # This is essentially doing "hg outgoing" to find all local-only |
|
225 | 225 | # commits. We will then check that the local-only commits don't |
|
226 | 226 | # have any changes to files that will be untracked. |
|
227 | 227 | unfi = repo.unfiltered() |
|
228 | 228 | outgoing = discovery.findcommonoutgoing(unfi, remote, commoninc=commoninc) |
|
229 | 229 | ui.status(_(b'looking for local changes to affected paths\n')) |
|
230 | 230 | progress = ui.makeprogress( |
|
231 | 231 | topic=_(b'changesets'), |
|
232 | 232 | unit=_(b'changesets'), |
|
233 | 233 | total=len(outgoing.missing) + len(outgoing.excluded), |
|
234 | 234 | ) |
|
235 | 235 | localnodes = [] |
|
236 | 236 | with progress: |
|
237 | 237 | for n in itertools.chain(outgoing.missing, outgoing.excluded): |
|
238 | 238 | progress.increment() |
|
239 | 239 | if any(oldmatch(f) and not newmatch(f) for f in unfi[n].files()): |
|
240 | 240 | localnodes.append(n) |
|
241 | 241 | revstostrip = unfi.revs(b'descendants(%ln)', localnodes) |
|
242 | 242 | hiddenrevs = repoview.filterrevs(repo, b'visible') |
|
243 | 243 | visibletostrip = list( |
|
244 | 244 | repo.changelog.node(r) for r in (revstostrip - hiddenrevs) |
|
245 | 245 | ) |
|
246 | 246 | if visibletostrip: |
|
247 | 247 | ui.status( |
|
248 | 248 | _( |
|
249 | 249 | b'The following changeset(s) or their ancestors have ' |
|
250 | 250 | b'local changes not on the remote:\n' |
|
251 | 251 | ) |
|
252 | 252 | ) |
|
253 | 253 | maxnodes = 10 |
|
254 | 254 | if ui.verbose or len(visibletostrip) <= maxnodes: |
|
255 | 255 | for n in visibletostrip: |
|
256 | 256 | ui.status(b'%s\n' % short(n)) |
|
257 | 257 | else: |
|
258 | 258 | for n in visibletostrip[:maxnodes]: |
|
259 | 259 | ui.status(b'%s\n' % short(n)) |
|
260 | 260 | ui.status( |
|
261 | 261 | _(b'...and %d more, use --verbose to list all\n') |
|
262 | 262 | % (len(visibletostrip) - maxnodes) |
|
263 | 263 | ) |
|
264 | 264 | if not force: |
|
265 | 265 | raise error.StateError( |
|
266 | 266 | _(b'local changes found'), |
|
267 | 267 | hint=_(b'use --force-delete-local-changes to ignore'), |
|
268 | 268 | ) |
|
269 | 269 | |
|
270 | 270 | with ui.uninterruptible(): |
|
271 | 271 | if revstostrip: |
|
272 | 272 | tostrip = [unfi.changelog.node(r) for r in revstostrip] |
|
273 | 273 | if repo[b'.'].node() in tostrip: |
|
274 | 274 | # stripping working copy, so move to a different commit first |
|
275 | 275 | urev = max( |
|
276 | 276 | repo.revs( |
|
277 | 277 | b'(::%n) - %ln + null', |
|
278 | 278 | repo[b'.'].node(), |
|
279 | 279 | visibletostrip, |
|
280 | 280 | ) |
|
281 | 281 | ) |
|
282 | 282 | hg.clean(repo, urev) |
|
283 | 283 | overrides = {(b'devel', b'strip-obsmarkers'): False} |
|
284 | 284 | if backup: |
|
285 | 285 | ui.status(_(b'moving unwanted changesets to backup\n')) |
|
286 | 286 | else: |
|
287 | 287 | ui.status(_(b'deleting unwanted changesets\n')) |
|
288 | 288 | with ui.configoverride(overrides, b'narrow'): |
|
289 | 289 | repair.strip(ui, unfi, tostrip, topic=b'narrow', backup=backup) |
|
290 | 290 | |
|
291 | 291 | todelete = [] |
|
292 |
for t, f |
|
|
292 | for t, f, size in repo.store.datafiles(): | |
|
293 | 293 | if f.startswith(b'data/'): |
|
294 | 294 | file = f[5:-2] |
|
295 | 295 | if not newmatch(file): |
|
296 | 296 | todelete.append(f) |
|
297 | 297 | elif f.startswith(b'meta/'): |
|
298 | 298 | dir = f[5:-13] |
|
299 | 299 | dirs = sorted(pathutil.dirs({dir})) + [dir] |
|
300 | 300 | include = True |
|
301 | 301 | for d in dirs: |
|
302 | 302 | visit = newmatch.visitdir(d) |
|
303 | 303 | if not visit: |
|
304 | 304 | include = False |
|
305 | 305 | break |
|
306 | 306 | if visit == b'all': |
|
307 | 307 | break |
|
308 | 308 | if not include: |
|
309 | 309 | todelete.append(f) |
|
310 | 310 | |
|
311 | 311 | repo.destroying() |
|
312 | 312 | |
|
313 | 313 | with repo.transaction(b'narrowing'): |
|
314 | 314 | # Update narrowspec before removing revlogs, so repo won't be |
|
315 | 315 | # corrupt in case of crash |
|
316 | 316 | repo.setnarrowpats(newincludes, newexcludes) |
|
317 | 317 | |
|
318 | 318 | for f in todelete: |
|
319 | 319 | ui.status(_(b'deleting %s\n') % f) |
|
320 | 320 | util.unlinkpath(repo.svfs.join(f)) |
|
321 | 321 | repo.store.markremoved(f) |
|
322 | 322 | |
|
323 | 323 | ui.status(_(b'deleting unwanted files from working copy\n')) |
|
324 | 324 | with repo.dirstate.parentchange(): |
|
325 | 325 | narrowspec.updateworkingcopy(repo, assumeclean=True) |
|
326 | 326 | narrowspec.copytoworkingcopy(repo) |
|
327 | 327 | |
|
328 | 328 | repo.destroyed() |
|
329 | 329 | |
|
330 | 330 | |
|
331 | 331 | def _widen( |
|
332 | 332 | ui, |
|
333 | 333 | repo, |
|
334 | 334 | remote, |
|
335 | 335 | commoninc, |
|
336 | 336 | oldincludes, |
|
337 | 337 | oldexcludes, |
|
338 | 338 | newincludes, |
|
339 | 339 | newexcludes, |
|
340 | 340 | ): |
|
341 | 341 | # for now we assume that if a server has ellipses enabled, we will be |
|
342 | 342 | # exchanging ellipses nodes. In future we should add ellipses as a client |
|
343 | 343 | # side requirement (maybe) to distinguish a client is shallow or not and |
|
344 | 344 | # then send that information to server whether we want ellipses or not. |
|
345 | 345 | # Theoretically a non-ellipses repo should be able to use narrow |
|
346 | 346 | # functionality from an ellipses enabled server |
|
347 | 347 | remotecap = remote.capabilities() |
|
348 | 348 | ellipsesremote = any( |
|
349 | 349 | cap in remotecap for cap in wireprototypes.SUPPORTED_ELLIPSESCAP |
|
350 | 350 | ) |
|
351 | 351 | |
|
352 | 352 | # check whether we are talking to a server which supports old version of |
|
353 | 353 | # ellipses capabilities |
|
354 | 354 | isoldellipses = ( |
|
355 | 355 | ellipsesremote |
|
356 | 356 | and wireprototypes.ELLIPSESCAP1 in remotecap |
|
357 | 357 | and wireprototypes.ELLIPSESCAP not in remotecap |
|
358 | 358 | ) |
|
359 | 359 | |
|
360 | 360 | def pullbundle2extraprepare_widen(orig, pullop, kwargs): |
|
361 | 361 | orig(pullop, kwargs) |
|
362 | 362 | # The old{in,ex}cludepats have already been set by orig() |
|
363 | 363 | kwargs[b'includepats'] = newincludes |
|
364 | 364 | kwargs[b'excludepats'] = newexcludes |
|
365 | 365 | |
|
366 | 366 | wrappedextraprepare = extensions.wrappedfunction( |
|
367 | 367 | exchange, b'_pullbundle2extraprepare', pullbundle2extraprepare_widen |
|
368 | 368 | ) |
|
369 | 369 | |
|
370 | 370 | # define a function that narrowbundle2 can call after creating the |
|
371 | 371 | # backup bundle, but before applying the bundle from the server |
|
372 | 372 | def setnewnarrowpats(): |
|
373 | 373 | repo.setnarrowpats(newincludes, newexcludes) |
|
374 | 374 | |
|
375 | 375 | repo.setnewnarrowpats = setnewnarrowpats |
|
376 | 376 | # silence the devel-warning of applying an empty changegroup |
|
377 | 377 | overrides = {(b'devel', b'all-warnings'): False} |
|
378 | 378 | |
|
379 | 379 | common = commoninc[0] |
|
380 | 380 | with ui.uninterruptible(): |
|
381 | 381 | if ellipsesremote: |
|
382 | 382 | ds = repo.dirstate |
|
383 | 383 | p1, p2 = ds.p1(), ds.p2() |
|
384 | 384 | with ds.parentchange(): |
|
385 | 385 | ds.setparents(repo.nullid, repo.nullid) |
|
386 | 386 | if isoldellipses: |
|
387 | 387 | with wrappedextraprepare: |
|
388 | 388 | exchange.pull(repo, remote, heads=common) |
|
389 | 389 | else: |
|
390 | 390 | known = [] |
|
391 | 391 | if ellipsesremote: |
|
392 | 392 | known = [ |
|
393 | 393 | ctx.node() |
|
394 | 394 | for ctx in repo.set(b'::%ln', common) |
|
395 | 395 | if ctx.node() != repo.nullid |
|
396 | 396 | ] |
|
397 | 397 | with remote.commandexecutor() as e: |
|
398 | 398 | bundle = e.callcommand( |
|
399 | 399 | b'narrow_widen', |
|
400 | 400 | { |
|
401 | 401 | b'oldincludes': oldincludes, |
|
402 | 402 | b'oldexcludes': oldexcludes, |
|
403 | 403 | b'newincludes': newincludes, |
|
404 | 404 | b'newexcludes': newexcludes, |
|
405 | 405 | b'cgversion': b'03', |
|
406 | 406 | b'commonheads': common, |
|
407 | 407 | b'known': known, |
|
408 | 408 | b'ellipses': ellipsesremote, |
|
409 | 409 | }, |
|
410 | 410 | ).result() |
|
411 | 411 | |
|
412 | 412 | trmanager = exchange.transactionmanager( |
|
413 | 413 | repo, b'widen', remote.url() |
|
414 | 414 | ) |
|
415 | 415 | with trmanager, repo.ui.configoverride(overrides, b'widen'): |
|
416 | 416 | op = bundle2.bundleoperation( |
|
417 | 417 | repo, trmanager.transaction, source=b'widen' |
|
418 | 418 | ) |
|
419 | 419 | # TODO: we should catch error.Abort here |
|
420 | 420 | bundle2.processbundle(repo, bundle, op=op) |
|
421 | 421 | |
|
422 | 422 | if ellipsesremote: |
|
423 | 423 | with ds.parentchange(): |
|
424 | 424 | ds.setparents(p1, p2) |
|
425 | 425 | |
|
426 | 426 | with repo.transaction(b'widening'), repo.dirstate.parentchange(): |
|
427 | 427 | repo.setnewnarrowpats() |
|
428 | 428 | narrowspec.updateworkingcopy(repo) |
|
429 | 429 | narrowspec.copytoworkingcopy(repo) |
|
430 | 430 | |
|
431 | 431 | |
|
432 | 432 | # TODO(rdamazio): Make new matcher format and update description |
|
433 | 433 | @command( |
|
434 | 434 | b'tracked', |
|
435 | 435 | [ |
|
436 | 436 | (b'', b'addinclude', [], _(b'new paths to include')), |
|
437 | 437 | (b'', b'removeinclude', [], _(b'old paths to no longer include')), |
|
438 | 438 | ( |
|
439 | 439 | b'', |
|
440 | 440 | b'auto-remove-includes', |
|
441 | 441 | False, |
|
442 | 442 | _(b'automatically choose unused includes to remove'), |
|
443 | 443 | ), |
|
444 | 444 | (b'', b'addexclude', [], _(b'new paths to exclude')), |
|
445 | 445 | (b'', b'import-rules', b'', _(b'import narrowspecs from a file')), |
|
446 | 446 | (b'', b'removeexclude', [], _(b'old paths to no longer exclude')), |
|
447 | 447 | ( |
|
448 | 448 | b'', |
|
449 | 449 | b'clear', |
|
450 | 450 | False, |
|
451 | 451 | _(b'whether to replace the existing narrowspec'), |
|
452 | 452 | ), |
|
453 | 453 | ( |
|
454 | 454 | b'', |
|
455 | 455 | b'force-delete-local-changes', |
|
456 | 456 | False, |
|
457 | 457 | _(b'forces deletion of local changes when narrowing'), |
|
458 | 458 | ), |
|
459 | 459 | ( |
|
460 | 460 | b'', |
|
461 | 461 | b'backup', |
|
462 | 462 | True, |
|
463 | 463 | _(b'back up local changes when narrowing'), |
|
464 | 464 | ), |
|
465 | 465 | ( |
|
466 | 466 | b'', |
|
467 | 467 | b'update-working-copy', |
|
468 | 468 | False, |
|
469 | 469 | _(b'update working copy when the store has changed'), |
|
470 | 470 | ), |
|
471 | 471 | ] |
|
472 | 472 | + commands.remoteopts, |
|
473 | 473 | _(b'[OPTIONS]... [REMOTE]'), |
|
474 | 474 | inferrepo=True, |
|
475 | 475 | helpcategory=command.CATEGORY_MAINTENANCE, |
|
476 | 476 | ) |
|
477 | 477 | def trackedcmd(ui, repo, remotepath=None, *pats, **opts): |
|
478 | 478 | """show or change the current narrowspec |
|
479 | 479 | |
|
480 | 480 | With no argument, shows the current narrowspec entries, one per line. Each |
|
481 | 481 | line will be prefixed with 'I' or 'X' for included or excluded patterns, |
|
482 | 482 | respectively. |
|
483 | 483 | |
|
484 | 484 | The narrowspec is comprised of expressions to match remote files and/or |
|
485 | 485 | directories that should be pulled into your client. |
|
486 | 486 | The narrowspec has *include* and *exclude* expressions, with excludes always |
|
487 | 487 | trumping includes: that is, if a file matches an exclude expression, it will |
|
488 | 488 | be excluded even if it also matches an include expression. |
|
489 | 489 | Excluding files that were never included has no effect. |
|
490 | 490 | |
|
491 | 491 | Each included or excluded entry is in the format described by |
|
492 | 492 | 'hg help patterns'. |
|
493 | 493 | |
|
494 | 494 | The options allow you to add or remove included and excluded expressions. |
|
495 | 495 | |
|
496 | 496 | If --clear is specified, then all previous includes and excludes are DROPPED |
|
497 | 497 | and replaced by the new ones specified to --addinclude and --addexclude. |
|
498 | 498 | If --clear is specified without any further options, the narrowspec will be |
|
499 | 499 | empty and will not match any files. |
|
500 | 500 | |
|
501 | 501 | If --auto-remove-includes is specified, then those includes that don't match |
|
502 | 502 | any files modified by currently visible local commits (those not shared by |
|
503 | 503 | the remote) will be added to the set of explicitly specified includes to |
|
504 | 504 | remove. |
|
505 | 505 | |
|
506 | 506 | --import-rules accepts a path to a file containing rules, allowing you to |
|
507 | 507 | add --addinclude, --addexclude rules in bulk. Like the other include and |
|
508 | 508 | exclude switches, the changes are applied immediately. |
|
509 | 509 | """ |
|
510 | 510 | opts = pycompat.byteskwargs(opts) |
|
511 | 511 | if requirements.NARROW_REQUIREMENT not in repo.requirements: |
|
512 | 512 | raise error.InputError( |
|
513 | 513 | _( |
|
514 | 514 | b'the tracked command is only supported on ' |
|
515 | 515 | b'repositories cloned with --narrow' |
|
516 | 516 | ) |
|
517 | 517 | ) |
|
518 | 518 | |
|
519 | 519 | # Before supporting, decide whether it "hg tracked --clear" should mean |
|
520 | 520 | # tracking no paths or all paths. |
|
521 | 521 | if opts[b'clear']: |
|
522 | 522 | raise error.InputError(_(b'the --clear option is not yet supported')) |
|
523 | 523 | |
|
524 | 524 | # import rules from a file |
|
525 | 525 | newrules = opts.get(b'import_rules') |
|
526 | 526 | if newrules: |
|
527 | 527 | try: |
|
528 | 528 | filepath = os.path.join(encoding.getcwd(), newrules) |
|
529 | 529 | fdata = util.readfile(filepath) |
|
530 | 530 | except IOError as inst: |
|
531 | 531 | raise error.StorageError( |
|
532 | 532 | _(b"cannot read narrowspecs from '%s': %s") |
|
533 | 533 | % (filepath, encoding.strtolocal(inst.strerror)) |
|
534 | 534 | ) |
|
535 | 535 | includepats, excludepats, profiles = sparse.parseconfig( |
|
536 | 536 | ui, fdata, b'narrow' |
|
537 | 537 | ) |
|
538 | 538 | if profiles: |
|
539 | 539 | raise error.InputError( |
|
540 | 540 | _( |
|
541 | 541 | b"including other spec files using '%include' " |
|
542 | 542 | b"is not supported in narrowspec" |
|
543 | 543 | ) |
|
544 | 544 | ) |
|
545 | 545 | opts[b'addinclude'].extend(includepats) |
|
546 | 546 | opts[b'addexclude'].extend(excludepats) |
|
547 | 547 | |
|
548 | 548 | addedincludes = narrowspec.parsepatterns(opts[b'addinclude']) |
|
549 | 549 | removedincludes = narrowspec.parsepatterns(opts[b'removeinclude']) |
|
550 | 550 | addedexcludes = narrowspec.parsepatterns(opts[b'addexclude']) |
|
551 | 551 | removedexcludes = narrowspec.parsepatterns(opts[b'removeexclude']) |
|
552 | 552 | autoremoveincludes = opts[b'auto_remove_includes'] |
|
553 | 553 | |
|
554 | 554 | update_working_copy = opts[b'update_working_copy'] |
|
555 | 555 | only_show = not ( |
|
556 | 556 | addedincludes |
|
557 | 557 | or removedincludes |
|
558 | 558 | or addedexcludes |
|
559 | 559 | or removedexcludes |
|
560 | 560 | or newrules |
|
561 | 561 | or autoremoveincludes |
|
562 | 562 | or update_working_copy |
|
563 | 563 | ) |
|
564 | 564 | |
|
565 | 565 | oldincludes, oldexcludes = repo.narrowpats |
|
566 | 566 | |
|
567 | 567 | # filter the user passed additions and deletions into actual additions and |
|
568 | 568 | # deletions of excludes and includes |
|
569 | 569 | addedincludes -= oldincludes |
|
570 | 570 | removedincludes &= oldincludes |
|
571 | 571 | addedexcludes -= oldexcludes |
|
572 | 572 | removedexcludes &= oldexcludes |
|
573 | 573 | |
|
574 | 574 | widening = addedincludes or removedexcludes |
|
575 | 575 | narrowing = removedincludes or addedexcludes |
|
576 | 576 | |
|
577 | 577 | # Only print the current narrowspec. |
|
578 | 578 | if only_show: |
|
579 | 579 | ui.pager(b'tracked') |
|
580 | 580 | fm = ui.formatter(b'narrow', opts) |
|
581 | 581 | for i in sorted(oldincludes): |
|
582 | 582 | fm.startitem() |
|
583 | 583 | fm.write(b'status', b'%s ', b'I', label=b'narrow.included') |
|
584 | 584 | fm.write(b'pat', b'%s\n', i, label=b'narrow.included') |
|
585 | 585 | for i in sorted(oldexcludes): |
|
586 | 586 | fm.startitem() |
|
587 | 587 | fm.write(b'status', b'%s ', b'X', label=b'narrow.excluded') |
|
588 | 588 | fm.write(b'pat', b'%s\n', i, label=b'narrow.excluded') |
|
589 | 589 | fm.end() |
|
590 | 590 | return 0 |
|
591 | 591 | |
|
592 | 592 | if update_working_copy: |
|
593 | 593 | with repo.wlock(), repo.lock(), repo.transaction( |
|
594 | 594 | b'narrow-wc' |
|
595 | 595 | ), repo.dirstate.parentchange(): |
|
596 | 596 | narrowspec.updateworkingcopy(repo) |
|
597 | 597 | narrowspec.copytoworkingcopy(repo) |
|
598 | 598 | return 0 |
|
599 | 599 | |
|
600 | 600 | if not (widening or narrowing or autoremoveincludes): |
|
601 | 601 | ui.status(_(b"nothing to widen or narrow\n")) |
|
602 | 602 | return 0 |
|
603 | 603 | |
|
604 | 604 | with repo.wlock(), repo.lock(): |
|
605 | 605 | cmdutil.bailifchanged(repo) |
|
606 | 606 | |
|
607 | 607 | # Find the revisions we have in common with the remote. These will |
|
608 | 608 | # be used for finding local-only changes for narrowing. They will |
|
609 | 609 | # also define the set of revisions to update for widening. |
|
610 | 610 | r = urlutil.get_unique_pull_path(b'tracked', repo, ui, remotepath) |
|
611 | 611 | url, branches = r |
|
612 | 612 | ui.status(_(b'comparing with %s\n') % urlutil.hidepassword(url)) |
|
613 | 613 | remote = hg.peer(repo, opts, url) |
|
614 | 614 | |
|
615 | 615 | try: |
|
616 | 616 | # check narrow support before doing anything if widening needs to be |
|
617 | 617 | # performed. In future we should also abort if client is ellipses and |
|
618 | 618 | # server does not support ellipses |
|
619 | 619 | if ( |
|
620 | 620 | widening |
|
621 | 621 | and wireprototypes.NARROWCAP not in remote.capabilities() |
|
622 | 622 | ): |
|
623 | 623 | raise error.Abort(_(b"server does not support narrow clones")) |
|
624 | 624 | |
|
625 | 625 | commoninc = discovery.findcommonincoming(repo, remote) |
|
626 | 626 | |
|
627 | 627 | if autoremoveincludes: |
|
628 | 628 | outgoing = discovery.findcommonoutgoing( |
|
629 | 629 | repo, remote, commoninc=commoninc |
|
630 | 630 | ) |
|
631 | 631 | ui.status(_(b'looking for unused includes to remove\n')) |
|
632 | 632 | localfiles = set() |
|
633 | 633 | for n in itertools.chain(outgoing.missing, outgoing.excluded): |
|
634 | 634 | localfiles.update(repo[n].files()) |
|
635 | 635 | suggestedremovals = [] |
|
636 | 636 | for include in sorted(oldincludes): |
|
637 | 637 | match = narrowspec.match(repo.root, [include], oldexcludes) |
|
638 | 638 | if not any(match(f) for f in localfiles): |
|
639 | 639 | suggestedremovals.append(include) |
|
640 | 640 | if suggestedremovals: |
|
641 | 641 | for s in suggestedremovals: |
|
642 | 642 | ui.status(b'%s\n' % s) |
|
643 | 643 | if ( |
|
644 | 644 | ui.promptchoice( |
|
645 | 645 | _( |
|
646 | 646 | b'remove these unused includes (yn)?' |
|
647 | 647 | b'$$ &Yes $$ &No' |
|
648 | 648 | ) |
|
649 | 649 | ) |
|
650 | 650 | == 0 |
|
651 | 651 | ): |
|
652 | 652 | removedincludes.update(suggestedremovals) |
|
653 | 653 | narrowing = True |
|
654 | 654 | else: |
|
655 | 655 | ui.status(_(b'found no unused includes\n')) |
|
656 | 656 | |
|
657 | 657 | if narrowing: |
|
658 | 658 | newincludes = oldincludes - removedincludes |
|
659 | 659 | newexcludes = oldexcludes | addedexcludes |
|
660 | 660 | _narrow( |
|
661 | 661 | ui, |
|
662 | 662 | repo, |
|
663 | 663 | remote, |
|
664 | 664 | commoninc, |
|
665 | 665 | oldincludes, |
|
666 | 666 | oldexcludes, |
|
667 | 667 | newincludes, |
|
668 | 668 | newexcludes, |
|
669 | 669 | opts[b'force_delete_local_changes'], |
|
670 | 670 | opts[b'backup'], |
|
671 | 671 | ) |
|
672 | 672 | # _narrow() updated the narrowspec and _widen() below needs to |
|
673 | 673 | # use the updated values as its base (otherwise removed includes |
|
674 | 674 | # and addedexcludes will be lost in the resulting narrowspec) |
|
675 | 675 | oldincludes = newincludes |
|
676 | 676 | oldexcludes = newexcludes |
|
677 | 677 | |
|
678 | 678 | if widening: |
|
679 | 679 | newincludes = oldincludes | addedincludes |
|
680 | 680 | newexcludes = oldexcludes - removedexcludes |
|
681 | 681 | _widen( |
|
682 | 682 | ui, |
|
683 | 683 | repo, |
|
684 | 684 | remote, |
|
685 | 685 | commoninc, |
|
686 | 686 | oldincludes, |
|
687 | 687 | oldexcludes, |
|
688 | 688 | newincludes, |
|
689 | 689 | newexcludes, |
|
690 | 690 | ) |
|
691 | 691 | finally: |
|
692 | 692 | remote.close() |
|
693 | 693 | |
|
694 | 694 | return 0 |
@@ -1,399 +1,399 b'' | |||
|
1 | 1 | from __future__ import absolute_import |
|
2 | 2 | |
|
3 | 3 | import threading |
|
4 | 4 | |
|
5 | 5 | from mercurial.node import ( |
|
6 | 6 | hex, |
|
7 | 7 | sha1nodeconstants, |
|
8 | 8 | ) |
|
9 | 9 | from mercurial.pycompat import getattr |
|
10 | 10 | from mercurial import ( |
|
11 | 11 | mdiff, |
|
12 | 12 | pycompat, |
|
13 | 13 | revlog, |
|
14 | 14 | ) |
|
15 | 15 | from . import ( |
|
16 | 16 | basestore, |
|
17 | 17 | constants, |
|
18 | 18 | shallowutil, |
|
19 | 19 | ) |
|
20 | 20 | |
|
21 | 21 | |
|
22 | 22 | class ChainIndicies(object): |
|
23 | 23 | """A static class for easy reference to the delta chain indicies.""" |
|
24 | 24 | |
|
25 | 25 | # The filename of this revision delta |
|
26 | 26 | NAME = 0 |
|
27 | 27 | # The mercurial file node for this revision delta |
|
28 | 28 | NODE = 1 |
|
29 | 29 | # The filename of the delta base's revision. This is useful when delta |
|
30 | 30 | # between different files (like in the case of a move or copy, we can delta |
|
31 | 31 | # against the original file content). |
|
32 | 32 | BASENAME = 2 |
|
33 | 33 | # The mercurial file node for the delta base revision. This is the nullid if |
|
34 | 34 | # this delta is a full text. |
|
35 | 35 | BASENODE = 3 |
|
36 | 36 | # The actual delta or full text data. |
|
37 | 37 | DATA = 4 |
|
38 | 38 | |
|
39 | 39 | |
|
40 | 40 | class unioncontentstore(basestore.baseunionstore): |
|
41 | 41 | def __init__(self, *args, **kwargs): |
|
42 | 42 | super(unioncontentstore, self).__init__(*args, **kwargs) |
|
43 | 43 | |
|
44 | 44 | self.stores = args |
|
45 | 45 | self.writestore = kwargs.get('writestore') |
|
46 | 46 | |
|
47 | 47 | # If allowincomplete==True then the union store can return partial |
|
48 | 48 | # delta chains, otherwise it will throw a KeyError if a full |
|
49 | 49 | # deltachain can't be found. |
|
50 | 50 | self.allowincomplete = kwargs.get('allowincomplete', False) |
|
51 | 51 | |
|
52 | 52 | def get(self, name, node): |
|
53 | 53 | """Fetches the full text revision contents of the given name+node pair. |
|
54 | 54 | If the full text doesn't exist, throws a KeyError. |
|
55 | 55 | |
|
56 | 56 | Under the hood, this uses getdeltachain() across all the stores to build |
|
57 | 57 | up a full chain to produce the full text. |
|
58 | 58 | """ |
|
59 | 59 | chain = self.getdeltachain(name, node) |
|
60 | 60 | |
|
61 | 61 | if chain[-1][ChainIndicies.BASENODE] != sha1nodeconstants.nullid: |
|
62 | 62 | # If we didn't receive a full chain, throw |
|
63 | 63 | raise KeyError((name, hex(node))) |
|
64 | 64 | |
|
65 | 65 | # The last entry in the chain is a full text, so we start our delta |
|
66 | 66 | # applies with that. |
|
67 | 67 | fulltext = chain.pop()[ChainIndicies.DATA] |
|
68 | 68 | |
|
69 | 69 | text = fulltext |
|
70 | 70 | while chain: |
|
71 | 71 | delta = chain.pop()[ChainIndicies.DATA] |
|
72 | 72 | text = mdiff.patches(text, [delta]) |
|
73 | 73 | |
|
74 | 74 | return text |
|
75 | 75 | |
|
76 | 76 | @basestore.baseunionstore.retriable |
|
77 | 77 | def getdelta(self, name, node): |
|
78 | 78 | """Return the single delta entry for the given name/node pair.""" |
|
79 | 79 | for store in self.stores: |
|
80 | 80 | try: |
|
81 | 81 | return store.getdelta(name, node) |
|
82 | 82 | except KeyError: |
|
83 | 83 | pass |
|
84 | 84 | |
|
85 | 85 | raise KeyError((name, hex(node))) |
|
86 | 86 | |
|
87 | 87 | def getdeltachain(self, name, node): |
|
88 | 88 | """Returns the deltachain for the given name/node pair. |
|
89 | 89 | |
|
90 | 90 | Returns an ordered list of: |
|
91 | 91 | |
|
92 | 92 | [(name, node, deltabasename, deltabasenode, deltacontent),...] |
|
93 | 93 | |
|
94 | 94 | where the chain is terminated by a full text entry with a nullid |
|
95 | 95 | deltabasenode. |
|
96 | 96 | """ |
|
97 | 97 | chain = self._getpartialchain(name, node) |
|
98 | 98 | while chain[-1][ChainIndicies.BASENODE] != sha1nodeconstants.nullid: |
|
99 | 99 | x, x, deltabasename, deltabasenode, x = chain[-1] |
|
100 | 100 | try: |
|
101 | 101 | morechain = self._getpartialchain(deltabasename, deltabasenode) |
|
102 | 102 | chain.extend(morechain) |
|
103 | 103 | except KeyError: |
|
104 | 104 | # If we allow incomplete chains, don't throw. |
|
105 | 105 | if not self.allowincomplete: |
|
106 | 106 | raise |
|
107 | 107 | break |
|
108 | 108 | |
|
109 | 109 | return chain |
|
110 | 110 | |
|
111 | 111 | @basestore.baseunionstore.retriable |
|
112 | 112 | def getmeta(self, name, node): |
|
113 | 113 | """Returns the metadata dict for given node.""" |
|
114 | 114 | for store in self.stores: |
|
115 | 115 | try: |
|
116 | 116 | return store.getmeta(name, node) |
|
117 | 117 | except KeyError: |
|
118 | 118 | pass |
|
119 | 119 | raise KeyError((name, hex(node))) |
|
120 | 120 | |
|
121 | 121 | def getmetrics(self): |
|
122 | 122 | metrics = [s.getmetrics() for s in self.stores] |
|
123 | 123 | return shallowutil.sumdicts(*metrics) |
|
124 | 124 | |
|
125 | 125 | @basestore.baseunionstore.retriable |
|
126 | 126 | def _getpartialchain(self, name, node): |
|
127 | 127 | """Returns a partial delta chain for the given name/node pair. |
|
128 | 128 | |
|
129 | 129 | A partial chain is a chain that may not be terminated in a full-text. |
|
130 | 130 | """ |
|
131 | 131 | for store in self.stores: |
|
132 | 132 | try: |
|
133 | 133 | return store.getdeltachain(name, node) |
|
134 | 134 | except KeyError: |
|
135 | 135 | pass |
|
136 | 136 | |
|
137 | 137 | raise KeyError((name, hex(node))) |
|
138 | 138 | |
|
139 | 139 | def add(self, name, node, data): |
|
140 | 140 | raise RuntimeError( |
|
141 | 141 | b"cannot add content only to remotefilelog contentstore" |
|
142 | 142 | ) |
|
143 | 143 | |
|
144 | 144 | def getmissing(self, keys): |
|
145 | 145 | missing = keys |
|
146 | 146 | for store in self.stores: |
|
147 | 147 | if missing: |
|
148 | 148 | missing = store.getmissing(missing) |
|
149 | 149 | return missing |
|
150 | 150 | |
|
151 | 151 | def addremotefilelognode(self, name, node, data): |
|
152 | 152 | if self.writestore: |
|
153 | 153 | self.writestore.addremotefilelognode(name, node, data) |
|
154 | 154 | else: |
|
155 | 155 | raise RuntimeError(b"no writable store configured") |
|
156 | 156 | |
|
157 | 157 | def markledger(self, ledger, options=None): |
|
158 | 158 | for store in self.stores: |
|
159 | 159 | store.markledger(ledger, options) |
|
160 | 160 | |
|
161 | 161 | |
|
162 | 162 | class remotefilelogcontentstore(basestore.basestore): |
|
163 | 163 | def __init__(self, *args, **kwargs): |
|
164 | 164 | super(remotefilelogcontentstore, self).__init__(*args, **kwargs) |
|
165 | 165 | self._threaddata = threading.local() |
|
166 | 166 | |
|
167 | 167 | def get(self, name, node): |
|
168 | 168 | # return raw revision text |
|
169 | 169 | data = self._getdata(name, node) |
|
170 | 170 | |
|
171 | 171 | offset, size, flags = shallowutil.parsesizeflags(data) |
|
172 | 172 | content = data[offset : offset + size] |
|
173 | 173 | |
|
174 | 174 | ancestormap = shallowutil.ancestormap(data) |
|
175 | 175 | p1, p2, linknode, copyfrom = ancestormap[node] |
|
176 | 176 | copyrev = None |
|
177 | 177 | if copyfrom: |
|
178 | 178 | copyrev = hex(p1) |
|
179 | 179 | |
|
180 | 180 | self._updatemetacache(node, size, flags) |
|
181 | 181 | |
|
182 | 182 | # lfs tracks renames in its own metadata, remove hg copy metadata, |
|
183 | 183 | # because copy metadata will be re-added by lfs flag processor. |
|
184 | 184 | if flags & revlog.REVIDX_EXTSTORED: |
|
185 | 185 | copyrev = copyfrom = None |
|
186 | 186 | revision = shallowutil.createrevlogtext(content, copyfrom, copyrev) |
|
187 | 187 | return revision |
|
188 | 188 | |
|
189 | 189 | def getdelta(self, name, node): |
|
190 | 190 | # Since remotefilelog content stores only contain full texts, just |
|
191 | 191 | # return that. |
|
192 | 192 | revision = self.get(name, node) |
|
193 | 193 | return ( |
|
194 | 194 | revision, |
|
195 | 195 | name, |
|
196 | 196 | sha1nodeconstants.nullid, |
|
197 | 197 | self.getmeta(name, node), |
|
198 | 198 | ) |
|
199 | 199 | |
|
200 | 200 | def getdeltachain(self, name, node): |
|
201 | 201 | # Since remotefilelog content stores just contain full texts, we return |
|
202 | 202 | # a fake delta chain that just consists of a single full text revision. |
|
203 | 203 | # The nullid in the deltabasenode slot indicates that the revision is a |
|
204 | 204 | # fulltext. |
|
205 | 205 | revision = self.get(name, node) |
|
206 | 206 | return [(name, node, None, sha1nodeconstants.nullid, revision)] |
|
207 | 207 | |
|
208 | 208 | def getmeta(self, name, node): |
|
209 | 209 | self._sanitizemetacache() |
|
210 | 210 | if node != self._threaddata.metacache[0]: |
|
211 | 211 | data = self._getdata(name, node) |
|
212 | 212 | offset, size, flags = shallowutil.parsesizeflags(data) |
|
213 | 213 | self._updatemetacache(node, size, flags) |
|
214 | 214 | return self._threaddata.metacache[1] |
|
215 | 215 | |
|
216 | 216 | def add(self, name, node, data): |
|
217 | 217 | raise RuntimeError( |
|
218 | 218 | b"cannot add content only to remotefilelog contentstore" |
|
219 | 219 | ) |
|
220 | 220 | |
|
221 | 221 | def _sanitizemetacache(self): |
|
222 | 222 | metacache = getattr(self._threaddata, 'metacache', None) |
|
223 | 223 | if metacache is None: |
|
224 | 224 | self._threaddata.metacache = (None, None) # (node, meta) |
|
225 | 225 | |
|
226 | 226 | def _updatemetacache(self, node, size, flags): |
|
227 | 227 | self._sanitizemetacache() |
|
228 | 228 | if node == self._threaddata.metacache[0]: |
|
229 | 229 | return |
|
230 | 230 | meta = {constants.METAKEYFLAG: flags, constants.METAKEYSIZE: size} |
|
231 | 231 | self._threaddata.metacache = (node, meta) |
|
232 | 232 | |
|
233 | 233 | |
|
234 | 234 | class remotecontentstore(object): |
|
235 | 235 | def __init__(self, ui, fileservice, shared): |
|
236 | 236 | self._fileservice = fileservice |
|
237 | 237 | # type(shared) is usually remotefilelogcontentstore |
|
238 | 238 | self._shared = shared |
|
239 | 239 | |
|
240 | 240 | def get(self, name, node): |
|
241 | 241 | self._fileservice.prefetch( |
|
242 | 242 | [(name, hex(node))], force=True, fetchdata=True |
|
243 | 243 | ) |
|
244 | 244 | return self._shared.get(name, node) |
|
245 | 245 | |
|
246 | 246 | def getdelta(self, name, node): |
|
247 | 247 | revision = self.get(name, node) |
|
248 | 248 | return ( |
|
249 | 249 | revision, |
|
250 | 250 | name, |
|
251 | 251 | sha1nodeconstants.nullid, |
|
252 | 252 | self._shared.getmeta(name, node), |
|
253 | 253 | ) |
|
254 | 254 | |
|
255 | 255 | def getdeltachain(self, name, node): |
|
256 | 256 | # Since our remote content stores just contain full texts, we return a |
|
257 | 257 | # fake delta chain that just consists of a single full text revision. |
|
258 | 258 | # The nullid in the deltabasenode slot indicates that the revision is a |
|
259 | 259 | # fulltext. |
|
260 | 260 | revision = self.get(name, node) |
|
261 | 261 | return [(name, node, None, sha1nodeconstants.nullid, revision)] |
|
262 | 262 | |
|
263 | 263 | def getmeta(self, name, node): |
|
264 | 264 | self._fileservice.prefetch( |
|
265 | 265 | [(name, hex(node))], force=True, fetchdata=True |
|
266 | 266 | ) |
|
267 | 267 | return self._shared.getmeta(name, node) |
|
268 | 268 | |
|
269 | 269 | def add(self, name, node, data): |
|
270 | 270 | raise RuntimeError(b"cannot add to a remote store") |
|
271 | 271 | |
|
272 | 272 | def getmissing(self, keys): |
|
273 | 273 | return keys |
|
274 | 274 | |
|
275 | 275 | def markledger(self, ledger, options=None): |
|
276 | 276 | pass |
|
277 | 277 | |
|
278 | 278 | |
|
279 | 279 | class manifestrevlogstore(object): |
|
280 | 280 | def __init__(self, repo): |
|
281 | 281 | self._store = repo.store |
|
282 | 282 | self._svfs = repo.svfs |
|
283 | 283 | self._revlogs = dict() |
|
284 | 284 | self._cl = revlog.revlog(self._svfs, radix=b'00changelog.i') |
|
285 | 285 | self._repackstartlinkrev = 0 |
|
286 | 286 | |
|
287 | 287 | def get(self, name, node): |
|
288 | 288 | return self._revlog(name).rawdata(node) |
|
289 | 289 | |
|
290 | 290 | def getdelta(self, name, node): |
|
291 | 291 | revision = self.get(name, node) |
|
292 | 292 | return revision, name, self._cl.nullid, self.getmeta(name, node) |
|
293 | 293 | |
|
294 | 294 | def getdeltachain(self, name, node): |
|
295 | 295 | revision = self.get(name, node) |
|
296 | 296 | return [(name, node, None, self._cl.nullid, revision)] |
|
297 | 297 | |
|
298 | 298 | def getmeta(self, name, node): |
|
299 | 299 | rl = self._revlog(name) |
|
300 | 300 | rev = rl.rev(node) |
|
301 | 301 | return { |
|
302 | 302 | constants.METAKEYFLAG: rl.flags(rev), |
|
303 | 303 | constants.METAKEYSIZE: rl.rawsize(rev), |
|
304 | 304 | } |
|
305 | 305 | |
|
306 | 306 | def getancestors(self, name, node, known=None): |
|
307 | 307 | if known is None: |
|
308 | 308 | known = set() |
|
309 | 309 | if node in known: |
|
310 | 310 | return [] |
|
311 | 311 | |
|
312 | 312 | rl = self._revlog(name) |
|
313 | 313 | ancestors = {} |
|
314 | 314 | missing = {node} |
|
315 | 315 | for ancrev in rl.ancestors([rl.rev(node)], inclusive=True): |
|
316 | 316 | ancnode = rl.node(ancrev) |
|
317 | 317 | missing.discard(ancnode) |
|
318 | 318 | |
|
319 | 319 | p1, p2 = rl.parents(ancnode) |
|
320 | 320 | if p1 != self._cl.nullid and p1 not in known: |
|
321 | 321 | missing.add(p1) |
|
322 | 322 | if p2 != self._cl.nullid and p2 not in known: |
|
323 | 323 | missing.add(p2) |
|
324 | 324 | |
|
325 | 325 | linknode = self._cl.node(rl.linkrev(ancrev)) |
|
326 | 326 | ancestors[rl.node(ancrev)] = (p1, p2, linknode, b'') |
|
327 | 327 | if not missing: |
|
328 | 328 | break |
|
329 | 329 | return ancestors |
|
330 | 330 | |
|
331 | 331 | def getnodeinfo(self, name, node): |
|
332 | 332 | cl = self._cl |
|
333 | 333 | rl = self._revlog(name) |
|
334 | 334 | parents = rl.parents(node) |
|
335 | 335 | linkrev = rl.linkrev(rl.rev(node)) |
|
336 | 336 | return (parents[0], parents[1], cl.node(linkrev), None) |
|
337 | 337 | |
|
338 | 338 | def add(self, *args): |
|
339 | 339 | raise RuntimeError(b"cannot add to a revlog store") |
|
340 | 340 | |
|
341 | 341 | def _revlog(self, name): |
|
342 | 342 | rl = self._revlogs.get(name) |
|
343 | 343 | if rl is None: |
|
344 | 344 | revlogname = b'00manifesttree' |
|
345 | 345 | if name != b'': |
|
346 | 346 | revlogname = b'meta/%s/00manifest' % name |
|
347 | 347 | rl = revlog.revlog(self._svfs, radix=revlogname) |
|
348 | 348 | self._revlogs[name] = rl |
|
349 | 349 | return rl |
|
350 | 350 | |
|
351 | 351 | def getmissing(self, keys): |
|
352 | 352 | missing = [] |
|
353 | 353 | for name, node in keys: |
|
354 | 354 | mfrevlog = self._revlog(name) |
|
355 | 355 | if node not in mfrevlog.nodemap: |
|
356 | 356 | missing.append((name, node)) |
|
357 | 357 | |
|
358 | 358 | return missing |
|
359 | 359 | |
|
360 | 360 | def setrepacklinkrevrange(self, startrev, endrev): |
|
361 | 361 | self._repackstartlinkrev = startrev |
|
362 | 362 | self._repackendlinkrev = endrev |
|
363 | 363 | |
|
364 | 364 | def markledger(self, ledger, options=None): |
|
365 | 365 | if options and options.get(constants.OPTION_PACKSONLY): |
|
366 | 366 | return |
|
367 | 367 | treename = b'' |
|
368 | 368 | rl = revlog.revlog(self._svfs, radix=b'00manifesttree') |
|
369 | 369 | startlinkrev = self._repackstartlinkrev |
|
370 | 370 | endlinkrev = self._repackendlinkrev |
|
371 | 371 | for rev in pycompat.xrange(len(rl) - 1, -1, -1): |
|
372 | 372 | linkrev = rl.linkrev(rev) |
|
373 | 373 | if linkrev < startlinkrev: |
|
374 | 374 | break |
|
375 | 375 | if linkrev > endlinkrev: |
|
376 | 376 | continue |
|
377 | 377 | node = rl.node(rev) |
|
378 | 378 | ledger.markdataentry(self, treename, node) |
|
379 | 379 | ledger.markhistoryentry(self, treename, node) |
|
380 | 380 | |
|
381 |
for t, path |
|
|
381 | for t, path, size in self._store.datafiles(): | |
|
382 | 382 | if path[:5] != b'meta/' or path[-2:] != b'.i': |
|
383 | 383 | continue |
|
384 | 384 | |
|
385 | 385 | treename = path[5 : -len(b'/00manifest')] |
|
386 | 386 | |
|
387 | 387 | rl = revlog.revlog(self._svfs, indexfile=path[:-2]) |
|
388 | 388 | for rev in pycompat.xrange(len(rl) - 1, -1, -1): |
|
389 | 389 | linkrev = rl.linkrev(rev) |
|
390 | 390 | if linkrev < startlinkrev: |
|
391 | 391 | break |
|
392 | 392 | if linkrev > endlinkrev: |
|
393 | 393 | continue |
|
394 | 394 | node = rl.node(rev) |
|
395 | 395 | ledger.markdataentry(self, treename, node) |
|
396 | 396 | ledger.markhistoryentry(self, treename, node) |
|
397 | 397 | |
|
398 | 398 | def cleanup(self, ledger): |
|
399 | 399 | pass |
@@ -1,441 +1,441 b'' | |||
|
1 | 1 | # remotefilelogserver.py - server logic for a remotefilelog server |
|
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 | from __future__ import absolute_import |
|
8 | 8 | |
|
9 | 9 | import errno |
|
10 | 10 | import os |
|
11 | 11 | import stat |
|
12 | 12 | import time |
|
13 | 13 | import zlib |
|
14 | 14 | |
|
15 | 15 | from mercurial.i18n import _ |
|
16 | 16 | from mercurial.node import bin, hex |
|
17 | 17 | from mercurial.pycompat import open |
|
18 | 18 | from mercurial import ( |
|
19 | 19 | changegroup, |
|
20 | 20 | changelog, |
|
21 | 21 | context, |
|
22 | 22 | error, |
|
23 | 23 | extensions, |
|
24 | 24 | match, |
|
25 | 25 | pycompat, |
|
26 | 26 | scmutil, |
|
27 | 27 | store, |
|
28 | 28 | streamclone, |
|
29 | 29 | util, |
|
30 | 30 | wireprotoserver, |
|
31 | 31 | wireprototypes, |
|
32 | 32 | wireprotov1server, |
|
33 | 33 | ) |
|
34 | 34 | from . import ( |
|
35 | 35 | constants, |
|
36 | 36 | shallowutil, |
|
37 | 37 | ) |
|
38 | 38 | |
|
39 | 39 | _sshv1server = wireprotoserver.sshv1protocolhandler |
|
40 | 40 | |
|
41 | 41 | |
|
42 | 42 | def setupserver(ui, repo): |
|
43 | 43 | """Sets up a normal Mercurial repo so it can serve files to shallow repos.""" |
|
44 | 44 | onetimesetup(ui) |
|
45 | 45 | |
|
46 | 46 | # don't send files to shallow clients during pulls |
|
47 | 47 | def generatefiles( |
|
48 | 48 | orig, self, changedfiles, linknodes, commonrevs, source, *args, **kwargs |
|
49 | 49 | ): |
|
50 | 50 | caps = self._bundlecaps or [] |
|
51 | 51 | if constants.BUNDLE2_CAPABLITY in caps: |
|
52 | 52 | # only send files that don't match the specified patterns |
|
53 | 53 | includepattern = None |
|
54 | 54 | excludepattern = None |
|
55 | 55 | for cap in self._bundlecaps or []: |
|
56 | 56 | if cap.startswith(b"includepattern="): |
|
57 | 57 | includepattern = cap[len(b"includepattern=") :].split(b'\0') |
|
58 | 58 | elif cap.startswith(b"excludepattern="): |
|
59 | 59 | excludepattern = cap[len(b"excludepattern=") :].split(b'\0') |
|
60 | 60 | |
|
61 | 61 | m = match.always() |
|
62 | 62 | if includepattern or excludepattern: |
|
63 | 63 | m = match.match( |
|
64 | 64 | repo.root, b'', None, includepattern, excludepattern |
|
65 | 65 | ) |
|
66 | 66 | |
|
67 | 67 | changedfiles = list([f for f in changedfiles if not m(f)]) |
|
68 | 68 | return orig( |
|
69 | 69 | self, changedfiles, linknodes, commonrevs, source, *args, **kwargs |
|
70 | 70 | ) |
|
71 | 71 | |
|
72 | 72 | extensions.wrapfunction( |
|
73 | 73 | changegroup.cgpacker, b'generatefiles', generatefiles |
|
74 | 74 | ) |
|
75 | 75 | |
|
76 | 76 | |
|
77 | 77 | onetime = False |
|
78 | 78 | |
|
79 | 79 | |
|
80 | 80 | def onetimesetup(ui): |
|
81 | 81 | """Configures the wireprotocol for both clients and servers.""" |
|
82 | 82 | global onetime |
|
83 | 83 | if onetime: |
|
84 | 84 | return |
|
85 | 85 | onetime = True |
|
86 | 86 | |
|
87 | 87 | # support file content requests |
|
88 | 88 | wireprotov1server.wireprotocommand( |
|
89 | 89 | b'x_rfl_getflogheads', b'path', permission=b'pull' |
|
90 | 90 | )(getflogheads) |
|
91 | 91 | wireprotov1server.wireprotocommand( |
|
92 | 92 | b'x_rfl_getfiles', b'', permission=b'pull' |
|
93 | 93 | )(getfiles) |
|
94 | 94 | wireprotov1server.wireprotocommand( |
|
95 | 95 | b'x_rfl_getfile', b'file node', permission=b'pull' |
|
96 | 96 | )(getfile) |
|
97 | 97 | |
|
98 | 98 | class streamstate(object): |
|
99 | 99 | match = None |
|
100 | 100 | shallowremote = False |
|
101 | 101 | noflatmf = False |
|
102 | 102 | |
|
103 | 103 | state = streamstate() |
|
104 | 104 | |
|
105 | 105 | def stream_out_shallow(repo, proto, other): |
|
106 | 106 | includepattern = None |
|
107 | 107 | excludepattern = None |
|
108 | 108 | raw = other.get(b'includepattern') |
|
109 | 109 | if raw: |
|
110 | 110 | includepattern = raw.split(b'\0') |
|
111 | 111 | raw = other.get(b'excludepattern') |
|
112 | 112 | if raw: |
|
113 | 113 | excludepattern = raw.split(b'\0') |
|
114 | 114 | |
|
115 | 115 | oldshallow = state.shallowremote |
|
116 | 116 | oldmatch = state.match |
|
117 | 117 | oldnoflatmf = state.noflatmf |
|
118 | 118 | try: |
|
119 | 119 | state.shallowremote = True |
|
120 | 120 | state.match = match.always() |
|
121 | 121 | state.noflatmf = other.get(b'noflatmanifest') == b'True' |
|
122 | 122 | if includepattern or excludepattern: |
|
123 | 123 | state.match = match.match( |
|
124 | 124 | repo.root, b'', None, includepattern, excludepattern |
|
125 | 125 | ) |
|
126 | 126 | streamres = wireprotov1server.stream(repo, proto) |
|
127 | 127 | |
|
128 | 128 | # Force the first value to execute, so the file list is computed |
|
129 | 129 | # within the try/finally scope |
|
130 | 130 | first = next(streamres.gen) |
|
131 | 131 | second = next(streamres.gen) |
|
132 | 132 | |
|
133 | 133 | def gen(): |
|
134 | 134 | yield first |
|
135 | 135 | yield second |
|
136 | 136 | for value in streamres.gen: |
|
137 | 137 | yield value |
|
138 | 138 | |
|
139 | 139 | return wireprototypes.streamres(gen()) |
|
140 | 140 | finally: |
|
141 | 141 | state.shallowremote = oldshallow |
|
142 | 142 | state.match = oldmatch |
|
143 | 143 | state.noflatmf = oldnoflatmf |
|
144 | 144 | |
|
145 | 145 | wireprotov1server.commands[b'stream_out_shallow'] = ( |
|
146 | 146 | stream_out_shallow, |
|
147 | 147 | b'*', |
|
148 | 148 | ) |
|
149 | 149 | |
|
150 | 150 | # don't clone filelogs to shallow clients |
|
151 | 151 | def _walkstreamfiles(orig, repo, matcher=None): |
|
152 | 152 | if state.shallowremote: |
|
153 | 153 | # if we are shallow ourselves, stream our local commits |
|
154 | 154 | if shallowutil.isenabled(repo): |
|
155 | 155 | striplen = len(repo.store.path) + 1 |
|
156 | 156 | readdir = repo.store.rawvfs.readdir |
|
157 | 157 | visit = [os.path.join(repo.store.path, b'data')] |
|
158 | 158 | while visit: |
|
159 | 159 | p = visit.pop() |
|
160 | 160 | for f, kind, st in readdir(p, stat=True): |
|
161 | 161 | fp = p + b'/' + f |
|
162 | 162 | if kind == stat.S_IFREG: |
|
163 | 163 | if not fp.endswith(b'.i') and not fp.endswith( |
|
164 | 164 | b'.d' |
|
165 | 165 | ): |
|
166 | 166 | n = util.pconvert(fp[striplen:]) |
|
167 | 167 | d = store.decodedir(n) |
|
168 | 168 | t = store.FILETYPE_OTHER |
|
169 |
yield (t, d, |
|
|
169 | yield (t, d, st.st_size) | |
|
170 | 170 | if kind == stat.S_IFDIR: |
|
171 | 171 | visit.append(fp) |
|
172 | 172 | |
|
173 | 173 | if scmutil.istreemanifest(repo): |
|
174 |
for (t, u |
|
|
174 | for (t, u, s) in repo.store.datafiles(): | |
|
175 | 175 | if u.startswith(b'meta/') and ( |
|
176 | 176 | u.endswith(b'.i') or u.endswith(b'.d') |
|
177 | 177 | ): |
|
178 |
yield (t, u, |
|
|
178 | yield (t, u, s) | |
|
179 | 179 | |
|
180 | 180 | # Return .d and .i files that do not match the shallow pattern |
|
181 | 181 | match = state.match |
|
182 | 182 | if match and not match.always(): |
|
183 |
for (t, u |
|
|
183 | for (t, u, s) in repo.store.datafiles(): | |
|
184 | 184 | f = u[5:-2] # trim data/... and .i/.d |
|
185 | 185 | if not state.match(f): |
|
186 |
yield (t, u, |
|
|
186 | yield (t, u, s) | |
|
187 | 187 | |
|
188 | 188 | for x in repo.store.topfiles(): |
|
189 | 189 | if state.noflatmf and x[1][:11] == b'00manifest.': |
|
190 | 190 | continue |
|
191 | 191 | yield x |
|
192 | 192 | |
|
193 | 193 | elif shallowutil.isenabled(repo): |
|
194 | 194 | # don't allow cloning from a shallow repo to a full repo |
|
195 | 195 | # since it would require fetching every version of every |
|
196 | 196 | # file in order to create the revlogs. |
|
197 | 197 | raise error.Abort( |
|
198 | 198 | _(b"Cannot clone from a shallow repo to a full repo.") |
|
199 | 199 | ) |
|
200 | 200 | else: |
|
201 | 201 | for x in orig(repo, matcher): |
|
202 | 202 | yield x |
|
203 | 203 | |
|
204 | 204 | extensions.wrapfunction(streamclone, b'_walkstreamfiles', _walkstreamfiles) |
|
205 | 205 | |
|
206 | 206 | # expose remotefilelog capabilities |
|
207 | 207 | def _capabilities(orig, repo, proto): |
|
208 | 208 | caps = orig(repo, proto) |
|
209 | 209 | if shallowutil.isenabled(repo) or ui.configbool( |
|
210 | 210 | b'remotefilelog', b'server' |
|
211 | 211 | ): |
|
212 | 212 | if isinstance(proto, _sshv1server): |
|
213 | 213 | # legacy getfiles method which only works over ssh |
|
214 | 214 | caps.append(constants.NETWORK_CAP_LEGACY_SSH_GETFILES) |
|
215 | 215 | caps.append(b'x_rfl_getflogheads') |
|
216 | 216 | caps.append(b'x_rfl_getfile') |
|
217 | 217 | return caps |
|
218 | 218 | |
|
219 | 219 | extensions.wrapfunction(wireprotov1server, b'_capabilities', _capabilities) |
|
220 | 220 | |
|
221 | 221 | def _adjustlinkrev(orig, self, *args, **kwargs): |
|
222 | 222 | # When generating file blobs, taking the real path is too slow on large |
|
223 | 223 | # repos, so force it to just return the linkrev directly. |
|
224 | 224 | repo = self._repo |
|
225 | 225 | if util.safehasattr(repo, b'forcelinkrev') and repo.forcelinkrev: |
|
226 | 226 | return self._filelog.linkrev(self._filelog.rev(self._filenode)) |
|
227 | 227 | return orig(self, *args, **kwargs) |
|
228 | 228 | |
|
229 | 229 | extensions.wrapfunction( |
|
230 | 230 | context.basefilectx, b'_adjustlinkrev', _adjustlinkrev |
|
231 | 231 | ) |
|
232 | 232 | |
|
233 | 233 | def _iscmd(orig, cmd): |
|
234 | 234 | if cmd == b'x_rfl_getfiles': |
|
235 | 235 | return False |
|
236 | 236 | return orig(cmd) |
|
237 | 237 | |
|
238 | 238 | extensions.wrapfunction(wireprotoserver, b'iscmd', _iscmd) |
|
239 | 239 | |
|
240 | 240 | |
|
241 | 241 | def _loadfileblob(repo, cachepath, path, node): |
|
242 | 242 | filecachepath = os.path.join(cachepath, path, hex(node)) |
|
243 | 243 | if not os.path.exists(filecachepath) or os.path.getsize(filecachepath) == 0: |
|
244 | 244 | filectx = repo.filectx(path, fileid=node) |
|
245 | 245 | if filectx.node() == repo.nullid: |
|
246 | 246 | repo.changelog = changelog.changelog(repo.svfs) |
|
247 | 247 | filectx = repo.filectx(path, fileid=node) |
|
248 | 248 | |
|
249 | 249 | text = createfileblob(filectx) |
|
250 | 250 | # TODO configurable compression engines |
|
251 | 251 | text = zlib.compress(text) |
|
252 | 252 | |
|
253 | 253 | # everything should be user & group read/writable |
|
254 | 254 | oldumask = os.umask(0o002) |
|
255 | 255 | try: |
|
256 | 256 | dirname = os.path.dirname(filecachepath) |
|
257 | 257 | if not os.path.exists(dirname): |
|
258 | 258 | try: |
|
259 | 259 | os.makedirs(dirname) |
|
260 | 260 | except OSError as ex: |
|
261 | 261 | if ex.errno != errno.EEXIST: |
|
262 | 262 | raise |
|
263 | 263 | |
|
264 | 264 | f = None |
|
265 | 265 | try: |
|
266 | 266 | f = util.atomictempfile(filecachepath, b"wb") |
|
267 | 267 | f.write(text) |
|
268 | 268 | except (IOError, OSError): |
|
269 | 269 | # Don't abort if the user only has permission to read, |
|
270 | 270 | # and not write. |
|
271 | 271 | pass |
|
272 | 272 | finally: |
|
273 | 273 | if f: |
|
274 | 274 | f.close() |
|
275 | 275 | finally: |
|
276 | 276 | os.umask(oldumask) |
|
277 | 277 | else: |
|
278 | 278 | with open(filecachepath, b"rb") as f: |
|
279 | 279 | text = f.read() |
|
280 | 280 | return text |
|
281 | 281 | |
|
282 | 282 | |
|
283 | 283 | def getflogheads(repo, proto, path): |
|
284 | 284 | """A server api for requesting a filelog's heads""" |
|
285 | 285 | flog = repo.file(path) |
|
286 | 286 | heads = flog.heads() |
|
287 | 287 | return b'\n'.join((hex(head) for head in heads if head != repo.nullid)) |
|
288 | 288 | |
|
289 | 289 | |
|
290 | 290 | def getfile(repo, proto, file, node): |
|
291 | 291 | """A server api for requesting a particular version of a file. Can be used |
|
292 | 292 | in batches to request many files at once. The return protocol is: |
|
293 | 293 | <errorcode>\0<data/errormsg> where <errorcode> is 0 for success or |
|
294 | 294 | non-zero for an error. |
|
295 | 295 | |
|
296 | 296 | data is a compressed blob with revlog flag and ancestors information. See |
|
297 | 297 | createfileblob for its content. |
|
298 | 298 | """ |
|
299 | 299 | if shallowutil.isenabled(repo): |
|
300 | 300 | return b'1\0' + _(b'cannot fetch remote files from shallow repo') |
|
301 | 301 | cachepath = repo.ui.config(b"remotefilelog", b"servercachepath") |
|
302 | 302 | if not cachepath: |
|
303 | 303 | cachepath = os.path.join(repo.path, b"remotefilelogcache") |
|
304 | 304 | node = bin(node.strip()) |
|
305 | 305 | if node == repo.nullid: |
|
306 | 306 | return b'0\0' |
|
307 | 307 | return b'0\0' + _loadfileblob(repo, cachepath, file, node) |
|
308 | 308 | |
|
309 | 309 | |
|
310 | 310 | def getfiles(repo, proto): |
|
311 | 311 | """A server api for requesting particular versions of particular files.""" |
|
312 | 312 | if shallowutil.isenabled(repo): |
|
313 | 313 | raise error.Abort(_(b'cannot fetch remote files from shallow repo')) |
|
314 | 314 | if not isinstance(proto, _sshv1server): |
|
315 | 315 | raise error.Abort(_(b'cannot fetch remote files over non-ssh protocol')) |
|
316 | 316 | |
|
317 | 317 | def streamer(): |
|
318 | 318 | fin = proto._fin |
|
319 | 319 | |
|
320 | 320 | cachepath = repo.ui.config(b"remotefilelog", b"servercachepath") |
|
321 | 321 | if not cachepath: |
|
322 | 322 | cachepath = os.path.join(repo.path, b"remotefilelogcache") |
|
323 | 323 | |
|
324 | 324 | while True: |
|
325 | 325 | request = fin.readline()[:-1] |
|
326 | 326 | if not request: |
|
327 | 327 | break |
|
328 | 328 | |
|
329 | 329 | node = bin(request[:40]) |
|
330 | 330 | if node == repo.nullid: |
|
331 | 331 | yield b'0\n' |
|
332 | 332 | continue |
|
333 | 333 | |
|
334 | 334 | path = request[40:] |
|
335 | 335 | |
|
336 | 336 | text = _loadfileblob(repo, cachepath, path, node) |
|
337 | 337 | |
|
338 | 338 | yield b'%d\n%s' % (len(text), text) |
|
339 | 339 | |
|
340 | 340 | # it would be better to only flush after processing a whole batch |
|
341 | 341 | # but currently we don't know if there are more requests coming |
|
342 | 342 | proto._fout.flush() |
|
343 | 343 | |
|
344 | 344 | return wireprototypes.streamres(streamer()) |
|
345 | 345 | |
|
346 | 346 | |
|
347 | 347 | def createfileblob(filectx): |
|
348 | 348 | """ |
|
349 | 349 | format: |
|
350 | 350 | v0: |
|
351 | 351 | str(len(rawtext)) + '\0' + rawtext + ancestortext |
|
352 | 352 | v1: |
|
353 | 353 | 'v1' + '\n' + metalist + '\0' + rawtext + ancestortext |
|
354 | 354 | metalist := metalist + '\n' + meta | meta |
|
355 | 355 | meta := sizemeta | flagmeta |
|
356 | 356 | sizemeta := METAKEYSIZE + str(len(rawtext)) |
|
357 | 357 | flagmeta := METAKEYFLAG + str(flag) |
|
358 | 358 | |
|
359 | 359 | note: sizemeta must exist. METAKEYFLAG and METAKEYSIZE must have a |
|
360 | 360 | length of 1. |
|
361 | 361 | """ |
|
362 | 362 | flog = filectx.filelog() |
|
363 | 363 | frev = filectx.filerev() |
|
364 | 364 | revlogflags = flog._revlog.flags(frev) |
|
365 | 365 | if revlogflags == 0: |
|
366 | 366 | # normal files |
|
367 | 367 | text = filectx.data() |
|
368 | 368 | else: |
|
369 | 369 | # lfs, read raw revision data |
|
370 | 370 | text = flog.rawdata(frev) |
|
371 | 371 | |
|
372 | 372 | repo = filectx._repo |
|
373 | 373 | |
|
374 | 374 | ancestors = [filectx] |
|
375 | 375 | |
|
376 | 376 | try: |
|
377 | 377 | repo.forcelinkrev = True |
|
378 | 378 | ancestors.extend([f for f in filectx.ancestors()]) |
|
379 | 379 | |
|
380 | 380 | ancestortext = b"" |
|
381 | 381 | for ancestorctx in ancestors: |
|
382 | 382 | parents = ancestorctx.parents() |
|
383 | 383 | p1 = repo.nullid |
|
384 | 384 | p2 = repo.nullid |
|
385 | 385 | if len(parents) > 0: |
|
386 | 386 | p1 = parents[0].filenode() |
|
387 | 387 | if len(parents) > 1: |
|
388 | 388 | p2 = parents[1].filenode() |
|
389 | 389 | |
|
390 | 390 | copyname = b"" |
|
391 | 391 | rename = ancestorctx.renamed() |
|
392 | 392 | if rename: |
|
393 | 393 | copyname = rename[0] |
|
394 | 394 | linknode = ancestorctx.node() |
|
395 | 395 | ancestortext += b"%s%s%s%s%s\0" % ( |
|
396 | 396 | ancestorctx.filenode(), |
|
397 | 397 | p1, |
|
398 | 398 | p2, |
|
399 | 399 | linknode, |
|
400 | 400 | copyname, |
|
401 | 401 | ) |
|
402 | 402 | finally: |
|
403 | 403 | repo.forcelinkrev = False |
|
404 | 404 | |
|
405 | 405 | header = shallowutil.buildfileblobheader(len(text), revlogflags) |
|
406 | 406 | |
|
407 | 407 | return b"%s\0%s%s" % (header, text, ancestortext) |
|
408 | 408 | |
|
409 | 409 | |
|
410 | 410 | def gcserver(ui, repo): |
|
411 | 411 | if not repo.ui.configbool(b"remotefilelog", b"server"): |
|
412 | 412 | return |
|
413 | 413 | |
|
414 | 414 | neededfiles = set() |
|
415 | 415 | heads = repo.revs(b"heads(tip~25000:) - null") |
|
416 | 416 | |
|
417 | 417 | cachepath = repo.vfs.join(b"remotefilelogcache") |
|
418 | 418 | for head in heads: |
|
419 | 419 | mf = repo[head].manifest() |
|
420 | 420 | for filename, filenode in pycompat.iteritems(mf): |
|
421 | 421 | filecachepath = os.path.join(cachepath, filename, hex(filenode)) |
|
422 | 422 | neededfiles.add(filecachepath) |
|
423 | 423 | |
|
424 | 424 | # delete unneeded older files |
|
425 | 425 | days = repo.ui.configint(b"remotefilelog", b"serverexpiration") |
|
426 | 426 | expiration = time.time() - (days * 24 * 60 * 60) |
|
427 | 427 | |
|
428 | 428 | progress = ui.makeprogress(_(b"removing old server cache"), unit=b"files") |
|
429 | 429 | progress.update(0) |
|
430 | 430 | for root, dirs, files in os.walk(cachepath): |
|
431 | 431 | for file in files: |
|
432 | 432 | filepath = os.path.join(root, file) |
|
433 | 433 | progress.increment() |
|
434 | 434 | if filepath in neededfiles: |
|
435 | 435 | continue |
|
436 | 436 | |
|
437 | 437 | stat = os.stat(filepath) |
|
438 | 438 | if stat.st_mtime < expiration: |
|
439 | 439 | os.remove(filepath) |
|
440 | 440 | |
|
441 | 441 | progress.complete() |
@@ -1,563 +1,563 b'' | |||
|
1 | 1 | # repair.py - functions for repository repair for mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2005, 2006 Chris Mason <mason@suse.com> |
|
4 | 4 | # Copyright 2007 Olivia Mackall |
|
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 | from __future__ import absolute_import |
|
10 | 10 | |
|
11 | 11 | import errno |
|
12 | 12 | |
|
13 | 13 | from .i18n import _ |
|
14 | 14 | from .node import ( |
|
15 | 15 | hex, |
|
16 | 16 | short, |
|
17 | 17 | ) |
|
18 | 18 | from . import ( |
|
19 | 19 | bundle2, |
|
20 | 20 | changegroup, |
|
21 | 21 | discovery, |
|
22 | 22 | error, |
|
23 | 23 | exchange, |
|
24 | 24 | obsolete, |
|
25 | 25 | obsutil, |
|
26 | 26 | pathutil, |
|
27 | 27 | phases, |
|
28 | 28 | pycompat, |
|
29 | 29 | requirements, |
|
30 | 30 | scmutil, |
|
31 | 31 | util, |
|
32 | 32 | ) |
|
33 | 33 | from .utils import ( |
|
34 | 34 | hashutil, |
|
35 | 35 | stringutil, |
|
36 | 36 | urlutil, |
|
37 | 37 | ) |
|
38 | 38 | |
|
39 | 39 | |
|
40 | 40 | def backupbundle( |
|
41 | 41 | repo, bases, heads, node, suffix, compress=True, obsolescence=True |
|
42 | 42 | ): |
|
43 | 43 | """create a bundle with the specified revisions as a backup""" |
|
44 | 44 | |
|
45 | 45 | backupdir = b"strip-backup" |
|
46 | 46 | vfs = repo.vfs |
|
47 | 47 | if not vfs.isdir(backupdir): |
|
48 | 48 | vfs.mkdir(backupdir) |
|
49 | 49 | |
|
50 | 50 | # Include a hash of all the nodes in the filename for uniqueness |
|
51 | 51 | allcommits = repo.set(b'%ln::%ln', bases, heads) |
|
52 | 52 | allhashes = sorted(c.hex() for c in allcommits) |
|
53 | 53 | totalhash = hashutil.sha1(b''.join(allhashes)).digest() |
|
54 | 54 | name = b"%s/%s-%s-%s.hg" % ( |
|
55 | 55 | backupdir, |
|
56 | 56 | short(node), |
|
57 | 57 | hex(totalhash[:4]), |
|
58 | 58 | suffix, |
|
59 | 59 | ) |
|
60 | 60 | |
|
61 | 61 | cgversion = changegroup.localversion(repo) |
|
62 | 62 | comp = None |
|
63 | 63 | if cgversion != b'01': |
|
64 | 64 | bundletype = b"HG20" |
|
65 | 65 | if compress: |
|
66 | 66 | comp = b'BZ' |
|
67 | 67 | elif compress: |
|
68 | 68 | bundletype = b"HG10BZ" |
|
69 | 69 | else: |
|
70 | 70 | bundletype = b"HG10UN" |
|
71 | 71 | |
|
72 | 72 | outgoing = discovery.outgoing(repo, missingroots=bases, ancestorsof=heads) |
|
73 | 73 | contentopts = { |
|
74 | 74 | b'cg.version': cgversion, |
|
75 | 75 | b'obsolescence': obsolescence, |
|
76 | 76 | b'phases': True, |
|
77 | 77 | } |
|
78 | 78 | return bundle2.writenewbundle( |
|
79 | 79 | repo.ui, |
|
80 | 80 | repo, |
|
81 | 81 | b'strip', |
|
82 | 82 | name, |
|
83 | 83 | bundletype, |
|
84 | 84 | outgoing, |
|
85 | 85 | contentopts, |
|
86 | 86 | vfs, |
|
87 | 87 | compression=comp, |
|
88 | 88 | ) |
|
89 | 89 | |
|
90 | 90 | |
|
91 | 91 | def _collectfiles(repo, striprev): |
|
92 | 92 | """find out the filelogs affected by the strip""" |
|
93 | 93 | files = set() |
|
94 | 94 | |
|
95 | 95 | for x in pycompat.xrange(striprev, len(repo)): |
|
96 | 96 | files.update(repo[x].files()) |
|
97 | 97 | |
|
98 | 98 | return sorted(files) |
|
99 | 99 | |
|
100 | 100 | |
|
101 | 101 | def _collectrevlog(revlog, striprev): |
|
102 | 102 | _, brokenset = revlog.getstrippoint(striprev) |
|
103 | 103 | return [revlog.linkrev(r) for r in brokenset] |
|
104 | 104 | |
|
105 | 105 | |
|
106 | 106 | def _collectbrokencsets(repo, files, striprev): |
|
107 | 107 | """return the changesets which will be broken by the truncation""" |
|
108 | 108 | s = set() |
|
109 | 109 | |
|
110 | 110 | for revlog in manifestrevlogs(repo): |
|
111 | 111 | s.update(_collectrevlog(revlog, striprev)) |
|
112 | 112 | for fname in files: |
|
113 | 113 | s.update(_collectrevlog(repo.file(fname), striprev)) |
|
114 | 114 | |
|
115 | 115 | return s |
|
116 | 116 | |
|
117 | 117 | |
|
118 | 118 | def strip(ui, repo, nodelist, backup=True, topic=b'backup'): |
|
119 | 119 | # This function requires the caller to lock the repo, but it operates |
|
120 | 120 | # within a transaction of its own, and thus requires there to be no current |
|
121 | 121 | # transaction when it is called. |
|
122 | 122 | if repo.currenttransaction() is not None: |
|
123 | 123 | raise error.ProgrammingError(b'cannot strip from inside a transaction') |
|
124 | 124 | |
|
125 | 125 | # Simple way to maintain backwards compatibility for this |
|
126 | 126 | # argument. |
|
127 | 127 | if backup in [b'none', b'strip']: |
|
128 | 128 | backup = False |
|
129 | 129 | |
|
130 | 130 | repo = repo.unfiltered() |
|
131 | 131 | repo.destroying() |
|
132 | 132 | vfs = repo.vfs |
|
133 | 133 | # load bookmark before changelog to avoid side effect from outdated |
|
134 | 134 | # changelog (see repo._refreshchangelog) |
|
135 | 135 | repo._bookmarks |
|
136 | 136 | cl = repo.changelog |
|
137 | 137 | |
|
138 | 138 | # TODO handle undo of merge sets |
|
139 | 139 | if isinstance(nodelist, bytes): |
|
140 | 140 | nodelist = [nodelist] |
|
141 | 141 | striplist = [cl.rev(node) for node in nodelist] |
|
142 | 142 | striprev = min(striplist) |
|
143 | 143 | |
|
144 | 144 | files = _collectfiles(repo, striprev) |
|
145 | 145 | saverevs = _collectbrokencsets(repo, files, striprev) |
|
146 | 146 | |
|
147 | 147 | # Some revisions with rev > striprev may not be descendants of striprev. |
|
148 | 148 | # We have to find these revisions and put them in a bundle, so that |
|
149 | 149 | # we can restore them after the truncations. |
|
150 | 150 | # To create the bundle we use repo.changegroupsubset which requires |
|
151 | 151 | # the list of heads and bases of the set of interesting revisions. |
|
152 | 152 | # (head = revision in the set that has no descendant in the set; |
|
153 | 153 | # base = revision in the set that has no ancestor in the set) |
|
154 | 154 | tostrip = set(striplist) |
|
155 | 155 | saveheads = set(saverevs) |
|
156 | 156 | for r in cl.revs(start=striprev + 1): |
|
157 | 157 | if any(p in tostrip for p in cl.parentrevs(r)): |
|
158 | 158 | tostrip.add(r) |
|
159 | 159 | |
|
160 | 160 | if r not in tostrip: |
|
161 | 161 | saverevs.add(r) |
|
162 | 162 | saveheads.difference_update(cl.parentrevs(r)) |
|
163 | 163 | saveheads.add(r) |
|
164 | 164 | saveheads = [cl.node(r) for r in saveheads] |
|
165 | 165 | |
|
166 | 166 | # compute base nodes |
|
167 | 167 | if saverevs: |
|
168 | 168 | descendants = set(cl.descendants(saverevs)) |
|
169 | 169 | saverevs.difference_update(descendants) |
|
170 | 170 | savebases = [cl.node(r) for r in saverevs] |
|
171 | 171 | stripbases = [cl.node(r) for r in tostrip] |
|
172 | 172 | |
|
173 | 173 | stripobsidx = obsmarkers = () |
|
174 | 174 | if repo.ui.configbool(b'devel', b'strip-obsmarkers'): |
|
175 | 175 | obsmarkers = obsutil.exclusivemarkers(repo, stripbases) |
|
176 | 176 | if obsmarkers: |
|
177 | 177 | stripobsidx = [ |
|
178 | 178 | i for i, m in enumerate(repo.obsstore) if m in obsmarkers |
|
179 | 179 | ] |
|
180 | 180 | |
|
181 | 181 | newbmtarget, updatebm = _bookmarkmovements(repo, tostrip) |
|
182 | 182 | |
|
183 | 183 | backupfile = None |
|
184 | 184 | node = nodelist[-1] |
|
185 | 185 | if backup: |
|
186 | 186 | backupfile = _createstripbackup(repo, stripbases, node, topic) |
|
187 | 187 | # create a changegroup for all the branches we need to keep |
|
188 | 188 | tmpbundlefile = None |
|
189 | 189 | if saveheads: |
|
190 | 190 | # do not compress temporary bundle if we remove it from disk later |
|
191 | 191 | # |
|
192 | 192 | # We do not include obsolescence, it might re-introduce prune markers |
|
193 | 193 | # we are trying to strip. This is harmless since the stripped markers |
|
194 | 194 | # are already backed up and we did not touched the markers for the |
|
195 | 195 | # saved changesets. |
|
196 | 196 | tmpbundlefile = backupbundle( |
|
197 | 197 | repo, |
|
198 | 198 | savebases, |
|
199 | 199 | saveheads, |
|
200 | 200 | node, |
|
201 | 201 | b'temp', |
|
202 | 202 | compress=False, |
|
203 | 203 | obsolescence=False, |
|
204 | 204 | ) |
|
205 | 205 | |
|
206 | 206 | with ui.uninterruptible(): |
|
207 | 207 | try: |
|
208 | 208 | with repo.transaction(b"strip") as tr: |
|
209 | 209 | # TODO this code violates the interface abstraction of the |
|
210 | 210 | # transaction and makes assumptions that file storage is |
|
211 | 211 | # using append-only files. We'll need some kind of storage |
|
212 | 212 | # API to handle stripping for us. |
|
213 | 213 | oldfiles = set(tr._offsetmap.keys()) |
|
214 | 214 | oldfiles.update(tr._newfiles) |
|
215 | 215 | |
|
216 | 216 | tr.startgroup() |
|
217 | 217 | cl.strip(striprev, tr) |
|
218 | 218 | stripmanifest(repo, striprev, tr, files) |
|
219 | 219 | |
|
220 | 220 | for fn in files: |
|
221 | 221 | repo.file(fn).strip(striprev, tr) |
|
222 | 222 | tr.endgroup() |
|
223 | 223 | |
|
224 | 224 | entries = tr.readjournal() |
|
225 | 225 | |
|
226 | 226 | for file, troffset in entries: |
|
227 | 227 | if file in oldfiles: |
|
228 | 228 | continue |
|
229 | 229 | with repo.svfs(file, b'a', checkambig=True) as fp: |
|
230 | 230 | fp.truncate(troffset) |
|
231 | 231 | if troffset == 0: |
|
232 | 232 | repo.store.markremoved(file) |
|
233 | 233 | |
|
234 | 234 | deleteobsmarkers(repo.obsstore, stripobsidx) |
|
235 | 235 | del repo.obsstore |
|
236 | 236 | repo.invalidatevolatilesets() |
|
237 | 237 | repo._phasecache.filterunknown(repo) |
|
238 | 238 | |
|
239 | 239 | if tmpbundlefile: |
|
240 | 240 | ui.note(_(b"adding branch\n")) |
|
241 | 241 | f = vfs.open(tmpbundlefile, b"rb") |
|
242 | 242 | gen = exchange.readbundle(ui, f, tmpbundlefile, vfs) |
|
243 | 243 | # silence internal shuffling chatter |
|
244 | 244 | maybe_silent = ( |
|
245 | 245 | repo.ui.silent() |
|
246 | 246 | if not repo.ui.verbose |
|
247 | 247 | else util.nullcontextmanager() |
|
248 | 248 | ) |
|
249 | 249 | with maybe_silent: |
|
250 | 250 | tmpbundleurl = b'bundle:' + vfs.join(tmpbundlefile) |
|
251 | 251 | txnname = b'strip' |
|
252 | 252 | if not isinstance(gen, bundle2.unbundle20): |
|
253 | 253 | txnname = b"strip\n%s" % urlutil.hidepassword( |
|
254 | 254 | tmpbundleurl |
|
255 | 255 | ) |
|
256 | 256 | with repo.transaction(txnname) as tr: |
|
257 | 257 | bundle2.applybundle( |
|
258 | 258 | repo, gen, tr, source=b'strip', url=tmpbundleurl |
|
259 | 259 | ) |
|
260 | 260 | f.close() |
|
261 | 261 | |
|
262 | 262 | with repo.transaction(b'repair') as tr: |
|
263 | 263 | bmchanges = [(m, repo[newbmtarget].node()) for m in updatebm] |
|
264 | 264 | repo._bookmarks.applychanges(repo, tr, bmchanges) |
|
265 | 265 | |
|
266 | 266 | # remove undo files |
|
267 | 267 | for undovfs, undofile in repo.undofiles(): |
|
268 | 268 | try: |
|
269 | 269 | undovfs.unlink(undofile) |
|
270 | 270 | except OSError as e: |
|
271 | 271 | if e.errno != errno.ENOENT: |
|
272 | 272 | ui.warn( |
|
273 | 273 | _(b'error removing %s: %s\n') |
|
274 | 274 | % ( |
|
275 | 275 | undovfs.join(undofile), |
|
276 | 276 | stringutil.forcebytestr(e), |
|
277 | 277 | ) |
|
278 | 278 | ) |
|
279 | 279 | |
|
280 | 280 | except: # re-raises |
|
281 | 281 | if backupfile: |
|
282 | 282 | ui.warn( |
|
283 | 283 | _(b"strip failed, backup bundle stored in '%s'\n") |
|
284 | 284 | % vfs.join(backupfile) |
|
285 | 285 | ) |
|
286 | 286 | if tmpbundlefile: |
|
287 | 287 | ui.warn( |
|
288 | 288 | _(b"strip failed, unrecovered changes stored in '%s'\n") |
|
289 | 289 | % vfs.join(tmpbundlefile) |
|
290 | 290 | ) |
|
291 | 291 | ui.warn( |
|
292 | 292 | _( |
|
293 | 293 | b"(fix the problem, then recover the changesets with " |
|
294 | 294 | b"\"hg unbundle '%s'\")\n" |
|
295 | 295 | ) |
|
296 | 296 | % vfs.join(tmpbundlefile) |
|
297 | 297 | ) |
|
298 | 298 | raise |
|
299 | 299 | else: |
|
300 | 300 | if tmpbundlefile: |
|
301 | 301 | # Remove temporary bundle only if there were no exceptions |
|
302 | 302 | vfs.unlink(tmpbundlefile) |
|
303 | 303 | |
|
304 | 304 | repo.destroyed() |
|
305 | 305 | # return the backup file path (or None if 'backup' was False) so |
|
306 | 306 | # extensions can use it |
|
307 | 307 | return backupfile |
|
308 | 308 | |
|
309 | 309 | |
|
310 | 310 | def softstrip(ui, repo, nodelist, backup=True, topic=b'backup'): |
|
311 | 311 | """perform a "soft" strip using the archived phase""" |
|
312 | 312 | tostrip = [c.node() for c in repo.set(b'sort(%ln::)', nodelist)] |
|
313 | 313 | if not tostrip: |
|
314 | 314 | return None |
|
315 | 315 | |
|
316 | 316 | backupfile = None |
|
317 | 317 | if backup: |
|
318 | 318 | node = tostrip[0] |
|
319 | 319 | backupfile = _createstripbackup(repo, tostrip, node, topic) |
|
320 | 320 | |
|
321 | 321 | newbmtarget, updatebm = _bookmarkmovements(repo, tostrip) |
|
322 | 322 | with repo.transaction(b'strip') as tr: |
|
323 | 323 | phases.retractboundary(repo, tr, phases.archived, tostrip) |
|
324 | 324 | bmchanges = [(m, repo[newbmtarget].node()) for m in updatebm] |
|
325 | 325 | repo._bookmarks.applychanges(repo, tr, bmchanges) |
|
326 | 326 | return backupfile |
|
327 | 327 | |
|
328 | 328 | |
|
329 | 329 | def _bookmarkmovements(repo, tostrip): |
|
330 | 330 | # compute necessary bookmark movement |
|
331 | 331 | bm = repo._bookmarks |
|
332 | 332 | updatebm = [] |
|
333 | 333 | for m in bm: |
|
334 | 334 | rev = repo[bm[m]].rev() |
|
335 | 335 | if rev in tostrip: |
|
336 | 336 | updatebm.append(m) |
|
337 | 337 | newbmtarget = None |
|
338 | 338 | # If we need to move bookmarks, compute bookmark |
|
339 | 339 | # targets. Otherwise we can skip doing this logic. |
|
340 | 340 | if updatebm: |
|
341 | 341 | # For a set s, max(parents(s) - s) is the same as max(heads(::s - s)), |
|
342 | 342 | # but is much faster |
|
343 | 343 | newbmtarget = repo.revs(b'max(parents(%ld) - (%ld))', tostrip, tostrip) |
|
344 | 344 | if newbmtarget: |
|
345 | 345 | newbmtarget = repo[newbmtarget.first()].node() |
|
346 | 346 | else: |
|
347 | 347 | newbmtarget = b'.' |
|
348 | 348 | return newbmtarget, updatebm |
|
349 | 349 | |
|
350 | 350 | |
|
351 | 351 | def _createstripbackup(repo, stripbases, node, topic): |
|
352 | 352 | # backup the changeset we are about to strip |
|
353 | 353 | vfs = repo.vfs |
|
354 | 354 | cl = repo.changelog |
|
355 | 355 | backupfile = backupbundle(repo, stripbases, cl.heads(), node, topic) |
|
356 | 356 | repo.ui.status(_(b"saved backup bundle to %s\n") % vfs.join(backupfile)) |
|
357 | 357 | repo.ui.log( |
|
358 | 358 | b"backupbundle", b"saved backup bundle to %s\n", vfs.join(backupfile) |
|
359 | 359 | ) |
|
360 | 360 | return backupfile |
|
361 | 361 | |
|
362 | 362 | |
|
363 | 363 | def safestriproots(ui, repo, nodes): |
|
364 | 364 | """return list of roots of nodes where descendants are covered by nodes""" |
|
365 | 365 | torev = repo.unfiltered().changelog.rev |
|
366 | 366 | revs = {torev(n) for n in nodes} |
|
367 | 367 | # tostrip = wanted - unsafe = wanted - ancestors(orphaned) |
|
368 | 368 | # orphaned = affected - wanted |
|
369 | 369 | # affected = descendants(roots(wanted)) |
|
370 | 370 | # wanted = revs |
|
371 | 371 | revset = b'%ld - ( ::( (roots(%ld):: and not _phase(%s)) -%ld) )' |
|
372 | 372 | tostrip = set(repo.revs(revset, revs, revs, phases.internal, revs)) |
|
373 | 373 | notstrip = revs - tostrip |
|
374 | 374 | if notstrip: |
|
375 | 375 | nodestr = b', '.join(sorted(short(repo[n].node()) for n in notstrip)) |
|
376 | 376 | ui.warn( |
|
377 | 377 | _(b'warning: orphaned descendants detected, not stripping %s\n') |
|
378 | 378 | % nodestr |
|
379 | 379 | ) |
|
380 | 380 | return [c.node() for c in repo.set(b'roots(%ld)', tostrip)] |
|
381 | 381 | |
|
382 | 382 | |
|
383 | 383 | class stripcallback(object): |
|
384 | 384 | """used as a transaction postclose callback""" |
|
385 | 385 | |
|
386 | 386 | def __init__(self, ui, repo, backup, topic): |
|
387 | 387 | self.ui = ui |
|
388 | 388 | self.repo = repo |
|
389 | 389 | self.backup = backup |
|
390 | 390 | self.topic = topic or b'backup' |
|
391 | 391 | self.nodelist = [] |
|
392 | 392 | |
|
393 | 393 | def addnodes(self, nodes): |
|
394 | 394 | self.nodelist.extend(nodes) |
|
395 | 395 | |
|
396 | 396 | def __call__(self, tr): |
|
397 | 397 | roots = safestriproots(self.ui, self.repo, self.nodelist) |
|
398 | 398 | if roots: |
|
399 | 399 | strip(self.ui, self.repo, roots, self.backup, self.topic) |
|
400 | 400 | |
|
401 | 401 | |
|
402 | 402 | def delayedstrip(ui, repo, nodelist, topic=None, backup=True): |
|
403 | 403 | """like strip, but works inside transaction and won't strip irreverent revs |
|
404 | 404 | |
|
405 | 405 | nodelist must explicitly contain all descendants. Otherwise a warning will |
|
406 | 406 | be printed that some nodes are not stripped. |
|
407 | 407 | |
|
408 | 408 | Will do a backup if `backup` is True. The last non-None "topic" will be |
|
409 | 409 | used as the backup topic name. The default backup topic name is "backup". |
|
410 | 410 | """ |
|
411 | 411 | tr = repo.currenttransaction() |
|
412 | 412 | if not tr: |
|
413 | 413 | nodes = safestriproots(ui, repo, nodelist) |
|
414 | 414 | return strip(ui, repo, nodes, backup=backup, topic=topic) |
|
415 | 415 | # transaction postclose callbacks are called in alphabet order. |
|
416 | 416 | # use '\xff' as prefix so we are likely to be called last. |
|
417 | 417 | callback = tr.getpostclose(b'\xffstrip') |
|
418 | 418 | if callback is None: |
|
419 | 419 | callback = stripcallback(ui, repo, backup=backup, topic=topic) |
|
420 | 420 | tr.addpostclose(b'\xffstrip', callback) |
|
421 | 421 | if topic: |
|
422 | 422 | callback.topic = topic |
|
423 | 423 | callback.addnodes(nodelist) |
|
424 | 424 | |
|
425 | 425 | |
|
426 | 426 | def stripmanifest(repo, striprev, tr, files): |
|
427 | 427 | for revlog in manifestrevlogs(repo): |
|
428 | 428 | revlog.strip(striprev, tr) |
|
429 | 429 | |
|
430 | 430 | |
|
431 | 431 | def manifestrevlogs(repo): |
|
432 | 432 | yield repo.manifestlog.getstorage(b'') |
|
433 | 433 | if scmutil.istreemanifest(repo): |
|
434 | 434 | # This logic is safe if treemanifest isn't enabled, but also |
|
435 | 435 | # pointless, so we skip it if treemanifest isn't enabled. |
|
436 |
for t, unencoded |
|
|
436 | for t, unencoded, size in repo.store.datafiles(): | |
|
437 | 437 | if unencoded.startswith(b'meta/') and unencoded.endswith( |
|
438 | 438 | b'00manifest.i' |
|
439 | 439 | ): |
|
440 | 440 | dir = unencoded[5:-12] |
|
441 | 441 | yield repo.manifestlog.getstorage(dir) |
|
442 | 442 | |
|
443 | 443 | |
|
444 | 444 | def rebuildfncache(ui, repo, only_data=False): |
|
445 | 445 | """Rebuilds the fncache file from repo history. |
|
446 | 446 | |
|
447 | 447 | Missing entries will be added. Extra entries will be removed. |
|
448 | 448 | """ |
|
449 | 449 | repo = repo.unfiltered() |
|
450 | 450 | |
|
451 | 451 | if requirements.FNCACHE_REQUIREMENT not in repo.requirements: |
|
452 | 452 | ui.warn( |
|
453 | 453 | _( |
|
454 | 454 | b'(not rebuilding fncache because repository does not ' |
|
455 | 455 | b'support fncache)\n' |
|
456 | 456 | ) |
|
457 | 457 | ) |
|
458 | 458 | return |
|
459 | 459 | |
|
460 | 460 | with repo.lock(): |
|
461 | 461 | fnc = repo.store.fncache |
|
462 | 462 | fnc.ensureloaded(warn=ui.warn) |
|
463 | 463 | |
|
464 | 464 | oldentries = set(fnc.entries) |
|
465 | 465 | newentries = set() |
|
466 | 466 | seenfiles = set() |
|
467 | 467 | |
|
468 | 468 | if only_data: |
|
469 | 469 | # Trust the listing of .i from the fncache, but not the .d. This is |
|
470 | 470 | # much faster, because we only need to stat every possible .d files, |
|
471 | 471 | # instead of reading the full changelog |
|
472 | 472 | for f in fnc: |
|
473 | 473 | if f[:5] == b'data/' and f[-2:] == b'.i': |
|
474 | 474 | seenfiles.add(f[5:-2]) |
|
475 | 475 | newentries.add(f) |
|
476 | 476 | dataf = f[:-2] + b'.d' |
|
477 | 477 | if repo.store._exists(dataf): |
|
478 | 478 | newentries.add(dataf) |
|
479 | 479 | else: |
|
480 | 480 | progress = ui.makeprogress( |
|
481 | 481 | _(b'rebuilding'), unit=_(b'changesets'), total=len(repo) |
|
482 | 482 | ) |
|
483 | 483 | for rev in repo: |
|
484 | 484 | progress.update(rev) |
|
485 | 485 | |
|
486 | 486 | ctx = repo[rev] |
|
487 | 487 | for f in ctx.files(): |
|
488 | 488 | # This is to minimize I/O. |
|
489 | 489 | if f in seenfiles: |
|
490 | 490 | continue |
|
491 | 491 | seenfiles.add(f) |
|
492 | 492 | |
|
493 | 493 | i = b'data/%s.i' % f |
|
494 | 494 | d = b'data/%s.d' % f |
|
495 | 495 | |
|
496 | 496 | if repo.store._exists(i): |
|
497 | 497 | newentries.add(i) |
|
498 | 498 | if repo.store._exists(d): |
|
499 | 499 | newentries.add(d) |
|
500 | 500 | |
|
501 | 501 | progress.complete() |
|
502 | 502 | |
|
503 | 503 | if requirements.TREEMANIFEST_REQUIREMENT in repo.requirements: |
|
504 | 504 | # This logic is safe if treemanifest isn't enabled, but also |
|
505 | 505 | # pointless, so we skip it if treemanifest isn't enabled. |
|
506 | 506 | for dir in pathutil.dirs(seenfiles): |
|
507 | 507 | i = b'meta/%s/00manifest.i' % dir |
|
508 | 508 | d = b'meta/%s/00manifest.d' % dir |
|
509 | 509 | |
|
510 | 510 | if repo.store._exists(i): |
|
511 | 511 | newentries.add(i) |
|
512 | 512 | if repo.store._exists(d): |
|
513 | 513 | newentries.add(d) |
|
514 | 514 | |
|
515 | 515 | addcount = len(newentries - oldentries) |
|
516 | 516 | removecount = len(oldentries - newentries) |
|
517 | 517 | for p in sorted(oldentries - newentries): |
|
518 | 518 | ui.write(_(b'removing %s\n') % p) |
|
519 | 519 | for p in sorted(newentries - oldentries): |
|
520 | 520 | ui.write(_(b'adding %s\n') % p) |
|
521 | 521 | |
|
522 | 522 | if addcount or removecount: |
|
523 | 523 | ui.write( |
|
524 | 524 | _(b'%d items added, %d removed from fncache\n') |
|
525 | 525 | % (addcount, removecount) |
|
526 | 526 | ) |
|
527 | 527 | fnc.entries = newentries |
|
528 | 528 | fnc._dirty = True |
|
529 | 529 | |
|
530 | 530 | with repo.transaction(b'fncache') as tr: |
|
531 | 531 | fnc.write(tr) |
|
532 | 532 | else: |
|
533 | 533 | ui.write(_(b'fncache already up to date\n')) |
|
534 | 534 | |
|
535 | 535 | |
|
536 | 536 | def deleteobsmarkers(obsstore, indices): |
|
537 | 537 | """Delete some obsmarkers from obsstore and return how many were deleted |
|
538 | 538 | |
|
539 | 539 | 'indices' is a list of ints which are the indices |
|
540 | 540 | of the markers to be deleted. |
|
541 | 541 | |
|
542 | 542 | Every invocation of this function completely rewrites the obsstore file, |
|
543 | 543 | skipping the markers we want to be removed. The new temporary file is |
|
544 | 544 | created, remaining markers are written there and on .close() this file |
|
545 | 545 | gets atomically renamed to obsstore, thus guaranteeing consistency.""" |
|
546 | 546 | if not indices: |
|
547 | 547 | # we don't want to rewrite the obsstore with the same content |
|
548 | 548 | return |
|
549 | 549 | |
|
550 | 550 | left = [] |
|
551 | 551 | current = obsstore._all |
|
552 | 552 | n = 0 |
|
553 | 553 | for i, m in enumerate(current): |
|
554 | 554 | if i in indices: |
|
555 | 555 | n += 1 |
|
556 | 556 | continue |
|
557 | 557 | left.append(m) |
|
558 | 558 | |
|
559 | 559 | newobsstorefile = obsstore.svfs(b'obsstore', b'w', atomictemp=True) |
|
560 | 560 | for bytes in obsolete.encodemarkers(left, True, obsstore._version): |
|
561 | 561 | newobsstorefile.write(bytes) |
|
562 | 562 | newobsstorefile.close() |
|
563 | 563 | return n |
@@ -1,886 +1,886 b'' | |||
|
1 | 1 | # censor code related to censoring revision |
|
2 | 2 | # coding: utf8 |
|
3 | 3 | # |
|
4 | 4 | # Copyright 2021 Pierre-Yves David <pierre-yves.david@octobus.net> |
|
5 | 5 | # Copyright 2015 Google, Inc <martinvonz@google.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 | import binascii |
|
11 | 11 | import contextlib |
|
12 | 12 | import os |
|
13 | 13 | import struct |
|
14 | 14 | |
|
15 | 15 | from ..node import ( |
|
16 | 16 | nullrev, |
|
17 | 17 | ) |
|
18 | 18 | from .constants import ( |
|
19 | 19 | COMP_MODE_PLAIN, |
|
20 | 20 | ENTRY_DATA_COMPRESSED_LENGTH, |
|
21 | 21 | ENTRY_DATA_COMPRESSION_MODE, |
|
22 | 22 | ENTRY_DATA_OFFSET, |
|
23 | 23 | ENTRY_DATA_UNCOMPRESSED_LENGTH, |
|
24 | 24 | ENTRY_DELTA_BASE, |
|
25 | 25 | ENTRY_LINK_REV, |
|
26 | 26 | ENTRY_NODE_ID, |
|
27 | 27 | ENTRY_PARENT_1, |
|
28 | 28 | ENTRY_PARENT_2, |
|
29 | 29 | ENTRY_SIDEDATA_COMPRESSED_LENGTH, |
|
30 | 30 | ENTRY_SIDEDATA_COMPRESSION_MODE, |
|
31 | 31 | ENTRY_SIDEDATA_OFFSET, |
|
32 | 32 | REVIDX_ISCENSORED, |
|
33 | 33 | REVLOGV0, |
|
34 | 34 | REVLOGV1, |
|
35 | 35 | ) |
|
36 | 36 | from ..i18n import _ |
|
37 | 37 | |
|
38 | 38 | from .. import ( |
|
39 | 39 | error, |
|
40 | 40 | mdiff, |
|
41 | 41 | pycompat, |
|
42 | 42 | revlogutils, |
|
43 | 43 | util, |
|
44 | 44 | ) |
|
45 | 45 | from ..utils import ( |
|
46 | 46 | storageutil, |
|
47 | 47 | ) |
|
48 | 48 | from . import ( |
|
49 | 49 | constants, |
|
50 | 50 | deltas, |
|
51 | 51 | ) |
|
52 | 52 | |
|
53 | 53 | |
|
54 | 54 | def v1_censor(rl, tr, censornode, tombstone=b''): |
|
55 | 55 | """censors a revision in a "version 1" revlog""" |
|
56 | 56 | assert rl._format_version == constants.REVLOGV1, rl._format_version |
|
57 | 57 | |
|
58 | 58 | # avoid cycle |
|
59 | 59 | from .. import revlog |
|
60 | 60 | |
|
61 | 61 | censorrev = rl.rev(censornode) |
|
62 | 62 | tombstone = storageutil.packmeta({b'censored': tombstone}, b'') |
|
63 | 63 | |
|
64 | 64 | # Rewriting the revlog in place is hard. Our strategy for censoring is |
|
65 | 65 | # to create a new revlog, copy all revisions to it, then replace the |
|
66 | 66 | # revlogs on transaction close. |
|
67 | 67 | # |
|
68 | 68 | # This is a bit dangerous. We could easily have a mismatch of state. |
|
69 | 69 | newrl = revlog.revlog( |
|
70 | 70 | rl.opener, |
|
71 | 71 | target=rl.target, |
|
72 | 72 | radix=rl.radix, |
|
73 | 73 | postfix=b'tmpcensored', |
|
74 | 74 | censorable=True, |
|
75 | 75 | ) |
|
76 | 76 | newrl._format_version = rl._format_version |
|
77 | 77 | newrl._format_flags = rl._format_flags |
|
78 | 78 | newrl._generaldelta = rl._generaldelta |
|
79 | 79 | newrl._parse_index = rl._parse_index |
|
80 | 80 | |
|
81 | 81 | for rev in rl.revs(): |
|
82 | 82 | node = rl.node(rev) |
|
83 | 83 | p1, p2 = rl.parents(node) |
|
84 | 84 | |
|
85 | 85 | if rev == censorrev: |
|
86 | 86 | newrl.addrawrevision( |
|
87 | 87 | tombstone, |
|
88 | 88 | tr, |
|
89 | 89 | rl.linkrev(censorrev), |
|
90 | 90 | p1, |
|
91 | 91 | p2, |
|
92 | 92 | censornode, |
|
93 | 93 | constants.REVIDX_ISCENSORED, |
|
94 | 94 | ) |
|
95 | 95 | |
|
96 | 96 | if newrl.deltaparent(rev) != nullrev: |
|
97 | 97 | m = _(b'censored revision stored as delta; cannot censor') |
|
98 | 98 | h = _( |
|
99 | 99 | b'censoring of revlogs is not fully implemented;' |
|
100 | 100 | b' please report this bug' |
|
101 | 101 | ) |
|
102 | 102 | raise error.Abort(m, hint=h) |
|
103 | 103 | continue |
|
104 | 104 | |
|
105 | 105 | if rl.iscensored(rev): |
|
106 | 106 | if rl.deltaparent(rev) != nullrev: |
|
107 | 107 | m = _( |
|
108 | 108 | b'cannot censor due to censored ' |
|
109 | 109 | b'revision having delta stored' |
|
110 | 110 | ) |
|
111 | 111 | raise error.Abort(m) |
|
112 | 112 | rawtext = rl._chunk(rev) |
|
113 | 113 | else: |
|
114 | 114 | rawtext = rl.rawdata(rev) |
|
115 | 115 | |
|
116 | 116 | newrl.addrawrevision( |
|
117 | 117 | rawtext, tr, rl.linkrev(rev), p1, p2, node, rl.flags(rev) |
|
118 | 118 | ) |
|
119 | 119 | |
|
120 | 120 | tr.addbackup(rl._indexfile, location=b'store') |
|
121 | 121 | if not rl._inline: |
|
122 | 122 | tr.addbackup(rl._datafile, location=b'store') |
|
123 | 123 | |
|
124 | 124 | rl.opener.rename(newrl._indexfile, rl._indexfile) |
|
125 | 125 | if not rl._inline: |
|
126 | 126 | rl.opener.rename(newrl._datafile, rl._datafile) |
|
127 | 127 | |
|
128 | 128 | rl.clearcaches() |
|
129 | 129 | rl._loadindex() |
|
130 | 130 | |
|
131 | 131 | |
|
132 | 132 | def v2_censor(revlog, tr, censornode, tombstone=b''): |
|
133 | 133 | """censors a revision in a "version 2" revlog""" |
|
134 | 134 | assert revlog._format_version != REVLOGV0, revlog._format_version |
|
135 | 135 | assert revlog._format_version != REVLOGV1, revlog._format_version |
|
136 | 136 | |
|
137 | 137 | censor_revs = {revlog.rev(censornode)} |
|
138 | 138 | _rewrite_v2(revlog, tr, censor_revs, tombstone) |
|
139 | 139 | |
|
140 | 140 | |
|
141 | 141 | def _rewrite_v2(revlog, tr, censor_revs, tombstone=b''): |
|
142 | 142 | """rewrite a revlog to censor some of its content |
|
143 | 143 | |
|
144 | 144 | General principle |
|
145 | 145 | |
|
146 | 146 | We create new revlog files (index/data/sidedata) to copy the content of |
|
147 | 147 | the existing data without the censored data. |
|
148 | 148 | |
|
149 | 149 | We need to recompute new delta for any revision that used the censored |
|
150 | 150 | revision as delta base. As the cumulative size of the new delta may be |
|
151 | 151 | large, we store them in a temporary file until they are stored in their |
|
152 | 152 | final destination. |
|
153 | 153 | |
|
154 | 154 | All data before the censored data can be blindly copied. The rest needs |
|
155 | 155 | to be copied as we go and the associated index entry needs adjustement. |
|
156 | 156 | """ |
|
157 | 157 | assert revlog._format_version != REVLOGV0, revlog._format_version |
|
158 | 158 | assert revlog._format_version != REVLOGV1, revlog._format_version |
|
159 | 159 | |
|
160 | 160 | old_index = revlog.index |
|
161 | 161 | docket = revlog._docket |
|
162 | 162 | |
|
163 | 163 | tombstone = storageutil.packmeta({b'censored': tombstone}, b'') |
|
164 | 164 | |
|
165 | 165 | first_excl_rev = min(censor_revs) |
|
166 | 166 | |
|
167 | 167 | first_excl_entry = revlog.index[first_excl_rev] |
|
168 | 168 | index_cutoff = revlog.index.entry_size * first_excl_rev |
|
169 | 169 | data_cutoff = first_excl_entry[ENTRY_DATA_OFFSET] >> 16 |
|
170 | 170 | sidedata_cutoff = revlog.sidedata_cut_off(first_excl_rev) |
|
171 | 171 | |
|
172 | 172 | with pycompat.unnamedtempfile(mode=b"w+b") as tmp_storage: |
|
173 | 173 | # rev β (new_base, data_start, data_end, compression_mode) |
|
174 | 174 | rewritten_entries = _precompute_rewritten_delta( |
|
175 | 175 | revlog, |
|
176 | 176 | old_index, |
|
177 | 177 | censor_revs, |
|
178 | 178 | tmp_storage, |
|
179 | 179 | ) |
|
180 | 180 | |
|
181 | 181 | all_files = _setup_new_files( |
|
182 | 182 | revlog, |
|
183 | 183 | index_cutoff, |
|
184 | 184 | data_cutoff, |
|
185 | 185 | sidedata_cutoff, |
|
186 | 186 | ) |
|
187 | 187 | |
|
188 | 188 | # we dont need to open the old index file since its content already |
|
189 | 189 | # exist in a usable form in `old_index`. |
|
190 | 190 | with all_files() as open_files: |
|
191 | 191 | ( |
|
192 | 192 | old_data_file, |
|
193 | 193 | old_sidedata_file, |
|
194 | 194 | new_index_file, |
|
195 | 195 | new_data_file, |
|
196 | 196 | new_sidedata_file, |
|
197 | 197 | ) = open_files |
|
198 | 198 | |
|
199 | 199 | # writing the censored revision |
|
200 | 200 | |
|
201 | 201 | # Writing all subsequent revisions |
|
202 | 202 | for rev in range(first_excl_rev, len(old_index)): |
|
203 | 203 | if rev in censor_revs: |
|
204 | 204 | _rewrite_censor( |
|
205 | 205 | revlog, |
|
206 | 206 | old_index, |
|
207 | 207 | open_files, |
|
208 | 208 | rev, |
|
209 | 209 | tombstone, |
|
210 | 210 | ) |
|
211 | 211 | else: |
|
212 | 212 | _rewrite_simple( |
|
213 | 213 | revlog, |
|
214 | 214 | old_index, |
|
215 | 215 | open_files, |
|
216 | 216 | rev, |
|
217 | 217 | rewritten_entries, |
|
218 | 218 | tmp_storage, |
|
219 | 219 | ) |
|
220 | 220 | docket.write(transaction=None, stripping=True) |
|
221 | 221 | |
|
222 | 222 | |
|
223 | 223 | def _precompute_rewritten_delta( |
|
224 | 224 | revlog, |
|
225 | 225 | old_index, |
|
226 | 226 | excluded_revs, |
|
227 | 227 | tmp_storage, |
|
228 | 228 | ): |
|
229 | 229 | """Compute new delta for revisions whose delta is based on revision that |
|
230 | 230 | will not survive as is. |
|
231 | 231 | |
|
232 | 232 | Return a mapping: {rev β (new_base, data_start, data_end, compression_mode)} |
|
233 | 233 | """ |
|
234 | 234 | dc = deltas.deltacomputer(revlog) |
|
235 | 235 | rewritten_entries = {} |
|
236 | 236 | first_excl_rev = min(excluded_revs) |
|
237 | 237 | with revlog._segmentfile._open_read() as dfh: |
|
238 | 238 | for rev in range(first_excl_rev, len(old_index)): |
|
239 | 239 | if rev in excluded_revs: |
|
240 | 240 | # this revision will be preserved as is, so we don't need to |
|
241 | 241 | # consider recomputing a delta. |
|
242 | 242 | continue |
|
243 | 243 | entry = old_index[rev] |
|
244 | 244 | if entry[ENTRY_DELTA_BASE] not in excluded_revs: |
|
245 | 245 | continue |
|
246 | 246 | # This is a revision that use the censored revision as the base |
|
247 | 247 | # for its delta. We need a need new deltas |
|
248 | 248 | if entry[ENTRY_DATA_UNCOMPRESSED_LENGTH] == 0: |
|
249 | 249 | # this revision is empty, we can delta against nullrev |
|
250 | 250 | rewritten_entries[rev] = (nullrev, 0, 0, COMP_MODE_PLAIN) |
|
251 | 251 | else: |
|
252 | 252 | |
|
253 | 253 | text = revlog.rawdata(rev, _df=dfh) |
|
254 | 254 | info = revlogutils.revisioninfo( |
|
255 | 255 | node=entry[ENTRY_NODE_ID], |
|
256 | 256 | p1=revlog.node(entry[ENTRY_PARENT_1]), |
|
257 | 257 | p2=revlog.node(entry[ENTRY_PARENT_2]), |
|
258 | 258 | btext=[text], |
|
259 | 259 | textlen=len(text), |
|
260 | 260 | cachedelta=None, |
|
261 | 261 | flags=entry[ENTRY_DATA_OFFSET] & 0xFFFF, |
|
262 | 262 | ) |
|
263 | 263 | d = dc.finddeltainfo( |
|
264 | 264 | info, dfh, excluded_bases=excluded_revs, target_rev=rev |
|
265 | 265 | ) |
|
266 | 266 | default_comp = revlog._docket.default_compression_header |
|
267 | 267 | comp_mode, d = deltas.delta_compression(default_comp, d) |
|
268 | 268 | # using `tell` is a bit lazy, but we are not here for speed |
|
269 | 269 | start = tmp_storage.tell() |
|
270 | 270 | tmp_storage.write(d.data[1]) |
|
271 | 271 | end = tmp_storage.tell() |
|
272 | 272 | rewritten_entries[rev] = (d.base, start, end, comp_mode) |
|
273 | 273 | return rewritten_entries |
|
274 | 274 | |
|
275 | 275 | |
|
276 | 276 | def _setup_new_files( |
|
277 | 277 | revlog, |
|
278 | 278 | index_cutoff, |
|
279 | 279 | data_cutoff, |
|
280 | 280 | sidedata_cutoff, |
|
281 | 281 | ): |
|
282 | 282 | """ |
|
283 | 283 | |
|
284 | 284 | return a context manager to open all the relevant files: |
|
285 | 285 | - old_data_file, |
|
286 | 286 | - old_sidedata_file, |
|
287 | 287 | - new_index_file, |
|
288 | 288 | - new_data_file, |
|
289 | 289 | - new_sidedata_file, |
|
290 | 290 | |
|
291 | 291 | The old_index_file is not here because it is accessed through the |
|
292 | 292 | `old_index` object if the caller function. |
|
293 | 293 | """ |
|
294 | 294 | docket = revlog._docket |
|
295 | 295 | old_index_filepath = revlog.opener.join(docket.index_filepath()) |
|
296 | 296 | old_data_filepath = revlog.opener.join(docket.data_filepath()) |
|
297 | 297 | old_sidedata_filepath = revlog.opener.join(docket.sidedata_filepath()) |
|
298 | 298 | |
|
299 | 299 | new_index_filepath = revlog.opener.join(docket.new_index_file()) |
|
300 | 300 | new_data_filepath = revlog.opener.join(docket.new_data_file()) |
|
301 | 301 | new_sidedata_filepath = revlog.opener.join(docket.new_sidedata_file()) |
|
302 | 302 | |
|
303 | 303 | util.copyfile(old_index_filepath, new_index_filepath, nb_bytes=index_cutoff) |
|
304 | 304 | util.copyfile(old_data_filepath, new_data_filepath, nb_bytes=data_cutoff) |
|
305 | 305 | util.copyfile( |
|
306 | 306 | old_sidedata_filepath, |
|
307 | 307 | new_sidedata_filepath, |
|
308 | 308 | nb_bytes=sidedata_cutoff, |
|
309 | 309 | ) |
|
310 | 310 | revlog.opener.register_file(docket.index_filepath()) |
|
311 | 311 | revlog.opener.register_file(docket.data_filepath()) |
|
312 | 312 | revlog.opener.register_file(docket.sidedata_filepath()) |
|
313 | 313 | |
|
314 | 314 | docket.index_end = index_cutoff |
|
315 | 315 | docket.data_end = data_cutoff |
|
316 | 316 | docket.sidedata_end = sidedata_cutoff |
|
317 | 317 | |
|
318 | 318 | # reload the revlog internal information |
|
319 | 319 | revlog.clearcaches() |
|
320 | 320 | revlog._loadindex(docket=docket) |
|
321 | 321 | |
|
322 | 322 | @contextlib.contextmanager |
|
323 | 323 | def all_files_opener(): |
|
324 | 324 | # hide opening in an helper function to please check-code, black |
|
325 | 325 | # and various python version at the same time |
|
326 | 326 | with open(old_data_filepath, 'rb') as old_data_file: |
|
327 | 327 | with open(old_sidedata_filepath, 'rb') as old_sidedata_file: |
|
328 | 328 | with open(new_index_filepath, 'r+b') as new_index_file: |
|
329 | 329 | with open(new_data_filepath, 'r+b') as new_data_file: |
|
330 | 330 | with open( |
|
331 | 331 | new_sidedata_filepath, 'r+b' |
|
332 | 332 | ) as new_sidedata_file: |
|
333 | 333 | new_index_file.seek(0, os.SEEK_END) |
|
334 | 334 | assert new_index_file.tell() == index_cutoff |
|
335 | 335 | new_data_file.seek(0, os.SEEK_END) |
|
336 | 336 | assert new_data_file.tell() == data_cutoff |
|
337 | 337 | new_sidedata_file.seek(0, os.SEEK_END) |
|
338 | 338 | assert new_sidedata_file.tell() == sidedata_cutoff |
|
339 | 339 | yield ( |
|
340 | 340 | old_data_file, |
|
341 | 341 | old_sidedata_file, |
|
342 | 342 | new_index_file, |
|
343 | 343 | new_data_file, |
|
344 | 344 | new_sidedata_file, |
|
345 | 345 | ) |
|
346 | 346 | |
|
347 | 347 | return all_files_opener |
|
348 | 348 | |
|
349 | 349 | |
|
350 | 350 | def _rewrite_simple( |
|
351 | 351 | revlog, |
|
352 | 352 | old_index, |
|
353 | 353 | all_files, |
|
354 | 354 | rev, |
|
355 | 355 | rewritten_entries, |
|
356 | 356 | tmp_storage, |
|
357 | 357 | ): |
|
358 | 358 | """append a normal revision to the index after the rewritten one(s)""" |
|
359 | 359 | ( |
|
360 | 360 | old_data_file, |
|
361 | 361 | old_sidedata_file, |
|
362 | 362 | new_index_file, |
|
363 | 363 | new_data_file, |
|
364 | 364 | new_sidedata_file, |
|
365 | 365 | ) = all_files |
|
366 | 366 | entry = old_index[rev] |
|
367 | 367 | flags = entry[ENTRY_DATA_OFFSET] & 0xFFFF |
|
368 | 368 | old_data_offset = entry[ENTRY_DATA_OFFSET] >> 16 |
|
369 | 369 | |
|
370 | 370 | if rev not in rewritten_entries: |
|
371 | 371 | old_data_file.seek(old_data_offset) |
|
372 | 372 | new_data_size = entry[ENTRY_DATA_COMPRESSED_LENGTH] |
|
373 | 373 | new_data = old_data_file.read(new_data_size) |
|
374 | 374 | data_delta_base = entry[ENTRY_DELTA_BASE] |
|
375 | 375 | d_comp_mode = entry[ENTRY_DATA_COMPRESSION_MODE] |
|
376 | 376 | else: |
|
377 | 377 | ( |
|
378 | 378 | data_delta_base, |
|
379 | 379 | start, |
|
380 | 380 | end, |
|
381 | 381 | d_comp_mode, |
|
382 | 382 | ) = rewritten_entries[rev] |
|
383 | 383 | new_data_size = end - start |
|
384 | 384 | tmp_storage.seek(start) |
|
385 | 385 | new_data = tmp_storage.read(new_data_size) |
|
386 | 386 | |
|
387 | 387 | # It might be faster to group continuous read/write operation, |
|
388 | 388 | # however, this is censor, an operation that is not focussed |
|
389 | 389 | # around stellar performance. So I have not written this |
|
390 | 390 | # optimisation yet. |
|
391 | 391 | new_data_offset = new_data_file.tell() |
|
392 | 392 | new_data_file.write(new_data) |
|
393 | 393 | |
|
394 | 394 | sidedata_size = entry[ENTRY_SIDEDATA_COMPRESSED_LENGTH] |
|
395 | 395 | new_sidedata_offset = new_sidedata_file.tell() |
|
396 | 396 | if 0 < sidedata_size: |
|
397 | 397 | old_sidedata_offset = entry[ENTRY_SIDEDATA_OFFSET] |
|
398 | 398 | old_sidedata_file.seek(old_sidedata_offset) |
|
399 | 399 | new_sidedata = old_sidedata_file.read(sidedata_size) |
|
400 | 400 | new_sidedata_file.write(new_sidedata) |
|
401 | 401 | |
|
402 | 402 | data_uncompressed_length = entry[ENTRY_DATA_UNCOMPRESSED_LENGTH] |
|
403 | 403 | sd_com_mode = entry[ENTRY_SIDEDATA_COMPRESSION_MODE] |
|
404 | 404 | assert data_delta_base <= rev, (data_delta_base, rev) |
|
405 | 405 | |
|
406 | 406 | new_entry = revlogutils.entry( |
|
407 | 407 | flags=flags, |
|
408 | 408 | data_offset=new_data_offset, |
|
409 | 409 | data_compressed_length=new_data_size, |
|
410 | 410 | data_uncompressed_length=data_uncompressed_length, |
|
411 | 411 | data_delta_base=data_delta_base, |
|
412 | 412 | link_rev=entry[ENTRY_LINK_REV], |
|
413 | 413 | parent_rev_1=entry[ENTRY_PARENT_1], |
|
414 | 414 | parent_rev_2=entry[ENTRY_PARENT_2], |
|
415 | 415 | node_id=entry[ENTRY_NODE_ID], |
|
416 | 416 | sidedata_offset=new_sidedata_offset, |
|
417 | 417 | sidedata_compressed_length=sidedata_size, |
|
418 | 418 | data_compression_mode=d_comp_mode, |
|
419 | 419 | sidedata_compression_mode=sd_com_mode, |
|
420 | 420 | ) |
|
421 | 421 | revlog.index.append(new_entry) |
|
422 | 422 | entry_bin = revlog.index.entry_binary(rev) |
|
423 | 423 | new_index_file.write(entry_bin) |
|
424 | 424 | |
|
425 | 425 | revlog._docket.index_end = new_index_file.tell() |
|
426 | 426 | revlog._docket.data_end = new_data_file.tell() |
|
427 | 427 | revlog._docket.sidedata_end = new_sidedata_file.tell() |
|
428 | 428 | |
|
429 | 429 | |
|
430 | 430 | def _rewrite_censor( |
|
431 | 431 | revlog, |
|
432 | 432 | old_index, |
|
433 | 433 | all_files, |
|
434 | 434 | rev, |
|
435 | 435 | tombstone, |
|
436 | 436 | ): |
|
437 | 437 | """rewrite and append a censored revision""" |
|
438 | 438 | ( |
|
439 | 439 | old_data_file, |
|
440 | 440 | old_sidedata_file, |
|
441 | 441 | new_index_file, |
|
442 | 442 | new_data_file, |
|
443 | 443 | new_sidedata_file, |
|
444 | 444 | ) = all_files |
|
445 | 445 | entry = old_index[rev] |
|
446 | 446 | |
|
447 | 447 | # XXX consider trying the default compression too |
|
448 | 448 | new_data_size = len(tombstone) |
|
449 | 449 | new_data_offset = new_data_file.tell() |
|
450 | 450 | new_data_file.write(tombstone) |
|
451 | 451 | |
|
452 | 452 | # we are not adding any sidedata as they might leak info about the censored version |
|
453 | 453 | |
|
454 | 454 | link_rev = entry[ENTRY_LINK_REV] |
|
455 | 455 | |
|
456 | 456 | p1 = entry[ENTRY_PARENT_1] |
|
457 | 457 | p2 = entry[ENTRY_PARENT_2] |
|
458 | 458 | |
|
459 | 459 | new_entry = revlogutils.entry( |
|
460 | 460 | flags=constants.REVIDX_ISCENSORED, |
|
461 | 461 | data_offset=new_data_offset, |
|
462 | 462 | data_compressed_length=new_data_size, |
|
463 | 463 | data_uncompressed_length=new_data_size, |
|
464 | 464 | data_delta_base=rev, |
|
465 | 465 | link_rev=link_rev, |
|
466 | 466 | parent_rev_1=p1, |
|
467 | 467 | parent_rev_2=p2, |
|
468 | 468 | node_id=entry[ENTRY_NODE_ID], |
|
469 | 469 | sidedata_offset=0, |
|
470 | 470 | sidedata_compressed_length=0, |
|
471 | 471 | data_compression_mode=COMP_MODE_PLAIN, |
|
472 | 472 | sidedata_compression_mode=COMP_MODE_PLAIN, |
|
473 | 473 | ) |
|
474 | 474 | revlog.index.append(new_entry) |
|
475 | 475 | entry_bin = revlog.index.entry_binary(rev) |
|
476 | 476 | new_index_file.write(entry_bin) |
|
477 | 477 | revlog._docket.index_end = new_index_file.tell() |
|
478 | 478 | revlog._docket.data_end = new_data_file.tell() |
|
479 | 479 | |
|
480 | 480 | |
|
481 | 481 | def _get_filename_from_filelog_index(path): |
|
482 | 482 | # Drop the extension and the `data/` prefix |
|
483 | 483 | path_part = path.rsplit(b'.', 1)[0].split(b'/', 1) |
|
484 | 484 | if len(path_part) < 2: |
|
485 | 485 | msg = _(b"cannot recognize filelog from filename: '%s'") |
|
486 | 486 | msg %= path |
|
487 | 487 | raise error.Abort(msg) |
|
488 | 488 | |
|
489 | 489 | return path_part[1] |
|
490 | 490 | |
|
491 | 491 | |
|
492 | 492 | def _filelog_from_filename(repo, path): |
|
493 | 493 | """Returns the filelog for the given `path`. Stolen from `engine.py`""" |
|
494 | 494 | |
|
495 | 495 | from .. import filelog # avoid cycle |
|
496 | 496 | |
|
497 | 497 | fl = filelog.filelog(repo.svfs, path) |
|
498 | 498 | return fl |
|
499 | 499 | |
|
500 | 500 | |
|
501 | 501 | def _write_swapped_parents(repo, rl, rev, offset, fp): |
|
502 | 502 | """Swaps p1 and p2 and overwrites the revlog entry for `rev` in `fp`""" |
|
503 | 503 | from ..pure import parsers # avoid cycle |
|
504 | 504 | |
|
505 | 505 | if repo._currentlock(repo._lockref) is None: |
|
506 | 506 | # Let's be paranoid about it |
|
507 | 507 | msg = "repo needs to be locked to rewrite parents" |
|
508 | 508 | raise error.ProgrammingError(msg) |
|
509 | 509 | |
|
510 | 510 | index_format = parsers.IndexObject.index_format |
|
511 | 511 | entry = rl.index[rev] |
|
512 | 512 | new_entry = list(entry) |
|
513 | 513 | new_entry[5], new_entry[6] = entry[6], entry[5] |
|
514 | 514 | packed = index_format.pack(*new_entry[:8]) |
|
515 | 515 | fp.seek(offset) |
|
516 | 516 | fp.write(packed) |
|
517 | 517 | |
|
518 | 518 | |
|
519 | 519 | def _reorder_filelog_parents(repo, fl, to_fix): |
|
520 | 520 | """ |
|
521 | 521 | Swaps p1 and p2 for all `to_fix` revisions of filelog `fl` and writes the |
|
522 | 522 | new version to disk, overwriting the old one with a rename. |
|
523 | 523 | """ |
|
524 | 524 | from ..pure import parsers # avoid cycle |
|
525 | 525 | |
|
526 | 526 | ui = repo.ui |
|
527 | 527 | assert len(to_fix) > 0 |
|
528 | 528 | rl = fl._revlog |
|
529 | 529 | if rl._format_version != constants.REVLOGV1: |
|
530 | 530 | msg = "expected version 1 revlog, got version '%d'" % rl._format_version |
|
531 | 531 | raise error.ProgrammingError(msg) |
|
532 | 532 | |
|
533 | 533 | index_file = rl._indexfile |
|
534 | 534 | new_file_path = index_file + b'.tmp-parents-fix' |
|
535 | 535 | repaired_msg = _(b"repaired revision %d of 'filelog %s'\n") |
|
536 | 536 | |
|
537 | 537 | with ui.uninterruptible(): |
|
538 | 538 | try: |
|
539 | 539 | util.copyfile( |
|
540 | 540 | rl.opener.join(index_file), |
|
541 | 541 | rl.opener.join(new_file_path), |
|
542 | 542 | checkambig=rl._checkambig, |
|
543 | 543 | ) |
|
544 | 544 | |
|
545 | 545 | with rl.opener(new_file_path, mode=b"r+") as fp: |
|
546 | 546 | if rl._inline: |
|
547 | 547 | index = parsers.InlinedIndexObject(fp.read()) |
|
548 | 548 | for rev in fl.revs(): |
|
549 | 549 | if rev in to_fix: |
|
550 | 550 | offset = index._calculate_index(rev) |
|
551 | 551 | _write_swapped_parents(repo, rl, rev, offset, fp) |
|
552 | 552 | ui.write(repaired_msg % (rev, index_file)) |
|
553 | 553 | else: |
|
554 | 554 | index_format = parsers.IndexObject.index_format |
|
555 | 555 | for rev in to_fix: |
|
556 | 556 | offset = rev * index_format.size |
|
557 | 557 | _write_swapped_parents(repo, rl, rev, offset, fp) |
|
558 | 558 | ui.write(repaired_msg % (rev, index_file)) |
|
559 | 559 | |
|
560 | 560 | rl.opener.rename(new_file_path, index_file) |
|
561 | 561 | rl.clearcaches() |
|
562 | 562 | rl._loadindex() |
|
563 | 563 | finally: |
|
564 | 564 | util.tryunlink(new_file_path) |
|
565 | 565 | |
|
566 | 566 | |
|
567 | 567 | def _is_revision_affected(fl, filerev, metadata_cache=None): |
|
568 | 568 | full_text = lambda: fl._revlog.rawdata(filerev) |
|
569 | 569 | parent_revs = lambda: fl._revlog.parentrevs(filerev) |
|
570 | 570 | return _is_revision_affected_inner( |
|
571 | 571 | full_text, parent_revs, filerev, metadata_cache |
|
572 | 572 | ) |
|
573 | 573 | |
|
574 | 574 | |
|
575 | 575 | def _is_revision_affected_inner( |
|
576 | 576 | full_text, |
|
577 | 577 | parents_revs, |
|
578 | 578 | filerev, |
|
579 | 579 | metadata_cache=None, |
|
580 | 580 | ): |
|
581 | 581 | """Mercurial currently (5.9rc0) uses `p1 == nullrev and p2 != nullrev` as a |
|
582 | 582 | special meaning compared to the reverse in the context of filelog-based |
|
583 | 583 | copytracing. issue6528 exists because new code assumed that parent ordering |
|
584 | 584 | didn't matter, so this detects if the revision contains metadata (since |
|
585 | 585 | it's only used for filelog-based copytracing) and its parents are in the |
|
586 | 586 | "wrong" order.""" |
|
587 | 587 | try: |
|
588 | 588 | raw_text = full_text() |
|
589 | 589 | except error.CensoredNodeError: |
|
590 | 590 | # We don't care about censored nodes as they never carry metadata |
|
591 | 591 | return False |
|
592 | 592 | has_meta = raw_text.startswith(b'\x01\n') |
|
593 | 593 | if metadata_cache is not None: |
|
594 | 594 | metadata_cache[filerev] = has_meta |
|
595 | 595 | if has_meta: |
|
596 | 596 | (p1, p2) = parents_revs() |
|
597 | 597 | if p1 != nullrev and p2 == nullrev: |
|
598 | 598 | return True |
|
599 | 599 | return False |
|
600 | 600 | |
|
601 | 601 | |
|
602 | 602 | def _is_revision_affected_fast(repo, fl, filerev, metadata_cache): |
|
603 | 603 | rl = fl._revlog |
|
604 | 604 | is_censored = lambda: rl.iscensored(filerev) |
|
605 | 605 | delta_base = lambda: rl.deltaparent(filerev) |
|
606 | 606 | delta = lambda: rl._chunk(filerev) |
|
607 | 607 | full_text = lambda: rl.rawdata(filerev) |
|
608 | 608 | parent_revs = lambda: rl.parentrevs(filerev) |
|
609 | 609 | return _is_revision_affected_fast_inner( |
|
610 | 610 | is_censored, |
|
611 | 611 | delta_base, |
|
612 | 612 | delta, |
|
613 | 613 | full_text, |
|
614 | 614 | parent_revs, |
|
615 | 615 | filerev, |
|
616 | 616 | metadata_cache, |
|
617 | 617 | ) |
|
618 | 618 | |
|
619 | 619 | |
|
620 | 620 | def _is_revision_affected_fast_inner( |
|
621 | 621 | is_censored, |
|
622 | 622 | delta_base, |
|
623 | 623 | delta, |
|
624 | 624 | full_text, |
|
625 | 625 | parent_revs, |
|
626 | 626 | filerev, |
|
627 | 627 | metadata_cache, |
|
628 | 628 | ): |
|
629 | 629 | """Optimization fast-path for `_is_revision_affected`. |
|
630 | 630 | |
|
631 | 631 | `metadata_cache` is a dict of `{rev: has_metadata}` which allows any |
|
632 | 632 | revision to check if its base has metadata, saving computation of the full |
|
633 | 633 | text, instead looking at the current delta. |
|
634 | 634 | |
|
635 | 635 | This optimization only works if the revisions are looked at in order.""" |
|
636 | 636 | |
|
637 | 637 | if is_censored(): |
|
638 | 638 | # Censored revisions don't contain metadata, so they cannot be affected |
|
639 | 639 | metadata_cache[filerev] = False |
|
640 | 640 | return False |
|
641 | 641 | |
|
642 | 642 | p1, p2 = parent_revs() |
|
643 | 643 | if p1 == nullrev or p2 != nullrev: |
|
644 | 644 | return False |
|
645 | 645 | |
|
646 | 646 | delta_parent = delta_base() |
|
647 | 647 | parent_has_metadata = metadata_cache.get(delta_parent) |
|
648 | 648 | if parent_has_metadata is None: |
|
649 | 649 | return _is_revision_affected_inner( |
|
650 | 650 | full_text, |
|
651 | 651 | parent_revs, |
|
652 | 652 | filerev, |
|
653 | 653 | metadata_cache, |
|
654 | 654 | ) |
|
655 | 655 | |
|
656 | 656 | chunk = delta() |
|
657 | 657 | if not len(chunk): |
|
658 | 658 | # No diff for this revision |
|
659 | 659 | return parent_has_metadata |
|
660 | 660 | |
|
661 | 661 | header_length = 12 |
|
662 | 662 | if len(chunk) < header_length: |
|
663 | 663 | raise error.Abort(_(b"patch cannot be decoded")) |
|
664 | 664 | |
|
665 | 665 | start, _end, _length = struct.unpack(b">lll", chunk[:header_length]) |
|
666 | 666 | |
|
667 | 667 | if start < 2: # len(b'\x01\n') == 2 |
|
668 | 668 | # This delta does *something* to the metadata marker (if any). |
|
669 | 669 | # Check it the slow way |
|
670 | 670 | is_affected = _is_revision_affected_inner( |
|
671 | 671 | full_text, |
|
672 | 672 | parent_revs, |
|
673 | 673 | filerev, |
|
674 | 674 | metadata_cache, |
|
675 | 675 | ) |
|
676 | 676 | return is_affected |
|
677 | 677 | |
|
678 | 678 | # The diff did not remove or add the metadata header, it's then in the same |
|
679 | 679 | # situation as its parent |
|
680 | 680 | metadata_cache[filerev] = parent_has_metadata |
|
681 | 681 | return parent_has_metadata |
|
682 | 682 | |
|
683 | 683 | |
|
684 | 684 | def _from_report(ui, repo, context, from_report, dry_run): |
|
685 | 685 | """ |
|
686 | 686 | Fix the revisions given in the `from_report` file, but still checks if the |
|
687 | 687 | revisions are indeed affected to prevent an unfortunate cyclic situation |
|
688 | 688 | where we'd swap well-ordered parents again. |
|
689 | 689 | |
|
690 | 690 | See the doc for `debug_fix_issue6528` for the format documentation. |
|
691 | 691 | """ |
|
692 | 692 | ui.write(_(b"loading report file '%s'\n") % from_report) |
|
693 | 693 | |
|
694 | 694 | with context(), open(from_report, mode='rb') as f: |
|
695 | 695 | for line in f.read().split(b'\n'): |
|
696 | 696 | if not line: |
|
697 | 697 | continue |
|
698 | 698 | filenodes, filename = line.split(b' ', 1) |
|
699 | 699 | fl = _filelog_from_filename(repo, filename) |
|
700 | 700 | to_fix = set( |
|
701 | 701 | fl.rev(binascii.unhexlify(n)) for n in filenodes.split(b',') |
|
702 | 702 | ) |
|
703 | 703 | excluded = set() |
|
704 | 704 | |
|
705 | 705 | for filerev in to_fix: |
|
706 | 706 | if _is_revision_affected(fl, filerev): |
|
707 | 707 | msg = b"found affected revision %d for filelog '%s'\n" |
|
708 | 708 | ui.warn(msg % (filerev, filename)) |
|
709 | 709 | else: |
|
710 | 710 | msg = _(b"revision %s of file '%s' is not affected\n") |
|
711 | 711 | msg %= (binascii.hexlify(fl.node(filerev)), filename) |
|
712 | 712 | ui.warn(msg) |
|
713 | 713 | excluded.add(filerev) |
|
714 | 714 | |
|
715 | 715 | to_fix = to_fix - excluded |
|
716 | 716 | if not to_fix: |
|
717 | 717 | msg = _(b"no affected revisions were found for '%s'\n") |
|
718 | 718 | ui.write(msg % filename) |
|
719 | 719 | continue |
|
720 | 720 | if not dry_run: |
|
721 | 721 | _reorder_filelog_parents(repo, fl, sorted(to_fix)) |
|
722 | 722 | |
|
723 | 723 | |
|
724 | 724 | def filter_delta_issue6528(revlog, deltas_iter): |
|
725 | 725 | """filter incomind deltas to repaire issue 6528 on the fly""" |
|
726 | 726 | metadata_cache = {} |
|
727 | 727 | |
|
728 | 728 | deltacomputer = deltas.deltacomputer(revlog) |
|
729 | 729 | |
|
730 | 730 | for rev, d in enumerate(deltas_iter, len(revlog)): |
|
731 | 731 | ( |
|
732 | 732 | node, |
|
733 | 733 | p1_node, |
|
734 | 734 | p2_node, |
|
735 | 735 | linknode, |
|
736 | 736 | deltabase, |
|
737 | 737 | delta, |
|
738 | 738 | flags, |
|
739 | 739 | sidedata, |
|
740 | 740 | ) = d |
|
741 | 741 | |
|
742 | 742 | if not revlog.index.has_node(deltabase): |
|
743 | 743 | raise error.LookupError( |
|
744 | 744 | deltabase, revlog.radix, _(b'unknown parent') |
|
745 | 745 | ) |
|
746 | 746 | base_rev = revlog.rev(deltabase) |
|
747 | 747 | if not revlog.index.has_node(p1_node): |
|
748 | 748 | raise error.LookupError(p1_node, revlog.radix, _(b'unknown parent')) |
|
749 | 749 | p1_rev = revlog.rev(p1_node) |
|
750 | 750 | if not revlog.index.has_node(p2_node): |
|
751 | 751 | raise error.LookupError(p2_node, revlog.radix, _(b'unknown parent')) |
|
752 | 752 | p2_rev = revlog.rev(p2_node) |
|
753 | 753 | |
|
754 | 754 | is_censored = lambda: bool(flags & REVIDX_ISCENSORED) |
|
755 | 755 | delta_base = lambda: revlog.rev(delta_base) |
|
756 | 756 | delta_base = lambda: base_rev |
|
757 | 757 | parent_revs = lambda: (p1_rev, p2_rev) |
|
758 | 758 | |
|
759 | 759 | def full_text(): |
|
760 | 760 | # note: being able to reuse the full text computation in the |
|
761 | 761 | # underlying addrevision would be useful however this is a bit too |
|
762 | 762 | # intrusive the for the "quick" issue6528 we are writing before the |
|
763 | 763 | # 5.8 release |
|
764 | 764 | textlen = mdiff.patchedsize(revlog.size(base_rev), delta) |
|
765 | 765 | |
|
766 | 766 | revinfo = revlogutils.revisioninfo( |
|
767 | 767 | node, |
|
768 | 768 | p1_node, |
|
769 | 769 | p2_node, |
|
770 | 770 | [None], |
|
771 | 771 | textlen, |
|
772 | 772 | (base_rev, delta), |
|
773 | 773 | flags, |
|
774 | 774 | ) |
|
775 | 775 | # cached by the global "writing" context |
|
776 | 776 | assert revlog._writinghandles is not None |
|
777 | 777 | if revlog._inline: |
|
778 | 778 | fh = revlog._writinghandles[0] |
|
779 | 779 | else: |
|
780 | 780 | fh = revlog._writinghandles[1] |
|
781 | 781 | return deltacomputer.buildtext(revinfo, fh) |
|
782 | 782 | |
|
783 | 783 | is_affected = _is_revision_affected_fast_inner( |
|
784 | 784 | is_censored, |
|
785 | 785 | delta_base, |
|
786 | 786 | lambda: delta, |
|
787 | 787 | full_text, |
|
788 | 788 | parent_revs, |
|
789 | 789 | rev, |
|
790 | 790 | metadata_cache, |
|
791 | 791 | ) |
|
792 | 792 | if is_affected: |
|
793 | 793 | d = ( |
|
794 | 794 | node, |
|
795 | 795 | p2_node, |
|
796 | 796 | p1_node, |
|
797 | 797 | linknode, |
|
798 | 798 | deltabase, |
|
799 | 799 | delta, |
|
800 | 800 | flags, |
|
801 | 801 | sidedata, |
|
802 | 802 | ) |
|
803 | 803 | yield d |
|
804 | 804 | |
|
805 | 805 | |
|
806 | 806 | def repair_issue6528( |
|
807 | 807 | ui, repo, dry_run=False, to_report=None, from_report=None, paranoid=False |
|
808 | 808 | ): |
|
809 | 809 | from .. import store # avoid cycle |
|
810 | 810 | |
|
811 | 811 | @contextlib.contextmanager |
|
812 | 812 | def context(): |
|
813 | 813 | if dry_run or to_report: # No need for locking |
|
814 | 814 | yield |
|
815 | 815 | else: |
|
816 | 816 | with repo.wlock(), repo.lock(): |
|
817 | 817 | yield |
|
818 | 818 | |
|
819 | 819 | if from_report: |
|
820 | 820 | return _from_report(ui, repo, context, from_report, dry_run) |
|
821 | 821 | |
|
822 | 822 | report_entries = [] |
|
823 | 823 | |
|
824 | 824 | with context(): |
|
825 | 825 | files = list( |
|
826 | 826 | (file_type, path) |
|
827 |
for (file_type, path |
|
|
827 | for (file_type, path, _s) in repo.store.datafiles() | |
|
828 | 828 | if path.endswith(b'.i') and file_type & store.FILEFLAGS_FILELOG |
|
829 | 829 | ) |
|
830 | 830 | |
|
831 | 831 | progress = ui.makeprogress( |
|
832 | 832 | _(b"looking for affected revisions"), |
|
833 | 833 | unit=_(b"filelogs"), |
|
834 | 834 | total=len(files), |
|
835 | 835 | ) |
|
836 | 836 | found_nothing = True |
|
837 | 837 | |
|
838 | 838 | for file_type, path in files: |
|
839 | 839 | if ( |
|
840 | 840 | not path.endswith(b'.i') |
|
841 | 841 | or not file_type & store.FILEFLAGS_FILELOG |
|
842 | 842 | ): |
|
843 | 843 | continue |
|
844 | 844 | progress.increment() |
|
845 | 845 | filename = _get_filename_from_filelog_index(path) |
|
846 | 846 | fl = _filelog_from_filename(repo, filename) |
|
847 | 847 | |
|
848 | 848 | # Set of filerevs (or hex filenodes if `to_report`) that need fixing |
|
849 | 849 | to_fix = set() |
|
850 | 850 | metadata_cache = {} |
|
851 | 851 | for filerev in fl.revs(): |
|
852 | 852 | affected = _is_revision_affected_fast( |
|
853 | 853 | repo, fl, filerev, metadata_cache |
|
854 | 854 | ) |
|
855 | 855 | if paranoid: |
|
856 | 856 | slow = _is_revision_affected(fl, filerev) |
|
857 | 857 | if slow != affected: |
|
858 | 858 | msg = _(b"paranoid check failed for '%s' at node %s") |
|
859 | 859 | node = binascii.hexlify(fl.node(filerev)) |
|
860 | 860 | raise error.Abort(msg % (filename, node)) |
|
861 | 861 | if affected: |
|
862 | 862 | msg = b"found affected revision %d for filelog '%s'\n" |
|
863 | 863 | ui.warn(msg % (filerev, path)) |
|
864 | 864 | found_nothing = False |
|
865 | 865 | if not dry_run: |
|
866 | 866 | if to_report: |
|
867 | 867 | to_fix.add(binascii.hexlify(fl.node(filerev))) |
|
868 | 868 | else: |
|
869 | 869 | to_fix.add(filerev) |
|
870 | 870 | |
|
871 | 871 | if to_fix: |
|
872 | 872 | to_fix = sorted(to_fix) |
|
873 | 873 | if to_report: |
|
874 | 874 | report_entries.append((filename, to_fix)) |
|
875 | 875 | else: |
|
876 | 876 | _reorder_filelog_parents(repo, fl, to_fix) |
|
877 | 877 | |
|
878 | 878 | if found_nothing: |
|
879 | 879 | ui.write(_(b"no affected revisions were found\n")) |
|
880 | 880 | |
|
881 | 881 | if to_report and report_entries: |
|
882 | 882 | with open(to_report, mode="wb") as f: |
|
883 | 883 | for path, to_fix in report_entries: |
|
884 | 884 | f.write(b"%s %s\n" % (b",".join(to_fix), path)) |
|
885 | 885 | |
|
886 | 886 | progress.complete() |
@@ -1,839 +1,850 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 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import errno |
|
11 | 11 | import functools |
|
12 | 12 | import os |
|
13 | 13 | import re |
|
14 | 14 | import stat |
|
15 | 15 | |
|
16 | 16 | from .i18n import _ |
|
17 | 17 | from .pycompat import getattr |
|
18 | 18 | from .node import hex |
|
19 | 19 | from . import ( |
|
20 | 20 | changelog, |
|
21 | 21 | error, |
|
22 | 22 | manifest, |
|
23 | 23 | policy, |
|
24 | 24 | pycompat, |
|
25 | 25 | util, |
|
26 | 26 | vfs as vfsmod, |
|
27 | 27 | ) |
|
28 | 28 | from .utils import hashutil |
|
29 | 29 | |
|
30 | 30 | parsers = policy.importmod('parsers') |
|
31 | 31 | # how much bytes should be read from fncache in one read |
|
32 | 32 | # It is done to prevent loading large fncache files into memory |
|
33 | 33 | fncache_chunksize = 10 ** 6 |
|
34 | 34 | |
|
35 | 35 | |
|
36 | 36 | def _matchtrackedpath(path, matcher): |
|
37 | 37 | """parses a fncache entry and returns whether the entry is tracking a path |
|
38 | 38 | matched by matcher or not. |
|
39 | 39 | |
|
40 | 40 | If matcher is None, returns True""" |
|
41 | 41 | |
|
42 | 42 | if matcher is None: |
|
43 | 43 | return True |
|
44 | 44 | path = decodedir(path) |
|
45 | 45 | if path.startswith(b'data/'): |
|
46 | 46 | return matcher(path[len(b'data/') : -len(b'.i')]) |
|
47 | 47 | elif path.startswith(b'meta/'): |
|
48 | 48 | return matcher.visitdir(path[len(b'meta/') : -len(b'/00manifest.i')]) |
|
49 | 49 | |
|
50 | 50 | raise error.ProgrammingError(b"cannot decode path %s" % path) |
|
51 | 51 | |
|
52 | 52 | |
|
53 | 53 | # This avoids a collision between a file named foo and a dir named |
|
54 | 54 | # foo.i or foo.d |
|
55 | 55 | def _encodedir(path): |
|
56 | 56 | """ |
|
57 | 57 | >>> _encodedir(b'data/foo.i') |
|
58 | 58 | 'data/foo.i' |
|
59 | 59 | >>> _encodedir(b'data/foo.i/bla.i') |
|
60 | 60 | 'data/foo.i.hg/bla.i' |
|
61 | 61 | >>> _encodedir(b'data/foo.i.hg/bla.i') |
|
62 | 62 | 'data/foo.i.hg.hg/bla.i' |
|
63 | 63 | >>> _encodedir(b'data/foo.i\\ndata/foo.i/bla.i\\ndata/foo.i.hg/bla.i\\n') |
|
64 | 64 | 'data/foo.i\\ndata/foo.i.hg/bla.i\\ndata/foo.i.hg.hg/bla.i\\n' |
|
65 | 65 | """ |
|
66 | 66 | return ( |
|
67 | 67 | path.replace(b".hg/", b".hg.hg/") |
|
68 | 68 | .replace(b".i/", b".i.hg/") |
|
69 | 69 | .replace(b".d/", b".d.hg/") |
|
70 | 70 | ) |
|
71 | 71 | |
|
72 | 72 | |
|
73 | 73 | encodedir = getattr(parsers, 'encodedir', _encodedir) |
|
74 | 74 | |
|
75 | 75 | |
|
76 | 76 | def decodedir(path): |
|
77 | 77 | """ |
|
78 | 78 | >>> decodedir(b'data/foo.i') |
|
79 | 79 | 'data/foo.i' |
|
80 | 80 | >>> decodedir(b'data/foo.i.hg/bla.i') |
|
81 | 81 | 'data/foo.i/bla.i' |
|
82 | 82 | >>> decodedir(b'data/foo.i.hg.hg/bla.i') |
|
83 | 83 | 'data/foo.i.hg/bla.i' |
|
84 | 84 | """ |
|
85 | 85 | if b".hg/" not in path: |
|
86 | 86 | return path |
|
87 | 87 | return ( |
|
88 | 88 | path.replace(b".d.hg/", b".d/") |
|
89 | 89 | .replace(b".i.hg/", b".i/") |
|
90 | 90 | .replace(b".hg.hg/", b".hg/") |
|
91 | 91 | ) |
|
92 | 92 | |
|
93 | 93 | |
|
94 | 94 | def _reserved(): |
|
95 | 95 | """characters that are problematic for filesystems |
|
96 | 96 | |
|
97 | 97 | * ascii escapes (0..31) |
|
98 | 98 | * ascii hi (126..255) |
|
99 | 99 | * windows specials |
|
100 | 100 | |
|
101 | 101 | these characters will be escaped by encodefunctions |
|
102 | 102 | """ |
|
103 | 103 | winreserved = [ord(x) for x in u'\\:*?"<>|'] |
|
104 | 104 | for x in range(32): |
|
105 | 105 | yield x |
|
106 | 106 | for x in range(126, 256): |
|
107 | 107 | yield x |
|
108 | 108 | for x in winreserved: |
|
109 | 109 | yield x |
|
110 | 110 | |
|
111 | 111 | |
|
112 | 112 | def _buildencodefun(): |
|
113 | 113 | """ |
|
114 | 114 | >>> enc, dec = _buildencodefun() |
|
115 | 115 | |
|
116 | 116 | >>> enc(b'nothing/special.txt') |
|
117 | 117 | 'nothing/special.txt' |
|
118 | 118 | >>> dec(b'nothing/special.txt') |
|
119 | 119 | 'nothing/special.txt' |
|
120 | 120 | |
|
121 | 121 | >>> enc(b'HELLO') |
|
122 | 122 | '_h_e_l_l_o' |
|
123 | 123 | >>> dec(b'_h_e_l_l_o') |
|
124 | 124 | 'HELLO' |
|
125 | 125 | |
|
126 | 126 | >>> enc(b'hello:world?') |
|
127 | 127 | 'hello~3aworld~3f' |
|
128 | 128 | >>> dec(b'hello~3aworld~3f') |
|
129 | 129 | 'hello:world?' |
|
130 | 130 | |
|
131 | 131 | >>> enc(b'the\\x07quick\\xADshot') |
|
132 | 132 | 'the~07quick~adshot' |
|
133 | 133 | >>> dec(b'the~07quick~adshot') |
|
134 | 134 | 'the\\x07quick\\xadshot' |
|
135 | 135 | """ |
|
136 | 136 | e = b'_' |
|
137 | 137 | xchr = pycompat.bytechr |
|
138 | 138 | asciistr = list(map(xchr, range(127))) |
|
139 | 139 | capitals = list(range(ord(b"A"), ord(b"Z") + 1)) |
|
140 | 140 | |
|
141 | 141 | cmap = {x: x for x in asciistr} |
|
142 | 142 | for x in _reserved(): |
|
143 | 143 | cmap[xchr(x)] = b"~%02x" % x |
|
144 | 144 | for x in capitals + [ord(e)]: |
|
145 | 145 | cmap[xchr(x)] = e + xchr(x).lower() |
|
146 | 146 | |
|
147 | 147 | dmap = {} |
|
148 | 148 | for k, v in pycompat.iteritems(cmap): |
|
149 | 149 | dmap[v] = k |
|
150 | 150 | |
|
151 | 151 | def decode(s): |
|
152 | 152 | i = 0 |
|
153 | 153 | while i < len(s): |
|
154 | 154 | for l in pycompat.xrange(1, 4): |
|
155 | 155 | try: |
|
156 | 156 | yield dmap[s[i : i + l]] |
|
157 | 157 | i += l |
|
158 | 158 | break |
|
159 | 159 | except KeyError: |
|
160 | 160 | pass |
|
161 | 161 | else: |
|
162 | 162 | raise KeyError |
|
163 | 163 | |
|
164 | 164 | return ( |
|
165 | 165 | lambda s: b''.join( |
|
166 | 166 | [cmap[s[c : c + 1]] for c in pycompat.xrange(len(s))] |
|
167 | 167 | ), |
|
168 | 168 | lambda s: b''.join(list(decode(s))), |
|
169 | 169 | ) |
|
170 | 170 | |
|
171 | 171 | |
|
172 | 172 | _encodefname, _decodefname = _buildencodefun() |
|
173 | 173 | |
|
174 | 174 | |
|
175 | 175 | def encodefilename(s): |
|
176 | 176 | """ |
|
177 | 177 | >>> encodefilename(b'foo.i/bar.d/bla.hg/hi:world?/HELLO') |
|
178 | 178 | 'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o' |
|
179 | 179 | """ |
|
180 | 180 | return _encodefname(encodedir(s)) |
|
181 | 181 | |
|
182 | 182 | |
|
183 | 183 | def decodefilename(s): |
|
184 | 184 | """ |
|
185 | 185 | >>> decodefilename(b'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o') |
|
186 | 186 | 'foo.i/bar.d/bla.hg/hi:world?/HELLO' |
|
187 | 187 | """ |
|
188 | 188 | return decodedir(_decodefname(s)) |
|
189 | 189 | |
|
190 | 190 | |
|
191 | 191 | def _buildlowerencodefun(): |
|
192 | 192 | """ |
|
193 | 193 | >>> f = _buildlowerencodefun() |
|
194 | 194 | >>> f(b'nothing/special.txt') |
|
195 | 195 | 'nothing/special.txt' |
|
196 | 196 | >>> f(b'HELLO') |
|
197 | 197 | 'hello' |
|
198 | 198 | >>> f(b'hello:world?') |
|
199 | 199 | 'hello~3aworld~3f' |
|
200 | 200 | >>> f(b'the\\x07quick\\xADshot') |
|
201 | 201 | 'the~07quick~adshot' |
|
202 | 202 | """ |
|
203 | 203 | xchr = pycompat.bytechr |
|
204 | 204 | cmap = {xchr(x): xchr(x) for x in pycompat.xrange(127)} |
|
205 | 205 | for x in _reserved(): |
|
206 | 206 | cmap[xchr(x)] = b"~%02x" % x |
|
207 | 207 | for x in range(ord(b"A"), ord(b"Z") + 1): |
|
208 | 208 | cmap[xchr(x)] = xchr(x).lower() |
|
209 | 209 | |
|
210 | 210 | def lowerencode(s): |
|
211 | 211 | return b"".join([cmap[c] for c in pycompat.iterbytestr(s)]) |
|
212 | 212 | |
|
213 | 213 | return lowerencode |
|
214 | 214 | |
|
215 | 215 | |
|
216 | 216 | lowerencode = getattr(parsers, 'lowerencode', None) or _buildlowerencodefun() |
|
217 | 217 | |
|
218 | 218 | # Windows reserved names: con, prn, aux, nul, com1..com9, lpt1..lpt9 |
|
219 | 219 | _winres3 = (b'aux', b'con', b'prn', b'nul') # length 3 |
|
220 | 220 | _winres4 = (b'com', b'lpt') # length 4 (with trailing 1..9) |
|
221 | 221 | |
|
222 | 222 | |
|
223 | 223 | def _auxencode(path, dotencode): |
|
224 | 224 | """ |
|
225 | 225 | Encodes filenames containing names reserved by Windows or which end in |
|
226 | 226 | period or space. Does not touch other single reserved characters c. |
|
227 | 227 | Specifically, c in '\\:*?"<>|' or ord(c) <= 31 are *not* encoded here. |
|
228 | 228 | Additionally encodes space or period at the beginning, if dotencode is |
|
229 | 229 | True. Parameter path is assumed to be all lowercase. |
|
230 | 230 | A segment only needs encoding if a reserved name appears as a |
|
231 | 231 | basename (e.g. "aux", "aux.foo"). A directory or file named "foo.aux" |
|
232 | 232 | doesn't need encoding. |
|
233 | 233 | |
|
234 | 234 | >>> s = b'.foo/aux.txt/txt.aux/con/prn/nul/foo.' |
|
235 | 235 | >>> _auxencode(s.split(b'/'), True) |
|
236 | 236 | ['~2efoo', 'au~78.txt', 'txt.aux', 'co~6e', 'pr~6e', 'nu~6c', 'foo~2e'] |
|
237 | 237 | >>> s = b'.com1com2/lpt9.lpt4.lpt1/conprn/com0/lpt0/foo.' |
|
238 | 238 | >>> _auxencode(s.split(b'/'), False) |
|
239 | 239 | ['.com1com2', 'lp~749.lpt4.lpt1', 'conprn', 'com0', 'lpt0', 'foo~2e'] |
|
240 | 240 | >>> _auxencode([b'foo. '], True) |
|
241 | 241 | ['foo.~20'] |
|
242 | 242 | >>> _auxencode([b' .foo'], True) |
|
243 | 243 | ['~20.foo'] |
|
244 | 244 | """ |
|
245 | 245 | for i, n in enumerate(path): |
|
246 | 246 | if not n: |
|
247 | 247 | continue |
|
248 | 248 | if dotencode and n[0] in b'. ': |
|
249 | 249 | n = b"~%02x" % ord(n[0:1]) + n[1:] |
|
250 | 250 | path[i] = n |
|
251 | 251 | else: |
|
252 | 252 | l = n.find(b'.') |
|
253 | 253 | if l == -1: |
|
254 | 254 | l = len(n) |
|
255 | 255 | if (l == 3 and n[:3] in _winres3) or ( |
|
256 | 256 | l == 4 |
|
257 | 257 | and n[3:4] <= b'9' |
|
258 | 258 | and n[3:4] >= b'1' |
|
259 | 259 | and n[:3] in _winres4 |
|
260 | 260 | ): |
|
261 | 261 | # encode third letter ('aux' -> 'au~78') |
|
262 | 262 | ec = b"~%02x" % ord(n[2:3]) |
|
263 | 263 | n = n[0:2] + ec + n[3:] |
|
264 | 264 | path[i] = n |
|
265 | 265 | if n[-1] in b'. ': |
|
266 | 266 | # encode last period or space ('foo...' -> 'foo..~2e') |
|
267 | 267 | path[i] = n[:-1] + b"~%02x" % ord(n[-1:]) |
|
268 | 268 | return path |
|
269 | 269 | |
|
270 | 270 | |
|
271 | 271 | _maxstorepathlen = 120 |
|
272 | 272 | _dirprefixlen = 8 |
|
273 | 273 | _maxshortdirslen = 8 * (_dirprefixlen + 1) - 4 |
|
274 | 274 | |
|
275 | 275 | |
|
276 | 276 | def _hashencode(path, dotencode): |
|
277 | 277 | digest = hex(hashutil.sha1(path).digest()) |
|
278 | 278 | le = lowerencode(path[5:]).split(b'/') # skips prefix 'data/' or 'meta/' |
|
279 | 279 | parts = _auxencode(le, dotencode) |
|
280 | 280 | basename = parts[-1] |
|
281 | 281 | _root, ext = os.path.splitext(basename) |
|
282 | 282 | sdirs = [] |
|
283 | 283 | sdirslen = 0 |
|
284 | 284 | for p in parts[:-1]: |
|
285 | 285 | d = p[:_dirprefixlen] |
|
286 | 286 | if d[-1] in b'. ': |
|
287 | 287 | # Windows can't access dirs ending in period or space |
|
288 | 288 | d = d[:-1] + b'_' |
|
289 | 289 | if sdirslen == 0: |
|
290 | 290 | t = len(d) |
|
291 | 291 | else: |
|
292 | 292 | t = sdirslen + 1 + len(d) |
|
293 | 293 | if t > _maxshortdirslen: |
|
294 | 294 | break |
|
295 | 295 | sdirs.append(d) |
|
296 | 296 | sdirslen = t |
|
297 | 297 | dirs = b'/'.join(sdirs) |
|
298 | 298 | if len(dirs) > 0: |
|
299 | 299 | dirs += b'/' |
|
300 | 300 | res = b'dh/' + dirs + digest + ext |
|
301 | 301 | spaceleft = _maxstorepathlen - len(res) |
|
302 | 302 | if spaceleft > 0: |
|
303 | 303 | filler = basename[:spaceleft] |
|
304 | 304 | res = b'dh/' + dirs + filler + digest + ext |
|
305 | 305 | return res |
|
306 | 306 | |
|
307 | 307 | |
|
308 | 308 | def _hybridencode(path, dotencode): |
|
309 | 309 | """encodes path with a length limit |
|
310 | 310 | |
|
311 | 311 | Encodes all paths that begin with 'data/', according to the following. |
|
312 | 312 | |
|
313 | 313 | Default encoding (reversible): |
|
314 | 314 | |
|
315 | 315 | Encodes all uppercase letters 'X' as '_x'. All reserved or illegal |
|
316 | 316 | characters are encoded as '~xx', where xx is the two digit hex code |
|
317 | 317 | of the character (see encodefilename). |
|
318 | 318 | Relevant path components consisting of Windows reserved filenames are |
|
319 | 319 | masked by encoding the third character ('aux' -> 'au~78', see _auxencode). |
|
320 | 320 | |
|
321 | 321 | Hashed encoding (not reversible): |
|
322 | 322 | |
|
323 | 323 | If the default-encoded path is longer than _maxstorepathlen, a |
|
324 | 324 | non-reversible hybrid hashing of the path is done instead. |
|
325 | 325 | This encoding uses up to _dirprefixlen characters of all directory |
|
326 | 326 | levels of the lowerencoded path, but not more levels than can fit into |
|
327 | 327 | _maxshortdirslen. |
|
328 | 328 | Then follows the filler followed by the sha digest of the full path. |
|
329 | 329 | The filler is the beginning of the basename of the lowerencoded path |
|
330 | 330 | (the basename is everything after the last path separator). The filler |
|
331 | 331 | is as long as possible, filling in characters from the basename until |
|
332 | 332 | the encoded path has _maxstorepathlen characters (or all chars of the |
|
333 | 333 | basename have been taken). |
|
334 | 334 | The extension (e.g. '.i' or '.d') is preserved. |
|
335 | 335 | |
|
336 | 336 | The string 'data/' at the beginning is replaced with 'dh/', if the hashed |
|
337 | 337 | encoding was used. |
|
338 | 338 | """ |
|
339 | 339 | path = encodedir(path) |
|
340 | 340 | ef = _encodefname(path).split(b'/') |
|
341 | 341 | res = b'/'.join(_auxencode(ef, dotencode)) |
|
342 | 342 | if len(res) > _maxstorepathlen: |
|
343 | 343 | res = _hashencode(path, dotencode) |
|
344 | 344 | return res |
|
345 | 345 | |
|
346 | 346 | |
|
347 | 347 | def _pathencode(path): |
|
348 | 348 | de = encodedir(path) |
|
349 | 349 | if len(path) > _maxstorepathlen: |
|
350 | 350 | return _hashencode(de, True) |
|
351 | 351 | ef = _encodefname(de).split(b'/') |
|
352 | 352 | res = b'/'.join(_auxencode(ef, True)) |
|
353 | 353 | if len(res) > _maxstorepathlen: |
|
354 | 354 | return _hashencode(de, True) |
|
355 | 355 | return res |
|
356 | 356 | |
|
357 | 357 | |
|
358 | 358 | _pathencode = getattr(parsers, 'pathencode', _pathencode) |
|
359 | 359 | |
|
360 | 360 | |
|
361 | 361 | def _plainhybridencode(f): |
|
362 | 362 | return _hybridencode(f, False) |
|
363 | 363 | |
|
364 | 364 | |
|
365 | 365 | def _calcmode(vfs): |
|
366 | 366 | try: |
|
367 | 367 | # files in .hg/ will be created using this mode |
|
368 | 368 | mode = vfs.stat().st_mode |
|
369 | 369 | # avoid some useless chmods |
|
370 | 370 | if (0o777 & ~util.umask) == (0o777 & mode): |
|
371 | 371 | mode = None |
|
372 | 372 | except OSError: |
|
373 | 373 | mode = None |
|
374 | 374 | return mode |
|
375 | 375 | |
|
376 | 376 | |
|
377 | 377 | _data = [ |
|
378 | 378 | b'bookmarks', |
|
379 | 379 | b'narrowspec', |
|
380 | 380 | b'data', |
|
381 | 381 | b'meta', |
|
382 | 382 | b'00manifest.d', |
|
383 | 383 | b'00manifest.i', |
|
384 | 384 | b'00changelog.d', |
|
385 | 385 | b'00changelog.i', |
|
386 | 386 | b'phaseroots', |
|
387 | 387 | b'obsstore', |
|
388 | 388 | b'requires', |
|
389 | 389 | ] |
|
390 | 390 | |
|
391 | 391 | REVLOG_FILES_MAIN_EXT = (b'.i', b'i.tmpcensored') |
|
392 | 392 | REVLOG_FILES_OTHER_EXT = ( |
|
393 | 393 | b'.idx', |
|
394 | 394 | b'.d', |
|
395 | 395 | b'.dat', |
|
396 | 396 | b'.n', |
|
397 | 397 | b'.nd', |
|
398 | 398 | b'.sda', |
|
399 | 399 | b'd.tmpcensored', |
|
400 | 400 | ) |
|
401 | 401 | # files that are "volatile" and might change between listing and streaming |
|
402 | 402 | # |
|
403 | 403 | # note: the ".nd" file are nodemap data and won't "change" but they might be |
|
404 | 404 | # deleted. |
|
405 | 405 | REVLOG_FILES_VOLATILE_EXT = (b'.n', b'.nd') |
|
406 | 406 | |
|
407 | 407 | # some exception to the above matching |
|
408 | 408 | # |
|
409 | 409 | # XXX This is currently not in use because of issue6542 |
|
410 | 410 | EXCLUDED = re.compile(b'.*undo\.[^/]+\.(nd?|i)$') |
|
411 | 411 | |
|
412 | 412 | |
|
413 | 413 | def is_revlog(f, kind, st): |
|
414 | 414 | if kind != stat.S_IFREG: |
|
415 | 415 | return None |
|
416 | 416 | return revlog_type(f) |
|
417 | 417 | |
|
418 | 418 | |
|
419 | 419 | def revlog_type(f): |
|
420 | 420 | # XXX we need to filter `undo.` created by the transaction here, however |
|
421 | 421 | # being naive about it also filter revlog for `undo.*` files, leading to |
|
422 | 422 | # issue6542. So we no longer use EXCLUDED. |
|
423 | 423 | if f.endswith(REVLOG_FILES_MAIN_EXT): |
|
424 | 424 | return FILEFLAGS_REVLOG_MAIN |
|
425 | 425 | elif f.endswith(REVLOG_FILES_OTHER_EXT): |
|
426 | 426 | t = FILETYPE_FILELOG_OTHER |
|
427 | 427 | if f.endswith(REVLOG_FILES_VOLATILE_EXT): |
|
428 | 428 | t |= FILEFLAGS_VOLATILE |
|
429 | 429 | return t |
|
430 | 430 | return None |
|
431 | 431 | |
|
432 | 432 | |
|
433 | 433 | # the file is part of changelog data |
|
434 | 434 | FILEFLAGS_CHANGELOG = 1 << 13 |
|
435 | 435 | # the file is part of manifest data |
|
436 | 436 | FILEFLAGS_MANIFESTLOG = 1 << 12 |
|
437 | 437 | # the file is part of filelog data |
|
438 | 438 | FILEFLAGS_FILELOG = 1 << 11 |
|
439 | 439 | # file that are not directly part of a revlog |
|
440 | 440 | FILEFLAGS_OTHER = 1 << 10 |
|
441 | 441 | |
|
442 | 442 | # the main entry point for a revlog |
|
443 | 443 | FILEFLAGS_REVLOG_MAIN = 1 << 1 |
|
444 | 444 | # a secondary file for a revlog |
|
445 | 445 | FILEFLAGS_REVLOG_OTHER = 1 << 0 |
|
446 | 446 | |
|
447 | 447 | # files that are "volatile" and might change between listing and streaming |
|
448 | 448 | FILEFLAGS_VOLATILE = 1 << 20 |
|
449 | 449 | |
|
450 | 450 | FILETYPE_CHANGELOG_MAIN = FILEFLAGS_CHANGELOG | FILEFLAGS_REVLOG_MAIN |
|
451 | 451 | FILETYPE_CHANGELOG_OTHER = FILEFLAGS_CHANGELOG | FILEFLAGS_REVLOG_OTHER |
|
452 | 452 | FILETYPE_MANIFESTLOG_MAIN = FILEFLAGS_MANIFESTLOG | FILEFLAGS_REVLOG_MAIN |
|
453 | 453 | FILETYPE_MANIFESTLOG_OTHER = FILEFLAGS_MANIFESTLOG | FILEFLAGS_REVLOG_OTHER |
|
454 | 454 | FILETYPE_FILELOG_MAIN = FILEFLAGS_FILELOG | FILEFLAGS_REVLOG_MAIN |
|
455 | 455 | FILETYPE_FILELOG_OTHER = FILEFLAGS_FILELOG | FILEFLAGS_REVLOG_OTHER |
|
456 | 456 | FILETYPE_OTHER = FILEFLAGS_OTHER |
|
457 | 457 | |
|
458 | 458 | |
|
459 | 459 | class basicstore(object): |
|
460 | 460 | '''base class for local repository stores''' |
|
461 | 461 | |
|
462 | 462 | def __init__(self, path, vfstype): |
|
463 | 463 | vfs = vfstype(path) |
|
464 | 464 | self.path = vfs.base |
|
465 | 465 | self.createmode = _calcmode(vfs) |
|
466 | 466 | vfs.createmode = self.createmode |
|
467 | 467 | self.rawvfs = vfs |
|
468 | 468 | self.vfs = vfsmod.filtervfs(vfs, encodedir) |
|
469 | 469 | self.opener = self.vfs |
|
470 | 470 | |
|
471 | 471 | def join(self, f): |
|
472 | 472 | return self.path + b'/' + encodedir(f) |
|
473 | 473 | |
|
474 | 474 | def _walk(self, relpath, recurse): |
|
475 |
'''yields ( |
|
|
475 | '''yields (revlog_type, unencoded, size)''' | |
|
476 | 476 | path = self.path |
|
477 | 477 | if relpath: |
|
478 | 478 | path += b'/' + relpath |
|
479 | 479 | striplen = len(self.path) + 1 |
|
480 | 480 | l = [] |
|
481 | 481 | if self.rawvfs.isdir(path): |
|
482 | 482 | visit = [path] |
|
483 | 483 | readdir = self.rawvfs.readdir |
|
484 | 484 | while visit: |
|
485 | 485 | p = visit.pop() |
|
486 | 486 | for f, kind, st in readdir(p, stat=True): |
|
487 | 487 | fp = p + b'/' + f |
|
488 | 488 | rl_type = is_revlog(f, kind, st) |
|
489 | 489 | if rl_type is not None: |
|
490 | 490 | n = util.pconvert(fp[striplen:]) |
|
491 |
l.append((rl_type, decodedir(n), |
|
|
491 | l.append((rl_type, decodedir(n), st.st_size)) | |
|
492 | 492 | elif kind == stat.S_IFDIR and recurse: |
|
493 | 493 | visit.append(fp) |
|
494 | 494 | l.sort() |
|
495 | 495 | return l |
|
496 | 496 | |
|
497 | 497 | def changelog(self, trypending, concurrencychecker=None): |
|
498 | 498 | return changelog.changelog( |
|
499 | 499 | self.vfs, |
|
500 | 500 | trypending=trypending, |
|
501 | 501 | concurrencychecker=concurrencychecker, |
|
502 | 502 | ) |
|
503 | 503 | |
|
504 | 504 | def manifestlog(self, repo, storenarrowmatch): |
|
505 | 505 | rootstore = manifest.manifestrevlog(repo.nodeconstants, self.vfs) |
|
506 | 506 | return manifest.manifestlog(self.vfs, repo, rootstore, storenarrowmatch) |
|
507 | 507 | |
|
508 | def datafiles(self, matcher=None): | |
|
508 | def datafiles(self, matcher=None, undecodable=None): | |
|
509 | """Like walk, but excluding the changelog and root manifest. | |
|
510 | ||
|
511 | When [undecodable] is None, revlogs names that can't be | |
|
512 | decoded cause an exception. When it is provided, it should | |
|
513 | be a list and the filenames that can't be decoded are added | |
|
514 | to it instead. This is very rarely needed.""" | |
|
509 | 515 | files = self._walk(b'data', True) + self._walk(b'meta', True) |
|
510 |
for (t, u, |
|
|
511 |
yield (FILEFLAGS_FILELOG | t, u, |
|
|
516 | for (t, u, s) in files: | |
|
517 | yield (FILEFLAGS_FILELOG | t, u, s) | |
|
512 | 518 | |
|
513 | 519 | def topfiles(self): |
|
514 | 520 | # yield manifest before changelog |
|
515 | 521 | files = reversed(self._walk(b'', False)) |
|
516 |
for (t, u, |
|
|
522 | for (t, u, s) in files: | |
|
517 | 523 | if u.startswith(b'00changelog'): |
|
518 |
yield (FILEFLAGS_CHANGELOG | t, u, |
|
|
524 | yield (FILEFLAGS_CHANGELOG | t, u, s) | |
|
519 | 525 | elif u.startswith(b'00manifest'): |
|
520 |
yield (FILEFLAGS_MANIFESTLOG | t, u, |
|
|
526 | yield (FILEFLAGS_MANIFESTLOG | t, u, s) | |
|
521 | 527 | else: |
|
522 |
yield (FILETYPE_OTHER | t, u, |
|
|
528 | yield (FILETYPE_OTHER | t, u, s) | |
|
523 | 529 | |
|
524 | 530 | def walk(self, matcher=None): |
|
525 | 531 | """return file related to data storage (ie: revlogs) |
|
526 | 532 | |
|
527 |
yields (file_type, unencoded, |
|
|
533 | yields (file_type, unencoded, size) | |
|
528 | 534 | |
|
529 | 535 | if a matcher is passed, storage files of only those tracked paths |
|
530 | 536 | are passed with matches the matcher |
|
531 | 537 | """ |
|
532 | 538 | # yield data files first |
|
533 | 539 | for x in self.datafiles(matcher): |
|
534 | 540 | yield x |
|
535 | 541 | for x in self.topfiles(): |
|
536 | 542 | yield x |
|
537 | 543 | |
|
538 | 544 | def copylist(self): |
|
539 | 545 | return _data |
|
540 | 546 | |
|
541 | 547 | def write(self, tr): |
|
542 | 548 | pass |
|
543 | 549 | |
|
544 | 550 | def invalidatecaches(self): |
|
545 | 551 | pass |
|
546 | 552 | |
|
547 | 553 | def markremoved(self, fn): |
|
548 | 554 | pass |
|
549 | 555 | |
|
550 | 556 | def __contains__(self, path): |
|
551 | 557 | '''Checks if the store contains path''' |
|
552 | 558 | path = b"/".join((b"data", path)) |
|
553 | 559 | # file? |
|
554 | 560 | if self.vfs.exists(path + b".i"): |
|
555 | 561 | return True |
|
556 | 562 | # dir? |
|
557 | 563 | if not path.endswith(b"/"): |
|
558 | 564 | path = path + b"/" |
|
559 | 565 | return self.vfs.exists(path) |
|
560 | 566 | |
|
561 | 567 | |
|
562 | 568 | class encodedstore(basicstore): |
|
563 | 569 | def __init__(self, path, vfstype): |
|
564 | 570 | vfs = vfstype(path + b'/store') |
|
565 | 571 | self.path = vfs.base |
|
566 | 572 | self.createmode = _calcmode(vfs) |
|
567 | 573 | vfs.createmode = self.createmode |
|
568 | 574 | self.rawvfs = vfs |
|
569 | 575 | self.vfs = vfsmod.filtervfs(vfs, encodefilename) |
|
570 | 576 | self.opener = self.vfs |
|
571 | 577 | |
|
572 | 578 | # note: topfiles would also need a decode phase. It is just that in |
|
573 | 579 | # practice we do not have any file outside of `data/` that needs encoding. |
|
574 | 580 | # However that might change so we should probably add a test and encoding |
|
575 | 581 | # decoding for it too. see issue6548 |
|
576 | 582 | |
|
577 | def datafiles(self, matcher=None): | |
|
578 |
for t, |
|
|
583 | def datafiles(self, matcher=None, undecodable=None): | |
|
584 | for t, f1, size in super(encodedstore, self).datafiles(): | |
|
579 | 585 | try: |
|
580 |
|
|
|
586 | f2 = decodefilename(f1) | |
|
581 | 587 | except KeyError: |
|
582 |
|
|
|
583 | if a is not None and not _matchtrackedpath(a, matcher): | |
|
588 | if undecodable is None: | |
|
589 | msg = _(b'undecodable revlog name %s') % f1 | |
|
590 | raise error.StorageError(msg) | |
|
591 | else: | |
|
592 | undecodable.append(f1) | |
|
593 | continue | |
|
594 | if not _matchtrackedpath(f2, matcher): | |
|
584 | 595 | continue |
|
585 |
yield t, |
|
|
596 | yield t, f2, size | |
|
586 | 597 | |
|
587 | 598 | def join(self, f): |
|
588 | 599 | return self.path + b'/' + encodefilename(f) |
|
589 | 600 | |
|
590 | 601 | def copylist(self): |
|
591 | 602 | return [b'requires', b'00changelog.i'] + [b'store/' + f for f in _data] |
|
592 | 603 | |
|
593 | 604 | |
|
594 | 605 | class fncache(object): |
|
595 | 606 | # the filename used to be partially encoded |
|
596 | 607 | # hence the encodedir/decodedir dance |
|
597 | 608 | def __init__(self, vfs): |
|
598 | 609 | self.vfs = vfs |
|
599 | 610 | self.entries = None |
|
600 | 611 | self._dirty = False |
|
601 | 612 | # set of new additions to fncache |
|
602 | 613 | self.addls = set() |
|
603 | 614 | |
|
604 | 615 | def ensureloaded(self, warn=None): |
|
605 | 616 | """read the fncache file if not already read. |
|
606 | 617 | |
|
607 | 618 | If the file on disk is corrupted, raise. If warn is provided, |
|
608 | 619 | warn and keep going instead.""" |
|
609 | 620 | if self.entries is None: |
|
610 | 621 | self._load(warn) |
|
611 | 622 | |
|
612 | 623 | def _load(self, warn=None): |
|
613 | 624 | '''fill the entries from the fncache file''' |
|
614 | 625 | self._dirty = False |
|
615 | 626 | try: |
|
616 | 627 | fp = self.vfs(b'fncache', mode=b'rb') |
|
617 | 628 | except IOError: |
|
618 | 629 | # skip nonexistent file |
|
619 | 630 | self.entries = set() |
|
620 | 631 | return |
|
621 | 632 | |
|
622 | 633 | self.entries = set() |
|
623 | 634 | chunk = b'' |
|
624 | 635 | for c in iter(functools.partial(fp.read, fncache_chunksize), b''): |
|
625 | 636 | chunk += c |
|
626 | 637 | try: |
|
627 | 638 | p = chunk.rindex(b'\n') |
|
628 | 639 | self.entries.update(decodedir(chunk[: p + 1]).splitlines()) |
|
629 | 640 | chunk = chunk[p + 1 :] |
|
630 | 641 | except ValueError: |
|
631 | 642 | # substring '\n' not found, maybe the entry is bigger than the |
|
632 | 643 | # chunksize, so let's keep iterating |
|
633 | 644 | pass |
|
634 | 645 | |
|
635 | 646 | if chunk: |
|
636 | 647 | msg = _(b"fncache does not ends with a newline") |
|
637 | 648 | if warn: |
|
638 | 649 | warn(msg + b'\n') |
|
639 | 650 | else: |
|
640 | 651 | raise error.Abort( |
|
641 | 652 | msg, |
|
642 | 653 | hint=_( |
|
643 | 654 | b"use 'hg debugrebuildfncache' to " |
|
644 | 655 | b"rebuild the fncache" |
|
645 | 656 | ), |
|
646 | 657 | ) |
|
647 | 658 | self._checkentries(fp, warn) |
|
648 | 659 | fp.close() |
|
649 | 660 | |
|
650 | 661 | def _checkentries(self, fp, warn): |
|
651 | 662 | """make sure there is no empty string in entries""" |
|
652 | 663 | if b'' in self.entries: |
|
653 | 664 | fp.seek(0) |
|
654 | 665 | for n, line in enumerate(util.iterfile(fp)): |
|
655 | 666 | if not line.rstrip(b'\n'): |
|
656 | 667 | t = _(b'invalid entry in fncache, line %d') % (n + 1) |
|
657 | 668 | if warn: |
|
658 | 669 | warn(t + b'\n') |
|
659 | 670 | else: |
|
660 | 671 | raise error.Abort(t) |
|
661 | 672 | |
|
662 | 673 | def write(self, tr): |
|
663 | 674 | if self._dirty: |
|
664 | 675 | assert self.entries is not None |
|
665 | 676 | self.entries = self.entries | self.addls |
|
666 | 677 | self.addls = set() |
|
667 | 678 | tr.addbackup(b'fncache') |
|
668 | 679 | fp = self.vfs(b'fncache', mode=b'wb', atomictemp=True) |
|
669 | 680 | if self.entries: |
|
670 | 681 | fp.write(encodedir(b'\n'.join(self.entries) + b'\n')) |
|
671 | 682 | fp.close() |
|
672 | 683 | self._dirty = False |
|
673 | 684 | if self.addls: |
|
674 | 685 | # if we have just new entries, let's append them to the fncache |
|
675 | 686 | tr.addbackup(b'fncache') |
|
676 | 687 | fp = self.vfs(b'fncache', mode=b'ab', atomictemp=True) |
|
677 | 688 | if self.addls: |
|
678 | 689 | fp.write(encodedir(b'\n'.join(self.addls) + b'\n')) |
|
679 | 690 | fp.close() |
|
680 | 691 | self.entries = None |
|
681 | 692 | self.addls = set() |
|
682 | 693 | |
|
683 | 694 | def add(self, fn): |
|
684 | 695 | if self.entries is None: |
|
685 | 696 | self._load() |
|
686 | 697 | if fn not in self.entries: |
|
687 | 698 | self.addls.add(fn) |
|
688 | 699 | |
|
689 | 700 | def remove(self, fn): |
|
690 | 701 | if self.entries is None: |
|
691 | 702 | self._load() |
|
692 | 703 | if fn in self.addls: |
|
693 | 704 | self.addls.remove(fn) |
|
694 | 705 | return |
|
695 | 706 | try: |
|
696 | 707 | self.entries.remove(fn) |
|
697 | 708 | self._dirty = True |
|
698 | 709 | except KeyError: |
|
699 | 710 | pass |
|
700 | 711 | |
|
701 | 712 | def __contains__(self, fn): |
|
702 | 713 | if fn in self.addls: |
|
703 | 714 | return True |
|
704 | 715 | if self.entries is None: |
|
705 | 716 | self._load() |
|
706 | 717 | return fn in self.entries |
|
707 | 718 | |
|
708 | 719 | def __iter__(self): |
|
709 | 720 | if self.entries is None: |
|
710 | 721 | self._load() |
|
711 | 722 | return iter(self.entries | self.addls) |
|
712 | 723 | |
|
713 | 724 | |
|
714 | 725 | class _fncachevfs(vfsmod.proxyvfs): |
|
715 | 726 | def __init__(self, vfs, fnc, encode): |
|
716 | 727 | vfsmod.proxyvfs.__init__(self, vfs) |
|
717 | 728 | self.fncache = fnc |
|
718 | 729 | self.encode = encode |
|
719 | 730 | |
|
720 | 731 | def __call__(self, path, mode=b'r', *args, **kw): |
|
721 | 732 | encoded = self.encode(path) |
|
722 | 733 | if mode not in (b'r', b'rb') and ( |
|
723 | 734 | path.startswith(b'data/') or path.startswith(b'meta/') |
|
724 | 735 | ): |
|
725 | 736 | # do not trigger a fncache load when adding a file that already is |
|
726 | 737 | # known to exist. |
|
727 | 738 | notload = self.fncache.entries is None and self.vfs.exists(encoded) |
|
728 | 739 | if notload and b'r+' in mode and not self.vfs.stat(encoded).st_size: |
|
729 | 740 | # when appending to an existing file, if the file has size zero, |
|
730 | 741 | # it should be considered as missing. Such zero-size files are |
|
731 | 742 | # the result of truncation when a transaction is aborted. |
|
732 | 743 | notload = False |
|
733 | 744 | if not notload: |
|
734 | 745 | self.fncache.add(path) |
|
735 | 746 | return self.vfs(encoded, mode, *args, **kw) |
|
736 | 747 | |
|
737 | 748 | def join(self, path): |
|
738 | 749 | if path: |
|
739 | 750 | return self.vfs.join(self.encode(path)) |
|
740 | 751 | else: |
|
741 | 752 | return self.vfs.join(path) |
|
742 | 753 | |
|
743 | 754 | def register_file(self, path): |
|
744 | 755 | """generic hook point to lets fncache steer its stew""" |
|
745 | 756 | if path.startswith(b'data/') or path.startswith(b'meta/'): |
|
746 | 757 | self.fncache.add(path) |
|
747 | 758 | |
|
748 | 759 | |
|
749 | 760 | class fncachestore(basicstore): |
|
750 | 761 | def __init__(self, path, vfstype, dotencode): |
|
751 | 762 | if dotencode: |
|
752 | 763 | encode = _pathencode |
|
753 | 764 | else: |
|
754 | 765 | encode = _plainhybridencode |
|
755 | 766 | self.encode = encode |
|
756 | 767 | vfs = vfstype(path + b'/store') |
|
757 | 768 | self.path = vfs.base |
|
758 | 769 | self.pathsep = self.path + b'/' |
|
759 | 770 | self.createmode = _calcmode(vfs) |
|
760 | 771 | vfs.createmode = self.createmode |
|
761 | 772 | self.rawvfs = vfs |
|
762 | 773 | fnc = fncache(vfs) |
|
763 | 774 | self.fncache = fnc |
|
764 | 775 | self.vfs = _fncachevfs(vfs, fnc, encode) |
|
765 | 776 | self.opener = self.vfs |
|
766 | 777 | |
|
767 | 778 | def join(self, f): |
|
768 | 779 | return self.pathsep + self.encode(f) |
|
769 | 780 | |
|
770 | 781 | def getsize(self, path): |
|
771 | 782 | return self.rawvfs.stat(path).st_size |
|
772 | 783 | |
|
773 | def datafiles(self, matcher=None): | |
|
784 | def datafiles(self, matcher=None, undecodable=None): | |
|
774 | 785 | for f in sorted(self.fncache): |
|
775 | 786 | if not _matchtrackedpath(f, matcher): |
|
776 | 787 | continue |
|
777 | 788 | ef = self.encode(f) |
|
778 | 789 | try: |
|
779 | 790 | t = revlog_type(f) |
|
780 | 791 | assert t is not None, f |
|
781 | 792 | t |= FILEFLAGS_FILELOG |
|
782 |
yield t, f, |
|
|
793 | yield t, f, self.getsize(ef) | |
|
783 | 794 | except OSError as err: |
|
784 | 795 | if err.errno != errno.ENOENT: |
|
785 | 796 | raise |
|
786 | 797 | |
|
787 | 798 | def copylist(self): |
|
788 | 799 | d = ( |
|
789 | 800 | b'bookmarks', |
|
790 | 801 | b'narrowspec', |
|
791 | 802 | b'data', |
|
792 | 803 | b'meta', |
|
793 | 804 | b'dh', |
|
794 | 805 | b'fncache', |
|
795 | 806 | b'phaseroots', |
|
796 | 807 | b'obsstore', |
|
797 | 808 | b'00manifest.d', |
|
798 | 809 | b'00manifest.i', |
|
799 | 810 | b'00changelog.d', |
|
800 | 811 | b'00changelog.i', |
|
801 | 812 | b'requires', |
|
802 | 813 | ) |
|
803 | 814 | return [b'requires', b'00changelog.i'] + [b'store/' + f for f in d] |
|
804 | 815 | |
|
805 | 816 | def write(self, tr): |
|
806 | 817 | self.fncache.write(tr) |
|
807 | 818 | |
|
808 | 819 | def invalidatecaches(self): |
|
809 | 820 | self.fncache.entries = None |
|
810 | 821 | self.fncache.addls = set() |
|
811 | 822 | |
|
812 | 823 | def markremoved(self, fn): |
|
813 | 824 | self.fncache.remove(fn) |
|
814 | 825 | |
|
815 | 826 | def _exists(self, f): |
|
816 | 827 | ef = self.encode(f) |
|
817 | 828 | try: |
|
818 | 829 | self.getsize(ef) |
|
819 | 830 | return True |
|
820 | 831 | except OSError as err: |
|
821 | 832 | if err.errno != errno.ENOENT: |
|
822 | 833 | raise |
|
823 | 834 | # nonexistent entry |
|
824 | 835 | return False |
|
825 | 836 | |
|
826 | 837 | def __contains__(self, path): |
|
827 | 838 | '''Checks if the store contains path''' |
|
828 | 839 | path = b"/".join((b"data", path)) |
|
829 | 840 | # check for files (exact match) |
|
830 | 841 | e = path + b'.i' |
|
831 | 842 | if e in self.fncache and self._exists(e): |
|
832 | 843 | return True |
|
833 | 844 | # now check for directories (prefix match) |
|
834 | 845 | if not path.endswith(b'/'): |
|
835 | 846 | path += b'/' |
|
836 | 847 | for e in self.fncache: |
|
837 | 848 | if e.startswith(path) and self._exists(e): |
|
838 | 849 | return True |
|
839 | 850 | return False |
@@ -1,918 +1,918 b'' | |||
|
1 | 1 | # streamclone.py - producing and consuming streaming repository data |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2015 Gregory Szorc <gregory.szorc@gmail.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 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import contextlib |
|
11 | 11 | import errno |
|
12 | 12 | import os |
|
13 | 13 | import struct |
|
14 | 14 | |
|
15 | 15 | from .i18n import _ |
|
16 | 16 | from .pycompat import open |
|
17 | 17 | from .interfaces import repository |
|
18 | 18 | from . import ( |
|
19 | 19 | bookmarks, |
|
20 | 20 | cacheutil, |
|
21 | 21 | error, |
|
22 | 22 | narrowspec, |
|
23 | 23 | phases, |
|
24 | 24 | pycompat, |
|
25 | 25 | requirements as requirementsmod, |
|
26 | 26 | scmutil, |
|
27 | 27 | store, |
|
28 | 28 | util, |
|
29 | 29 | ) |
|
30 | 30 | from .utils import ( |
|
31 | 31 | stringutil, |
|
32 | 32 | ) |
|
33 | 33 | |
|
34 | 34 | |
|
35 | 35 | def canperformstreamclone(pullop, bundle2=False): |
|
36 | 36 | """Whether it is possible to perform a streaming clone as part of pull. |
|
37 | 37 | |
|
38 | 38 | ``bundle2`` will cause the function to consider stream clone through |
|
39 | 39 | bundle2 and only through bundle2. |
|
40 | 40 | |
|
41 | 41 | Returns a tuple of (supported, requirements). ``supported`` is True if |
|
42 | 42 | streaming clone is supported and False otherwise. ``requirements`` is |
|
43 | 43 | a set of repo requirements from the remote, or ``None`` if stream clone |
|
44 | 44 | isn't supported. |
|
45 | 45 | """ |
|
46 | 46 | repo = pullop.repo |
|
47 | 47 | remote = pullop.remote |
|
48 | 48 | |
|
49 | 49 | bundle2supported = False |
|
50 | 50 | if pullop.canusebundle2: |
|
51 | 51 | if b'v2' in pullop.remotebundle2caps.get(b'stream', []): |
|
52 | 52 | bundle2supported = True |
|
53 | 53 | # else |
|
54 | 54 | # Server doesn't support bundle2 stream clone or doesn't support |
|
55 | 55 | # the versions we support. Fall back and possibly allow legacy. |
|
56 | 56 | |
|
57 | 57 | # Ensures legacy code path uses available bundle2. |
|
58 | 58 | if bundle2supported and not bundle2: |
|
59 | 59 | return False, None |
|
60 | 60 | # Ensures bundle2 doesn't try to do a stream clone if it isn't supported. |
|
61 | 61 | elif bundle2 and not bundle2supported: |
|
62 | 62 | return False, None |
|
63 | 63 | |
|
64 | 64 | # Streaming clone only works on empty repositories. |
|
65 | 65 | if len(repo): |
|
66 | 66 | return False, None |
|
67 | 67 | |
|
68 | 68 | # Streaming clone only works if all data is being requested. |
|
69 | 69 | if pullop.heads: |
|
70 | 70 | return False, None |
|
71 | 71 | |
|
72 | 72 | streamrequested = pullop.streamclonerequested |
|
73 | 73 | |
|
74 | 74 | # If we don't have a preference, let the server decide for us. This |
|
75 | 75 | # likely only comes into play in LANs. |
|
76 | 76 | if streamrequested is None: |
|
77 | 77 | # The server can advertise whether to prefer streaming clone. |
|
78 | 78 | streamrequested = remote.capable(b'stream-preferred') |
|
79 | 79 | |
|
80 | 80 | if not streamrequested: |
|
81 | 81 | return False, None |
|
82 | 82 | |
|
83 | 83 | # In order for stream clone to work, the client has to support all the |
|
84 | 84 | # requirements advertised by the server. |
|
85 | 85 | # |
|
86 | 86 | # The server advertises its requirements via the "stream" and "streamreqs" |
|
87 | 87 | # capability. "stream" (a value-less capability) is advertised if and only |
|
88 | 88 | # if the only requirement is "revlogv1." Else, the "streamreqs" capability |
|
89 | 89 | # is advertised and contains a comma-delimited list of requirements. |
|
90 | 90 | requirements = set() |
|
91 | 91 | if remote.capable(b'stream'): |
|
92 | 92 | requirements.add(requirementsmod.REVLOGV1_REQUIREMENT) |
|
93 | 93 | else: |
|
94 | 94 | streamreqs = remote.capable(b'streamreqs') |
|
95 | 95 | # This is weird and shouldn't happen with modern servers. |
|
96 | 96 | if not streamreqs: |
|
97 | 97 | pullop.repo.ui.warn( |
|
98 | 98 | _( |
|
99 | 99 | b'warning: stream clone requested but server has them ' |
|
100 | 100 | b'disabled\n' |
|
101 | 101 | ) |
|
102 | 102 | ) |
|
103 | 103 | return False, None |
|
104 | 104 | |
|
105 | 105 | streamreqs = set(streamreqs.split(b',')) |
|
106 | 106 | # Server requires something we don't support. Bail. |
|
107 | 107 | missingreqs = streamreqs - repo.supportedformats |
|
108 | 108 | if missingreqs: |
|
109 | 109 | pullop.repo.ui.warn( |
|
110 | 110 | _( |
|
111 | 111 | b'warning: stream clone requested but client is missing ' |
|
112 | 112 | b'requirements: %s\n' |
|
113 | 113 | ) |
|
114 | 114 | % b', '.join(sorted(missingreqs)) |
|
115 | 115 | ) |
|
116 | 116 | pullop.repo.ui.warn( |
|
117 | 117 | _( |
|
118 | 118 | b'(see https://www.mercurial-scm.org/wiki/MissingRequirement ' |
|
119 | 119 | b'for more information)\n' |
|
120 | 120 | ) |
|
121 | 121 | ) |
|
122 | 122 | return False, None |
|
123 | 123 | requirements = streamreqs |
|
124 | 124 | |
|
125 | 125 | return True, requirements |
|
126 | 126 | |
|
127 | 127 | |
|
128 | 128 | def maybeperformlegacystreamclone(pullop): |
|
129 | 129 | """Possibly perform a legacy stream clone operation. |
|
130 | 130 | |
|
131 | 131 | Legacy stream clones are performed as part of pull but before all other |
|
132 | 132 | operations. |
|
133 | 133 | |
|
134 | 134 | A legacy stream clone will not be performed if a bundle2 stream clone is |
|
135 | 135 | supported. |
|
136 | 136 | """ |
|
137 | 137 | from . import localrepo |
|
138 | 138 | |
|
139 | 139 | supported, requirements = canperformstreamclone(pullop) |
|
140 | 140 | |
|
141 | 141 | if not supported: |
|
142 | 142 | return |
|
143 | 143 | |
|
144 | 144 | repo = pullop.repo |
|
145 | 145 | remote = pullop.remote |
|
146 | 146 | |
|
147 | 147 | # Save remote branchmap. We will use it later to speed up branchcache |
|
148 | 148 | # creation. |
|
149 | 149 | rbranchmap = None |
|
150 | 150 | if remote.capable(b'branchmap'): |
|
151 | 151 | with remote.commandexecutor() as e: |
|
152 | 152 | rbranchmap = e.callcommand(b'branchmap', {}).result() |
|
153 | 153 | |
|
154 | 154 | repo.ui.status(_(b'streaming all changes\n')) |
|
155 | 155 | |
|
156 | 156 | with remote.commandexecutor() as e: |
|
157 | 157 | fp = e.callcommand(b'stream_out', {}).result() |
|
158 | 158 | |
|
159 | 159 | # TODO strictly speaking, this code should all be inside the context |
|
160 | 160 | # manager because the context manager is supposed to ensure all wire state |
|
161 | 161 | # is flushed when exiting. But the legacy peers don't do this, so it |
|
162 | 162 | # doesn't matter. |
|
163 | 163 | l = fp.readline() |
|
164 | 164 | try: |
|
165 | 165 | resp = int(l) |
|
166 | 166 | except ValueError: |
|
167 | 167 | raise error.ResponseError( |
|
168 | 168 | _(b'unexpected response from remote server:'), l |
|
169 | 169 | ) |
|
170 | 170 | if resp == 1: |
|
171 | 171 | raise error.Abort(_(b'operation forbidden by server')) |
|
172 | 172 | elif resp == 2: |
|
173 | 173 | raise error.Abort(_(b'locking the remote repository failed')) |
|
174 | 174 | elif resp != 0: |
|
175 | 175 | raise error.Abort(_(b'the server sent an unknown error code')) |
|
176 | 176 | |
|
177 | 177 | l = fp.readline() |
|
178 | 178 | try: |
|
179 | 179 | filecount, bytecount = map(int, l.split(b' ', 1)) |
|
180 | 180 | except (ValueError, TypeError): |
|
181 | 181 | raise error.ResponseError( |
|
182 | 182 | _(b'unexpected response from remote server:'), l |
|
183 | 183 | ) |
|
184 | 184 | |
|
185 | 185 | with repo.lock(): |
|
186 | 186 | consumev1(repo, fp, filecount, bytecount) |
|
187 | 187 | |
|
188 | 188 | # new requirements = old non-format requirements + |
|
189 | 189 | # new format-related remote requirements |
|
190 | 190 | # requirements from the streamed-in repository |
|
191 | 191 | repo.requirements = requirements | ( |
|
192 | 192 | repo.requirements - repo.supportedformats |
|
193 | 193 | ) |
|
194 | 194 | repo.svfs.options = localrepo.resolvestorevfsoptions( |
|
195 | 195 | repo.ui, repo.requirements, repo.features |
|
196 | 196 | ) |
|
197 | 197 | scmutil.writereporequirements(repo) |
|
198 | 198 | |
|
199 | 199 | if rbranchmap: |
|
200 | 200 | repo._branchcaches.replace(repo, rbranchmap) |
|
201 | 201 | |
|
202 | 202 | repo.invalidate() |
|
203 | 203 | |
|
204 | 204 | |
|
205 | 205 | def allowservergeneration(repo): |
|
206 | 206 | """Whether streaming clones are allowed from the server.""" |
|
207 | 207 | if repository.REPO_FEATURE_STREAM_CLONE not in repo.features: |
|
208 | 208 | return False |
|
209 | 209 | |
|
210 | 210 | if not repo.ui.configbool(b'server', b'uncompressed', untrusted=True): |
|
211 | 211 | return False |
|
212 | 212 | |
|
213 | 213 | # The way stream clone works makes it impossible to hide secret changesets. |
|
214 | 214 | # So don't allow this by default. |
|
215 | 215 | secret = phases.hassecret(repo) |
|
216 | 216 | if secret: |
|
217 | 217 | return repo.ui.configbool(b'server', b'uncompressedallowsecret') |
|
218 | 218 | |
|
219 | 219 | return True |
|
220 | 220 | |
|
221 | 221 | |
|
222 | 222 | # This is it's own function so extensions can override it. |
|
223 | 223 | def _walkstreamfiles(repo, matcher=None): |
|
224 | 224 | return repo.store.walk(matcher) |
|
225 | 225 | |
|
226 | 226 | |
|
227 | 227 | def generatev1(repo): |
|
228 | 228 | """Emit content for version 1 of a streaming clone. |
|
229 | 229 | |
|
230 | 230 | This returns a 3-tuple of (file count, byte size, data iterator). |
|
231 | 231 | |
|
232 | 232 | The data iterator consists of N entries for each file being transferred. |
|
233 | 233 | Each file entry starts as a line with the file name and integer size |
|
234 | 234 | delimited by a null byte. |
|
235 | 235 | |
|
236 | 236 | The raw file data follows. Following the raw file data is the next file |
|
237 | 237 | entry, or EOF. |
|
238 | 238 | |
|
239 | 239 | When used on the wire protocol, an additional line indicating protocol |
|
240 | 240 | success will be prepended to the stream. This function is not responsible |
|
241 | 241 | for adding it. |
|
242 | 242 | |
|
243 | 243 | This function will obtain a repository lock to ensure a consistent view of |
|
244 | 244 | the store is captured. It therefore may raise LockError. |
|
245 | 245 | """ |
|
246 | 246 | entries = [] |
|
247 | 247 | total_bytes = 0 |
|
248 | 248 | # Get consistent snapshot of repo, lock during scan. |
|
249 | 249 | with repo.lock(): |
|
250 | 250 | repo.ui.debug(b'scanning\n') |
|
251 |
for file_type, name |
|
|
251 | for file_type, name, size in _walkstreamfiles(repo): | |
|
252 | 252 | if size: |
|
253 | 253 | entries.append((name, size)) |
|
254 | 254 | total_bytes += size |
|
255 | 255 | _test_sync_point_walk_1(repo) |
|
256 | 256 | _test_sync_point_walk_2(repo) |
|
257 | 257 | |
|
258 | 258 | repo.ui.debug( |
|
259 | 259 | b'%d files, %d bytes to transfer\n' % (len(entries), total_bytes) |
|
260 | 260 | ) |
|
261 | 261 | |
|
262 | 262 | svfs = repo.svfs |
|
263 | 263 | debugflag = repo.ui.debugflag |
|
264 | 264 | |
|
265 | 265 | def emitrevlogdata(): |
|
266 | 266 | for name, size in entries: |
|
267 | 267 | if debugflag: |
|
268 | 268 | repo.ui.debug(b'sending %s (%d bytes)\n' % (name, size)) |
|
269 | 269 | # partially encode name over the wire for backwards compat |
|
270 | 270 | yield b'%s\0%d\n' % (store.encodedir(name), size) |
|
271 | 271 | # auditing at this stage is both pointless (paths are already |
|
272 | 272 | # trusted by the local repo) and expensive |
|
273 | 273 | with svfs(name, b'rb', auditpath=False) as fp: |
|
274 | 274 | if size <= 65536: |
|
275 | 275 | yield fp.read(size) |
|
276 | 276 | else: |
|
277 | 277 | for chunk in util.filechunkiter(fp, limit=size): |
|
278 | 278 | yield chunk |
|
279 | 279 | |
|
280 | 280 | return len(entries), total_bytes, emitrevlogdata() |
|
281 | 281 | |
|
282 | 282 | |
|
283 | 283 | def generatev1wireproto(repo): |
|
284 | 284 | """Emit content for version 1 of streaming clone suitable for the wire. |
|
285 | 285 | |
|
286 | 286 | This is the data output from ``generatev1()`` with 2 header lines. The |
|
287 | 287 | first line indicates overall success. The 2nd contains the file count and |
|
288 | 288 | byte size of payload. |
|
289 | 289 | |
|
290 | 290 | The success line contains "0" for success, "1" for stream generation not |
|
291 | 291 | allowed, and "2" for error locking the repository (possibly indicating |
|
292 | 292 | a permissions error for the server process). |
|
293 | 293 | """ |
|
294 | 294 | if not allowservergeneration(repo): |
|
295 | 295 | yield b'1\n' |
|
296 | 296 | return |
|
297 | 297 | |
|
298 | 298 | try: |
|
299 | 299 | filecount, bytecount, it = generatev1(repo) |
|
300 | 300 | except error.LockError: |
|
301 | 301 | yield b'2\n' |
|
302 | 302 | return |
|
303 | 303 | |
|
304 | 304 | # Indicates successful response. |
|
305 | 305 | yield b'0\n' |
|
306 | 306 | yield b'%d %d\n' % (filecount, bytecount) |
|
307 | 307 | for chunk in it: |
|
308 | 308 | yield chunk |
|
309 | 309 | |
|
310 | 310 | |
|
311 | 311 | def generatebundlev1(repo, compression=b'UN'): |
|
312 | 312 | """Emit content for version 1 of a stream clone bundle. |
|
313 | 313 | |
|
314 | 314 | The first 4 bytes of the output ("HGS1") denote this as stream clone |
|
315 | 315 | bundle version 1. |
|
316 | 316 | |
|
317 | 317 | The next 2 bytes indicate the compression type. Only "UN" is currently |
|
318 | 318 | supported. |
|
319 | 319 | |
|
320 | 320 | The next 16 bytes are two 64-bit big endian unsigned integers indicating |
|
321 | 321 | file count and byte count, respectively. |
|
322 | 322 | |
|
323 | 323 | The next 2 bytes is a 16-bit big endian unsigned short declaring the length |
|
324 | 324 | of the requirements string, including a trailing \0. The following N bytes |
|
325 | 325 | are the requirements string, which is ASCII containing a comma-delimited |
|
326 | 326 | list of repo requirements that are needed to support the data. |
|
327 | 327 | |
|
328 | 328 | The remaining content is the output of ``generatev1()`` (which may be |
|
329 | 329 | compressed in the future). |
|
330 | 330 | |
|
331 | 331 | Returns a tuple of (requirements, data generator). |
|
332 | 332 | """ |
|
333 | 333 | if compression != b'UN': |
|
334 | 334 | raise ValueError(b'we do not support the compression argument yet') |
|
335 | 335 | |
|
336 | 336 | requirements = repo.requirements & repo.supportedformats |
|
337 | 337 | requires = b','.join(sorted(requirements)) |
|
338 | 338 | |
|
339 | 339 | def gen(): |
|
340 | 340 | yield b'HGS1' |
|
341 | 341 | yield compression |
|
342 | 342 | |
|
343 | 343 | filecount, bytecount, it = generatev1(repo) |
|
344 | 344 | repo.ui.status( |
|
345 | 345 | _(b'writing %d bytes for %d files\n') % (bytecount, filecount) |
|
346 | 346 | ) |
|
347 | 347 | |
|
348 | 348 | yield struct.pack(b'>QQ', filecount, bytecount) |
|
349 | 349 | yield struct.pack(b'>H', len(requires) + 1) |
|
350 | 350 | yield requires + b'\0' |
|
351 | 351 | |
|
352 | 352 | # This is where we'll add compression in the future. |
|
353 | 353 | assert compression == b'UN' |
|
354 | 354 | |
|
355 | 355 | progress = repo.ui.makeprogress( |
|
356 | 356 | _(b'bundle'), total=bytecount, unit=_(b'bytes') |
|
357 | 357 | ) |
|
358 | 358 | progress.update(0) |
|
359 | 359 | |
|
360 | 360 | for chunk in it: |
|
361 | 361 | progress.increment(step=len(chunk)) |
|
362 | 362 | yield chunk |
|
363 | 363 | |
|
364 | 364 | progress.complete() |
|
365 | 365 | |
|
366 | 366 | return requirements, gen() |
|
367 | 367 | |
|
368 | 368 | |
|
369 | 369 | def consumev1(repo, fp, filecount, bytecount): |
|
370 | 370 | """Apply the contents from version 1 of a streaming clone file handle. |
|
371 | 371 | |
|
372 | 372 | This takes the output from "stream_out" and applies it to the specified |
|
373 | 373 | repository. |
|
374 | 374 | |
|
375 | 375 | Like "stream_out," the status line added by the wire protocol is not |
|
376 | 376 | handled by this function. |
|
377 | 377 | """ |
|
378 | 378 | with repo.lock(): |
|
379 | 379 | repo.ui.status( |
|
380 | 380 | _(b'%d files to transfer, %s of data\n') |
|
381 | 381 | % (filecount, util.bytecount(bytecount)) |
|
382 | 382 | ) |
|
383 | 383 | progress = repo.ui.makeprogress( |
|
384 | 384 | _(b'clone'), total=bytecount, unit=_(b'bytes') |
|
385 | 385 | ) |
|
386 | 386 | progress.update(0) |
|
387 | 387 | start = util.timer() |
|
388 | 388 | |
|
389 | 389 | # TODO: get rid of (potential) inconsistency |
|
390 | 390 | # |
|
391 | 391 | # If transaction is started and any @filecache property is |
|
392 | 392 | # changed at this point, it causes inconsistency between |
|
393 | 393 | # in-memory cached property and streamclone-ed file on the |
|
394 | 394 | # disk. Nested transaction prevents transaction scope "clone" |
|
395 | 395 | # below from writing in-memory changes out at the end of it, |
|
396 | 396 | # even though in-memory changes are discarded at the end of it |
|
397 | 397 | # regardless of transaction nesting. |
|
398 | 398 | # |
|
399 | 399 | # But transaction nesting can't be simply prohibited, because |
|
400 | 400 | # nesting occurs also in ordinary case (e.g. enabling |
|
401 | 401 | # clonebundles). |
|
402 | 402 | |
|
403 | 403 | with repo.transaction(b'clone'): |
|
404 | 404 | with repo.svfs.backgroundclosing(repo.ui, expectedcount=filecount): |
|
405 | 405 | for i in pycompat.xrange(filecount): |
|
406 | 406 | # XXX doesn't support '\n' or '\r' in filenames |
|
407 | 407 | l = fp.readline() |
|
408 | 408 | try: |
|
409 | 409 | name, size = l.split(b'\0', 1) |
|
410 | 410 | size = int(size) |
|
411 | 411 | except (ValueError, TypeError): |
|
412 | 412 | raise error.ResponseError( |
|
413 | 413 | _(b'unexpected response from remote server:'), l |
|
414 | 414 | ) |
|
415 | 415 | if repo.ui.debugflag: |
|
416 | 416 | repo.ui.debug( |
|
417 | 417 | b'adding %s (%s)\n' % (name, util.bytecount(size)) |
|
418 | 418 | ) |
|
419 | 419 | # for backwards compat, name was partially encoded |
|
420 | 420 | path = store.decodedir(name) |
|
421 | 421 | with repo.svfs(path, b'w', backgroundclose=True) as ofp: |
|
422 | 422 | for chunk in util.filechunkiter(fp, limit=size): |
|
423 | 423 | progress.increment(step=len(chunk)) |
|
424 | 424 | ofp.write(chunk) |
|
425 | 425 | |
|
426 | 426 | # force @filecache properties to be reloaded from |
|
427 | 427 | # streamclone-ed file at next access |
|
428 | 428 | repo.invalidate(clearfilecache=True) |
|
429 | 429 | |
|
430 | 430 | elapsed = util.timer() - start |
|
431 | 431 | if elapsed <= 0: |
|
432 | 432 | elapsed = 0.001 |
|
433 | 433 | progress.complete() |
|
434 | 434 | repo.ui.status( |
|
435 | 435 | _(b'transferred %s in %.1f seconds (%s/sec)\n') |
|
436 | 436 | % ( |
|
437 | 437 | util.bytecount(bytecount), |
|
438 | 438 | elapsed, |
|
439 | 439 | util.bytecount(bytecount / elapsed), |
|
440 | 440 | ) |
|
441 | 441 | ) |
|
442 | 442 | |
|
443 | 443 | |
|
444 | 444 | def readbundle1header(fp): |
|
445 | 445 | compression = fp.read(2) |
|
446 | 446 | if compression != b'UN': |
|
447 | 447 | raise error.Abort( |
|
448 | 448 | _( |
|
449 | 449 | b'only uncompressed stream clone bundles are ' |
|
450 | 450 | b'supported; got %s' |
|
451 | 451 | ) |
|
452 | 452 | % compression |
|
453 | 453 | ) |
|
454 | 454 | |
|
455 | 455 | filecount, bytecount = struct.unpack(b'>QQ', fp.read(16)) |
|
456 | 456 | requireslen = struct.unpack(b'>H', fp.read(2))[0] |
|
457 | 457 | requires = fp.read(requireslen) |
|
458 | 458 | |
|
459 | 459 | if not requires.endswith(b'\0'): |
|
460 | 460 | raise error.Abort( |
|
461 | 461 | _( |
|
462 | 462 | b'malformed stream clone bundle: ' |
|
463 | 463 | b'requirements not properly encoded' |
|
464 | 464 | ) |
|
465 | 465 | ) |
|
466 | 466 | |
|
467 | 467 | requirements = set(requires.rstrip(b'\0').split(b',')) |
|
468 | 468 | |
|
469 | 469 | return filecount, bytecount, requirements |
|
470 | 470 | |
|
471 | 471 | |
|
472 | 472 | def applybundlev1(repo, fp): |
|
473 | 473 | """Apply the content from a stream clone bundle version 1. |
|
474 | 474 | |
|
475 | 475 | We assume the 4 byte header has been read and validated and the file handle |
|
476 | 476 | is at the 2 byte compression identifier. |
|
477 | 477 | """ |
|
478 | 478 | if len(repo): |
|
479 | 479 | raise error.Abort( |
|
480 | 480 | _(b'cannot apply stream clone bundle on non-empty repo') |
|
481 | 481 | ) |
|
482 | 482 | |
|
483 | 483 | filecount, bytecount, requirements = readbundle1header(fp) |
|
484 | 484 | missingreqs = requirements - repo.supportedformats |
|
485 | 485 | if missingreqs: |
|
486 | 486 | raise error.Abort( |
|
487 | 487 | _(b'unable to apply stream clone: unsupported format: %s') |
|
488 | 488 | % b', '.join(sorted(missingreqs)) |
|
489 | 489 | ) |
|
490 | 490 | |
|
491 | 491 | consumev1(repo, fp, filecount, bytecount) |
|
492 | 492 | |
|
493 | 493 | |
|
494 | 494 | class streamcloneapplier(object): |
|
495 | 495 | """Class to manage applying streaming clone bundles. |
|
496 | 496 | |
|
497 | 497 | We need to wrap ``applybundlev1()`` in a dedicated type to enable bundle |
|
498 | 498 | readers to perform bundle type-specific functionality. |
|
499 | 499 | """ |
|
500 | 500 | |
|
501 | 501 | def __init__(self, fh): |
|
502 | 502 | self._fh = fh |
|
503 | 503 | |
|
504 | 504 | def apply(self, repo): |
|
505 | 505 | return applybundlev1(repo, self._fh) |
|
506 | 506 | |
|
507 | 507 | |
|
508 | 508 | # type of file to stream |
|
509 | 509 | _fileappend = 0 # append only file |
|
510 | 510 | _filefull = 1 # full snapshot file |
|
511 | 511 | |
|
512 | 512 | # Source of the file |
|
513 | 513 | _srcstore = b's' # store (svfs) |
|
514 | 514 | _srccache = b'c' # cache (cache) |
|
515 | 515 | |
|
516 | 516 | # This is it's own function so extensions can override it. |
|
517 | 517 | def _walkstreamfullstorefiles(repo): |
|
518 | 518 | """list snapshot file from the store""" |
|
519 | 519 | fnames = [] |
|
520 | 520 | if not repo.publishing(): |
|
521 | 521 | fnames.append(b'phaseroots') |
|
522 | 522 | return fnames |
|
523 | 523 | |
|
524 | 524 | |
|
525 | 525 | def _filterfull(entry, copy, vfsmap): |
|
526 | 526 | """actually copy the snapshot files""" |
|
527 | 527 | src, name, ftype, data = entry |
|
528 | 528 | if ftype != _filefull: |
|
529 | 529 | return entry |
|
530 | 530 | return (src, name, ftype, copy(vfsmap[src].join(name))) |
|
531 | 531 | |
|
532 | 532 | |
|
533 | 533 | @contextlib.contextmanager |
|
534 | 534 | def maketempcopies(): |
|
535 | 535 | """return a function to temporary copy file""" |
|
536 | 536 | files = [] |
|
537 | 537 | try: |
|
538 | 538 | |
|
539 | 539 | def copy(src): |
|
540 | 540 | fd, dst = pycompat.mkstemp() |
|
541 | 541 | os.close(fd) |
|
542 | 542 | files.append(dst) |
|
543 | 543 | util.copyfiles(src, dst, hardlink=True) |
|
544 | 544 | return dst |
|
545 | 545 | |
|
546 | 546 | yield copy |
|
547 | 547 | finally: |
|
548 | 548 | for tmp in files: |
|
549 | 549 | util.tryunlink(tmp) |
|
550 | 550 | |
|
551 | 551 | |
|
552 | 552 | def _makemap(repo): |
|
553 | 553 | """make a (src -> vfs) map for the repo""" |
|
554 | 554 | vfsmap = { |
|
555 | 555 | _srcstore: repo.svfs, |
|
556 | 556 | _srccache: repo.cachevfs, |
|
557 | 557 | } |
|
558 | 558 | # we keep repo.vfs out of the on purpose, ther are too many danger there |
|
559 | 559 | # (eg: .hg/hgrc) |
|
560 | 560 | assert repo.vfs not in vfsmap.values() |
|
561 | 561 | |
|
562 | 562 | return vfsmap |
|
563 | 563 | |
|
564 | 564 | |
|
565 | 565 | def _emit2(repo, entries, totalfilesize): |
|
566 | 566 | """actually emit the stream bundle""" |
|
567 | 567 | vfsmap = _makemap(repo) |
|
568 | 568 | # we keep repo.vfs out of the on purpose, ther are too many danger there |
|
569 | 569 | # (eg: .hg/hgrc), |
|
570 | 570 | # |
|
571 | 571 | # this assert is duplicated (from _makemap) as author might think this is |
|
572 | 572 | # fine, while this is really not fine. |
|
573 | 573 | if repo.vfs in vfsmap.values(): |
|
574 | 574 | raise error.ProgrammingError( |
|
575 | 575 | b'repo.vfs must not be added to vfsmap for security reasons' |
|
576 | 576 | ) |
|
577 | 577 | |
|
578 | 578 | progress = repo.ui.makeprogress( |
|
579 | 579 | _(b'bundle'), total=totalfilesize, unit=_(b'bytes') |
|
580 | 580 | ) |
|
581 | 581 | progress.update(0) |
|
582 | 582 | with maketempcopies() as copy, progress: |
|
583 | 583 | # copy is delayed until we are in the try |
|
584 | 584 | entries = [_filterfull(e, copy, vfsmap) for e in entries] |
|
585 | 585 | yield None # this release the lock on the repository |
|
586 | 586 | totalbytecount = 0 |
|
587 | 587 | |
|
588 | 588 | for src, name, ftype, data in entries: |
|
589 | 589 | vfs = vfsmap[src] |
|
590 | 590 | yield src |
|
591 | 591 | yield util.uvarintencode(len(name)) |
|
592 | 592 | if ftype == _fileappend: |
|
593 | 593 | fp = vfs(name) |
|
594 | 594 | size = data |
|
595 | 595 | elif ftype == _filefull: |
|
596 | 596 | fp = open(data, b'rb') |
|
597 | 597 | size = util.fstat(fp).st_size |
|
598 | 598 | bytecount = 0 |
|
599 | 599 | try: |
|
600 | 600 | yield util.uvarintencode(size) |
|
601 | 601 | yield name |
|
602 | 602 | if size <= 65536: |
|
603 | 603 | chunks = (fp.read(size),) |
|
604 | 604 | else: |
|
605 | 605 | chunks = util.filechunkiter(fp, limit=size) |
|
606 | 606 | for chunk in chunks: |
|
607 | 607 | bytecount += len(chunk) |
|
608 | 608 | totalbytecount += len(chunk) |
|
609 | 609 | progress.update(totalbytecount) |
|
610 | 610 | yield chunk |
|
611 | 611 | if bytecount != size: |
|
612 | 612 | # Would most likely be caused by a race due to `hg strip` or |
|
613 | 613 | # a revlog split |
|
614 | 614 | raise error.Abort( |
|
615 | 615 | _( |
|
616 | 616 | b'clone could only read %d bytes from %s, but ' |
|
617 | 617 | b'expected %d bytes' |
|
618 | 618 | ) |
|
619 | 619 | % (bytecount, name, size) |
|
620 | 620 | ) |
|
621 | 621 | finally: |
|
622 | 622 | fp.close() |
|
623 | 623 | |
|
624 | 624 | |
|
625 | 625 | def _test_sync_point_walk_1(repo): |
|
626 | 626 | """a function for synchronisation during tests""" |
|
627 | 627 | |
|
628 | 628 | |
|
629 | 629 | def _test_sync_point_walk_2(repo): |
|
630 | 630 | """a function for synchronisation during tests""" |
|
631 | 631 | |
|
632 | 632 | |
|
633 | 633 | def _v2_walk(repo, includes, excludes, includeobsmarkers): |
|
634 | 634 | """emit a seris of files information useful to clone a repo |
|
635 | 635 | |
|
636 | 636 | return (entries, totalfilesize) |
|
637 | 637 | |
|
638 | 638 | entries is a list of tuple (vfs-key, file-path, file-type, size) |
|
639 | 639 | |
|
640 | 640 | - `vfs-key`: is a key to the right vfs to write the file (see _makemap) |
|
641 | 641 | - `name`: file path of the file to copy (to be feed to the vfss) |
|
642 | 642 | - `file-type`: do this file need to be copied with the source lock ? |
|
643 | 643 | - `size`: the size of the file (or None) |
|
644 | 644 | """ |
|
645 | 645 | assert repo._currentlock(repo._lockref) is not None |
|
646 | 646 | entries = [] |
|
647 | 647 | totalfilesize = 0 |
|
648 | 648 | |
|
649 | 649 | matcher = None |
|
650 | 650 | if includes or excludes: |
|
651 | 651 | matcher = narrowspec.match(repo.root, includes, excludes) |
|
652 | 652 | |
|
653 |
for rl_type, name |
|
|
653 | for rl_type, name, size in _walkstreamfiles(repo, matcher): | |
|
654 | 654 | if size: |
|
655 | 655 | ft = _fileappend |
|
656 | 656 | if rl_type & store.FILEFLAGS_VOLATILE: |
|
657 | 657 | ft = _filefull |
|
658 | 658 | entries.append((_srcstore, name, ft, size)) |
|
659 | 659 | totalfilesize += size |
|
660 | 660 | for name in _walkstreamfullstorefiles(repo): |
|
661 | 661 | if repo.svfs.exists(name): |
|
662 | 662 | totalfilesize += repo.svfs.lstat(name).st_size |
|
663 | 663 | entries.append((_srcstore, name, _filefull, None)) |
|
664 | 664 | if includeobsmarkers and repo.svfs.exists(b'obsstore'): |
|
665 | 665 | totalfilesize += repo.svfs.lstat(b'obsstore').st_size |
|
666 | 666 | entries.append((_srcstore, b'obsstore', _filefull, None)) |
|
667 | 667 | for name in cacheutil.cachetocopy(repo): |
|
668 | 668 | if repo.cachevfs.exists(name): |
|
669 | 669 | totalfilesize += repo.cachevfs.lstat(name).st_size |
|
670 | 670 | entries.append((_srccache, name, _filefull, None)) |
|
671 | 671 | return entries, totalfilesize |
|
672 | 672 | |
|
673 | 673 | |
|
674 | 674 | def generatev2(repo, includes, excludes, includeobsmarkers): |
|
675 | 675 | """Emit content for version 2 of a streaming clone. |
|
676 | 676 | |
|
677 | 677 | the data stream consists the following entries: |
|
678 | 678 | 1) A char representing the file destination (eg: store or cache) |
|
679 | 679 | 2) A varint containing the length of the filename |
|
680 | 680 | 3) A varint containing the length of file data |
|
681 | 681 | 4) N bytes containing the filename (the internal, store-agnostic form) |
|
682 | 682 | 5) N bytes containing the file data |
|
683 | 683 | |
|
684 | 684 | Returns a 3-tuple of (file count, file size, data iterator). |
|
685 | 685 | """ |
|
686 | 686 | |
|
687 | 687 | with repo.lock(): |
|
688 | 688 | |
|
689 | 689 | repo.ui.debug(b'scanning\n') |
|
690 | 690 | |
|
691 | 691 | entries, totalfilesize = _v2_walk( |
|
692 | 692 | repo, |
|
693 | 693 | includes=includes, |
|
694 | 694 | excludes=excludes, |
|
695 | 695 | includeobsmarkers=includeobsmarkers, |
|
696 | 696 | ) |
|
697 | 697 | |
|
698 | 698 | chunks = _emit2(repo, entries, totalfilesize) |
|
699 | 699 | first = next(chunks) |
|
700 | 700 | assert first is None |
|
701 | 701 | _test_sync_point_walk_1(repo) |
|
702 | 702 | _test_sync_point_walk_2(repo) |
|
703 | 703 | |
|
704 | 704 | return len(entries), totalfilesize, chunks |
|
705 | 705 | |
|
706 | 706 | |
|
707 | 707 | @contextlib.contextmanager |
|
708 | 708 | def nested(*ctxs): |
|
709 | 709 | this = ctxs[0] |
|
710 | 710 | rest = ctxs[1:] |
|
711 | 711 | with this: |
|
712 | 712 | if rest: |
|
713 | 713 | with nested(*rest): |
|
714 | 714 | yield |
|
715 | 715 | else: |
|
716 | 716 | yield |
|
717 | 717 | |
|
718 | 718 | |
|
719 | 719 | def consumev2(repo, fp, filecount, filesize): |
|
720 | 720 | """Apply the contents from a version 2 streaming clone. |
|
721 | 721 | |
|
722 | 722 | Data is read from an object that only needs to provide a ``read(size)`` |
|
723 | 723 | method. |
|
724 | 724 | """ |
|
725 | 725 | with repo.lock(): |
|
726 | 726 | repo.ui.status( |
|
727 | 727 | _(b'%d files to transfer, %s of data\n') |
|
728 | 728 | % (filecount, util.bytecount(filesize)) |
|
729 | 729 | ) |
|
730 | 730 | |
|
731 | 731 | start = util.timer() |
|
732 | 732 | progress = repo.ui.makeprogress( |
|
733 | 733 | _(b'clone'), total=filesize, unit=_(b'bytes') |
|
734 | 734 | ) |
|
735 | 735 | progress.update(0) |
|
736 | 736 | |
|
737 | 737 | vfsmap = _makemap(repo) |
|
738 | 738 | # we keep repo.vfs out of the on purpose, ther are too many danger |
|
739 | 739 | # there (eg: .hg/hgrc), |
|
740 | 740 | # |
|
741 | 741 | # this assert is duplicated (from _makemap) as author might think this |
|
742 | 742 | # is fine, while this is really not fine. |
|
743 | 743 | if repo.vfs in vfsmap.values(): |
|
744 | 744 | raise error.ProgrammingError( |
|
745 | 745 | b'repo.vfs must not be added to vfsmap for security reasons' |
|
746 | 746 | ) |
|
747 | 747 | |
|
748 | 748 | with repo.transaction(b'clone'): |
|
749 | 749 | ctxs = (vfs.backgroundclosing(repo.ui) for vfs in vfsmap.values()) |
|
750 | 750 | with nested(*ctxs): |
|
751 | 751 | for i in range(filecount): |
|
752 | 752 | src = util.readexactly(fp, 1) |
|
753 | 753 | vfs = vfsmap[src] |
|
754 | 754 | namelen = util.uvarintdecodestream(fp) |
|
755 | 755 | datalen = util.uvarintdecodestream(fp) |
|
756 | 756 | |
|
757 | 757 | name = util.readexactly(fp, namelen) |
|
758 | 758 | |
|
759 | 759 | if repo.ui.debugflag: |
|
760 | 760 | repo.ui.debug( |
|
761 | 761 | b'adding [%s] %s (%s)\n' |
|
762 | 762 | % (src, name, util.bytecount(datalen)) |
|
763 | 763 | ) |
|
764 | 764 | |
|
765 | 765 | with vfs(name, b'w') as ofp: |
|
766 | 766 | for chunk in util.filechunkiter(fp, limit=datalen): |
|
767 | 767 | progress.increment(step=len(chunk)) |
|
768 | 768 | ofp.write(chunk) |
|
769 | 769 | |
|
770 | 770 | # force @filecache properties to be reloaded from |
|
771 | 771 | # streamclone-ed file at next access |
|
772 | 772 | repo.invalidate(clearfilecache=True) |
|
773 | 773 | |
|
774 | 774 | elapsed = util.timer() - start |
|
775 | 775 | if elapsed <= 0: |
|
776 | 776 | elapsed = 0.001 |
|
777 | 777 | repo.ui.status( |
|
778 | 778 | _(b'transferred %s in %.1f seconds (%s/sec)\n') |
|
779 | 779 | % ( |
|
780 | 780 | util.bytecount(progress.pos), |
|
781 | 781 | elapsed, |
|
782 | 782 | util.bytecount(progress.pos / elapsed), |
|
783 | 783 | ) |
|
784 | 784 | ) |
|
785 | 785 | progress.complete() |
|
786 | 786 | |
|
787 | 787 | |
|
788 | 788 | def applybundlev2(repo, fp, filecount, filesize, requirements): |
|
789 | 789 | from . import localrepo |
|
790 | 790 | |
|
791 | 791 | missingreqs = [r for r in requirements if r not in repo.supported] |
|
792 | 792 | if missingreqs: |
|
793 | 793 | raise error.Abort( |
|
794 | 794 | _(b'unable to apply stream clone: unsupported format: %s') |
|
795 | 795 | % b', '.join(sorted(missingreqs)) |
|
796 | 796 | ) |
|
797 | 797 | |
|
798 | 798 | consumev2(repo, fp, filecount, filesize) |
|
799 | 799 | |
|
800 | 800 | # new requirements = old non-format requirements + |
|
801 | 801 | # new format-related remote requirements |
|
802 | 802 | # requirements from the streamed-in repository |
|
803 | 803 | repo.requirements = set(requirements) | ( |
|
804 | 804 | repo.requirements - repo.supportedformats |
|
805 | 805 | ) |
|
806 | 806 | repo.svfs.options = localrepo.resolvestorevfsoptions( |
|
807 | 807 | repo.ui, repo.requirements, repo.features |
|
808 | 808 | ) |
|
809 | 809 | scmutil.writereporequirements(repo) |
|
810 | 810 | |
|
811 | 811 | |
|
812 | 812 | def _copy_files(src_vfs_map, dst_vfs_map, entries, progress): |
|
813 | 813 | hardlink = [True] |
|
814 | 814 | |
|
815 | 815 | def copy_used(): |
|
816 | 816 | hardlink[0] = False |
|
817 | 817 | progress.topic = _(b'copying') |
|
818 | 818 | |
|
819 | 819 | for k, path, size in entries: |
|
820 | 820 | src_vfs = src_vfs_map[k] |
|
821 | 821 | dst_vfs = dst_vfs_map[k] |
|
822 | 822 | src_path = src_vfs.join(path) |
|
823 | 823 | dst_path = dst_vfs.join(path) |
|
824 | 824 | dirname = dst_vfs.dirname(path) |
|
825 | 825 | if not dst_vfs.exists(dirname): |
|
826 | 826 | dst_vfs.makedirs(dirname) |
|
827 | 827 | dst_vfs.register_file(path) |
|
828 | 828 | # XXX we could use the #nb_bytes argument. |
|
829 | 829 | util.copyfile( |
|
830 | 830 | src_path, |
|
831 | 831 | dst_path, |
|
832 | 832 | hardlink=hardlink[0], |
|
833 | 833 | no_hardlink_cb=copy_used, |
|
834 | 834 | check_fs_hardlink=False, |
|
835 | 835 | ) |
|
836 | 836 | progress.increment() |
|
837 | 837 | return hardlink[0] |
|
838 | 838 | |
|
839 | 839 | |
|
840 | 840 | def local_copy(src_repo, dest_repo): |
|
841 | 841 | """copy all content from one local repository to another |
|
842 | 842 | |
|
843 | 843 | This is useful for local clone""" |
|
844 | 844 | src_store_requirements = { |
|
845 | 845 | r |
|
846 | 846 | for r in src_repo.requirements |
|
847 | 847 | if r not in requirementsmod.WORKING_DIR_REQUIREMENTS |
|
848 | 848 | } |
|
849 | 849 | dest_store_requirements = { |
|
850 | 850 | r |
|
851 | 851 | for r in dest_repo.requirements |
|
852 | 852 | if r not in requirementsmod.WORKING_DIR_REQUIREMENTS |
|
853 | 853 | } |
|
854 | 854 | assert src_store_requirements == dest_store_requirements |
|
855 | 855 | |
|
856 | 856 | with dest_repo.lock(): |
|
857 | 857 | with src_repo.lock(): |
|
858 | 858 | |
|
859 | 859 | # bookmark is not integrated to the streaming as it might use the |
|
860 | 860 | # `repo.vfs` and they are too many sentitive data accessible |
|
861 | 861 | # through `repo.vfs` to expose it to streaming clone. |
|
862 | 862 | src_book_vfs = bookmarks.bookmarksvfs(src_repo) |
|
863 | 863 | srcbookmarks = src_book_vfs.join(b'bookmarks') |
|
864 | 864 | bm_count = 0 |
|
865 | 865 | if os.path.exists(srcbookmarks): |
|
866 | 866 | bm_count = 1 |
|
867 | 867 | |
|
868 | 868 | entries, totalfilesize = _v2_walk( |
|
869 | 869 | src_repo, |
|
870 | 870 | includes=None, |
|
871 | 871 | excludes=None, |
|
872 | 872 | includeobsmarkers=True, |
|
873 | 873 | ) |
|
874 | 874 | src_vfs_map = _makemap(src_repo) |
|
875 | 875 | dest_vfs_map = _makemap(dest_repo) |
|
876 | 876 | progress = src_repo.ui.makeprogress( |
|
877 | 877 | topic=_(b'linking'), |
|
878 | 878 | total=len(entries) + bm_count, |
|
879 | 879 | unit=_(b'files'), |
|
880 | 880 | ) |
|
881 | 881 | # copy files |
|
882 | 882 | # |
|
883 | 883 | # We could copy the full file while the source repository is locked |
|
884 | 884 | # and the other one without the lock. However, in the linking case, |
|
885 | 885 | # this would also requires checks that nobody is appending any data |
|
886 | 886 | # to the files while we do the clone, so this is not done yet. We |
|
887 | 887 | # could do this blindly when copying files. |
|
888 | 888 | files = ((k, path, size) for k, path, ftype, size in entries) |
|
889 | 889 | hardlink = _copy_files(src_vfs_map, dest_vfs_map, files, progress) |
|
890 | 890 | |
|
891 | 891 | # copy bookmarks over |
|
892 | 892 | if bm_count: |
|
893 | 893 | dst_book_vfs = bookmarks.bookmarksvfs(dest_repo) |
|
894 | 894 | dstbookmarks = dst_book_vfs.join(b'bookmarks') |
|
895 | 895 | util.copyfile(srcbookmarks, dstbookmarks) |
|
896 | 896 | progress.complete() |
|
897 | 897 | if hardlink: |
|
898 | 898 | msg = b'linked %d files\n' |
|
899 | 899 | else: |
|
900 | 900 | msg = b'copied %d files\n' |
|
901 | 901 | src_repo.ui.debug(msg % (len(entries) + bm_count)) |
|
902 | 902 | |
|
903 | 903 | with dest_repo.transaction(b"localclone") as tr: |
|
904 | 904 | dest_repo.store.write(tr) |
|
905 | 905 | |
|
906 | 906 | # clean up transaction file as they do not make sense |
|
907 | 907 | undo_files = [(dest_repo.svfs, b'undo.backupfiles')] |
|
908 | 908 | undo_files.extend(dest_repo.undofiles()) |
|
909 | 909 | for undovfs, undofile in undo_files: |
|
910 | 910 | try: |
|
911 | 911 | undovfs.unlink(undofile) |
|
912 | 912 | except OSError as e: |
|
913 | 913 | if e.errno != errno.ENOENT: |
|
914 | 914 | msg = _(b'error removing %s: %s\n') |
|
915 | 915 | path = undovfs.join(undofile) |
|
916 | 916 | e_msg = stringutil.forcebytestr(e) |
|
917 | 917 | msg %= (path, e_msg) |
|
918 | 918 | dest_repo.ui.warn(msg) |
@@ -1,649 +1,649 b'' | |||
|
1 | 1 | # upgrade.py - functions for in place upgrade of Mercurial repository |
|
2 | 2 | # |
|
3 | 3 | # Copyright (c) 2016-present, Gregory Szorc |
|
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 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import stat |
|
11 | 11 | |
|
12 | 12 | from ..i18n import _ |
|
13 | 13 | from ..pycompat import getattr |
|
14 | 14 | from .. import ( |
|
15 | 15 | changelog, |
|
16 | 16 | error, |
|
17 | 17 | filelog, |
|
18 | 18 | manifest, |
|
19 | 19 | metadata, |
|
20 | 20 | pycompat, |
|
21 | 21 | requirements, |
|
22 | 22 | scmutil, |
|
23 | 23 | store, |
|
24 | 24 | util, |
|
25 | 25 | vfs as vfsmod, |
|
26 | 26 | ) |
|
27 | 27 | from ..revlogutils import ( |
|
28 | 28 | constants as revlogconst, |
|
29 | 29 | flagutil, |
|
30 | 30 | nodemap, |
|
31 | 31 | sidedata as sidedatamod, |
|
32 | 32 | ) |
|
33 | 33 | from . import actions as upgrade_actions |
|
34 | 34 | |
|
35 | 35 | |
|
36 | 36 | def get_sidedata_helpers(srcrepo, dstrepo): |
|
37 | 37 | use_w = srcrepo.ui.configbool(b'experimental', b'worker.repository-upgrade') |
|
38 | 38 | sequential = pycompat.iswindows or not use_w |
|
39 | 39 | if not sequential: |
|
40 | 40 | srcrepo.register_sidedata_computer( |
|
41 | 41 | revlogconst.KIND_CHANGELOG, |
|
42 | 42 | sidedatamod.SD_FILES, |
|
43 | 43 | (sidedatamod.SD_FILES,), |
|
44 | 44 | metadata._get_worker_sidedata_adder(srcrepo, dstrepo), |
|
45 | 45 | flagutil.REVIDX_HASCOPIESINFO, |
|
46 | 46 | replace=True, |
|
47 | 47 | ) |
|
48 | 48 | return sidedatamod.get_sidedata_helpers(srcrepo, dstrepo._wanted_sidedata) |
|
49 | 49 | |
|
50 | 50 | |
|
51 | 51 | def _revlogfrompath(repo, rl_type, path): |
|
52 | 52 | """Obtain a revlog from a repo path. |
|
53 | 53 | |
|
54 | 54 | An instance of the appropriate class is returned. |
|
55 | 55 | """ |
|
56 | 56 | if rl_type & store.FILEFLAGS_CHANGELOG: |
|
57 | 57 | return changelog.changelog(repo.svfs) |
|
58 | 58 | elif rl_type & store.FILEFLAGS_MANIFESTLOG: |
|
59 | 59 | mandir = b'' |
|
60 | 60 | if b'/' in path: |
|
61 | 61 | mandir = path.rsplit(b'/', 1)[0] |
|
62 | 62 | return manifest.manifestrevlog( |
|
63 | 63 | repo.nodeconstants, repo.svfs, tree=mandir |
|
64 | 64 | ) |
|
65 | 65 | else: |
|
66 | 66 | # drop the extension and the `data/` prefix |
|
67 | 67 | path_part = path.rsplit(b'.', 1)[0].split(b'/', 1) |
|
68 | 68 | if len(path_part) < 2: |
|
69 | 69 | msg = _(b'cannot recognize revlog from filename: %s') |
|
70 | 70 | msg %= path |
|
71 | 71 | raise error.Abort(msg) |
|
72 | 72 | path = path_part[1] |
|
73 | 73 | return filelog.filelog(repo.svfs, path) |
|
74 | 74 | |
|
75 | 75 | |
|
76 | 76 | def _copyrevlog(tr, destrepo, oldrl, rl_type, unencodedname): |
|
77 | 77 | """copy all relevant files for `oldrl` into `destrepo` store |
|
78 | 78 | |
|
79 | 79 | Files are copied "as is" without any transformation. The copy is performed |
|
80 | 80 | without extra checks. Callers are responsible for making sure the copied |
|
81 | 81 | content is compatible with format of the destination repository. |
|
82 | 82 | """ |
|
83 | 83 | oldrl = getattr(oldrl, '_revlog', oldrl) |
|
84 | 84 | newrl = _revlogfrompath(destrepo, rl_type, unencodedname) |
|
85 | 85 | newrl = getattr(newrl, '_revlog', newrl) |
|
86 | 86 | |
|
87 | 87 | oldvfs = oldrl.opener |
|
88 | 88 | newvfs = newrl.opener |
|
89 | 89 | oldindex = oldvfs.join(oldrl._indexfile) |
|
90 | 90 | newindex = newvfs.join(newrl._indexfile) |
|
91 | 91 | olddata = oldvfs.join(oldrl._datafile) |
|
92 | 92 | newdata = newvfs.join(newrl._datafile) |
|
93 | 93 | |
|
94 | 94 | with newvfs(newrl._indexfile, b'w'): |
|
95 | 95 | pass # create all the directories |
|
96 | 96 | |
|
97 | 97 | util.copyfile(oldindex, newindex) |
|
98 | 98 | copydata = oldrl.opener.exists(oldrl._datafile) |
|
99 | 99 | if copydata: |
|
100 | 100 | util.copyfile(olddata, newdata) |
|
101 | 101 | |
|
102 | 102 | if rl_type & store.FILEFLAGS_FILELOG: |
|
103 | 103 | destrepo.svfs.fncache.add(unencodedname) |
|
104 | 104 | if copydata: |
|
105 | 105 | destrepo.svfs.fncache.add(unencodedname[:-2] + b'.d') |
|
106 | 106 | |
|
107 | 107 | |
|
108 | 108 | UPGRADE_CHANGELOG = b"changelog" |
|
109 | 109 | UPGRADE_MANIFEST = b"manifest" |
|
110 | 110 | UPGRADE_FILELOGS = b"all-filelogs" |
|
111 | 111 | |
|
112 | 112 | UPGRADE_ALL_REVLOGS = frozenset( |
|
113 | 113 | [UPGRADE_CHANGELOG, UPGRADE_MANIFEST, UPGRADE_FILELOGS] |
|
114 | 114 | ) |
|
115 | 115 | |
|
116 | 116 | |
|
117 | 117 | def matchrevlog(revlogfilter, rl_type): |
|
118 | 118 | """check if a revlog is selected for cloning. |
|
119 | 119 | |
|
120 | 120 | In other words, are there any updates which need to be done on revlog |
|
121 | 121 | or it can be blindly copied. |
|
122 | 122 | |
|
123 | 123 | The store entry is checked against the passed filter""" |
|
124 | 124 | if rl_type & store.FILEFLAGS_CHANGELOG: |
|
125 | 125 | return UPGRADE_CHANGELOG in revlogfilter |
|
126 | 126 | elif rl_type & store.FILEFLAGS_MANIFESTLOG: |
|
127 | 127 | return UPGRADE_MANIFEST in revlogfilter |
|
128 | 128 | assert rl_type & store.FILEFLAGS_FILELOG |
|
129 | 129 | return UPGRADE_FILELOGS in revlogfilter |
|
130 | 130 | |
|
131 | 131 | |
|
132 | 132 | def _perform_clone( |
|
133 | 133 | ui, |
|
134 | 134 | dstrepo, |
|
135 | 135 | tr, |
|
136 | 136 | old_revlog, |
|
137 | 137 | rl_type, |
|
138 | 138 | unencoded, |
|
139 | 139 | upgrade_op, |
|
140 | 140 | sidedata_helpers, |
|
141 | 141 | oncopiedrevision, |
|
142 | 142 | ): |
|
143 | 143 | """returns the new revlog object created""" |
|
144 | 144 | newrl = None |
|
145 | 145 | if matchrevlog(upgrade_op.revlogs_to_process, rl_type): |
|
146 | 146 | ui.note( |
|
147 | 147 | _(b'cloning %d revisions from %s\n') % (len(old_revlog), unencoded) |
|
148 | 148 | ) |
|
149 | 149 | newrl = _revlogfrompath(dstrepo, rl_type, unencoded) |
|
150 | 150 | old_revlog.clone( |
|
151 | 151 | tr, |
|
152 | 152 | newrl, |
|
153 | 153 | addrevisioncb=oncopiedrevision, |
|
154 | 154 | deltareuse=upgrade_op.delta_reuse_mode, |
|
155 | 155 | forcedeltabothparents=upgrade_op.force_re_delta_both_parents, |
|
156 | 156 | sidedata_helpers=sidedata_helpers, |
|
157 | 157 | ) |
|
158 | 158 | else: |
|
159 | 159 | msg = _(b'blindly copying %s containing %i revisions\n') |
|
160 | 160 | ui.note(msg % (unencoded, len(old_revlog))) |
|
161 | 161 | _copyrevlog(tr, dstrepo, old_revlog, rl_type, unencoded) |
|
162 | 162 | |
|
163 | 163 | newrl = _revlogfrompath(dstrepo, rl_type, unencoded) |
|
164 | 164 | return newrl |
|
165 | 165 | |
|
166 | 166 | |
|
167 | 167 | def _clonerevlogs( |
|
168 | 168 | ui, |
|
169 | 169 | srcrepo, |
|
170 | 170 | dstrepo, |
|
171 | 171 | tr, |
|
172 | 172 | upgrade_op, |
|
173 | 173 | ): |
|
174 | 174 | """Copy revlogs between 2 repos.""" |
|
175 | 175 | revcount = 0 |
|
176 | 176 | srcsize = 0 |
|
177 | 177 | srcrawsize = 0 |
|
178 | 178 | dstsize = 0 |
|
179 | 179 | fcount = 0 |
|
180 | 180 | frevcount = 0 |
|
181 | 181 | fsrcsize = 0 |
|
182 | 182 | frawsize = 0 |
|
183 | 183 | fdstsize = 0 |
|
184 | 184 | mcount = 0 |
|
185 | 185 | mrevcount = 0 |
|
186 | 186 | msrcsize = 0 |
|
187 | 187 | mrawsize = 0 |
|
188 | 188 | mdstsize = 0 |
|
189 | 189 | crevcount = 0 |
|
190 | 190 | csrcsize = 0 |
|
191 | 191 | crawsize = 0 |
|
192 | 192 | cdstsize = 0 |
|
193 | 193 | |
|
194 | 194 | alldatafiles = list(srcrepo.store.walk()) |
|
195 | 195 | # mapping of data files which needs to be cloned |
|
196 | 196 | # key is unencoded filename |
|
197 | 197 | # value is revlog_object_from_srcrepo |
|
198 | 198 | manifests = {} |
|
199 | 199 | changelogs = {} |
|
200 | 200 | filelogs = {} |
|
201 | 201 | |
|
202 | 202 | # Perform a pass to collect metadata. This validates we can open all |
|
203 | 203 | # source files and allows a unified progress bar to be displayed. |
|
204 |
for rl_type, unencoded, |
|
|
204 | for rl_type, unencoded, size in alldatafiles: | |
|
205 | 205 | if not rl_type & store.FILEFLAGS_REVLOG_MAIN: |
|
206 | 206 | continue |
|
207 | 207 | |
|
208 | 208 | # the store.walk function will wrongly pickup transaction backup and |
|
209 | 209 | # get confused. As a quick fix for 5.9 release, we ignore those. |
|
210 | 210 | # (this is not a module constants because it seems better to keep the |
|
211 | 211 | # hack together) |
|
212 | 212 | skip_undo = ( |
|
213 | 213 | b'undo.backup.00changelog.i', |
|
214 | 214 | b'undo.backup.00manifest.i', |
|
215 | 215 | ) |
|
216 | 216 | if unencoded in skip_undo: |
|
217 | 217 | continue |
|
218 | 218 | |
|
219 | 219 | rl = _revlogfrompath(srcrepo, rl_type, unencoded) |
|
220 | 220 | |
|
221 | 221 | info = rl.storageinfo( |
|
222 | 222 | exclusivefiles=True, |
|
223 | 223 | revisionscount=True, |
|
224 | 224 | trackedsize=True, |
|
225 | 225 | storedsize=True, |
|
226 | 226 | ) |
|
227 | 227 | |
|
228 | 228 | revcount += info[b'revisionscount'] or 0 |
|
229 | 229 | datasize = info[b'storedsize'] or 0 |
|
230 | 230 | rawsize = info[b'trackedsize'] or 0 |
|
231 | 231 | |
|
232 | 232 | srcsize += datasize |
|
233 | 233 | srcrawsize += rawsize |
|
234 | 234 | |
|
235 | 235 | # This is for the separate progress bars. |
|
236 | 236 | if rl_type & store.FILEFLAGS_CHANGELOG: |
|
237 | 237 | changelogs[unencoded] = (rl_type, rl) |
|
238 | 238 | crevcount += len(rl) |
|
239 | 239 | csrcsize += datasize |
|
240 | 240 | crawsize += rawsize |
|
241 | 241 | elif rl_type & store.FILEFLAGS_MANIFESTLOG: |
|
242 | 242 | manifests[unencoded] = (rl_type, rl) |
|
243 | 243 | mcount += 1 |
|
244 | 244 | mrevcount += len(rl) |
|
245 | 245 | msrcsize += datasize |
|
246 | 246 | mrawsize += rawsize |
|
247 | 247 | elif rl_type & store.FILEFLAGS_FILELOG: |
|
248 | 248 | filelogs[unencoded] = (rl_type, rl) |
|
249 | 249 | fcount += 1 |
|
250 | 250 | frevcount += len(rl) |
|
251 | 251 | fsrcsize += datasize |
|
252 | 252 | frawsize += rawsize |
|
253 | 253 | else: |
|
254 | 254 | error.ProgrammingError(b'unknown revlog type') |
|
255 | 255 | |
|
256 | 256 | if not revcount: |
|
257 | 257 | return |
|
258 | 258 | |
|
259 | 259 | ui.status( |
|
260 | 260 | _( |
|
261 | 261 | b'migrating %d total revisions (%d in filelogs, %d in manifests, ' |
|
262 | 262 | b'%d in changelog)\n' |
|
263 | 263 | ) |
|
264 | 264 | % (revcount, frevcount, mrevcount, crevcount) |
|
265 | 265 | ) |
|
266 | 266 | ui.status( |
|
267 | 267 | _(b'migrating %s in store; %s tracked data\n') |
|
268 | 268 | % ((util.bytecount(srcsize), util.bytecount(srcrawsize))) |
|
269 | 269 | ) |
|
270 | 270 | |
|
271 | 271 | # Used to keep track of progress. |
|
272 | 272 | progress = None |
|
273 | 273 | |
|
274 | 274 | def oncopiedrevision(rl, rev, node): |
|
275 | 275 | progress.increment() |
|
276 | 276 | |
|
277 | 277 | sidedata_helpers = get_sidedata_helpers(srcrepo, dstrepo) |
|
278 | 278 | |
|
279 | 279 | # Migrating filelogs |
|
280 | 280 | ui.status( |
|
281 | 281 | _( |
|
282 | 282 | b'migrating %d filelogs containing %d revisions ' |
|
283 | 283 | b'(%s in store; %s tracked data)\n' |
|
284 | 284 | ) |
|
285 | 285 | % ( |
|
286 | 286 | fcount, |
|
287 | 287 | frevcount, |
|
288 | 288 | util.bytecount(fsrcsize), |
|
289 | 289 | util.bytecount(frawsize), |
|
290 | 290 | ) |
|
291 | 291 | ) |
|
292 | 292 | progress = srcrepo.ui.makeprogress(_(b'file revisions'), total=frevcount) |
|
293 | 293 | for unencoded, (rl_type, oldrl) in sorted(filelogs.items()): |
|
294 | 294 | newrl = _perform_clone( |
|
295 | 295 | ui, |
|
296 | 296 | dstrepo, |
|
297 | 297 | tr, |
|
298 | 298 | oldrl, |
|
299 | 299 | rl_type, |
|
300 | 300 | unencoded, |
|
301 | 301 | upgrade_op, |
|
302 | 302 | sidedata_helpers, |
|
303 | 303 | oncopiedrevision, |
|
304 | 304 | ) |
|
305 | 305 | info = newrl.storageinfo(storedsize=True) |
|
306 | 306 | fdstsize += info[b'storedsize'] or 0 |
|
307 | 307 | ui.status( |
|
308 | 308 | _( |
|
309 | 309 | b'finished migrating %d filelog revisions across %d ' |
|
310 | 310 | b'filelogs; change in size: %s\n' |
|
311 | 311 | ) |
|
312 | 312 | % (frevcount, fcount, util.bytecount(fdstsize - fsrcsize)) |
|
313 | 313 | ) |
|
314 | 314 | |
|
315 | 315 | # Migrating manifests |
|
316 | 316 | ui.status( |
|
317 | 317 | _( |
|
318 | 318 | b'migrating %d manifests containing %d revisions ' |
|
319 | 319 | b'(%s in store; %s tracked data)\n' |
|
320 | 320 | ) |
|
321 | 321 | % ( |
|
322 | 322 | mcount, |
|
323 | 323 | mrevcount, |
|
324 | 324 | util.bytecount(msrcsize), |
|
325 | 325 | util.bytecount(mrawsize), |
|
326 | 326 | ) |
|
327 | 327 | ) |
|
328 | 328 | if progress: |
|
329 | 329 | progress.complete() |
|
330 | 330 | progress = srcrepo.ui.makeprogress( |
|
331 | 331 | _(b'manifest revisions'), total=mrevcount |
|
332 | 332 | ) |
|
333 | 333 | for unencoded, (rl_type, oldrl) in sorted(manifests.items()): |
|
334 | 334 | newrl = _perform_clone( |
|
335 | 335 | ui, |
|
336 | 336 | dstrepo, |
|
337 | 337 | tr, |
|
338 | 338 | oldrl, |
|
339 | 339 | rl_type, |
|
340 | 340 | unencoded, |
|
341 | 341 | upgrade_op, |
|
342 | 342 | sidedata_helpers, |
|
343 | 343 | oncopiedrevision, |
|
344 | 344 | ) |
|
345 | 345 | info = newrl.storageinfo(storedsize=True) |
|
346 | 346 | mdstsize += info[b'storedsize'] or 0 |
|
347 | 347 | ui.status( |
|
348 | 348 | _( |
|
349 | 349 | b'finished migrating %d manifest revisions across %d ' |
|
350 | 350 | b'manifests; change in size: %s\n' |
|
351 | 351 | ) |
|
352 | 352 | % (mrevcount, mcount, util.bytecount(mdstsize - msrcsize)) |
|
353 | 353 | ) |
|
354 | 354 | |
|
355 | 355 | # Migrating changelog |
|
356 | 356 | ui.status( |
|
357 | 357 | _( |
|
358 | 358 | b'migrating changelog containing %d revisions ' |
|
359 | 359 | b'(%s in store; %s tracked data)\n' |
|
360 | 360 | ) |
|
361 | 361 | % ( |
|
362 | 362 | crevcount, |
|
363 | 363 | util.bytecount(csrcsize), |
|
364 | 364 | util.bytecount(crawsize), |
|
365 | 365 | ) |
|
366 | 366 | ) |
|
367 | 367 | if progress: |
|
368 | 368 | progress.complete() |
|
369 | 369 | progress = srcrepo.ui.makeprogress( |
|
370 | 370 | _(b'changelog revisions'), total=crevcount |
|
371 | 371 | ) |
|
372 | 372 | for unencoded, (rl_type, oldrl) in sorted(changelogs.items()): |
|
373 | 373 | newrl = _perform_clone( |
|
374 | 374 | ui, |
|
375 | 375 | dstrepo, |
|
376 | 376 | tr, |
|
377 | 377 | oldrl, |
|
378 | 378 | rl_type, |
|
379 | 379 | unencoded, |
|
380 | 380 | upgrade_op, |
|
381 | 381 | sidedata_helpers, |
|
382 | 382 | oncopiedrevision, |
|
383 | 383 | ) |
|
384 | 384 | info = newrl.storageinfo(storedsize=True) |
|
385 | 385 | cdstsize += info[b'storedsize'] or 0 |
|
386 | 386 | progress.complete() |
|
387 | 387 | ui.status( |
|
388 | 388 | _( |
|
389 | 389 | b'finished migrating %d changelog revisions; change in size: ' |
|
390 | 390 | b'%s\n' |
|
391 | 391 | ) |
|
392 | 392 | % (crevcount, util.bytecount(cdstsize - csrcsize)) |
|
393 | 393 | ) |
|
394 | 394 | |
|
395 | 395 | dstsize = fdstsize + mdstsize + cdstsize |
|
396 | 396 | ui.status( |
|
397 | 397 | _( |
|
398 | 398 | b'finished migrating %d total revisions; total change in store ' |
|
399 | 399 | b'size: %s\n' |
|
400 | 400 | ) |
|
401 | 401 | % (revcount, util.bytecount(dstsize - srcsize)) |
|
402 | 402 | ) |
|
403 | 403 | |
|
404 | 404 | |
|
405 | 405 | def _files_to_copy_post_revlog_clone(srcrepo): |
|
406 | 406 | """yields files which should be copied to destination after revlogs |
|
407 | 407 | are cloned""" |
|
408 | 408 | for path, kind, st in sorted(srcrepo.store.vfs.readdir(b'', stat=True)): |
|
409 | 409 | # don't copy revlogs as they are already cloned |
|
410 | 410 | if store.revlog_type(path) is not None: |
|
411 | 411 | continue |
|
412 | 412 | # Skip transaction related files. |
|
413 | 413 | if path.startswith(b'undo'): |
|
414 | 414 | continue |
|
415 | 415 | # Only copy regular files. |
|
416 | 416 | if kind != stat.S_IFREG: |
|
417 | 417 | continue |
|
418 | 418 | # Skip other skipped files. |
|
419 | 419 | if path in (b'lock', b'fncache'): |
|
420 | 420 | continue |
|
421 | 421 | # TODO: should we skip cache too? |
|
422 | 422 | |
|
423 | 423 | yield path |
|
424 | 424 | |
|
425 | 425 | |
|
426 | 426 | def _replacestores(currentrepo, upgradedrepo, backupvfs, upgrade_op): |
|
427 | 427 | """Replace the stores after current repository is upgraded |
|
428 | 428 | |
|
429 | 429 | Creates a backup of current repository store at backup path |
|
430 | 430 | Replaces upgraded store files in current repo from upgraded one |
|
431 | 431 | |
|
432 | 432 | Arguments: |
|
433 | 433 | currentrepo: repo object of current repository |
|
434 | 434 | upgradedrepo: repo object of the upgraded data |
|
435 | 435 | backupvfs: vfs object for the backup path |
|
436 | 436 | upgrade_op: upgrade operation object |
|
437 | 437 | to be used to decide what all is upgraded |
|
438 | 438 | """ |
|
439 | 439 | # TODO: don't blindly rename everything in store |
|
440 | 440 | # There can be upgrades where store is not touched at all |
|
441 | 441 | if upgrade_op.backup_store: |
|
442 | 442 | util.rename(currentrepo.spath, backupvfs.join(b'store')) |
|
443 | 443 | else: |
|
444 | 444 | currentrepo.vfs.rmtree(b'store', forcibly=True) |
|
445 | 445 | util.rename(upgradedrepo.spath, currentrepo.spath) |
|
446 | 446 | |
|
447 | 447 | |
|
448 | 448 | def finishdatamigration(ui, srcrepo, dstrepo, requirements): |
|
449 | 449 | """Hook point for extensions to perform additional actions during upgrade. |
|
450 | 450 | |
|
451 | 451 | This function is called after revlogs and store files have been copied but |
|
452 | 452 | before the new store is swapped into the original location. |
|
453 | 453 | """ |
|
454 | 454 | |
|
455 | 455 | |
|
456 | 456 | def upgrade(ui, srcrepo, dstrepo, upgrade_op): |
|
457 | 457 | """Do the low-level work of upgrading a repository. |
|
458 | 458 | |
|
459 | 459 | The upgrade is effectively performed as a copy between a source |
|
460 | 460 | repository and a temporary destination repository. |
|
461 | 461 | |
|
462 | 462 | The source repository is unmodified for as long as possible so the |
|
463 | 463 | upgrade can abort at any time without causing loss of service for |
|
464 | 464 | readers and without corrupting the source repository. |
|
465 | 465 | """ |
|
466 | 466 | assert srcrepo.currentwlock() |
|
467 | 467 | assert dstrepo.currentwlock() |
|
468 | 468 | backuppath = None |
|
469 | 469 | backupvfs = None |
|
470 | 470 | |
|
471 | 471 | ui.status( |
|
472 | 472 | _( |
|
473 | 473 | b'(it is safe to interrupt this process any time before ' |
|
474 | 474 | b'data migration completes)\n' |
|
475 | 475 | ) |
|
476 | 476 | ) |
|
477 | 477 | |
|
478 | 478 | if upgrade_actions.dirstatev2 in upgrade_op.upgrade_actions: |
|
479 | 479 | ui.status(_(b'upgrading to dirstate-v2 from v1\n')) |
|
480 | 480 | upgrade_dirstate(ui, srcrepo, upgrade_op, b'v1', b'v2') |
|
481 | 481 | upgrade_op.upgrade_actions.remove(upgrade_actions.dirstatev2) |
|
482 | 482 | |
|
483 | 483 | if upgrade_actions.dirstatev2 in upgrade_op.removed_actions: |
|
484 | 484 | ui.status(_(b'downgrading from dirstate-v2 to v1\n')) |
|
485 | 485 | upgrade_dirstate(ui, srcrepo, upgrade_op, b'v2', b'v1') |
|
486 | 486 | upgrade_op.removed_actions.remove(upgrade_actions.dirstatev2) |
|
487 | 487 | |
|
488 | 488 | if not (upgrade_op.upgrade_actions or upgrade_op.removed_actions): |
|
489 | 489 | return |
|
490 | 490 | |
|
491 | 491 | if upgrade_op.requirements_only: |
|
492 | 492 | ui.status(_(b'upgrading repository requirements\n')) |
|
493 | 493 | scmutil.writereporequirements(srcrepo, upgrade_op.new_requirements) |
|
494 | 494 | # if there is only one action and that is persistent nodemap upgrade |
|
495 | 495 | # directly write the nodemap file and update requirements instead of going |
|
496 | 496 | # through the whole cloning process |
|
497 | 497 | elif ( |
|
498 | 498 | len(upgrade_op.upgrade_actions) == 1 |
|
499 | 499 | and b'persistent-nodemap' in upgrade_op.upgrade_actions_names |
|
500 | 500 | and not upgrade_op.removed_actions |
|
501 | 501 | ): |
|
502 | 502 | ui.status( |
|
503 | 503 | _(b'upgrading repository to use persistent nodemap feature\n') |
|
504 | 504 | ) |
|
505 | 505 | with srcrepo.transaction(b'upgrade') as tr: |
|
506 | 506 | unfi = srcrepo.unfiltered() |
|
507 | 507 | cl = unfi.changelog |
|
508 | 508 | nodemap.persist_nodemap(tr, cl, force=True) |
|
509 | 509 | # we want to directly operate on the underlying revlog to force |
|
510 | 510 | # create a nodemap file. This is fine since this is upgrade code |
|
511 | 511 | # and it heavily relies on repository being revlog based |
|
512 | 512 | # hence accessing private attributes can be justified |
|
513 | 513 | nodemap.persist_nodemap( |
|
514 | 514 | tr, unfi.manifestlog._rootstore._revlog, force=True |
|
515 | 515 | ) |
|
516 | 516 | scmutil.writereporequirements(srcrepo, upgrade_op.new_requirements) |
|
517 | 517 | elif ( |
|
518 | 518 | len(upgrade_op.removed_actions) == 1 |
|
519 | 519 | and [ |
|
520 | 520 | x |
|
521 | 521 | for x in upgrade_op.removed_actions |
|
522 | 522 | if x.name == b'persistent-nodemap' |
|
523 | 523 | ] |
|
524 | 524 | and not upgrade_op.upgrade_actions |
|
525 | 525 | ): |
|
526 | 526 | ui.status( |
|
527 | 527 | _(b'downgrading repository to not use persistent nodemap feature\n') |
|
528 | 528 | ) |
|
529 | 529 | with srcrepo.transaction(b'upgrade') as tr: |
|
530 | 530 | unfi = srcrepo.unfiltered() |
|
531 | 531 | cl = unfi.changelog |
|
532 | 532 | nodemap.delete_nodemap(tr, srcrepo, cl) |
|
533 | 533 | # check comment 20 lines above for accessing private attributes |
|
534 | 534 | nodemap.delete_nodemap( |
|
535 | 535 | tr, srcrepo, unfi.manifestlog._rootstore._revlog |
|
536 | 536 | ) |
|
537 | 537 | scmutil.writereporequirements(srcrepo, upgrade_op.new_requirements) |
|
538 | 538 | else: |
|
539 | 539 | with dstrepo.transaction(b'upgrade') as tr: |
|
540 | 540 | _clonerevlogs( |
|
541 | 541 | ui, |
|
542 | 542 | srcrepo, |
|
543 | 543 | dstrepo, |
|
544 | 544 | tr, |
|
545 | 545 | upgrade_op, |
|
546 | 546 | ) |
|
547 | 547 | |
|
548 | 548 | # Now copy other files in the store directory. |
|
549 | 549 | for p in _files_to_copy_post_revlog_clone(srcrepo): |
|
550 | 550 | srcrepo.ui.status(_(b'copying %s\n') % p) |
|
551 | 551 | src = srcrepo.store.rawvfs.join(p) |
|
552 | 552 | dst = dstrepo.store.rawvfs.join(p) |
|
553 | 553 | util.copyfile(src, dst, copystat=True) |
|
554 | 554 | |
|
555 | 555 | finishdatamigration(ui, srcrepo, dstrepo, requirements) |
|
556 | 556 | |
|
557 | 557 | ui.status(_(b'data fully upgraded in a temporary repository\n')) |
|
558 | 558 | |
|
559 | 559 | if upgrade_op.backup_store: |
|
560 | 560 | backuppath = pycompat.mkdtemp( |
|
561 | 561 | prefix=b'upgradebackup.', dir=srcrepo.path |
|
562 | 562 | ) |
|
563 | 563 | backupvfs = vfsmod.vfs(backuppath) |
|
564 | 564 | |
|
565 | 565 | # Make a backup of requires file first, as it is the first to be modified. |
|
566 | 566 | util.copyfile( |
|
567 | 567 | srcrepo.vfs.join(b'requires'), backupvfs.join(b'requires') |
|
568 | 568 | ) |
|
569 | 569 | |
|
570 | 570 | # We install an arbitrary requirement that clients must not support |
|
571 | 571 | # as a mechanism to lock out new clients during the data swap. This is |
|
572 | 572 | # better than allowing a client to continue while the repository is in |
|
573 | 573 | # an inconsistent state. |
|
574 | 574 | ui.status( |
|
575 | 575 | _( |
|
576 | 576 | b'marking source repository as being upgraded; clients will be ' |
|
577 | 577 | b'unable to read from repository\n' |
|
578 | 578 | ) |
|
579 | 579 | ) |
|
580 | 580 | scmutil.writereporequirements( |
|
581 | 581 | srcrepo, srcrepo.requirements | {b'upgradeinprogress'} |
|
582 | 582 | ) |
|
583 | 583 | |
|
584 | 584 | ui.status(_(b'starting in-place swap of repository data\n')) |
|
585 | 585 | if upgrade_op.backup_store: |
|
586 | 586 | ui.status( |
|
587 | 587 | _(b'replaced files will be backed up at %s\n') % backuppath |
|
588 | 588 | ) |
|
589 | 589 | |
|
590 | 590 | # Now swap in the new store directory. Doing it as a rename should make |
|
591 | 591 | # the operation nearly instantaneous and atomic (at least in well-behaved |
|
592 | 592 | # environments). |
|
593 | 593 | ui.status(_(b'replacing store...\n')) |
|
594 | 594 | tstart = util.timer() |
|
595 | 595 | _replacestores(srcrepo, dstrepo, backupvfs, upgrade_op) |
|
596 | 596 | elapsed = util.timer() - tstart |
|
597 | 597 | ui.status( |
|
598 | 598 | _( |
|
599 | 599 | b'store replacement complete; repository was inconsistent for ' |
|
600 | 600 | b'%0.1fs\n' |
|
601 | 601 | ) |
|
602 | 602 | % elapsed |
|
603 | 603 | ) |
|
604 | 604 | |
|
605 | 605 | # We first write the requirements file. Any new requirements will lock |
|
606 | 606 | # out legacy clients. |
|
607 | 607 | ui.status( |
|
608 | 608 | _( |
|
609 | 609 | b'finalizing requirements file and making repository readable ' |
|
610 | 610 | b'again\n' |
|
611 | 611 | ) |
|
612 | 612 | ) |
|
613 | 613 | scmutil.writereporequirements(srcrepo, upgrade_op.new_requirements) |
|
614 | 614 | |
|
615 | 615 | if upgrade_op.backup_store: |
|
616 | 616 | # The lock file from the old store won't be removed because nothing has a |
|
617 | 617 | # reference to its new location. So clean it up manually. Alternatively, we |
|
618 | 618 | # could update srcrepo.svfs and other variables to point to the new |
|
619 | 619 | # location. This is simpler. |
|
620 | 620 | assert backupvfs is not None # help pytype |
|
621 | 621 | backupvfs.unlink(b'store/lock') |
|
622 | 622 | |
|
623 | 623 | return backuppath |
|
624 | 624 | |
|
625 | 625 | |
|
626 | 626 | def upgrade_dirstate(ui, srcrepo, upgrade_op, old, new): |
|
627 | 627 | if upgrade_op.backup_store: |
|
628 | 628 | backuppath = pycompat.mkdtemp( |
|
629 | 629 | prefix=b'upgradebackup.', dir=srcrepo.path |
|
630 | 630 | ) |
|
631 | 631 | ui.status(_(b'replaced files will be backed up at %s\n') % backuppath) |
|
632 | 632 | backupvfs = vfsmod.vfs(backuppath) |
|
633 | 633 | util.copyfile( |
|
634 | 634 | srcrepo.vfs.join(b'requires'), backupvfs.join(b'requires') |
|
635 | 635 | ) |
|
636 | 636 | util.copyfile( |
|
637 | 637 | srcrepo.vfs.join(b'dirstate'), backupvfs.join(b'dirstate') |
|
638 | 638 | ) |
|
639 | 639 | |
|
640 | 640 | assert srcrepo.dirstate._use_dirstate_v2 == (old == b'v2') |
|
641 | 641 | srcrepo.dirstate._map._use_dirstate_tree = True |
|
642 | 642 | srcrepo.dirstate._map.preload() |
|
643 | 643 | srcrepo.dirstate._use_dirstate_v2 = new == b'v2' |
|
644 | 644 | srcrepo.dirstate._map._use_dirstate_v2 = srcrepo.dirstate._use_dirstate_v2 |
|
645 | 645 | srcrepo.dirstate._dirty = True |
|
646 | 646 | srcrepo.vfs.unlink(b'dirstate') |
|
647 | 647 | srcrepo.dirstate.write(None) |
|
648 | 648 | |
|
649 | 649 | scmutil.writereporequirements(srcrepo, upgrade_op.new_requirements) |
@@ -1,586 +1,588 b'' | |||
|
1 | 1 | # verify.py - repository integrity checking for Mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2006, 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 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import os |
|
11 | 11 | |
|
12 | 12 | from .i18n import _ |
|
13 | 13 | from .node import short |
|
14 | 14 | from .utils import stringutil |
|
15 | 15 | |
|
16 | 16 | from . import ( |
|
17 | 17 | error, |
|
18 | 18 | pycompat, |
|
19 | 19 | revlog, |
|
20 | 20 | util, |
|
21 | 21 | ) |
|
22 | 22 | |
|
23 | 23 | VERIFY_DEFAULT = 0 |
|
24 | 24 | VERIFY_FULL = 1 |
|
25 | 25 | |
|
26 | 26 | |
|
27 | 27 | def verify(repo, level=None): |
|
28 | 28 | with repo.lock(): |
|
29 | 29 | v = verifier(repo, level) |
|
30 | 30 | return v.verify() |
|
31 | 31 | |
|
32 | 32 | |
|
33 | 33 | def _normpath(f): |
|
34 | 34 | # under hg < 2.4, convert didn't sanitize paths properly, so a |
|
35 | 35 | # converted repo may contain repeated slashes |
|
36 | 36 | while b'//' in f: |
|
37 | 37 | f = f.replace(b'//', b'/') |
|
38 | 38 | return f |
|
39 | 39 | |
|
40 | 40 | |
|
41 | 41 | HINT_FNCACHE = _( |
|
42 | 42 | b'hint: run "hg debugrebuildfncache" to recover from corrupt fncache\n' |
|
43 | 43 | ) |
|
44 | 44 | |
|
45 | 45 | WARN_PARENT_DIR_UNKNOWN_REV = _( |
|
46 | 46 | b"parent-directory manifest refers to unknown revision %s" |
|
47 | 47 | ) |
|
48 | 48 | |
|
49 | 49 | WARN_UNKNOWN_COPY_SOURCE = _( |
|
50 | 50 | b"warning: copy source of '%s' not in parents of %s" |
|
51 | 51 | ) |
|
52 | 52 | |
|
53 | 53 | WARN_NULLID_COPY_SOURCE = _( |
|
54 | 54 | b"warning: %s@%s: copy source revision is nullid %s:%s\n" |
|
55 | 55 | ) |
|
56 | 56 | |
|
57 | 57 | |
|
58 | 58 | class verifier(object): |
|
59 | 59 | def __init__(self, repo, level=None): |
|
60 | 60 | self.repo = repo.unfiltered() |
|
61 | 61 | self.ui = repo.ui |
|
62 | 62 | self.match = repo.narrowmatch() |
|
63 | 63 | if level is None: |
|
64 | 64 | level = VERIFY_DEFAULT |
|
65 | 65 | self._level = level |
|
66 | 66 | self.badrevs = set() |
|
67 | 67 | self.errors = 0 |
|
68 | 68 | self.warnings = 0 |
|
69 | 69 | self.havecl = len(repo.changelog) > 0 |
|
70 | 70 | self.havemf = len(repo.manifestlog.getstorage(b'')) > 0 |
|
71 | 71 | self.revlogv1 = repo.changelog._format_version != revlog.REVLOGV0 |
|
72 | 72 | self.lrugetctx = util.lrucachefunc(repo.unfiltered().__getitem__) |
|
73 | 73 | self.refersmf = False |
|
74 | 74 | self.fncachewarned = False |
|
75 | 75 | # developer config: verify.skipflags |
|
76 | 76 | self.skipflags = repo.ui.configint(b'verify', b'skipflags') |
|
77 | 77 | self.warnorphanstorefiles = True |
|
78 | 78 | |
|
79 | 79 | def _warn(self, msg): |
|
80 | 80 | """record a "warning" level issue""" |
|
81 | 81 | self.ui.warn(msg + b"\n") |
|
82 | 82 | self.warnings += 1 |
|
83 | 83 | |
|
84 | 84 | def _err(self, linkrev, msg, filename=None): |
|
85 | 85 | """record a "error" level issue""" |
|
86 | 86 | if linkrev is not None: |
|
87 | 87 | self.badrevs.add(linkrev) |
|
88 | 88 | linkrev = b"%d" % linkrev |
|
89 | 89 | else: |
|
90 | 90 | linkrev = b'?' |
|
91 | 91 | msg = b"%s: %s" % (linkrev, msg) |
|
92 | 92 | if filename: |
|
93 | 93 | msg = b"%s@%s" % (filename, msg) |
|
94 | 94 | self.ui.warn(b" " + msg + b"\n") |
|
95 | 95 | self.errors += 1 |
|
96 | 96 | |
|
97 | 97 | def _exc(self, linkrev, msg, inst, filename=None): |
|
98 | 98 | """record exception raised during the verify process""" |
|
99 | 99 | fmsg = stringutil.forcebytestr(inst) |
|
100 | 100 | if not fmsg: |
|
101 | 101 | fmsg = pycompat.byterepr(inst) |
|
102 | 102 | self._err(linkrev, b"%s: %s" % (msg, fmsg), filename) |
|
103 | 103 | |
|
104 | 104 | def _checkrevlog(self, obj, name, linkrev): |
|
105 | 105 | """verify high level property of a revlog |
|
106 | 106 | |
|
107 | 107 | - revlog is present, |
|
108 | 108 | - revlog is non-empty, |
|
109 | 109 | - sizes (index and data) are correct, |
|
110 | 110 | - revlog's format version is correct. |
|
111 | 111 | """ |
|
112 | 112 | if not len(obj) and (self.havecl or self.havemf): |
|
113 | 113 | self._err(linkrev, _(b"empty or missing %s") % name) |
|
114 | 114 | return |
|
115 | 115 | |
|
116 | 116 | d = obj.checksize() |
|
117 | 117 | if d[0]: |
|
118 | 118 | self._err(None, _(b"data length off by %d bytes") % d[0], name) |
|
119 | 119 | if d[1]: |
|
120 | 120 | self._err(None, _(b"index contains %d extra bytes") % d[1], name) |
|
121 | 121 | |
|
122 | 122 | if obj._format_version != revlog.REVLOGV0: |
|
123 | 123 | if not self.revlogv1: |
|
124 | 124 | self._warn(_(b"warning: `%s' uses revlog format 1") % name) |
|
125 | 125 | elif self.revlogv1: |
|
126 | 126 | self._warn(_(b"warning: `%s' uses revlog format 0") % name) |
|
127 | 127 | |
|
128 | 128 | def _checkentry(self, obj, i, node, seen, linkrevs, f): |
|
129 | 129 | """verify a single revlog entry |
|
130 | 130 | |
|
131 | 131 | arguments are: |
|
132 | 132 | - obj: the source revlog |
|
133 | 133 | - i: the revision number |
|
134 | 134 | - node: the revision node id |
|
135 | 135 | - seen: nodes previously seen for this revlog |
|
136 | 136 | - linkrevs: [changelog-revisions] introducing "node" |
|
137 | 137 | - f: string label ("changelog", "manifest", or filename) |
|
138 | 138 | |
|
139 | 139 | Performs the following checks: |
|
140 | 140 | - linkrev points to an existing changelog revision, |
|
141 | 141 | - linkrev points to a changelog revision that introduces this revision, |
|
142 | 142 | - linkrev points to the lowest of these changesets, |
|
143 | 143 | - both parents exist in the revlog, |
|
144 | 144 | - the revision is not duplicated. |
|
145 | 145 | |
|
146 | 146 | Return the linkrev of the revision (or None for changelog's revisions). |
|
147 | 147 | """ |
|
148 | 148 | lr = obj.linkrev(obj.rev(node)) |
|
149 | 149 | if lr < 0 or (self.havecl and lr not in linkrevs): |
|
150 | 150 | if lr < 0 or lr >= len(self.repo.changelog): |
|
151 | 151 | msg = _(b"rev %d points to nonexistent changeset %d") |
|
152 | 152 | else: |
|
153 | 153 | msg = _(b"rev %d points to unexpected changeset %d") |
|
154 | 154 | self._err(None, msg % (i, lr), f) |
|
155 | 155 | if linkrevs: |
|
156 | 156 | if f and len(linkrevs) > 1: |
|
157 | 157 | try: |
|
158 | 158 | # attempt to filter down to real linkrevs |
|
159 | 159 | linkrevs = [] |
|
160 | 160 | for lr in linkrevs: |
|
161 | 161 | if self.lrugetctx(lr)[f].filenode() == node: |
|
162 | 162 | linkrevs.append(lr) |
|
163 | 163 | except Exception: |
|
164 | 164 | pass |
|
165 | 165 | msg = _(b" (expected %s)") |
|
166 | 166 | msg %= b" ".join(map(pycompat.bytestr, linkrevs)) |
|
167 | 167 | self._warn(msg) |
|
168 | 168 | lr = None # can't be trusted |
|
169 | 169 | |
|
170 | 170 | try: |
|
171 | 171 | p1, p2 = obj.parents(node) |
|
172 | 172 | if p1 not in seen and p1 != self.repo.nullid: |
|
173 | 173 | msg = _(b"unknown parent 1 %s of %s") % (short(p1), short(node)) |
|
174 | 174 | self._err(lr, msg, f) |
|
175 | 175 | if p2 not in seen and p2 != self.repo.nullid: |
|
176 | 176 | msg = _(b"unknown parent 2 %s of %s") % (short(p2), short(node)) |
|
177 | 177 | self._err(lr, msg, f) |
|
178 | 178 | except Exception as inst: |
|
179 | 179 | self._exc(lr, _(b"checking parents of %s") % short(node), inst, f) |
|
180 | 180 | |
|
181 | 181 | if node in seen: |
|
182 | 182 | self._err(lr, _(b"duplicate revision %d (%d)") % (i, seen[node]), f) |
|
183 | 183 | seen[node] = i |
|
184 | 184 | return lr |
|
185 | 185 | |
|
186 | 186 | def verify(self): |
|
187 | 187 | """verify the content of the Mercurial repository |
|
188 | 188 | |
|
189 | 189 | This method run all verifications, displaying issues as they are found. |
|
190 | 190 | |
|
191 | 191 | return 1 if any error have been encountered, 0 otherwise.""" |
|
192 | 192 | # initial validation and generic report |
|
193 | 193 | repo = self.repo |
|
194 | 194 | ui = repo.ui |
|
195 | 195 | if not repo.url().startswith(b'file:'): |
|
196 | 196 | raise error.Abort(_(b"cannot verify bundle or remote repos")) |
|
197 | 197 | |
|
198 | 198 | if os.path.exists(repo.sjoin(b"journal")): |
|
199 | 199 | ui.warn(_(b"abandoned transaction found - run hg recover\n")) |
|
200 | 200 | |
|
201 | 201 | if ui.verbose or not self.revlogv1: |
|
202 | 202 | ui.status( |
|
203 | 203 | _(b"repository uses revlog format %d\n") |
|
204 | 204 | % (self.revlogv1 and 1 or 0) |
|
205 | 205 | ) |
|
206 | 206 | |
|
207 | 207 | # data verification |
|
208 | 208 | mflinkrevs, filelinkrevs = self._verifychangelog() |
|
209 | 209 | filenodes = self._verifymanifest(mflinkrevs) |
|
210 | 210 | del mflinkrevs |
|
211 | 211 | self._crosscheckfiles(filelinkrevs, filenodes) |
|
212 | 212 | totalfiles, filerevisions = self._verifyfiles(filenodes, filelinkrevs) |
|
213 | 213 | |
|
214 | 214 | # final report |
|
215 | 215 | ui.status( |
|
216 | 216 | _(b"checked %d changesets with %d changes to %d files\n") |
|
217 | 217 | % (len(repo.changelog), filerevisions, totalfiles) |
|
218 | 218 | ) |
|
219 | 219 | if self.warnings: |
|
220 | 220 | ui.warn(_(b"%d warnings encountered!\n") % self.warnings) |
|
221 | 221 | if self.fncachewarned: |
|
222 | 222 | ui.warn(HINT_FNCACHE) |
|
223 | 223 | if self.errors: |
|
224 | 224 | ui.warn(_(b"%d integrity errors encountered!\n") % self.errors) |
|
225 | 225 | if self.badrevs: |
|
226 | 226 | msg = _(b"(first damaged changeset appears to be %d)\n") |
|
227 | 227 | msg %= min(self.badrevs) |
|
228 | 228 | ui.warn(msg) |
|
229 | 229 | return 1 |
|
230 | 230 | return 0 |
|
231 | 231 | |
|
232 | 232 | def _verifychangelog(self): |
|
233 | 233 | """verify the changelog of a repository |
|
234 | 234 | |
|
235 | 235 | The following checks are performed: |
|
236 | 236 | - all of `_checkrevlog` checks, |
|
237 | 237 | - all of `_checkentry` checks (for each revisions), |
|
238 | 238 | - each revision can be read. |
|
239 | 239 | |
|
240 | 240 | The function returns some of the data observed in the changesets as a |
|
241 | 241 | (mflinkrevs, filelinkrevs) tuples: |
|
242 | 242 | - mflinkrevs: is a { manifest-node -> [changelog-rev] } mapping |
|
243 | 243 | - filelinkrevs: is a { file-path -> [changelog-rev] } mapping |
|
244 | 244 | |
|
245 | 245 | If a matcher was specified, filelinkrevs will only contains matched |
|
246 | 246 | files. |
|
247 | 247 | """ |
|
248 | 248 | ui = self.ui |
|
249 | 249 | repo = self.repo |
|
250 | 250 | match = self.match |
|
251 | 251 | cl = repo.changelog |
|
252 | 252 | |
|
253 | 253 | ui.status(_(b"checking changesets\n")) |
|
254 | 254 | mflinkrevs = {} |
|
255 | 255 | filelinkrevs = {} |
|
256 | 256 | seen = {} |
|
257 | 257 | self._checkrevlog(cl, b"changelog", 0) |
|
258 | 258 | progress = ui.makeprogress( |
|
259 | 259 | _(b'checking'), unit=_(b'changesets'), total=len(repo) |
|
260 | 260 | ) |
|
261 | 261 | for i in repo: |
|
262 | 262 | progress.update(i) |
|
263 | 263 | n = cl.node(i) |
|
264 | 264 | self._checkentry(cl, i, n, seen, [i], b"changelog") |
|
265 | 265 | |
|
266 | 266 | try: |
|
267 | 267 | changes = cl.read(n) |
|
268 | 268 | if changes[0] != self.repo.nullid: |
|
269 | 269 | mflinkrevs.setdefault(changes[0], []).append(i) |
|
270 | 270 | self.refersmf = True |
|
271 | 271 | for f in changes[3]: |
|
272 | 272 | if match(f): |
|
273 | 273 | filelinkrevs.setdefault(_normpath(f), []).append(i) |
|
274 | 274 | except Exception as inst: |
|
275 | 275 | self.refersmf = True |
|
276 | 276 | self._exc(i, _(b"unpacking changeset %s") % short(n), inst) |
|
277 | 277 | progress.complete() |
|
278 | 278 | return mflinkrevs, filelinkrevs |
|
279 | 279 | |
|
280 | 280 | def _verifymanifest( |
|
281 | 281 | self, mflinkrevs, dir=b"", storefiles=None, subdirprogress=None |
|
282 | 282 | ): |
|
283 | 283 | """verify the manifestlog content |
|
284 | 284 | |
|
285 | 285 | Inputs: |
|
286 | 286 | - mflinkrevs: a {manifest-node -> [changelog-revisions]} mapping |
|
287 | 287 | - dir: a subdirectory to check (for tree manifest repo) |
|
288 | 288 | - storefiles: set of currently "orphan" files. |
|
289 | 289 | - subdirprogress: a progress object |
|
290 | 290 | |
|
291 | 291 | This function checks: |
|
292 | 292 | * all of `_checkrevlog` checks (for all manifest related revlogs) |
|
293 | 293 | * all of `_checkentry` checks (for all manifest related revisions) |
|
294 | 294 | * nodes for subdirectory exists in the sub-directory manifest |
|
295 | 295 | * each manifest entries have a file path |
|
296 | 296 | * each manifest node refered in mflinkrevs exist in the manifest log |
|
297 | 297 | |
|
298 | 298 | If tree manifest is in use and a matchers is specified, only the |
|
299 | 299 | sub-directories matching it will be verified. |
|
300 | 300 | |
|
301 | 301 | return a two level mapping: |
|
302 | 302 | {"path" -> { filenode -> changelog-revision}} |
|
303 | 303 | |
|
304 | 304 | This mapping primarily contains entries for every files in the |
|
305 | 305 | repository. In addition, when tree-manifest is used, it also contains |
|
306 | 306 | sub-directory entries. |
|
307 | 307 | |
|
308 | 308 | If a matcher is provided, only matching paths will be included. |
|
309 | 309 | """ |
|
310 | 310 | repo = self.repo |
|
311 | 311 | ui = self.ui |
|
312 | 312 | match = self.match |
|
313 | 313 | mfl = self.repo.manifestlog |
|
314 | 314 | mf = mfl.getstorage(dir) |
|
315 | 315 | |
|
316 | 316 | if not dir: |
|
317 | 317 | self.ui.status(_(b"checking manifests\n")) |
|
318 | 318 | |
|
319 | 319 | filenodes = {} |
|
320 | 320 | subdirnodes = {} |
|
321 | 321 | seen = {} |
|
322 | 322 | label = b"manifest" |
|
323 | 323 | if dir: |
|
324 | 324 | label = dir |
|
325 | 325 | revlogfiles = mf.files() |
|
326 | 326 | storefiles.difference_update(revlogfiles) |
|
327 | 327 | if subdirprogress: # should be true since we're in a subdirectory |
|
328 | 328 | subdirprogress.increment() |
|
329 | 329 | if self.refersmf: |
|
330 | 330 | # Do not check manifest if there are only changelog entries with |
|
331 | 331 | # null manifests. |
|
332 | 332 | self._checkrevlog(mf._revlog, label, 0) |
|
333 | 333 | progress = ui.makeprogress( |
|
334 | 334 | _(b'checking'), unit=_(b'manifests'), total=len(mf) |
|
335 | 335 | ) |
|
336 | 336 | for i in mf: |
|
337 | 337 | if not dir: |
|
338 | 338 | progress.update(i) |
|
339 | 339 | n = mf.node(i) |
|
340 | 340 | lr = self._checkentry(mf, i, n, seen, mflinkrevs.get(n, []), label) |
|
341 | 341 | if n in mflinkrevs: |
|
342 | 342 | del mflinkrevs[n] |
|
343 | 343 | elif dir: |
|
344 | 344 | msg = _(b"%s not in parent-directory manifest") % short(n) |
|
345 | 345 | self._err(lr, msg, label) |
|
346 | 346 | else: |
|
347 | 347 | self._err(lr, _(b"%s not in changesets") % short(n), label) |
|
348 | 348 | |
|
349 | 349 | try: |
|
350 | 350 | mfdelta = mfl.get(dir, n).readdelta(shallow=True) |
|
351 | 351 | for f, fn, fl in mfdelta.iterentries(): |
|
352 | 352 | if not f: |
|
353 | 353 | self._err(lr, _(b"entry without name in manifest")) |
|
354 | 354 | elif f == b"/dev/null": # ignore this in very old repos |
|
355 | 355 | continue |
|
356 | 356 | fullpath = dir + _normpath(f) |
|
357 | 357 | if fl == b't': |
|
358 | 358 | if not match.visitdir(fullpath): |
|
359 | 359 | continue |
|
360 | 360 | sdn = subdirnodes.setdefault(fullpath + b'/', {}) |
|
361 | 361 | sdn.setdefault(fn, []).append(lr) |
|
362 | 362 | else: |
|
363 | 363 | if not match(fullpath): |
|
364 | 364 | continue |
|
365 | 365 | filenodes.setdefault(fullpath, {}).setdefault(fn, lr) |
|
366 | 366 | except Exception as inst: |
|
367 | 367 | self._exc(lr, _(b"reading delta %s") % short(n), inst, label) |
|
368 | 368 | if self._level >= VERIFY_FULL: |
|
369 | 369 | try: |
|
370 | 370 | # Various issues can affect manifest. So we read each full |
|
371 | 371 | # text from storage. This triggers the checks from the core |
|
372 | 372 | # code (eg: hash verification, filename are ordered, etc.) |
|
373 | 373 | mfdelta = mfl.get(dir, n).read() |
|
374 | 374 | except Exception as inst: |
|
375 | 375 | msg = _(b"reading full manifest %s") % short(n) |
|
376 | 376 | self._exc(lr, msg, inst, label) |
|
377 | 377 | |
|
378 | 378 | if not dir: |
|
379 | 379 | progress.complete() |
|
380 | 380 | |
|
381 | 381 | if self.havemf: |
|
382 | 382 | # since we delete entry in `mflinkrevs` during iteration, any |
|
383 | 383 | # remaining entries are "missing". We need to issue errors for them. |
|
384 | 384 | changesetpairs = [(c, m) for m in mflinkrevs for c in mflinkrevs[m]] |
|
385 | 385 | for c, m in sorted(changesetpairs): |
|
386 | 386 | if dir: |
|
387 | 387 | self._err(c, WARN_PARENT_DIR_UNKNOWN_REV % short(m), label) |
|
388 | 388 | else: |
|
389 | 389 | msg = _(b"changeset refers to unknown revision %s") |
|
390 | 390 | msg %= short(m) |
|
391 | 391 | self._err(c, msg, label) |
|
392 | 392 | |
|
393 | 393 | if not dir and subdirnodes: |
|
394 | 394 | self.ui.status(_(b"checking directory manifests\n")) |
|
395 | 395 | storefiles = set() |
|
396 | 396 | subdirs = set() |
|
397 | 397 | revlogv1 = self.revlogv1 |
|
398 | for t, f, f2, size in repo.store.datafiles(): | |
|
399 | if not f: | |
|
400 | self._err(None, _(b"cannot decode filename '%s'") % f2) | |
|
401 | elif (size > 0 or not revlogv1) and f.startswith(b'meta/'): | |
|
398 | undecodable = [] | |
|
399 | for t, f, size in repo.store.datafiles(undecodable=undecodable): | |
|
400 | if (size > 0 or not revlogv1) and f.startswith(b'meta/'): | |
|
402 | 401 | storefiles.add(_normpath(f)) |
|
403 | 402 | subdirs.add(os.path.dirname(f)) |
|
403 | for f in undecodable: | |
|
404 | self._err(None, _(b"cannot decode filename '%s'") % f) | |
|
404 | 405 | subdirprogress = ui.makeprogress( |
|
405 | 406 | _(b'checking'), unit=_(b'manifests'), total=len(subdirs) |
|
406 | 407 | ) |
|
407 | 408 | |
|
408 | 409 | for subdir, linkrevs in pycompat.iteritems(subdirnodes): |
|
409 | 410 | subdirfilenodes = self._verifymanifest( |
|
410 | 411 | linkrevs, subdir, storefiles, subdirprogress |
|
411 | 412 | ) |
|
412 | 413 | for f, onefilenodes in pycompat.iteritems(subdirfilenodes): |
|
413 | 414 | filenodes.setdefault(f, {}).update(onefilenodes) |
|
414 | 415 | |
|
415 | 416 | if not dir and subdirnodes: |
|
416 | 417 | assert subdirprogress is not None # help pytype |
|
417 | 418 | subdirprogress.complete() |
|
418 | 419 | if self.warnorphanstorefiles: |
|
419 | 420 | for f in sorted(storefiles): |
|
420 | 421 | self._warn(_(b"warning: orphan data file '%s'") % f) |
|
421 | 422 | |
|
422 | 423 | return filenodes |
|
423 | 424 | |
|
424 | 425 | def _crosscheckfiles(self, filelinkrevs, filenodes): |
|
425 | 426 | repo = self.repo |
|
426 | 427 | ui = self.ui |
|
427 | 428 | ui.status(_(b"crosschecking files in changesets and manifests\n")) |
|
428 | 429 | |
|
429 | 430 | total = len(filelinkrevs) + len(filenodes) |
|
430 | 431 | progress = ui.makeprogress( |
|
431 | 432 | _(b'crosschecking'), unit=_(b'files'), total=total |
|
432 | 433 | ) |
|
433 | 434 | if self.havemf: |
|
434 | 435 | for f in sorted(filelinkrevs): |
|
435 | 436 | progress.increment() |
|
436 | 437 | if f not in filenodes: |
|
437 | 438 | lr = filelinkrevs[f][0] |
|
438 | 439 | self._err(lr, _(b"in changeset but not in manifest"), f) |
|
439 | 440 | |
|
440 | 441 | if self.havecl: |
|
441 | 442 | for f in sorted(filenodes): |
|
442 | 443 | progress.increment() |
|
443 | 444 | if f not in filelinkrevs: |
|
444 | 445 | try: |
|
445 | 446 | fl = repo.file(f) |
|
446 | 447 | lr = min([fl.linkrev(fl.rev(n)) for n in filenodes[f]]) |
|
447 | 448 | except Exception: |
|
448 | 449 | lr = None |
|
449 | 450 | self._err(lr, _(b"in manifest but not in changeset"), f) |
|
450 | 451 | |
|
451 | 452 | progress.complete() |
|
452 | 453 | |
|
453 | 454 | def _verifyfiles(self, filenodes, filelinkrevs): |
|
454 | 455 | repo = self.repo |
|
455 | 456 | ui = self.ui |
|
456 | 457 | lrugetctx = self.lrugetctx |
|
457 | 458 | revlogv1 = self.revlogv1 |
|
458 | 459 | havemf = self.havemf |
|
459 | 460 | ui.status(_(b"checking files\n")) |
|
460 | 461 | |
|
461 | 462 | storefiles = set() |
|
462 | for rl_type, f, f2, size in repo.store.datafiles(): | |
|
463 | if not f: | |
|
464 | self._err(None, _(b"cannot decode filename '%s'") % f2) | |
|
465 | elif (size > 0 or not revlogv1) and f.startswith(b'data/'): | |
|
463 | undecodable = [] | |
|
464 | for t, f, size in repo.store.datafiles(undecodable=undecodable): | |
|
465 | if (size > 0 or not revlogv1) and f.startswith(b'data/'): | |
|
466 | 466 | storefiles.add(_normpath(f)) |
|
467 | for f in undecodable: | |
|
468 | self._err(None, _(b"cannot decode filename '%s'") % f) | |
|
467 | 469 | |
|
468 | 470 | state = { |
|
469 | 471 | # TODO this assumes revlog storage for changelog. |
|
470 | 472 | b'expectedversion': self.repo.changelog._format_version, |
|
471 | 473 | b'skipflags': self.skipflags, |
|
472 | 474 | # experimental config: censor.policy |
|
473 | 475 | b'erroroncensored': ui.config(b'censor', b'policy') == b'abort', |
|
474 | 476 | } |
|
475 | 477 | |
|
476 | 478 | files = sorted(set(filenodes) | set(filelinkrevs)) |
|
477 | 479 | revisions = 0 |
|
478 | 480 | progress = ui.makeprogress( |
|
479 | 481 | _(b'checking'), unit=_(b'files'), total=len(files) |
|
480 | 482 | ) |
|
481 | 483 | for i, f in enumerate(files): |
|
482 | 484 | progress.update(i, item=f) |
|
483 | 485 | try: |
|
484 | 486 | linkrevs = filelinkrevs[f] |
|
485 | 487 | except KeyError: |
|
486 | 488 | # in manifest but not in changelog |
|
487 | 489 | linkrevs = [] |
|
488 | 490 | |
|
489 | 491 | if linkrevs: |
|
490 | 492 | lr = linkrevs[0] |
|
491 | 493 | else: |
|
492 | 494 | lr = None |
|
493 | 495 | |
|
494 | 496 | try: |
|
495 | 497 | fl = repo.file(f) |
|
496 | 498 | except error.StorageError as e: |
|
497 | 499 | self._err(lr, _(b"broken revlog! (%s)") % e, f) |
|
498 | 500 | continue |
|
499 | 501 | |
|
500 | 502 | for ff in fl.files(): |
|
501 | 503 | try: |
|
502 | 504 | storefiles.remove(ff) |
|
503 | 505 | except KeyError: |
|
504 | 506 | if self.warnorphanstorefiles: |
|
505 | 507 | msg = _(b" warning: revlog '%s' not in fncache!") |
|
506 | 508 | self._warn(msg % ff) |
|
507 | 509 | self.fncachewarned = True |
|
508 | 510 | |
|
509 | 511 | if not len(fl) and (self.havecl or self.havemf): |
|
510 | 512 | self._err(lr, _(b"empty or missing %s") % f) |
|
511 | 513 | else: |
|
512 | 514 | # Guard against implementations not setting this. |
|
513 | 515 | state[b'skipread'] = set() |
|
514 | 516 | state[b'safe_renamed'] = set() |
|
515 | 517 | |
|
516 | 518 | for problem in fl.verifyintegrity(state): |
|
517 | 519 | if problem.node is not None: |
|
518 | 520 | linkrev = fl.linkrev(fl.rev(problem.node)) |
|
519 | 521 | else: |
|
520 | 522 | linkrev = None |
|
521 | 523 | |
|
522 | 524 | if problem.warning: |
|
523 | 525 | self._warn(problem.warning) |
|
524 | 526 | elif problem.error: |
|
525 | 527 | linkrev_msg = linkrev if linkrev is not None else lr |
|
526 | 528 | self._err(linkrev_msg, problem.error, f) |
|
527 | 529 | else: |
|
528 | 530 | raise error.ProgrammingError( |
|
529 | 531 | b'problem instance does not set warning or error ' |
|
530 | 532 | b'attribute: %s' % problem.msg |
|
531 | 533 | ) |
|
532 | 534 | |
|
533 | 535 | seen = {} |
|
534 | 536 | for i in fl: |
|
535 | 537 | revisions += 1 |
|
536 | 538 | n = fl.node(i) |
|
537 | 539 | lr = self._checkentry(fl, i, n, seen, linkrevs, f) |
|
538 | 540 | if f in filenodes: |
|
539 | 541 | if havemf and n not in filenodes[f]: |
|
540 | 542 | self._err(lr, _(b"%s not in manifests") % (short(n)), f) |
|
541 | 543 | else: |
|
542 | 544 | del filenodes[f][n] |
|
543 | 545 | |
|
544 | 546 | if n in state[b'skipread'] and n not in state[b'safe_renamed']: |
|
545 | 547 | continue |
|
546 | 548 | |
|
547 | 549 | # check renames |
|
548 | 550 | try: |
|
549 | 551 | # This requires resolving fulltext (at least on revlogs, |
|
550 | 552 | # though not with LFS revisions). We may want |
|
551 | 553 | # ``verifyintegrity()`` to pass a set of nodes with |
|
552 | 554 | # rename metadata as an optimization. |
|
553 | 555 | rp = fl.renamed(n) |
|
554 | 556 | if rp: |
|
555 | 557 | if lr is not None and ui.verbose: |
|
556 | 558 | ctx = lrugetctx(lr) |
|
557 | 559 | if not any(rp[0] in pctx for pctx in ctx.parents()): |
|
558 | 560 | self._warn(WARN_UNKNOWN_COPY_SOURCE % (f, ctx)) |
|
559 | 561 | fl2 = repo.file(rp[0]) |
|
560 | 562 | if not len(fl2): |
|
561 | 563 | m = _(b"empty or missing copy source revlog %s:%s") |
|
562 | 564 | self._err(lr, m % (rp[0], short(rp[1])), f) |
|
563 | 565 | elif rp[1] == self.repo.nullid: |
|
564 | 566 | msg = WARN_NULLID_COPY_SOURCE |
|
565 | 567 | msg %= (f, lr, rp[0], short(rp[1])) |
|
566 | 568 | ui.note(msg) |
|
567 | 569 | else: |
|
568 | 570 | fl2.rev(rp[1]) |
|
569 | 571 | except Exception as inst: |
|
570 | 572 | self._exc( |
|
571 | 573 | lr, _(b"checking rename of %s") % short(n), inst, f |
|
572 | 574 | ) |
|
573 | 575 | |
|
574 | 576 | # cross-check |
|
575 | 577 | if f in filenodes: |
|
576 | 578 | fns = [(v, k) for k, v in pycompat.iteritems(filenodes[f])] |
|
577 | 579 | for lr, node in sorted(fns): |
|
578 | 580 | msg = _(b"manifest refers to unknown revision %s") |
|
579 | 581 | self._err(lr, msg % short(node), f) |
|
580 | 582 | progress.complete() |
|
581 | 583 | |
|
582 | 584 | if self.warnorphanstorefiles: |
|
583 | 585 | for f in sorted(storefiles): |
|
584 | 586 | self._warn(_(b"warning: orphan data file '%s'") % f) |
|
585 | 587 | |
|
586 | 588 | return len(files), revisions |
@@ -1,1613 +1,1613 b'' | |||
|
1 | 1 | # Copyright 21 May 2005 - (c) 2005 Jake Edge <jake@edge2.net> |
|
2 | 2 | # Copyright 2005-2007 Olivia Mackall <olivia@selenic.com> |
|
3 | 3 | # |
|
4 | 4 | # This software may be used and distributed according to the terms of the |
|
5 | 5 | # GNU General Public License version 2 or any later version. |
|
6 | 6 | |
|
7 | 7 | from __future__ import absolute_import |
|
8 | 8 | |
|
9 | 9 | import collections |
|
10 | 10 | import contextlib |
|
11 | 11 | |
|
12 | 12 | from .i18n import _ |
|
13 | 13 | from .node import hex |
|
14 | 14 | from . import ( |
|
15 | 15 | discovery, |
|
16 | 16 | encoding, |
|
17 | 17 | error, |
|
18 | 18 | match as matchmod, |
|
19 | 19 | narrowspec, |
|
20 | 20 | pycompat, |
|
21 | 21 | streamclone, |
|
22 | 22 | templatefilters, |
|
23 | 23 | util, |
|
24 | 24 | wireprotoframing, |
|
25 | 25 | wireprototypes, |
|
26 | 26 | ) |
|
27 | 27 | from .interfaces import util as interfaceutil |
|
28 | 28 | from .utils import ( |
|
29 | 29 | cborutil, |
|
30 | 30 | hashutil, |
|
31 | 31 | stringutil, |
|
32 | 32 | ) |
|
33 | 33 | |
|
34 | 34 | FRAMINGTYPE = b'application/mercurial-exp-framing-0006' |
|
35 | 35 | |
|
36 | 36 | HTTP_WIREPROTO_V2 = wireprototypes.HTTP_WIREPROTO_V2 |
|
37 | 37 | |
|
38 | 38 | COMMANDS = wireprototypes.commanddict() |
|
39 | 39 | |
|
40 | 40 | # Value inserted into cache key computation function. Change the value to |
|
41 | 41 | # force new cache keys for every command request. This should be done when |
|
42 | 42 | # there is a change to how caching works, etc. |
|
43 | 43 | GLOBAL_CACHE_VERSION = 1 |
|
44 | 44 | |
|
45 | 45 | |
|
46 | 46 | def handlehttpv2request(rctx, req, res, checkperm, urlparts): |
|
47 | 47 | from .hgweb import common as hgwebcommon |
|
48 | 48 | |
|
49 | 49 | # URL space looks like: <permissions>/<command>, where <permission> can |
|
50 | 50 | # be ``ro`` or ``rw`` to signal read-only or read-write, respectively. |
|
51 | 51 | |
|
52 | 52 | # Root URL does nothing meaningful... yet. |
|
53 | 53 | if not urlparts: |
|
54 | 54 | res.status = b'200 OK' |
|
55 | 55 | res.headers[b'Content-Type'] = b'text/plain' |
|
56 | 56 | res.setbodybytes(_(b'HTTP version 2 API handler')) |
|
57 | 57 | return |
|
58 | 58 | |
|
59 | 59 | if len(urlparts) == 1: |
|
60 | 60 | res.status = b'404 Not Found' |
|
61 | 61 | res.headers[b'Content-Type'] = b'text/plain' |
|
62 | 62 | res.setbodybytes( |
|
63 | 63 | _(b'do not know how to process %s\n') % req.dispatchpath |
|
64 | 64 | ) |
|
65 | 65 | return |
|
66 | 66 | |
|
67 | 67 | permission, command = urlparts[0:2] |
|
68 | 68 | |
|
69 | 69 | if permission not in (b'ro', b'rw'): |
|
70 | 70 | res.status = b'404 Not Found' |
|
71 | 71 | res.headers[b'Content-Type'] = b'text/plain' |
|
72 | 72 | res.setbodybytes(_(b'unknown permission: %s') % permission) |
|
73 | 73 | return |
|
74 | 74 | |
|
75 | 75 | if req.method != b'POST': |
|
76 | 76 | res.status = b'405 Method Not Allowed' |
|
77 | 77 | res.headers[b'Allow'] = b'POST' |
|
78 | 78 | res.setbodybytes(_(b'commands require POST requests')) |
|
79 | 79 | return |
|
80 | 80 | |
|
81 | 81 | # At some point we'll want to use our own API instead of recycling the |
|
82 | 82 | # behavior of version 1 of the wire protocol... |
|
83 | 83 | # TODO return reasonable responses - not responses that overload the |
|
84 | 84 | # HTTP status line message for error reporting. |
|
85 | 85 | try: |
|
86 | 86 | checkperm(rctx, req, b'pull' if permission == b'ro' else b'push') |
|
87 | 87 | except hgwebcommon.ErrorResponse as e: |
|
88 | 88 | res.status = hgwebcommon.statusmessage( |
|
89 | 89 | e.code, stringutil.forcebytestr(e) |
|
90 | 90 | ) |
|
91 | 91 | for k, v in e.headers: |
|
92 | 92 | res.headers[k] = v |
|
93 | 93 | res.setbodybytes(b'permission denied') |
|
94 | 94 | return |
|
95 | 95 | |
|
96 | 96 | # We have a special endpoint to reflect the request back at the client. |
|
97 | 97 | if command == b'debugreflect': |
|
98 | 98 | _processhttpv2reflectrequest(rctx.repo.ui, rctx.repo, req, res) |
|
99 | 99 | return |
|
100 | 100 | |
|
101 | 101 | # Extra commands that we handle that aren't really wire protocol |
|
102 | 102 | # commands. Think extra hard before making this hackery available to |
|
103 | 103 | # extension. |
|
104 | 104 | extracommands = {b'multirequest'} |
|
105 | 105 | |
|
106 | 106 | if command not in COMMANDS and command not in extracommands: |
|
107 | 107 | res.status = b'404 Not Found' |
|
108 | 108 | res.headers[b'Content-Type'] = b'text/plain' |
|
109 | 109 | res.setbodybytes(_(b'unknown wire protocol command: %s\n') % command) |
|
110 | 110 | return |
|
111 | 111 | |
|
112 | 112 | repo = rctx.repo |
|
113 | 113 | ui = repo.ui |
|
114 | 114 | |
|
115 | 115 | proto = httpv2protocolhandler(req, ui) |
|
116 | 116 | |
|
117 | 117 | if ( |
|
118 | 118 | not COMMANDS.commandavailable(command, proto) |
|
119 | 119 | and command not in extracommands |
|
120 | 120 | ): |
|
121 | 121 | res.status = b'404 Not Found' |
|
122 | 122 | res.headers[b'Content-Type'] = b'text/plain' |
|
123 | 123 | res.setbodybytes(_(b'invalid wire protocol command: %s') % command) |
|
124 | 124 | return |
|
125 | 125 | |
|
126 | 126 | # TODO consider cases where proxies may add additional Accept headers. |
|
127 | 127 | if req.headers.get(b'Accept') != FRAMINGTYPE: |
|
128 | 128 | res.status = b'406 Not Acceptable' |
|
129 | 129 | res.headers[b'Content-Type'] = b'text/plain' |
|
130 | 130 | res.setbodybytes( |
|
131 | 131 | _(b'client MUST specify Accept header with value: %s\n') |
|
132 | 132 | % FRAMINGTYPE |
|
133 | 133 | ) |
|
134 | 134 | return |
|
135 | 135 | |
|
136 | 136 | if req.headers.get(b'Content-Type') != FRAMINGTYPE: |
|
137 | 137 | res.status = b'415 Unsupported Media Type' |
|
138 | 138 | # TODO we should send a response with appropriate media type, |
|
139 | 139 | # since client does Accept it. |
|
140 | 140 | res.headers[b'Content-Type'] = b'text/plain' |
|
141 | 141 | res.setbodybytes( |
|
142 | 142 | _(b'client MUST send Content-Type header with value: %s\n') |
|
143 | 143 | % FRAMINGTYPE |
|
144 | 144 | ) |
|
145 | 145 | return |
|
146 | 146 | |
|
147 | 147 | _processhttpv2request(ui, repo, req, res, permission, command, proto) |
|
148 | 148 | |
|
149 | 149 | |
|
150 | 150 | def _processhttpv2reflectrequest(ui, repo, req, res): |
|
151 | 151 | """Reads unified frame protocol request and dumps out state to client. |
|
152 | 152 | |
|
153 | 153 | This special endpoint can be used to help debug the wire protocol. |
|
154 | 154 | |
|
155 | 155 | Instead of routing the request through the normal dispatch mechanism, |
|
156 | 156 | we instead read all frames, decode them, and feed them into our state |
|
157 | 157 | tracker. We then dump the log of all that activity back out to the |
|
158 | 158 | client. |
|
159 | 159 | """ |
|
160 | 160 | # Reflection APIs have a history of being abused, accidentally disclosing |
|
161 | 161 | # sensitive data, etc. So we have a config knob. |
|
162 | 162 | if not ui.configbool(b'experimental', b'web.api.debugreflect'): |
|
163 | 163 | res.status = b'404 Not Found' |
|
164 | 164 | res.headers[b'Content-Type'] = b'text/plain' |
|
165 | 165 | res.setbodybytes(_(b'debugreflect service not available')) |
|
166 | 166 | return |
|
167 | 167 | |
|
168 | 168 | # We assume we have a unified framing protocol request body. |
|
169 | 169 | |
|
170 | 170 | reactor = wireprotoframing.serverreactor(ui) |
|
171 | 171 | states = [] |
|
172 | 172 | |
|
173 | 173 | while True: |
|
174 | 174 | frame = wireprotoframing.readframe(req.bodyfh) |
|
175 | 175 | |
|
176 | 176 | if not frame: |
|
177 | 177 | states.append(b'received: <no frame>') |
|
178 | 178 | break |
|
179 | 179 | |
|
180 | 180 | states.append( |
|
181 | 181 | b'received: %d %d %d %s' |
|
182 | 182 | % (frame.typeid, frame.flags, frame.requestid, frame.payload) |
|
183 | 183 | ) |
|
184 | 184 | |
|
185 | 185 | action, meta = reactor.onframerecv(frame) |
|
186 | 186 | states.append(templatefilters.json((action, meta))) |
|
187 | 187 | |
|
188 | 188 | action, meta = reactor.oninputeof() |
|
189 | 189 | meta[b'action'] = action |
|
190 | 190 | states.append(templatefilters.json(meta)) |
|
191 | 191 | |
|
192 | 192 | res.status = b'200 OK' |
|
193 | 193 | res.headers[b'Content-Type'] = b'text/plain' |
|
194 | 194 | res.setbodybytes(b'\n'.join(states)) |
|
195 | 195 | |
|
196 | 196 | |
|
197 | 197 | def _processhttpv2request(ui, repo, req, res, authedperm, reqcommand, proto): |
|
198 | 198 | """Post-validation handler for HTTPv2 requests. |
|
199 | 199 | |
|
200 | 200 | Called when the HTTP request contains unified frame-based protocol |
|
201 | 201 | frames for evaluation. |
|
202 | 202 | """ |
|
203 | 203 | # TODO Some HTTP clients are full duplex and can receive data before |
|
204 | 204 | # the entire request is transmitted. Figure out a way to indicate support |
|
205 | 205 | # for that so we can opt into full duplex mode. |
|
206 | 206 | reactor = wireprotoframing.serverreactor(ui, deferoutput=True) |
|
207 | 207 | seencommand = False |
|
208 | 208 | |
|
209 | 209 | outstream = None |
|
210 | 210 | |
|
211 | 211 | while True: |
|
212 | 212 | frame = wireprotoframing.readframe(req.bodyfh) |
|
213 | 213 | if not frame: |
|
214 | 214 | break |
|
215 | 215 | |
|
216 | 216 | action, meta = reactor.onframerecv(frame) |
|
217 | 217 | |
|
218 | 218 | if action == b'wantframe': |
|
219 | 219 | # Need more data before we can do anything. |
|
220 | 220 | continue |
|
221 | 221 | elif action == b'runcommand': |
|
222 | 222 | # Defer creating output stream because we need to wait for |
|
223 | 223 | # protocol settings frames so proper encoding can be applied. |
|
224 | 224 | if not outstream: |
|
225 | 225 | outstream = reactor.makeoutputstream() |
|
226 | 226 | |
|
227 | 227 | sentoutput = _httpv2runcommand( |
|
228 | 228 | ui, |
|
229 | 229 | repo, |
|
230 | 230 | req, |
|
231 | 231 | res, |
|
232 | 232 | authedperm, |
|
233 | 233 | reqcommand, |
|
234 | 234 | reactor, |
|
235 | 235 | outstream, |
|
236 | 236 | meta, |
|
237 | 237 | issubsequent=seencommand, |
|
238 | 238 | ) |
|
239 | 239 | |
|
240 | 240 | if sentoutput: |
|
241 | 241 | return |
|
242 | 242 | |
|
243 | 243 | seencommand = True |
|
244 | 244 | |
|
245 | 245 | elif action == b'error': |
|
246 | 246 | # TODO define proper error mechanism. |
|
247 | 247 | res.status = b'200 OK' |
|
248 | 248 | res.headers[b'Content-Type'] = b'text/plain' |
|
249 | 249 | res.setbodybytes(meta[b'message'] + b'\n') |
|
250 | 250 | return |
|
251 | 251 | else: |
|
252 | 252 | raise error.ProgrammingError( |
|
253 | 253 | b'unhandled action from frame processor: %s' % action |
|
254 | 254 | ) |
|
255 | 255 | |
|
256 | 256 | action, meta = reactor.oninputeof() |
|
257 | 257 | if action == b'sendframes': |
|
258 | 258 | # We assume we haven't started sending the response yet. If we're |
|
259 | 259 | # wrong, the response type will raise an exception. |
|
260 | 260 | res.status = b'200 OK' |
|
261 | 261 | res.headers[b'Content-Type'] = FRAMINGTYPE |
|
262 | 262 | res.setbodygen(meta[b'framegen']) |
|
263 | 263 | elif action == b'noop': |
|
264 | 264 | pass |
|
265 | 265 | else: |
|
266 | 266 | raise error.ProgrammingError( |
|
267 | 267 | b'unhandled action from frame processor: %s' % action |
|
268 | 268 | ) |
|
269 | 269 | |
|
270 | 270 | |
|
271 | 271 | def _httpv2runcommand( |
|
272 | 272 | ui, |
|
273 | 273 | repo, |
|
274 | 274 | req, |
|
275 | 275 | res, |
|
276 | 276 | authedperm, |
|
277 | 277 | reqcommand, |
|
278 | 278 | reactor, |
|
279 | 279 | outstream, |
|
280 | 280 | command, |
|
281 | 281 | issubsequent, |
|
282 | 282 | ): |
|
283 | 283 | """Dispatch a wire protocol command made from HTTPv2 requests. |
|
284 | 284 | |
|
285 | 285 | The authenticated permission (``authedperm``) along with the original |
|
286 | 286 | command from the URL (``reqcommand``) are passed in. |
|
287 | 287 | """ |
|
288 | 288 | # We already validated that the session has permissions to perform the |
|
289 | 289 | # actions in ``authedperm``. In the unified frame protocol, the canonical |
|
290 | 290 | # command to run is expressed in a frame. However, the URL also requested |
|
291 | 291 | # to run a specific command. We need to be careful that the command we |
|
292 | 292 | # run doesn't have permissions requirements greater than what was granted |
|
293 | 293 | # by ``authedperm``. |
|
294 | 294 | # |
|
295 | 295 | # Our rule for this is we only allow one command per HTTP request and |
|
296 | 296 | # that command must match the command in the URL. However, we make |
|
297 | 297 | # an exception for the ``multirequest`` URL. This URL is allowed to |
|
298 | 298 | # execute multiple commands. We double check permissions of each command |
|
299 | 299 | # as it is invoked to ensure there is no privilege escalation. |
|
300 | 300 | # TODO consider allowing multiple commands to regular command URLs |
|
301 | 301 | # iff each command is the same. |
|
302 | 302 | |
|
303 | 303 | proto = httpv2protocolhandler(req, ui, args=command[b'args']) |
|
304 | 304 | |
|
305 | 305 | if reqcommand == b'multirequest': |
|
306 | 306 | if not COMMANDS.commandavailable(command[b'command'], proto): |
|
307 | 307 | # TODO proper error mechanism |
|
308 | 308 | res.status = b'200 OK' |
|
309 | 309 | res.headers[b'Content-Type'] = b'text/plain' |
|
310 | 310 | res.setbodybytes( |
|
311 | 311 | _(b'wire protocol command not available: %s') |
|
312 | 312 | % command[b'command'] |
|
313 | 313 | ) |
|
314 | 314 | return True |
|
315 | 315 | |
|
316 | 316 | # TODO don't use assert here, since it may be elided by -O. |
|
317 | 317 | assert authedperm in (b'ro', b'rw') |
|
318 | 318 | wirecommand = COMMANDS[command[b'command']] |
|
319 | 319 | assert wirecommand.permission in (b'push', b'pull') |
|
320 | 320 | |
|
321 | 321 | if authedperm == b'ro' and wirecommand.permission != b'pull': |
|
322 | 322 | # TODO proper error mechanism |
|
323 | 323 | res.status = b'403 Forbidden' |
|
324 | 324 | res.headers[b'Content-Type'] = b'text/plain' |
|
325 | 325 | res.setbodybytes( |
|
326 | 326 | _(b'insufficient permissions to execute command: %s') |
|
327 | 327 | % command[b'command'] |
|
328 | 328 | ) |
|
329 | 329 | return True |
|
330 | 330 | |
|
331 | 331 | # TODO should we also call checkperm() here? Maybe not if we're going |
|
332 | 332 | # to overhaul that API. The granted scope from the URL check should |
|
333 | 333 | # be good enough. |
|
334 | 334 | |
|
335 | 335 | else: |
|
336 | 336 | # Don't allow multiple commands outside of ``multirequest`` URL. |
|
337 | 337 | if issubsequent: |
|
338 | 338 | # TODO proper error mechanism |
|
339 | 339 | res.status = b'200 OK' |
|
340 | 340 | res.headers[b'Content-Type'] = b'text/plain' |
|
341 | 341 | res.setbodybytes( |
|
342 | 342 | _(b'multiple commands cannot be issued to this URL') |
|
343 | 343 | ) |
|
344 | 344 | return True |
|
345 | 345 | |
|
346 | 346 | if reqcommand != command[b'command']: |
|
347 | 347 | # TODO define proper error mechanism |
|
348 | 348 | res.status = b'200 OK' |
|
349 | 349 | res.headers[b'Content-Type'] = b'text/plain' |
|
350 | 350 | res.setbodybytes(_(b'command in frame must match command in URL')) |
|
351 | 351 | return True |
|
352 | 352 | |
|
353 | 353 | res.status = b'200 OK' |
|
354 | 354 | res.headers[b'Content-Type'] = FRAMINGTYPE |
|
355 | 355 | |
|
356 | 356 | try: |
|
357 | 357 | objs = dispatch(repo, proto, command[b'command'], command[b'redirect']) |
|
358 | 358 | |
|
359 | 359 | action, meta = reactor.oncommandresponsereadyobjects( |
|
360 | 360 | outstream, command[b'requestid'], objs |
|
361 | 361 | ) |
|
362 | 362 | |
|
363 | 363 | except error.WireprotoCommandError as e: |
|
364 | 364 | action, meta = reactor.oncommanderror( |
|
365 | 365 | outstream, command[b'requestid'], e.message, e.messageargs |
|
366 | 366 | ) |
|
367 | 367 | |
|
368 | 368 | except Exception as e: |
|
369 | 369 | action, meta = reactor.onservererror( |
|
370 | 370 | outstream, |
|
371 | 371 | command[b'requestid'], |
|
372 | 372 | _(b'exception when invoking command: %s') |
|
373 | 373 | % stringutil.forcebytestr(e), |
|
374 | 374 | ) |
|
375 | 375 | |
|
376 | 376 | if action == b'sendframes': |
|
377 | 377 | res.setbodygen(meta[b'framegen']) |
|
378 | 378 | return True |
|
379 | 379 | elif action == b'noop': |
|
380 | 380 | return False |
|
381 | 381 | else: |
|
382 | 382 | raise error.ProgrammingError( |
|
383 | 383 | b'unhandled event from reactor: %s' % action |
|
384 | 384 | ) |
|
385 | 385 | |
|
386 | 386 | |
|
387 | 387 | def getdispatchrepo(repo, proto, command): |
|
388 | 388 | viewconfig = repo.ui.config(b'server', b'view') |
|
389 | 389 | return repo.filtered(viewconfig) |
|
390 | 390 | |
|
391 | 391 | |
|
392 | 392 | def dispatch(repo, proto, command, redirect): |
|
393 | 393 | """Run a wire protocol command. |
|
394 | 394 | |
|
395 | 395 | Returns an iterable of objects that will be sent to the client. |
|
396 | 396 | """ |
|
397 | 397 | repo = getdispatchrepo(repo, proto, command) |
|
398 | 398 | |
|
399 | 399 | entry = COMMANDS[command] |
|
400 | 400 | func = entry.func |
|
401 | 401 | spec = entry.args |
|
402 | 402 | |
|
403 | 403 | args = proto.getargs(spec) |
|
404 | 404 | |
|
405 | 405 | # There is some duplicate boilerplate code here for calling the command and |
|
406 | 406 | # emitting objects. It is either that or a lot of indented code that looks |
|
407 | 407 | # like a pyramid (since there are a lot of code paths that result in not |
|
408 | 408 | # using the cacher). |
|
409 | 409 | callcommand = lambda: func(repo, proto, **pycompat.strkwargs(args)) |
|
410 | 410 | |
|
411 | 411 | # Request is not cacheable. Don't bother instantiating a cacher. |
|
412 | 412 | if not entry.cachekeyfn: |
|
413 | 413 | for o in callcommand(): |
|
414 | 414 | yield o |
|
415 | 415 | return |
|
416 | 416 | |
|
417 | 417 | if redirect: |
|
418 | 418 | redirecttargets = redirect[b'targets'] |
|
419 | 419 | redirecthashes = redirect[b'hashes'] |
|
420 | 420 | else: |
|
421 | 421 | redirecttargets = [] |
|
422 | 422 | redirecthashes = [] |
|
423 | 423 | |
|
424 | 424 | cacher = makeresponsecacher( |
|
425 | 425 | repo, |
|
426 | 426 | proto, |
|
427 | 427 | command, |
|
428 | 428 | args, |
|
429 | 429 | cborutil.streamencode, |
|
430 | 430 | redirecttargets=redirecttargets, |
|
431 | 431 | redirecthashes=redirecthashes, |
|
432 | 432 | ) |
|
433 | 433 | |
|
434 | 434 | # But we have no cacher. Do default handling. |
|
435 | 435 | if not cacher: |
|
436 | 436 | for o in callcommand(): |
|
437 | 437 | yield o |
|
438 | 438 | return |
|
439 | 439 | |
|
440 | 440 | with cacher: |
|
441 | 441 | cachekey = entry.cachekeyfn( |
|
442 | 442 | repo, proto, cacher, **pycompat.strkwargs(args) |
|
443 | 443 | ) |
|
444 | 444 | |
|
445 | 445 | # No cache key or the cacher doesn't like it. Do default handling. |
|
446 | 446 | if cachekey is None or not cacher.setcachekey(cachekey): |
|
447 | 447 | for o in callcommand(): |
|
448 | 448 | yield o |
|
449 | 449 | return |
|
450 | 450 | |
|
451 | 451 | # Serve it from the cache, if possible. |
|
452 | 452 | cached = cacher.lookup() |
|
453 | 453 | |
|
454 | 454 | if cached: |
|
455 | 455 | for o in cached[b'objs']: |
|
456 | 456 | yield o |
|
457 | 457 | return |
|
458 | 458 | |
|
459 | 459 | # Else call the command and feed its output into the cacher, allowing |
|
460 | 460 | # the cacher to buffer/mutate objects as it desires. |
|
461 | 461 | for o in callcommand(): |
|
462 | 462 | for o in cacher.onobject(o): |
|
463 | 463 | yield o |
|
464 | 464 | |
|
465 | 465 | for o in cacher.onfinished(): |
|
466 | 466 | yield o |
|
467 | 467 | |
|
468 | 468 | |
|
469 | 469 | @interfaceutil.implementer(wireprototypes.baseprotocolhandler) |
|
470 | 470 | class httpv2protocolhandler(object): |
|
471 | 471 | def __init__(self, req, ui, args=None): |
|
472 | 472 | self._req = req |
|
473 | 473 | self._ui = ui |
|
474 | 474 | self._args = args |
|
475 | 475 | |
|
476 | 476 | @property |
|
477 | 477 | def name(self): |
|
478 | 478 | return HTTP_WIREPROTO_V2 |
|
479 | 479 | |
|
480 | 480 | def getargs(self, args): |
|
481 | 481 | # First look for args that were passed but aren't registered on this |
|
482 | 482 | # command. |
|
483 | 483 | extra = set(self._args) - set(args) |
|
484 | 484 | if extra: |
|
485 | 485 | raise error.WireprotoCommandError( |
|
486 | 486 | b'unsupported argument to command: %s' |
|
487 | 487 | % b', '.join(sorted(extra)) |
|
488 | 488 | ) |
|
489 | 489 | |
|
490 | 490 | # And look for required arguments that are missing. |
|
491 | 491 | missing = {a for a in args if args[a][b'required']} - set(self._args) |
|
492 | 492 | |
|
493 | 493 | if missing: |
|
494 | 494 | raise error.WireprotoCommandError( |
|
495 | 495 | b'missing required arguments: %s' % b', '.join(sorted(missing)) |
|
496 | 496 | ) |
|
497 | 497 | |
|
498 | 498 | # Now derive the arguments to pass to the command, taking into |
|
499 | 499 | # account the arguments specified by the client. |
|
500 | 500 | data = {} |
|
501 | 501 | for k, meta in sorted(args.items()): |
|
502 | 502 | # This argument wasn't passed by the client. |
|
503 | 503 | if k not in self._args: |
|
504 | 504 | data[k] = meta[b'default']() |
|
505 | 505 | continue |
|
506 | 506 | |
|
507 | 507 | v = self._args[k] |
|
508 | 508 | |
|
509 | 509 | # Sets may be expressed as lists. Silently normalize. |
|
510 | 510 | if meta[b'type'] == b'set' and isinstance(v, list): |
|
511 | 511 | v = set(v) |
|
512 | 512 | |
|
513 | 513 | # TODO consider more/stronger type validation. |
|
514 | 514 | |
|
515 | 515 | data[k] = v |
|
516 | 516 | |
|
517 | 517 | return data |
|
518 | 518 | |
|
519 | 519 | def getprotocaps(self): |
|
520 | 520 | # Protocol capabilities are currently not implemented for HTTP V2. |
|
521 | 521 | return set() |
|
522 | 522 | |
|
523 | 523 | def getpayload(self): |
|
524 | 524 | raise NotImplementedError |
|
525 | 525 | |
|
526 | 526 | @contextlib.contextmanager |
|
527 | 527 | def mayberedirectstdio(self): |
|
528 | 528 | raise NotImplementedError |
|
529 | 529 | |
|
530 | 530 | def client(self): |
|
531 | 531 | raise NotImplementedError |
|
532 | 532 | |
|
533 | 533 | def addcapabilities(self, repo, caps): |
|
534 | 534 | return caps |
|
535 | 535 | |
|
536 | 536 | def checkperm(self, perm): |
|
537 | 537 | raise NotImplementedError |
|
538 | 538 | |
|
539 | 539 | |
|
540 | 540 | def httpv2apidescriptor(req, repo): |
|
541 | 541 | proto = httpv2protocolhandler(req, repo.ui) |
|
542 | 542 | |
|
543 | 543 | return _capabilitiesv2(repo, proto) |
|
544 | 544 | |
|
545 | 545 | |
|
546 | 546 | def _capabilitiesv2(repo, proto): |
|
547 | 547 | """Obtain the set of capabilities for version 2 transports. |
|
548 | 548 | |
|
549 | 549 | These capabilities are distinct from the capabilities for version 1 |
|
550 | 550 | transports. |
|
551 | 551 | """ |
|
552 | 552 | caps = { |
|
553 | 553 | b'commands': {}, |
|
554 | 554 | b'framingmediatypes': [FRAMINGTYPE], |
|
555 | 555 | b'pathfilterprefixes': set(narrowspec.VALID_PREFIXES), |
|
556 | 556 | } |
|
557 | 557 | |
|
558 | 558 | for command, entry in COMMANDS.items(): |
|
559 | 559 | args = {} |
|
560 | 560 | |
|
561 | 561 | for arg, meta in entry.args.items(): |
|
562 | 562 | args[arg] = { |
|
563 | 563 | # TODO should this be a normalized type using CBOR's |
|
564 | 564 | # terminology? |
|
565 | 565 | b'type': meta[b'type'], |
|
566 | 566 | b'required': meta[b'required'], |
|
567 | 567 | } |
|
568 | 568 | |
|
569 | 569 | if not meta[b'required']: |
|
570 | 570 | args[arg][b'default'] = meta[b'default']() |
|
571 | 571 | |
|
572 | 572 | if meta[b'validvalues']: |
|
573 | 573 | args[arg][b'validvalues'] = meta[b'validvalues'] |
|
574 | 574 | |
|
575 | 575 | # TODO this type of check should be defined in a per-command callback. |
|
576 | 576 | if ( |
|
577 | 577 | command == b'rawstorefiledata' |
|
578 | 578 | and not streamclone.allowservergeneration(repo) |
|
579 | 579 | ): |
|
580 | 580 | continue |
|
581 | 581 | |
|
582 | 582 | caps[b'commands'][command] = { |
|
583 | 583 | b'args': args, |
|
584 | 584 | b'permissions': [entry.permission], |
|
585 | 585 | } |
|
586 | 586 | |
|
587 | 587 | if entry.extracapabilitiesfn: |
|
588 | 588 | extracaps = entry.extracapabilitiesfn(repo, proto) |
|
589 | 589 | caps[b'commands'][command].update(extracaps) |
|
590 | 590 | |
|
591 | 591 | caps[b'rawrepoformats'] = sorted(repo.requirements & repo.supportedformats) |
|
592 | 592 | |
|
593 | 593 | targets = getadvertisedredirecttargets(repo, proto) |
|
594 | 594 | if targets: |
|
595 | 595 | caps[b'redirect'] = { |
|
596 | 596 | b'targets': [], |
|
597 | 597 | b'hashes': [b'sha256', b'sha1'], |
|
598 | 598 | } |
|
599 | 599 | |
|
600 | 600 | for target in targets: |
|
601 | 601 | entry = { |
|
602 | 602 | b'name': target[b'name'], |
|
603 | 603 | b'protocol': target[b'protocol'], |
|
604 | 604 | b'uris': target[b'uris'], |
|
605 | 605 | } |
|
606 | 606 | |
|
607 | 607 | for key in (b'snirequired', b'tlsversions'): |
|
608 | 608 | if key in target: |
|
609 | 609 | entry[key] = target[key] |
|
610 | 610 | |
|
611 | 611 | caps[b'redirect'][b'targets'].append(entry) |
|
612 | 612 | |
|
613 | 613 | return proto.addcapabilities(repo, caps) |
|
614 | 614 | |
|
615 | 615 | |
|
616 | 616 | def getadvertisedredirecttargets(repo, proto): |
|
617 | 617 | """Obtain a list of content redirect targets. |
|
618 | 618 | |
|
619 | 619 | Returns a list containing potential redirect targets that will be |
|
620 | 620 | advertised in capabilities data. Each dict MUST have the following |
|
621 | 621 | keys: |
|
622 | 622 | |
|
623 | 623 | name |
|
624 | 624 | The name of this redirect target. This is the identifier clients use |
|
625 | 625 | to refer to a target. It is transferred as part of every command |
|
626 | 626 | request. |
|
627 | 627 | |
|
628 | 628 | protocol |
|
629 | 629 | Network protocol used by this target. Typically this is the string |
|
630 | 630 | in front of the ``://`` in a URL. e.g. ``https``. |
|
631 | 631 | |
|
632 | 632 | uris |
|
633 | 633 | List of representative URIs for this target. Clients can use the |
|
634 | 634 | URIs to test parsing for compatibility or for ordering preference |
|
635 | 635 | for which target to use. |
|
636 | 636 | |
|
637 | 637 | The following optional keys are recognized: |
|
638 | 638 | |
|
639 | 639 | snirequired |
|
640 | 640 | Bool indicating if Server Name Indication (SNI) is required to |
|
641 | 641 | connect to this target. |
|
642 | 642 | |
|
643 | 643 | tlsversions |
|
644 | 644 | List of bytes indicating which TLS versions are supported by this |
|
645 | 645 | target. |
|
646 | 646 | |
|
647 | 647 | By default, clients reflect the target order advertised by servers |
|
648 | 648 | and servers will use the first client-advertised target when picking |
|
649 | 649 | a redirect target. So targets should be advertised in the order the |
|
650 | 650 | server prefers they be used. |
|
651 | 651 | """ |
|
652 | 652 | return [] |
|
653 | 653 | |
|
654 | 654 | |
|
655 | 655 | def wireprotocommand( |
|
656 | 656 | name, |
|
657 | 657 | args=None, |
|
658 | 658 | permission=b'push', |
|
659 | 659 | cachekeyfn=None, |
|
660 | 660 | extracapabilitiesfn=None, |
|
661 | 661 | ): |
|
662 | 662 | """Decorator to declare a wire protocol command. |
|
663 | 663 | |
|
664 | 664 | ``name`` is the name of the wire protocol command being provided. |
|
665 | 665 | |
|
666 | 666 | ``args`` is a dict defining arguments accepted by the command. Keys are |
|
667 | 667 | the argument name. Values are dicts with the following keys: |
|
668 | 668 | |
|
669 | 669 | ``type`` |
|
670 | 670 | The argument data type. Must be one of the following string |
|
671 | 671 | literals: ``bytes``, ``int``, ``list``, ``dict``, ``set``, |
|
672 | 672 | or ``bool``. |
|
673 | 673 | |
|
674 | 674 | ``default`` |
|
675 | 675 | A callable returning the default value for this argument. If not |
|
676 | 676 | specified, ``None`` will be the default value. |
|
677 | 677 | |
|
678 | 678 | ``example`` |
|
679 | 679 | An example value for this argument. |
|
680 | 680 | |
|
681 | 681 | ``validvalues`` |
|
682 | 682 | Set of recognized values for this argument. |
|
683 | 683 | |
|
684 | 684 | ``permission`` defines the permission type needed to run this command. |
|
685 | 685 | Can be ``push`` or ``pull``. These roughly map to read-write and read-only, |
|
686 | 686 | respectively. Default is to assume command requires ``push`` permissions |
|
687 | 687 | because otherwise commands not declaring their permissions could modify |
|
688 | 688 | a repository that is supposed to be read-only. |
|
689 | 689 | |
|
690 | 690 | ``cachekeyfn`` defines an optional callable that can derive the |
|
691 | 691 | cache key for this request. |
|
692 | 692 | |
|
693 | 693 | ``extracapabilitiesfn`` defines an optional callable that defines extra |
|
694 | 694 | command capabilities/parameters that are advertised next to the command |
|
695 | 695 | in the capabilities data structure describing the server. The callable |
|
696 | 696 | receives as arguments the repository and protocol objects. It returns |
|
697 | 697 | a dict of extra fields to add to the command descriptor. |
|
698 | 698 | |
|
699 | 699 | Wire protocol commands are generators of objects to be serialized and |
|
700 | 700 | sent to the client. |
|
701 | 701 | |
|
702 | 702 | If a command raises an uncaught exception, this will be translated into |
|
703 | 703 | a command error. |
|
704 | 704 | |
|
705 | 705 | All commands can opt in to being cacheable by defining a function |
|
706 | 706 | (``cachekeyfn``) that is called to derive a cache key. This function |
|
707 | 707 | receives the same arguments as the command itself plus a ``cacher`` |
|
708 | 708 | argument containing the active cacher for the request and returns a bytes |
|
709 | 709 | containing the key in a cache the response to this command may be cached |
|
710 | 710 | under. |
|
711 | 711 | """ |
|
712 | 712 | transports = { |
|
713 | 713 | k for k, v in wireprototypes.TRANSPORTS.items() if v[b'version'] == 2 |
|
714 | 714 | } |
|
715 | 715 | |
|
716 | 716 | if permission not in (b'push', b'pull'): |
|
717 | 717 | raise error.ProgrammingError( |
|
718 | 718 | b'invalid wire protocol permission; ' |
|
719 | 719 | b'got %s; expected "push" or "pull"' % permission |
|
720 | 720 | ) |
|
721 | 721 | |
|
722 | 722 | if args is None: |
|
723 | 723 | args = {} |
|
724 | 724 | |
|
725 | 725 | if not isinstance(args, dict): |
|
726 | 726 | raise error.ProgrammingError( |
|
727 | 727 | b'arguments for version 2 commands must be declared as dicts' |
|
728 | 728 | ) |
|
729 | 729 | |
|
730 | 730 | for arg, meta in args.items(): |
|
731 | 731 | if arg == b'*': |
|
732 | 732 | raise error.ProgrammingError( |
|
733 | 733 | b'* argument name not allowed on version 2 commands' |
|
734 | 734 | ) |
|
735 | 735 | |
|
736 | 736 | if not isinstance(meta, dict): |
|
737 | 737 | raise error.ProgrammingError( |
|
738 | 738 | b'arguments for version 2 commands ' |
|
739 | 739 | b'must declare metadata as a dict' |
|
740 | 740 | ) |
|
741 | 741 | |
|
742 | 742 | if b'type' not in meta: |
|
743 | 743 | raise error.ProgrammingError( |
|
744 | 744 | b'%s argument for command %s does not ' |
|
745 | 745 | b'declare type field' % (arg, name) |
|
746 | 746 | ) |
|
747 | 747 | |
|
748 | 748 | if meta[b'type'] not in ( |
|
749 | 749 | b'bytes', |
|
750 | 750 | b'int', |
|
751 | 751 | b'list', |
|
752 | 752 | b'dict', |
|
753 | 753 | b'set', |
|
754 | 754 | b'bool', |
|
755 | 755 | ): |
|
756 | 756 | raise error.ProgrammingError( |
|
757 | 757 | b'%s argument for command %s has ' |
|
758 | 758 | b'illegal type: %s' % (arg, name, meta[b'type']) |
|
759 | 759 | ) |
|
760 | 760 | |
|
761 | 761 | if b'example' not in meta: |
|
762 | 762 | raise error.ProgrammingError( |
|
763 | 763 | b'%s argument for command %s does not ' |
|
764 | 764 | b'declare example field' % (arg, name) |
|
765 | 765 | ) |
|
766 | 766 | |
|
767 | 767 | meta[b'required'] = b'default' not in meta |
|
768 | 768 | |
|
769 | 769 | meta.setdefault(b'default', lambda: None) |
|
770 | 770 | meta.setdefault(b'validvalues', None) |
|
771 | 771 | |
|
772 | 772 | def register(func): |
|
773 | 773 | if name in COMMANDS: |
|
774 | 774 | raise error.ProgrammingError( |
|
775 | 775 | b'%s command already registered for version 2' % name |
|
776 | 776 | ) |
|
777 | 777 | |
|
778 | 778 | COMMANDS[name] = wireprototypes.commandentry( |
|
779 | 779 | func, |
|
780 | 780 | args=args, |
|
781 | 781 | transports=transports, |
|
782 | 782 | permission=permission, |
|
783 | 783 | cachekeyfn=cachekeyfn, |
|
784 | 784 | extracapabilitiesfn=extracapabilitiesfn, |
|
785 | 785 | ) |
|
786 | 786 | |
|
787 | 787 | return func |
|
788 | 788 | |
|
789 | 789 | return register |
|
790 | 790 | |
|
791 | 791 | |
|
792 | 792 | def makecommandcachekeyfn(command, localversion=None, allargs=False): |
|
793 | 793 | """Construct a cache key derivation function with common features. |
|
794 | 794 | |
|
795 | 795 | By default, the cache key is a hash of: |
|
796 | 796 | |
|
797 | 797 | * The command name. |
|
798 | 798 | * A global cache version number. |
|
799 | 799 | * A local cache version number (passed via ``localversion``). |
|
800 | 800 | * All the arguments passed to the command. |
|
801 | 801 | * The media type used. |
|
802 | 802 | * Wire protocol version string. |
|
803 | 803 | * The repository path. |
|
804 | 804 | """ |
|
805 | 805 | if not allargs: |
|
806 | 806 | raise error.ProgrammingError( |
|
807 | 807 | b'only allargs=True is currently supported' |
|
808 | 808 | ) |
|
809 | 809 | |
|
810 | 810 | if localversion is None: |
|
811 | 811 | raise error.ProgrammingError(b'must set localversion argument value') |
|
812 | 812 | |
|
813 | 813 | def cachekeyfn(repo, proto, cacher, **args): |
|
814 | 814 | spec = COMMANDS[command] |
|
815 | 815 | |
|
816 | 816 | # Commands that mutate the repo can not be cached. |
|
817 | 817 | if spec.permission == b'push': |
|
818 | 818 | return None |
|
819 | 819 | |
|
820 | 820 | # TODO config option to disable caching. |
|
821 | 821 | |
|
822 | 822 | # Our key derivation strategy is to construct a data structure |
|
823 | 823 | # holding everything that could influence cacheability and to hash |
|
824 | 824 | # the CBOR representation of that. Using CBOR seems like it might |
|
825 | 825 | # be overkill. However, simpler hashing mechanisms are prone to |
|
826 | 826 | # duplicate input issues. e.g. if you just concatenate two values, |
|
827 | 827 | # "foo"+"bar" is identical to "fo"+"obar". Using CBOR provides |
|
828 | 828 | # "padding" between values and prevents these problems. |
|
829 | 829 | |
|
830 | 830 | # Seed the hash with various data. |
|
831 | 831 | state = { |
|
832 | 832 | # To invalidate all cache keys. |
|
833 | 833 | b'globalversion': GLOBAL_CACHE_VERSION, |
|
834 | 834 | # More granular cache key invalidation. |
|
835 | 835 | b'localversion': localversion, |
|
836 | 836 | # Cache keys are segmented by command. |
|
837 | 837 | b'command': command, |
|
838 | 838 | # Throw in the media type and API version strings so changes |
|
839 | 839 | # to exchange semantics invalid cache. |
|
840 | 840 | b'mediatype': FRAMINGTYPE, |
|
841 | 841 | b'version': HTTP_WIREPROTO_V2, |
|
842 | 842 | # So same requests for different repos don't share cache keys. |
|
843 | 843 | b'repo': repo.root, |
|
844 | 844 | } |
|
845 | 845 | |
|
846 | 846 | # The arguments passed to us will have already been normalized. |
|
847 | 847 | # Default values will be set, etc. This is important because it |
|
848 | 848 | # means that it doesn't matter if clients send an explicit argument |
|
849 | 849 | # or rely on the default value: it will all normalize to the same |
|
850 | 850 | # set of arguments on the server and therefore the same cache key. |
|
851 | 851 | # |
|
852 | 852 | # Arguments by their very nature must support being encoded to CBOR. |
|
853 | 853 | # And the CBOR encoder is deterministic. So we hash the arguments |
|
854 | 854 | # by feeding the CBOR of their representation into the hasher. |
|
855 | 855 | if allargs: |
|
856 | 856 | state[b'args'] = pycompat.byteskwargs(args) |
|
857 | 857 | |
|
858 | 858 | cacher.adjustcachekeystate(state) |
|
859 | 859 | |
|
860 | 860 | hasher = hashutil.sha1() |
|
861 | 861 | for chunk in cborutil.streamencode(state): |
|
862 | 862 | hasher.update(chunk) |
|
863 | 863 | |
|
864 | 864 | return pycompat.sysbytes(hasher.hexdigest()) |
|
865 | 865 | |
|
866 | 866 | return cachekeyfn |
|
867 | 867 | |
|
868 | 868 | |
|
869 | 869 | def makeresponsecacher( |
|
870 | 870 | repo, proto, command, args, objencoderfn, redirecttargets, redirecthashes |
|
871 | 871 | ): |
|
872 | 872 | """Construct a cacher for a cacheable command. |
|
873 | 873 | |
|
874 | 874 | Returns an ``iwireprotocolcommandcacher`` instance. |
|
875 | 875 | |
|
876 | 876 | Extensions can monkeypatch this function to provide custom caching |
|
877 | 877 | backends. |
|
878 | 878 | """ |
|
879 | 879 | return None |
|
880 | 880 | |
|
881 | 881 | |
|
882 | 882 | def resolvenodes(repo, revisions): |
|
883 | 883 | """Resolve nodes from a revisions specifier data structure.""" |
|
884 | 884 | cl = repo.changelog |
|
885 | 885 | clhasnode = cl.hasnode |
|
886 | 886 | |
|
887 | 887 | seen = set() |
|
888 | 888 | nodes = [] |
|
889 | 889 | |
|
890 | 890 | if not isinstance(revisions, list): |
|
891 | 891 | raise error.WireprotoCommandError( |
|
892 | 892 | b'revisions must be defined as an array' |
|
893 | 893 | ) |
|
894 | 894 | |
|
895 | 895 | for spec in revisions: |
|
896 | 896 | if b'type' not in spec: |
|
897 | 897 | raise error.WireprotoCommandError( |
|
898 | 898 | b'type key not present in revision specifier' |
|
899 | 899 | ) |
|
900 | 900 | |
|
901 | 901 | typ = spec[b'type'] |
|
902 | 902 | |
|
903 | 903 | if typ == b'changesetexplicit': |
|
904 | 904 | if b'nodes' not in spec: |
|
905 | 905 | raise error.WireprotoCommandError( |
|
906 | 906 | b'nodes key not present in changesetexplicit revision ' |
|
907 | 907 | b'specifier' |
|
908 | 908 | ) |
|
909 | 909 | |
|
910 | 910 | for node in spec[b'nodes']: |
|
911 | 911 | if node not in seen: |
|
912 | 912 | nodes.append(node) |
|
913 | 913 | seen.add(node) |
|
914 | 914 | |
|
915 | 915 | elif typ == b'changesetexplicitdepth': |
|
916 | 916 | for key in (b'nodes', b'depth'): |
|
917 | 917 | if key not in spec: |
|
918 | 918 | raise error.WireprotoCommandError( |
|
919 | 919 | b'%s key not present in changesetexplicitdepth revision ' |
|
920 | 920 | b'specifier', |
|
921 | 921 | (key,), |
|
922 | 922 | ) |
|
923 | 923 | |
|
924 | 924 | for rev in repo.revs( |
|
925 | 925 | b'ancestors(%ln, %s)', spec[b'nodes'], spec[b'depth'] - 1 |
|
926 | 926 | ): |
|
927 | 927 | node = cl.node(rev) |
|
928 | 928 | |
|
929 | 929 | if node not in seen: |
|
930 | 930 | nodes.append(node) |
|
931 | 931 | seen.add(node) |
|
932 | 932 | |
|
933 | 933 | elif typ == b'changesetdagrange': |
|
934 | 934 | for key in (b'roots', b'heads'): |
|
935 | 935 | if key not in spec: |
|
936 | 936 | raise error.WireprotoCommandError( |
|
937 | 937 | b'%s key not present in changesetdagrange revision ' |
|
938 | 938 | b'specifier', |
|
939 | 939 | (key,), |
|
940 | 940 | ) |
|
941 | 941 | |
|
942 | 942 | if not spec[b'heads']: |
|
943 | 943 | raise error.WireprotoCommandError( |
|
944 | 944 | b'heads key in changesetdagrange cannot be empty' |
|
945 | 945 | ) |
|
946 | 946 | |
|
947 | 947 | if spec[b'roots']: |
|
948 | 948 | common = [n for n in spec[b'roots'] if clhasnode(n)] |
|
949 | 949 | else: |
|
950 | 950 | common = [repo.nullid] |
|
951 | 951 | |
|
952 | 952 | for n in discovery.outgoing(repo, common, spec[b'heads']).missing: |
|
953 | 953 | if n not in seen: |
|
954 | 954 | nodes.append(n) |
|
955 | 955 | seen.add(n) |
|
956 | 956 | |
|
957 | 957 | else: |
|
958 | 958 | raise error.WireprotoCommandError( |
|
959 | 959 | b'unknown revision specifier type: %s', (typ,) |
|
960 | 960 | ) |
|
961 | 961 | |
|
962 | 962 | return nodes |
|
963 | 963 | |
|
964 | 964 | |
|
965 | 965 | @wireprotocommand(b'branchmap', permission=b'pull') |
|
966 | 966 | def branchmapv2(repo, proto): |
|
967 | 967 | yield { |
|
968 | 968 | encoding.fromlocal(k): v |
|
969 | 969 | for k, v in pycompat.iteritems(repo.branchmap()) |
|
970 | 970 | } |
|
971 | 971 | |
|
972 | 972 | |
|
973 | 973 | @wireprotocommand(b'capabilities', permission=b'pull') |
|
974 | 974 | def capabilitiesv2(repo, proto): |
|
975 | 975 | yield _capabilitiesv2(repo, proto) |
|
976 | 976 | |
|
977 | 977 | |
|
978 | 978 | @wireprotocommand( |
|
979 | 979 | b'changesetdata', |
|
980 | 980 | args={ |
|
981 | 981 | b'revisions': { |
|
982 | 982 | b'type': b'list', |
|
983 | 983 | b'example': [ |
|
984 | 984 | { |
|
985 | 985 | b'type': b'changesetexplicit', |
|
986 | 986 | b'nodes': [b'abcdef...'], |
|
987 | 987 | } |
|
988 | 988 | ], |
|
989 | 989 | }, |
|
990 | 990 | b'fields': { |
|
991 | 991 | b'type': b'set', |
|
992 | 992 | b'default': set, |
|
993 | 993 | b'example': {b'parents', b'revision'}, |
|
994 | 994 | b'validvalues': {b'bookmarks', b'parents', b'phase', b'revision'}, |
|
995 | 995 | }, |
|
996 | 996 | }, |
|
997 | 997 | permission=b'pull', |
|
998 | 998 | ) |
|
999 | 999 | def changesetdata(repo, proto, revisions, fields): |
|
1000 | 1000 | # TODO look for unknown fields and abort when they can't be serviced. |
|
1001 | 1001 | # This could probably be validated by dispatcher using validvalues. |
|
1002 | 1002 | |
|
1003 | 1003 | cl = repo.changelog |
|
1004 | 1004 | outgoing = resolvenodes(repo, revisions) |
|
1005 | 1005 | publishing = repo.publishing() |
|
1006 | 1006 | |
|
1007 | 1007 | if outgoing: |
|
1008 | 1008 | repo.hook(b'preoutgoing', throw=True, source=b'serve') |
|
1009 | 1009 | |
|
1010 | 1010 | yield { |
|
1011 | 1011 | b'totalitems': len(outgoing), |
|
1012 | 1012 | } |
|
1013 | 1013 | |
|
1014 | 1014 | # The phases of nodes already transferred to the client may have changed |
|
1015 | 1015 | # since the client last requested data. We send phase-only records |
|
1016 | 1016 | # for these revisions, if requested. |
|
1017 | 1017 | # TODO actually do this. We'll probably want to emit phase heads |
|
1018 | 1018 | # in the ancestry set of the outgoing revisions. This will ensure |
|
1019 | 1019 | # that phase updates within that set are seen. |
|
1020 | 1020 | if b'phase' in fields: |
|
1021 | 1021 | pass |
|
1022 | 1022 | |
|
1023 | 1023 | nodebookmarks = {} |
|
1024 | 1024 | for mark, node in repo._bookmarks.items(): |
|
1025 | 1025 | nodebookmarks.setdefault(node, set()).add(mark) |
|
1026 | 1026 | |
|
1027 | 1027 | # It is already topologically sorted by revision number. |
|
1028 | 1028 | for node in outgoing: |
|
1029 | 1029 | d = { |
|
1030 | 1030 | b'node': node, |
|
1031 | 1031 | } |
|
1032 | 1032 | |
|
1033 | 1033 | if b'parents' in fields: |
|
1034 | 1034 | d[b'parents'] = cl.parents(node) |
|
1035 | 1035 | |
|
1036 | 1036 | if b'phase' in fields: |
|
1037 | 1037 | if publishing: |
|
1038 | 1038 | d[b'phase'] = b'public' |
|
1039 | 1039 | else: |
|
1040 | 1040 | ctx = repo[node] |
|
1041 | 1041 | d[b'phase'] = ctx.phasestr() |
|
1042 | 1042 | |
|
1043 | 1043 | if b'bookmarks' in fields and node in nodebookmarks: |
|
1044 | 1044 | d[b'bookmarks'] = sorted(nodebookmarks[node]) |
|
1045 | 1045 | del nodebookmarks[node] |
|
1046 | 1046 | |
|
1047 | 1047 | followingmeta = [] |
|
1048 | 1048 | followingdata = [] |
|
1049 | 1049 | |
|
1050 | 1050 | if b'revision' in fields: |
|
1051 | 1051 | revisiondata = cl.revision(node) |
|
1052 | 1052 | followingmeta.append((b'revision', len(revisiondata))) |
|
1053 | 1053 | followingdata.append(revisiondata) |
|
1054 | 1054 | |
|
1055 | 1055 | # TODO make it possible for extensions to wrap a function or register |
|
1056 | 1056 | # a handler to service custom fields. |
|
1057 | 1057 | |
|
1058 | 1058 | if followingmeta: |
|
1059 | 1059 | d[b'fieldsfollowing'] = followingmeta |
|
1060 | 1060 | |
|
1061 | 1061 | yield d |
|
1062 | 1062 | |
|
1063 | 1063 | for extra in followingdata: |
|
1064 | 1064 | yield extra |
|
1065 | 1065 | |
|
1066 | 1066 | # If requested, send bookmarks from nodes that didn't have revision |
|
1067 | 1067 | # data sent so receiver is aware of any bookmark updates. |
|
1068 | 1068 | if b'bookmarks' in fields: |
|
1069 | 1069 | for node, marks in sorted(pycompat.iteritems(nodebookmarks)): |
|
1070 | 1070 | yield { |
|
1071 | 1071 | b'node': node, |
|
1072 | 1072 | b'bookmarks': sorted(marks), |
|
1073 | 1073 | } |
|
1074 | 1074 | |
|
1075 | 1075 | |
|
1076 | 1076 | class FileAccessError(Exception): |
|
1077 | 1077 | """Represents an error accessing a specific file.""" |
|
1078 | 1078 | |
|
1079 | 1079 | def __init__(self, path, msg, args): |
|
1080 | 1080 | self.path = path |
|
1081 | 1081 | self.msg = msg |
|
1082 | 1082 | self.args = args |
|
1083 | 1083 | |
|
1084 | 1084 | |
|
1085 | 1085 | def getfilestore(repo, proto, path): |
|
1086 | 1086 | """Obtain a file storage object for use with wire protocol. |
|
1087 | 1087 | |
|
1088 | 1088 | Exists as a standalone function so extensions can monkeypatch to add |
|
1089 | 1089 | access control. |
|
1090 | 1090 | """ |
|
1091 | 1091 | # This seems to work even if the file doesn't exist. So catch |
|
1092 | 1092 | # "empty" files and return an error. |
|
1093 | 1093 | fl = repo.file(path) |
|
1094 | 1094 | |
|
1095 | 1095 | if not len(fl): |
|
1096 | 1096 | raise FileAccessError(path, b'unknown file: %s', (path,)) |
|
1097 | 1097 | |
|
1098 | 1098 | return fl |
|
1099 | 1099 | |
|
1100 | 1100 | |
|
1101 | 1101 | def emitfilerevisions(repo, path, revisions, linknodes, fields): |
|
1102 | 1102 | for revision in revisions: |
|
1103 | 1103 | d = { |
|
1104 | 1104 | b'node': revision.node, |
|
1105 | 1105 | } |
|
1106 | 1106 | |
|
1107 | 1107 | if b'parents' in fields: |
|
1108 | 1108 | d[b'parents'] = [revision.p1node, revision.p2node] |
|
1109 | 1109 | |
|
1110 | 1110 | if b'linknode' in fields: |
|
1111 | 1111 | d[b'linknode'] = linknodes[revision.node] |
|
1112 | 1112 | |
|
1113 | 1113 | followingmeta = [] |
|
1114 | 1114 | followingdata = [] |
|
1115 | 1115 | |
|
1116 | 1116 | if b'revision' in fields: |
|
1117 | 1117 | if revision.revision is not None: |
|
1118 | 1118 | followingmeta.append((b'revision', len(revision.revision))) |
|
1119 | 1119 | followingdata.append(revision.revision) |
|
1120 | 1120 | else: |
|
1121 | 1121 | d[b'deltabasenode'] = revision.basenode |
|
1122 | 1122 | followingmeta.append((b'delta', len(revision.delta))) |
|
1123 | 1123 | followingdata.append(revision.delta) |
|
1124 | 1124 | |
|
1125 | 1125 | if followingmeta: |
|
1126 | 1126 | d[b'fieldsfollowing'] = followingmeta |
|
1127 | 1127 | |
|
1128 | 1128 | yield d |
|
1129 | 1129 | |
|
1130 | 1130 | for extra in followingdata: |
|
1131 | 1131 | yield extra |
|
1132 | 1132 | |
|
1133 | 1133 | |
|
1134 | 1134 | def makefilematcher(repo, pathfilter): |
|
1135 | 1135 | """Construct a matcher from a path filter dict.""" |
|
1136 | 1136 | |
|
1137 | 1137 | # Validate values. |
|
1138 | 1138 | if pathfilter: |
|
1139 | 1139 | for key in (b'include', b'exclude'): |
|
1140 | 1140 | for pattern in pathfilter.get(key, []): |
|
1141 | 1141 | if not pattern.startswith((b'path:', b'rootfilesin:')): |
|
1142 | 1142 | raise error.WireprotoCommandError( |
|
1143 | 1143 | b'%s pattern must begin with `path:` or `rootfilesin:`; ' |
|
1144 | 1144 | b'got %s', |
|
1145 | 1145 | (key, pattern), |
|
1146 | 1146 | ) |
|
1147 | 1147 | |
|
1148 | 1148 | if pathfilter: |
|
1149 | 1149 | matcher = matchmod.match( |
|
1150 | 1150 | repo.root, |
|
1151 | 1151 | b'', |
|
1152 | 1152 | include=pathfilter.get(b'include', []), |
|
1153 | 1153 | exclude=pathfilter.get(b'exclude', []), |
|
1154 | 1154 | ) |
|
1155 | 1155 | else: |
|
1156 | 1156 | matcher = matchmod.match(repo.root, b'') |
|
1157 | 1157 | |
|
1158 | 1158 | # Requested patterns could include files not in the local store. So |
|
1159 | 1159 | # filter those out. |
|
1160 | 1160 | return repo.narrowmatch(matcher) |
|
1161 | 1161 | |
|
1162 | 1162 | |
|
1163 | 1163 | @wireprotocommand( |
|
1164 | 1164 | b'filedata', |
|
1165 | 1165 | args={ |
|
1166 | 1166 | b'haveparents': { |
|
1167 | 1167 | b'type': b'bool', |
|
1168 | 1168 | b'default': lambda: False, |
|
1169 | 1169 | b'example': True, |
|
1170 | 1170 | }, |
|
1171 | 1171 | b'nodes': { |
|
1172 | 1172 | b'type': b'list', |
|
1173 | 1173 | b'example': [b'0123456...'], |
|
1174 | 1174 | }, |
|
1175 | 1175 | b'fields': { |
|
1176 | 1176 | b'type': b'set', |
|
1177 | 1177 | b'default': set, |
|
1178 | 1178 | b'example': {b'parents', b'revision'}, |
|
1179 | 1179 | b'validvalues': {b'parents', b'revision', b'linknode'}, |
|
1180 | 1180 | }, |
|
1181 | 1181 | b'path': { |
|
1182 | 1182 | b'type': b'bytes', |
|
1183 | 1183 | b'example': b'foo.txt', |
|
1184 | 1184 | }, |
|
1185 | 1185 | }, |
|
1186 | 1186 | permission=b'pull', |
|
1187 | 1187 | # TODO censoring a file revision won't invalidate the cache. |
|
1188 | 1188 | # Figure out a way to take censoring into account when deriving |
|
1189 | 1189 | # the cache key. |
|
1190 | 1190 | cachekeyfn=makecommandcachekeyfn(b'filedata', 1, allargs=True), |
|
1191 | 1191 | ) |
|
1192 | 1192 | def filedata(repo, proto, haveparents, nodes, fields, path): |
|
1193 | 1193 | # TODO this API allows access to file revisions that are attached to |
|
1194 | 1194 | # secret changesets. filesdata does not have this problem. Maybe this |
|
1195 | 1195 | # API should be deleted? |
|
1196 | 1196 | |
|
1197 | 1197 | try: |
|
1198 | 1198 | # Extensions may wish to access the protocol handler. |
|
1199 | 1199 | store = getfilestore(repo, proto, path) |
|
1200 | 1200 | except FileAccessError as e: |
|
1201 | 1201 | raise error.WireprotoCommandError(e.msg, e.args) |
|
1202 | 1202 | |
|
1203 | 1203 | clnode = repo.changelog.node |
|
1204 | 1204 | linknodes = {} |
|
1205 | 1205 | |
|
1206 | 1206 | # Validate requested nodes. |
|
1207 | 1207 | for node in nodes: |
|
1208 | 1208 | try: |
|
1209 | 1209 | store.rev(node) |
|
1210 | 1210 | except error.LookupError: |
|
1211 | 1211 | raise error.WireprotoCommandError( |
|
1212 | 1212 | b'unknown file node: %s', (hex(node),) |
|
1213 | 1213 | ) |
|
1214 | 1214 | |
|
1215 | 1215 | # TODO by creating the filectx against a specific file revision |
|
1216 | 1216 | # instead of changeset, linkrev() is always used. This is wrong for |
|
1217 | 1217 | # cases where linkrev() may refer to a hidden changeset. But since this |
|
1218 | 1218 | # API doesn't know anything about changesets, we're not sure how to |
|
1219 | 1219 | # disambiguate the linknode. Perhaps we should delete this API? |
|
1220 | 1220 | fctx = repo.filectx(path, fileid=node) |
|
1221 | 1221 | linknodes[node] = clnode(fctx.introrev()) |
|
1222 | 1222 | |
|
1223 | 1223 | revisions = store.emitrevisions( |
|
1224 | 1224 | nodes, |
|
1225 | 1225 | revisiondata=b'revision' in fields, |
|
1226 | 1226 | assumehaveparentrevisions=haveparents, |
|
1227 | 1227 | ) |
|
1228 | 1228 | |
|
1229 | 1229 | yield { |
|
1230 | 1230 | b'totalitems': len(nodes), |
|
1231 | 1231 | } |
|
1232 | 1232 | |
|
1233 | 1233 | for o in emitfilerevisions(repo, path, revisions, linknodes, fields): |
|
1234 | 1234 | yield o |
|
1235 | 1235 | |
|
1236 | 1236 | |
|
1237 | 1237 | def filesdatacapabilities(repo, proto): |
|
1238 | 1238 | batchsize = repo.ui.configint( |
|
1239 | 1239 | b'experimental', b'server.filesdata.recommended-batch-size' |
|
1240 | 1240 | ) |
|
1241 | 1241 | return { |
|
1242 | 1242 | b'recommendedbatchsize': batchsize, |
|
1243 | 1243 | } |
|
1244 | 1244 | |
|
1245 | 1245 | |
|
1246 | 1246 | @wireprotocommand( |
|
1247 | 1247 | b'filesdata', |
|
1248 | 1248 | args={ |
|
1249 | 1249 | b'haveparents': { |
|
1250 | 1250 | b'type': b'bool', |
|
1251 | 1251 | b'default': lambda: False, |
|
1252 | 1252 | b'example': True, |
|
1253 | 1253 | }, |
|
1254 | 1254 | b'fields': { |
|
1255 | 1255 | b'type': b'set', |
|
1256 | 1256 | b'default': set, |
|
1257 | 1257 | b'example': {b'parents', b'revision'}, |
|
1258 | 1258 | b'validvalues': { |
|
1259 | 1259 | b'firstchangeset', |
|
1260 | 1260 | b'linknode', |
|
1261 | 1261 | b'parents', |
|
1262 | 1262 | b'revision', |
|
1263 | 1263 | }, |
|
1264 | 1264 | }, |
|
1265 | 1265 | b'pathfilter': { |
|
1266 | 1266 | b'type': b'dict', |
|
1267 | 1267 | b'default': lambda: None, |
|
1268 | 1268 | b'example': {b'include': [b'path:tests']}, |
|
1269 | 1269 | }, |
|
1270 | 1270 | b'revisions': { |
|
1271 | 1271 | b'type': b'list', |
|
1272 | 1272 | b'example': [ |
|
1273 | 1273 | { |
|
1274 | 1274 | b'type': b'changesetexplicit', |
|
1275 | 1275 | b'nodes': [b'abcdef...'], |
|
1276 | 1276 | } |
|
1277 | 1277 | ], |
|
1278 | 1278 | }, |
|
1279 | 1279 | }, |
|
1280 | 1280 | permission=b'pull', |
|
1281 | 1281 | # TODO censoring a file revision won't invalidate the cache. |
|
1282 | 1282 | # Figure out a way to take censoring into account when deriving |
|
1283 | 1283 | # the cache key. |
|
1284 | 1284 | cachekeyfn=makecommandcachekeyfn(b'filesdata', 1, allargs=True), |
|
1285 | 1285 | extracapabilitiesfn=filesdatacapabilities, |
|
1286 | 1286 | ) |
|
1287 | 1287 | def filesdata(repo, proto, haveparents, fields, pathfilter, revisions): |
|
1288 | 1288 | # TODO This should operate on a repo that exposes obsolete changesets. There |
|
1289 | 1289 | # is a race between a client making a push that obsoletes a changeset and |
|
1290 | 1290 | # another client fetching files data for that changeset. If a client has a |
|
1291 | 1291 | # changeset, it should probably be allowed to access files data for that |
|
1292 | 1292 | # changeset. |
|
1293 | 1293 | |
|
1294 | 1294 | outgoing = resolvenodes(repo, revisions) |
|
1295 | 1295 | filematcher = makefilematcher(repo, pathfilter) |
|
1296 | 1296 | |
|
1297 | 1297 | # path -> {fnode: linknode} |
|
1298 | 1298 | fnodes = collections.defaultdict(dict) |
|
1299 | 1299 | |
|
1300 | 1300 | # We collect the set of relevant file revisions by iterating the changeset |
|
1301 | 1301 | # revisions and either walking the set of files recorded in the changeset |
|
1302 | 1302 | # or by walking the manifest at that revision. There is probably room for a |
|
1303 | 1303 | # storage-level API to request this data, as it can be expensive to compute |
|
1304 | 1304 | # and would benefit from caching or alternate storage from what revlogs |
|
1305 | 1305 | # provide. |
|
1306 | 1306 | for node in outgoing: |
|
1307 | 1307 | ctx = repo[node] |
|
1308 | 1308 | mctx = ctx.manifestctx() |
|
1309 | 1309 | md = mctx.read() |
|
1310 | 1310 | |
|
1311 | 1311 | if haveparents: |
|
1312 | 1312 | checkpaths = ctx.files() |
|
1313 | 1313 | else: |
|
1314 | 1314 | checkpaths = md.keys() |
|
1315 | 1315 | |
|
1316 | 1316 | for path in checkpaths: |
|
1317 | 1317 | fnode = md[path] |
|
1318 | 1318 | |
|
1319 | 1319 | if path in fnodes and fnode in fnodes[path]: |
|
1320 | 1320 | continue |
|
1321 | 1321 | |
|
1322 | 1322 | if not filematcher(path): |
|
1323 | 1323 | continue |
|
1324 | 1324 | |
|
1325 | 1325 | fnodes[path].setdefault(fnode, node) |
|
1326 | 1326 | |
|
1327 | 1327 | yield { |
|
1328 | 1328 | b'totalpaths': len(fnodes), |
|
1329 | 1329 | b'totalitems': sum(len(v) for v in fnodes.values()), |
|
1330 | 1330 | } |
|
1331 | 1331 | |
|
1332 | 1332 | for path, filenodes in sorted(fnodes.items()): |
|
1333 | 1333 | try: |
|
1334 | 1334 | store = getfilestore(repo, proto, path) |
|
1335 | 1335 | except FileAccessError as e: |
|
1336 | 1336 | raise error.WireprotoCommandError(e.msg, e.args) |
|
1337 | 1337 | |
|
1338 | 1338 | yield { |
|
1339 | 1339 | b'path': path, |
|
1340 | 1340 | b'totalitems': len(filenodes), |
|
1341 | 1341 | } |
|
1342 | 1342 | |
|
1343 | 1343 | revisions = store.emitrevisions( |
|
1344 | 1344 | filenodes.keys(), |
|
1345 | 1345 | revisiondata=b'revision' in fields, |
|
1346 | 1346 | assumehaveparentrevisions=haveparents, |
|
1347 | 1347 | ) |
|
1348 | 1348 | |
|
1349 | 1349 | for o in emitfilerevisions(repo, path, revisions, filenodes, fields): |
|
1350 | 1350 | yield o |
|
1351 | 1351 | |
|
1352 | 1352 | |
|
1353 | 1353 | @wireprotocommand( |
|
1354 | 1354 | b'heads', |
|
1355 | 1355 | args={ |
|
1356 | 1356 | b'publiconly': { |
|
1357 | 1357 | b'type': b'bool', |
|
1358 | 1358 | b'default': lambda: False, |
|
1359 | 1359 | b'example': False, |
|
1360 | 1360 | }, |
|
1361 | 1361 | }, |
|
1362 | 1362 | permission=b'pull', |
|
1363 | 1363 | ) |
|
1364 | 1364 | def headsv2(repo, proto, publiconly): |
|
1365 | 1365 | if publiconly: |
|
1366 | 1366 | repo = repo.filtered(b'immutable') |
|
1367 | 1367 | |
|
1368 | 1368 | yield repo.heads() |
|
1369 | 1369 | |
|
1370 | 1370 | |
|
1371 | 1371 | @wireprotocommand( |
|
1372 | 1372 | b'known', |
|
1373 | 1373 | args={ |
|
1374 | 1374 | b'nodes': { |
|
1375 | 1375 | b'type': b'list', |
|
1376 | 1376 | b'default': list, |
|
1377 | 1377 | b'example': [b'deadbeef'], |
|
1378 | 1378 | }, |
|
1379 | 1379 | }, |
|
1380 | 1380 | permission=b'pull', |
|
1381 | 1381 | ) |
|
1382 | 1382 | def knownv2(repo, proto, nodes): |
|
1383 | 1383 | result = b''.join(b'1' if n else b'0' for n in repo.known(nodes)) |
|
1384 | 1384 | yield result |
|
1385 | 1385 | |
|
1386 | 1386 | |
|
1387 | 1387 | @wireprotocommand( |
|
1388 | 1388 | b'listkeys', |
|
1389 | 1389 | args={ |
|
1390 | 1390 | b'namespace': { |
|
1391 | 1391 | b'type': b'bytes', |
|
1392 | 1392 | b'example': b'ns', |
|
1393 | 1393 | }, |
|
1394 | 1394 | }, |
|
1395 | 1395 | permission=b'pull', |
|
1396 | 1396 | ) |
|
1397 | 1397 | def listkeysv2(repo, proto, namespace): |
|
1398 | 1398 | keys = repo.listkeys(encoding.tolocal(namespace)) |
|
1399 | 1399 | keys = { |
|
1400 | 1400 | encoding.fromlocal(k): encoding.fromlocal(v) |
|
1401 | 1401 | for k, v in pycompat.iteritems(keys) |
|
1402 | 1402 | } |
|
1403 | 1403 | |
|
1404 | 1404 | yield keys |
|
1405 | 1405 | |
|
1406 | 1406 | |
|
1407 | 1407 | @wireprotocommand( |
|
1408 | 1408 | b'lookup', |
|
1409 | 1409 | args={ |
|
1410 | 1410 | b'key': { |
|
1411 | 1411 | b'type': b'bytes', |
|
1412 | 1412 | b'example': b'foo', |
|
1413 | 1413 | }, |
|
1414 | 1414 | }, |
|
1415 | 1415 | permission=b'pull', |
|
1416 | 1416 | ) |
|
1417 | 1417 | def lookupv2(repo, proto, key): |
|
1418 | 1418 | key = encoding.tolocal(key) |
|
1419 | 1419 | |
|
1420 | 1420 | # TODO handle exception. |
|
1421 | 1421 | node = repo.lookup(key) |
|
1422 | 1422 | |
|
1423 | 1423 | yield node |
|
1424 | 1424 | |
|
1425 | 1425 | |
|
1426 | 1426 | def manifestdatacapabilities(repo, proto): |
|
1427 | 1427 | batchsize = repo.ui.configint( |
|
1428 | 1428 | b'experimental', b'server.manifestdata.recommended-batch-size' |
|
1429 | 1429 | ) |
|
1430 | 1430 | |
|
1431 | 1431 | return { |
|
1432 | 1432 | b'recommendedbatchsize': batchsize, |
|
1433 | 1433 | } |
|
1434 | 1434 | |
|
1435 | 1435 | |
|
1436 | 1436 | @wireprotocommand( |
|
1437 | 1437 | b'manifestdata', |
|
1438 | 1438 | args={ |
|
1439 | 1439 | b'nodes': { |
|
1440 | 1440 | b'type': b'list', |
|
1441 | 1441 | b'example': [b'0123456...'], |
|
1442 | 1442 | }, |
|
1443 | 1443 | b'haveparents': { |
|
1444 | 1444 | b'type': b'bool', |
|
1445 | 1445 | b'default': lambda: False, |
|
1446 | 1446 | b'example': True, |
|
1447 | 1447 | }, |
|
1448 | 1448 | b'fields': { |
|
1449 | 1449 | b'type': b'set', |
|
1450 | 1450 | b'default': set, |
|
1451 | 1451 | b'example': {b'parents', b'revision'}, |
|
1452 | 1452 | b'validvalues': {b'parents', b'revision'}, |
|
1453 | 1453 | }, |
|
1454 | 1454 | b'tree': { |
|
1455 | 1455 | b'type': b'bytes', |
|
1456 | 1456 | b'example': b'', |
|
1457 | 1457 | }, |
|
1458 | 1458 | }, |
|
1459 | 1459 | permission=b'pull', |
|
1460 | 1460 | cachekeyfn=makecommandcachekeyfn(b'manifestdata', 1, allargs=True), |
|
1461 | 1461 | extracapabilitiesfn=manifestdatacapabilities, |
|
1462 | 1462 | ) |
|
1463 | 1463 | def manifestdata(repo, proto, haveparents, nodes, fields, tree): |
|
1464 | 1464 | store = repo.manifestlog.getstorage(tree) |
|
1465 | 1465 | |
|
1466 | 1466 | # Validate the node is known and abort on unknown revisions. |
|
1467 | 1467 | for node in nodes: |
|
1468 | 1468 | try: |
|
1469 | 1469 | store.rev(node) |
|
1470 | 1470 | except error.LookupError: |
|
1471 | 1471 | raise error.WireprotoCommandError(b'unknown node: %s', (node,)) |
|
1472 | 1472 | |
|
1473 | 1473 | revisions = store.emitrevisions( |
|
1474 | 1474 | nodes, |
|
1475 | 1475 | revisiondata=b'revision' in fields, |
|
1476 | 1476 | assumehaveparentrevisions=haveparents, |
|
1477 | 1477 | ) |
|
1478 | 1478 | |
|
1479 | 1479 | yield { |
|
1480 | 1480 | b'totalitems': len(nodes), |
|
1481 | 1481 | } |
|
1482 | 1482 | |
|
1483 | 1483 | for revision in revisions: |
|
1484 | 1484 | d = { |
|
1485 | 1485 | b'node': revision.node, |
|
1486 | 1486 | } |
|
1487 | 1487 | |
|
1488 | 1488 | if b'parents' in fields: |
|
1489 | 1489 | d[b'parents'] = [revision.p1node, revision.p2node] |
|
1490 | 1490 | |
|
1491 | 1491 | followingmeta = [] |
|
1492 | 1492 | followingdata = [] |
|
1493 | 1493 | |
|
1494 | 1494 | if b'revision' in fields: |
|
1495 | 1495 | if revision.revision is not None: |
|
1496 | 1496 | followingmeta.append((b'revision', len(revision.revision))) |
|
1497 | 1497 | followingdata.append(revision.revision) |
|
1498 | 1498 | else: |
|
1499 | 1499 | d[b'deltabasenode'] = revision.basenode |
|
1500 | 1500 | followingmeta.append((b'delta', len(revision.delta))) |
|
1501 | 1501 | followingdata.append(revision.delta) |
|
1502 | 1502 | |
|
1503 | 1503 | if followingmeta: |
|
1504 | 1504 | d[b'fieldsfollowing'] = followingmeta |
|
1505 | 1505 | |
|
1506 | 1506 | yield d |
|
1507 | 1507 | |
|
1508 | 1508 | for extra in followingdata: |
|
1509 | 1509 | yield extra |
|
1510 | 1510 | |
|
1511 | 1511 | |
|
1512 | 1512 | @wireprotocommand( |
|
1513 | 1513 | b'pushkey', |
|
1514 | 1514 | args={ |
|
1515 | 1515 | b'namespace': { |
|
1516 | 1516 | b'type': b'bytes', |
|
1517 | 1517 | b'example': b'ns', |
|
1518 | 1518 | }, |
|
1519 | 1519 | b'key': { |
|
1520 | 1520 | b'type': b'bytes', |
|
1521 | 1521 | b'example': b'key', |
|
1522 | 1522 | }, |
|
1523 | 1523 | b'old': { |
|
1524 | 1524 | b'type': b'bytes', |
|
1525 | 1525 | b'example': b'old', |
|
1526 | 1526 | }, |
|
1527 | 1527 | b'new': { |
|
1528 | 1528 | b'type': b'bytes', |
|
1529 | 1529 | b'example': b'new', |
|
1530 | 1530 | }, |
|
1531 | 1531 | }, |
|
1532 | 1532 | permission=b'push', |
|
1533 | 1533 | ) |
|
1534 | 1534 | def pushkeyv2(repo, proto, namespace, key, old, new): |
|
1535 | 1535 | # TODO handle ui output redirection |
|
1536 | 1536 | yield repo.pushkey( |
|
1537 | 1537 | encoding.tolocal(namespace), |
|
1538 | 1538 | encoding.tolocal(key), |
|
1539 | 1539 | encoding.tolocal(old), |
|
1540 | 1540 | encoding.tolocal(new), |
|
1541 | 1541 | ) |
|
1542 | 1542 | |
|
1543 | 1543 | |
|
1544 | 1544 | @wireprotocommand( |
|
1545 | 1545 | b'rawstorefiledata', |
|
1546 | 1546 | args={ |
|
1547 | 1547 | b'files': { |
|
1548 | 1548 | b'type': b'list', |
|
1549 | 1549 | b'example': [b'changelog', b'manifestlog'], |
|
1550 | 1550 | }, |
|
1551 | 1551 | b'pathfilter': { |
|
1552 | 1552 | b'type': b'list', |
|
1553 | 1553 | b'default': lambda: None, |
|
1554 | 1554 | b'example': {b'include': [b'path:tests']}, |
|
1555 | 1555 | }, |
|
1556 | 1556 | }, |
|
1557 | 1557 | permission=b'pull', |
|
1558 | 1558 | ) |
|
1559 | 1559 | def rawstorefiledata(repo, proto, files, pathfilter): |
|
1560 | 1560 | if not streamclone.allowservergeneration(repo): |
|
1561 | 1561 | raise error.WireprotoCommandError(b'stream clone is disabled') |
|
1562 | 1562 | |
|
1563 | 1563 | # TODO support dynamically advertising what store files "sets" are |
|
1564 | 1564 | # available. For now, we support changelog, manifestlog, and files. |
|
1565 | 1565 | files = set(files) |
|
1566 | 1566 | allowedfiles = {b'changelog', b'manifestlog'} |
|
1567 | 1567 | |
|
1568 | 1568 | unsupported = files - allowedfiles |
|
1569 | 1569 | if unsupported: |
|
1570 | 1570 | raise error.WireprotoCommandError( |
|
1571 | 1571 | b'unknown file type: %s', (b', '.join(sorted(unsupported)),) |
|
1572 | 1572 | ) |
|
1573 | 1573 | |
|
1574 | 1574 | with repo.lock(): |
|
1575 | 1575 | topfiles = list(repo.store.topfiles()) |
|
1576 | 1576 | |
|
1577 | 1577 | sendfiles = [] |
|
1578 | 1578 | totalsize = 0 |
|
1579 | 1579 | |
|
1580 | 1580 | # TODO this is a bunch of storage layer interface abstractions because |
|
1581 | 1581 | # it assumes revlogs. |
|
1582 |
for rl_type, name, |
|
|
1582 | for rl_type, name, size in topfiles: | |
|
1583 | 1583 | # XXX use the `rl_type` for that |
|
1584 | 1584 | if b'changelog' in files and name.startswith(b'00changelog'): |
|
1585 | 1585 | pass |
|
1586 | 1586 | elif b'manifestlog' in files and name.startswith(b'00manifest'): |
|
1587 | 1587 | pass |
|
1588 | 1588 | else: |
|
1589 | 1589 | continue |
|
1590 | 1590 | |
|
1591 | 1591 | sendfiles.append((b'store', name, size)) |
|
1592 | 1592 | totalsize += size |
|
1593 | 1593 | |
|
1594 | 1594 | yield { |
|
1595 | 1595 | b'filecount': len(sendfiles), |
|
1596 | 1596 | b'totalsize': totalsize, |
|
1597 | 1597 | } |
|
1598 | 1598 | |
|
1599 | 1599 | for location, name, size in sendfiles: |
|
1600 | 1600 | yield { |
|
1601 | 1601 | b'location': location, |
|
1602 | 1602 | b'path': name, |
|
1603 | 1603 | b'size': size, |
|
1604 | 1604 | } |
|
1605 | 1605 | |
|
1606 | 1606 | # We have to use a closure for this to ensure the context manager is |
|
1607 | 1607 | # closed only after sending the final chunk. |
|
1608 | 1608 | def getfiledata(): |
|
1609 | 1609 | with repo.svfs(name, b'rb', auditpath=False) as fh: |
|
1610 | 1610 | for chunk in util.filechunkiter(fh, limit=size): |
|
1611 | 1611 | yield chunk |
|
1612 | 1612 | |
|
1613 | 1613 | yield wireprototypes.indefinitebytestringresponse(getfiledata()) |
@@ -1,741 +1,745 b'' | |||
|
1 | 1 | # simplestorerepo.py - Extension that swaps in alternate repository storage. |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2018 Gregory Szorc <gregory.szorc@gmail.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 | # To use this with the test suite: |
|
9 | 9 | # |
|
10 | 10 | # $ HGREPOFEATURES="simplestore" ./run-tests.py \ |
|
11 | 11 | # --extra-config-opt extensions.simplestore=`pwd`/simplestorerepo.py |
|
12 | 12 | |
|
13 | 13 | from __future__ import absolute_import |
|
14 | 14 | |
|
15 | 15 | import stat |
|
16 | 16 | |
|
17 | 17 | from mercurial.i18n import _ |
|
18 | 18 | from mercurial.node import ( |
|
19 | 19 | bin, |
|
20 | 20 | hex, |
|
21 | 21 | nullrev, |
|
22 | 22 | ) |
|
23 | 23 | from mercurial.thirdparty import attr |
|
24 | 24 | from mercurial import ( |
|
25 | 25 | ancestor, |
|
26 | 26 | bundlerepo, |
|
27 | 27 | error, |
|
28 | 28 | extensions, |
|
29 | 29 | localrepo, |
|
30 | 30 | mdiff, |
|
31 | 31 | pycompat, |
|
32 | 32 | revlog, |
|
33 | 33 | store, |
|
34 | 34 | verify, |
|
35 | 35 | ) |
|
36 | 36 | from mercurial.interfaces import ( |
|
37 | 37 | repository, |
|
38 | 38 | util as interfaceutil, |
|
39 | 39 | ) |
|
40 | 40 | from mercurial.utils import ( |
|
41 | 41 | cborutil, |
|
42 | 42 | storageutil, |
|
43 | 43 | ) |
|
44 | 44 | from mercurial.revlogutils import flagutil |
|
45 | 45 | |
|
46 | 46 | # Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for |
|
47 | 47 | # extensions which SHIP WITH MERCURIAL. Non-mainline extensions should |
|
48 | 48 | # be specifying the version(s) of Mercurial they are tested with, or |
|
49 | 49 | # leave the attribute unspecified. |
|
50 | 50 | testedwith = b'ships-with-hg-core' |
|
51 | 51 | |
|
52 | 52 | REQUIREMENT = b'testonly-simplestore' |
|
53 | 53 | |
|
54 | 54 | |
|
55 | 55 | def validatenode(node): |
|
56 | 56 | if isinstance(node, int): |
|
57 | 57 | raise ValueError('expected node; got int') |
|
58 | 58 | |
|
59 | 59 | if len(node) != 20: |
|
60 | 60 | raise ValueError('expected 20 byte node') |
|
61 | 61 | |
|
62 | 62 | |
|
63 | 63 | def validaterev(rev): |
|
64 | 64 | if not isinstance(rev, int): |
|
65 | 65 | raise ValueError('expected int') |
|
66 | 66 | |
|
67 | 67 | |
|
68 | 68 | class simplestoreerror(error.StorageError): |
|
69 | 69 | pass |
|
70 | 70 | |
|
71 | 71 | |
|
72 | 72 | @interfaceutil.implementer(repository.irevisiondelta) |
|
73 | 73 | @attr.s(slots=True) |
|
74 | 74 | class simplestorerevisiondelta(object): |
|
75 | 75 | node = attr.ib() |
|
76 | 76 | p1node = attr.ib() |
|
77 | 77 | p2node = attr.ib() |
|
78 | 78 | basenode = attr.ib() |
|
79 | 79 | flags = attr.ib() |
|
80 | 80 | baserevisionsize = attr.ib() |
|
81 | 81 | revision = attr.ib() |
|
82 | 82 | delta = attr.ib() |
|
83 | 83 | linknode = attr.ib(default=None) |
|
84 | 84 | |
|
85 | 85 | |
|
86 | 86 | @interfaceutil.implementer(repository.iverifyproblem) |
|
87 | 87 | @attr.s(frozen=True) |
|
88 | 88 | class simplefilestoreproblem(object): |
|
89 | 89 | warning = attr.ib(default=None) |
|
90 | 90 | error = attr.ib(default=None) |
|
91 | 91 | node = attr.ib(default=None) |
|
92 | 92 | |
|
93 | 93 | |
|
94 | 94 | @interfaceutil.implementer(repository.ifilestorage) |
|
95 | 95 | class filestorage(object): |
|
96 | 96 | """Implements storage for a tracked path. |
|
97 | 97 | |
|
98 | 98 | Data is stored in the VFS in a directory corresponding to the tracked |
|
99 | 99 | path. |
|
100 | 100 | |
|
101 | 101 | Index data is stored in an ``index`` file using CBOR. |
|
102 | 102 | |
|
103 | 103 | Fulltext data is stored in files having names of the node. |
|
104 | 104 | """ |
|
105 | 105 | |
|
106 | 106 | _flagserrorclass = simplestoreerror |
|
107 | 107 | |
|
108 | 108 | def __init__(self, repo, svfs, path): |
|
109 | 109 | self.nullid = repo.nullid |
|
110 | 110 | self._repo = repo |
|
111 | 111 | self._svfs = svfs |
|
112 | 112 | self._path = path |
|
113 | 113 | |
|
114 | 114 | self._storepath = b'/'.join([b'data', path]) |
|
115 | 115 | self._indexpath = b'/'.join([self._storepath, b'index']) |
|
116 | 116 | |
|
117 | 117 | indexdata = self._svfs.tryread(self._indexpath) |
|
118 | 118 | if indexdata: |
|
119 | 119 | indexdata = cborutil.decodeall(indexdata) |
|
120 | 120 | |
|
121 | 121 | self._indexdata = indexdata or [] |
|
122 | 122 | self._indexbynode = {} |
|
123 | 123 | self._indexbyrev = {} |
|
124 | 124 | self._index = [] |
|
125 | 125 | self._refreshindex() |
|
126 | 126 | |
|
127 | 127 | self._flagprocessors = dict(flagutil.flagprocessors) |
|
128 | 128 | |
|
129 | 129 | def _refreshindex(self): |
|
130 | 130 | self._indexbynode.clear() |
|
131 | 131 | self._indexbyrev.clear() |
|
132 | 132 | self._index = [] |
|
133 | 133 | |
|
134 | 134 | for i, entry in enumerate(self._indexdata): |
|
135 | 135 | self._indexbynode[entry[b'node']] = entry |
|
136 | 136 | self._indexbyrev[i] = entry |
|
137 | 137 | |
|
138 | 138 | self._indexbynode[self._repo.nullid] = { |
|
139 | 139 | b'node': self._repo.nullid, |
|
140 | 140 | b'p1': self._repo.nullid, |
|
141 | 141 | b'p2': self._repo.nullid, |
|
142 | 142 | b'linkrev': nullrev, |
|
143 | 143 | b'flags': 0, |
|
144 | 144 | } |
|
145 | 145 | |
|
146 | 146 | self._indexbyrev[nullrev] = { |
|
147 | 147 | b'node': self._repo.nullid, |
|
148 | 148 | b'p1': self._repo.nullid, |
|
149 | 149 | b'p2': self._repo.nullid, |
|
150 | 150 | b'linkrev': nullrev, |
|
151 | 151 | b'flags': 0, |
|
152 | 152 | } |
|
153 | 153 | |
|
154 | 154 | for i, entry in enumerate(self._indexdata): |
|
155 | 155 | p1rev, p2rev = self.parentrevs(self.rev(entry[b'node'])) |
|
156 | 156 | |
|
157 | 157 | # start, length, rawsize, chainbase, linkrev, p1, p2, node |
|
158 | 158 | self._index.append( |
|
159 | 159 | (0, 0, 0, -1, entry[b'linkrev'], p1rev, p2rev, entry[b'node']) |
|
160 | 160 | ) |
|
161 | 161 | |
|
162 | 162 | self._index.append((0, 0, 0, -1, -1, -1, -1, self._repo.nullid)) |
|
163 | 163 | |
|
164 | 164 | def __len__(self): |
|
165 | 165 | return len(self._indexdata) |
|
166 | 166 | |
|
167 | 167 | def __iter__(self): |
|
168 | 168 | return iter(range(len(self))) |
|
169 | 169 | |
|
170 | 170 | def revs(self, start=0, stop=None): |
|
171 | 171 | step = 1 |
|
172 | 172 | if stop is not None: |
|
173 | 173 | if start > stop: |
|
174 | 174 | step = -1 |
|
175 | 175 | |
|
176 | 176 | stop += step |
|
177 | 177 | else: |
|
178 | 178 | stop = len(self) |
|
179 | 179 | |
|
180 | 180 | return range(start, stop, step) |
|
181 | 181 | |
|
182 | 182 | def parents(self, node): |
|
183 | 183 | validatenode(node) |
|
184 | 184 | |
|
185 | 185 | if node not in self._indexbynode: |
|
186 | 186 | raise KeyError('unknown node') |
|
187 | 187 | |
|
188 | 188 | entry = self._indexbynode[node] |
|
189 | 189 | |
|
190 | 190 | return entry[b'p1'], entry[b'p2'] |
|
191 | 191 | |
|
192 | 192 | def parentrevs(self, rev): |
|
193 | 193 | p1, p2 = self.parents(self._indexbyrev[rev][b'node']) |
|
194 | 194 | return self.rev(p1), self.rev(p2) |
|
195 | 195 | |
|
196 | 196 | def rev(self, node): |
|
197 | 197 | validatenode(node) |
|
198 | 198 | |
|
199 | 199 | try: |
|
200 | 200 | self._indexbynode[node] |
|
201 | 201 | except KeyError: |
|
202 | 202 | raise error.LookupError(node, self._indexpath, _('no node')) |
|
203 | 203 | |
|
204 | 204 | for rev, entry in self._indexbyrev.items(): |
|
205 | 205 | if entry[b'node'] == node: |
|
206 | 206 | return rev |
|
207 | 207 | |
|
208 | 208 | raise error.ProgrammingError(b'this should not occur') |
|
209 | 209 | |
|
210 | 210 | def node(self, rev): |
|
211 | 211 | validaterev(rev) |
|
212 | 212 | |
|
213 | 213 | return self._indexbyrev[rev][b'node'] |
|
214 | 214 | |
|
215 | 215 | def hasnode(self, node): |
|
216 | 216 | validatenode(node) |
|
217 | 217 | return node in self._indexbynode |
|
218 | 218 | |
|
219 | 219 | def censorrevision(self, tr, censornode, tombstone=b''): |
|
220 | 220 | raise NotImplementedError('TODO') |
|
221 | 221 | |
|
222 | 222 | def lookup(self, node): |
|
223 | 223 | if isinstance(node, int): |
|
224 | 224 | return self.node(node) |
|
225 | 225 | |
|
226 | 226 | if len(node) == 20: |
|
227 | 227 | self.rev(node) |
|
228 | 228 | return node |
|
229 | 229 | |
|
230 | 230 | try: |
|
231 | 231 | rev = int(node) |
|
232 | 232 | if '%d' % rev != node: |
|
233 | 233 | raise ValueError |
|
234 | 234 | |
|
235 | 235 | if rev < 0: |
|
236 | 236 | rev = len(self) + rev |
|
237 | 237 | if rev < 0 or rev >= len(self): |
|
238 | 238 | raise ValueError |
|
239 | 239 | |
|
240 | 240 | return self.node(rev) |
|
241 | 241 | except (ValueError, OverflowError): |
|
242 | 242 | pass |
|
243 | 243 | |
|
244 | 244 | if len(node) == 40: |
|
245 | 245 | try: |
|
246 | 246 | rawnode = bin(node) |
|
247 | 247 | self.rev(rawnode) |
|
248 | 248 | return rawnode |
|
249 | 249 | except TypeError: |
|
250 | 250 | pass |
|
251 | 251 | |
|
252 | 252 | raise error.LookupError(node, self._path, _('invalid lookup input')) |
|
253 | 253 | |
|
254 | 254 | def linkrev(self, rev): |
|
255 | 255 | validaterev(rev) |
|
256 | 256 | |
|
257 | 257 | return self._indexbyrev[rev][b'linkrev'] |
|
258 | 258 | |
|
259 | 259 | def _flags(self, rev): |
|
260 | 260 | validaterev(rev) |
|
261 | 261 | |
|
262 | 262 | return self._indexbyrev[rev][b'flags'] |
|
263 | 263 | |
|
264 | 264 | def _candelta(self, baserev, rev): |
|
265 | 265 | validaterev(baserev) |
|
266 | 266 | validaterev(rev) |
|
267 | 267 | |
|
268 | 268 | if (self._flags(baserev) & revlog.REVIDX_RAWTEXT_CHANGING_FLAGS) or ( |
|
269 | 269 | self._flags(rev) & revlog.REVIDX_RAWTEXT_CHANGING_FLAGS |
|
270 | 270 | ): |
|
271 | 271 | return False |
|
272 | 272 | |
|
273 | 273 | return True |
|
274 | 274 | |
|
275 | 275 | def checkhash(self, text, node, p1=None, p2=None, rev=None): |
|
276 | 276 | if p1 is None and p2 is None: |
|
277 | 277 | p1, p2 = self.parents(node) |
|
278 | 278 | if node != storageutil.hashrevisionsha1(text, p1, p2): |
|
279 | 279 | raise simplestoreerror( |
|
280 | 280 | _("integrity check failed on %s") % self._path |
|
281 | 281 | ) |
|
282 | 282 | |
|
283 | 283 | def revision(self, nodeorrev, raw=False): |
|
284 | 284 | if isinstance(nodeorrev, int): |
|
285 | 285 | node = self.node(nodeorrev) |
|
286 | 286 | else: |
|
287 | 287 | node = nodeorrev |
|
288 | 288 | validatenode(node) |
|
289 | 289 | |
|
290 | 290 | if node == self._repo.nullid: |
|
291 | 291 | return b'' |
|
292 | 292 | |
|
293 | 293 | rev = self.rev(node) |
|
294 | 294 | flags = self._flags(rev) |
|
295 | 295 | |
|
296 | 296 | path = b'/'.join([self._storepath, hex(node)]) |
|
297 | 297 | rawtext = self._svfs.read(path) |
|
298 | 298 | |
|
299 | 299 | if raw: |
|
300 | 300 | validatehash = flagutil.processflagsraw(self, rawtext, flags) |
|
301 | 301 | text = rawtext |
|
302 | 302 | else: |
|
303 | 303 | r = flagutil.processflagsread(self, rawtext, flags) |
|
304 | 304 | text, validatehash = r |
|
305 | 305 | if validatehash: |
|
306 | 306 | self.checkhash(text, node, rev=rev) |
|
307 | 307 | |
|
308 | 308 | return text |
|
309 | 309 | |
|
310 | 310 | def rawdata(self, nodeorrev): |
|
311 | 311 | return self.revision(raw=True) |
|
312 | 312 | |
|
313 | 313 | def read(self, node): |
|
314 | 314 | validatenode(node) |
|
315 | 315 | |
|
316 | 316 | revision = self.revision(node) |
|
317 | 317 | |
|
318 | 318 | if not revision.startswith(b'\1\n'): |
|
319 | 319 | return revision |
|
320 | 320 | |
|
321 | 321 | start = revision.index(b'\1\n', 2) |
|
322 | 322 | return revision[start + 2 :] |
|
323 | 323 | |
|
324 | 324 | def renamed(self, node): |
|
325 | 325 | validatenode(node) |
|
326 | 326 | |
|
327 | 327 | if self.parents(node)[0] != self._repo.nullid: |
|
328 | 328 | return False |
|
329 | 329 | |
|
330 | 330 | fulltext = self.revision(node) |
|
331 | 331 | m = storageutil.parsemeta(fulltext)[0] |
|
332 | 332 | |
|
333 | 333 | if m and 'copy' in m: |
|
334 | 334 | return m['copy'], bin(m['copyrev']) |
|
335 | 335 | |
|
336 | 336 | return False |
|
337 | 337 | |
|
338 | 338 | def cmp(self, node, text): |
|
339 | 339 | validatenode(node) |
|
340 | 340 | |
|
341 | 341 | t = text |
|
342 | 342 | |
|
343 | 343 | if text.startswith(b'\1\n'): |
|
344 | 344 | t = b'\1\n\1\n' + text |
|
345 | 345 | |
|
346 | 346 | p1, p2 = self.parents(node) |
|
347 | 347 | |
|
348 | 348 | if storageutil.hashrevisionsha1(t, p1, p2) == node: |
|
349 | 349 | return False |
|
350 | 350 | |
|
351 | 351 | if self.iscensored(self.rev(node)): |
|
352 | 352 | return text != b'' |
|
353 | 353 | |
|
354 | 354 | if self.renamed(node): |
|
355 | 355 | t2 = self.read(node) |
|
356 | 356 | return t2 != text |
|
357 | 357 | |
|
358 | 358 | return True |
|
359 | 359 | |
|
360 | 360 | def size(self, rev): |
|
361 | 361 | validaterev(rev) |
|
362 | 362 | |
|
363 | 363 | node = self._indexbyrev[rev][b'node'] |
|
364 | 364 | |
|
365 | 365 | if self.renamed(node): |
|
366 | 366 | return len(self.read(node)) |
|
367 | 367 | |
|
368 | 368 | if self.iscensored(rev): |
|
369 | 369 | return 0 |
|
370 | 370 | |
|
371 | 371 | return len(self.revision(node)) |
|
372 | 372 | |
|
373 | 373 | def iscensored(self, rev): |
|
374 | 374 | validaterev(rev) |
|
375 | 375 | |
|
376 | 376 | return self._flags(rev) & repository.REVISION_FLAG_CENSORED |
|
377 | 377 | |
|
378 | 378 | def commonancestorsheads(self, a, b): |
|
379 | 379 | validatenode(a) |
|
380 | 380 | validatenode(b) |
|
381 | 381 | |
|
382 | 382 | a = self.rev(a) |
|
383 | 383 | b = self.rev(b) |
|
384 | 384 | |
|
385 | 385 | ancestors = ancestor.commonancestorsheads(self.parentrevs, a, b) |
|
386 | 386 | return pycompat.maplist(self.node, ancestors) |
|
387 | 387 | |
|
388 | 388 | def descendants(self, revs): |
|
389 | 389 | # This is a copy of revlog.descendants() |
|
390 | 390 | first = min(revs) |
|
391 | 391 | if first == nullrev: |
|
392 | 392 | for i in self: |
|
393 | 393 | yield i |
|
394 | 394 | return |
|
395 | 395 | |
|
396 | 396 | seen = set(revs) |
|
397 | 397 | for i in self.revs(start=first + 1): |
|
398 | 398 | for x in self.parentrevs(i): |
|
399 | 399 | if x != nullrev and x in seen: |
|
400 | 400 | seen.add(i) |
|
401 | 401 | yield i |
|
402 | 402 | break |
|
403 | 403 | |
|
404 | 404 | # Required by verify. |
|
405 | 405 | def files(self): |
|
406 | 406 | entries = self._svfs.listdir(self._storepath) |
|
407 | 407 | |
|
408 | 408 | # Strip out undo.backup.* files created as part of transaction |
|
409 | 409 | # recording. |
|
410 | 410 | entries = [f for f in entries if not f.startswith('undo.backup.')] |
|
411 | 411 | |
|
412 | 412 | return [b'/'.join((self._storepath, f)) for f in entries] |
|
413 | 413 | |
|
414 | 414 | def storageinfo( |
|
415 | 415 | self, |
|
416 | 416 | exclusivefiles=False, |
|
417 | 417 | sharedfiles=False, |
|
418 | 418 | revisionscount=False, |
|
419 | 419 | trackedsize=False, |
|
420 | 420 | storedsize=False, |
|
421 | 421 | ): |
|
422 | 422 | # TODO do a real implementation of this |
|
423 | 423 | return { |
|
424 | 424 | 'exclusivefiles': [], |
|
425 | 425 | 'sharedfiles': [], |
|
426 | 426 | 'revisionscount': len(self), |
|
427 | 427 | 'trackedsize': 0, |
|
428 | 428 | 'storedsize': None, |
|
429 | 429 | } |
|
430 | 430 | |
|
431 | 431 | def verifyintegrity(self, state): |
|
432 | 432 | state['skipread'] = set() |
|
433 | 433 | for rev in self: |
|
434 | 434 | node = self.node(rev) |
|
435 | 435 | try: |
|
436 | 436 | self.revision(node) |
|
437 | 437 | except Exception as e: |
|
438 | 438 | yield simplefilestoreproblem( |
|
439 | 439 | error='unpacking %s: %s' % (node, e), node=node |
|
440 | 440 | ) |
|
441 | 441 | state['skipread'].add(node) |
|
442 | 442 | |
|
443 | 443 | def emitrevisions( |
|
444 | 444 | self, |
|
445 | 445 | nodes, |
|
446 | 446 | nodesorder=None, |
|
447 | 447 | revisiondata=False, |
|
448 | 448 | assumehaveparentrevisions=False, |
|
449 | 449 | deltamode=repository.CG_DELTAMODE_STD, |
|
450 | 450 | sidedata_helpers=None, |
|
451 | 451 | ): |
|
452 | 452 | # TODO this will probably break on some ordering options. |
|
453 | 453 | nodes = [n for n in nodes if n != self._repo.nullid] |
|
454 | 454 | if not nodes: |
|
455 | 455 | return |
|
456 | 456 | for delta in storageutil.emitrevisions( |
|
457 | 457 | self, |
|
458 | 458 | nodes, |
|
459 | 459 | nodesorder, |
|
460 | 460 | simplestorerevisiondelta, |
|
461 | 461 | revisiondata=revisiondata, |
|
462 | 462 | assumehaveparentrevisions=assumehaveparentrevisions, |
|
463 | 463 | deltamode=deltamode, |
|
464 | 464 | sidedata_helpers=sidedata_helpers, |
|
465 | 465 | ): |
|
466 | 466 | yield delta |
|
467 | 467 | |
|
468 | 468 | def add(self, text, meta, transaction, linkrev, p1, p2): |
|
469 | 469 | if meta or text.startswith(b'\1\n'): |
|
470 | 470 | text = storageutil.packmeta(meta, text) |
|
471 | 471 | |
|
472 | 472 | return self.addrevision(text, transaction, linkrev, p1, p2) |
|
473 | 473 | |
|
474 | 474 | def addrevision( |
|
475 | 475 | self, |
|
476 | 476 | text, |
|
477 | 477 | transaction, |
|
478 | 478 | linkrev, |
|
479 | 479 | p1, |
|
480 | 480 | p2, |
|
481 | 481 | node=None, |
|
482 | 482 | flags=revlog.REVIDX_DEFAULT_FLAGS, |
|
483 | 483 | cachedelta=None, |
|
484 | 484 | ): |
|
485 | 485 | validatenode(p1) |
|
486 | 486 | validatenode(p2) |
|
487 | 487 | |
|
488 | 488 | if flags: |
|
489 | 489 | node = node or storageutil.hashrevisionsha1(text, p1, p2) |
|
490 | 490 | |
|
491 | 491 | rawtext, validatehash = flagutil.processflagswrite(self, text, flags) |
|
492 | 492 | |
|
493 | 493 | node = node or storageutil.hashrevisionsha1(text, p1, p2) |
|
494 | 494 | |
|
495 | 495 | if node in self._indexbynode: |
|
496 | 496 | return node |
|
497 | 497 | |
|
498 | 498 | if validatehash: |
|
499 | 499 | self.checkhash(rawtext, node, p1=p1, p2=p2) |
|
500 | 500 | |
|
501 | 501 | return self._addrawrevision( |
|
502 | 502 | node, rawtext, transaction, linkrev, p1, p2, flags |
|
503 | 503 | ) |
|
504 | 504 | |
|
505 | 505 | def _addrawrevision(self, node, rawtext, transaction, link, p1, p2, flags): |
|
506 | 506 | transaction.addbackup(self._indexpath) |
|
507 | 507 | |
|
508 | 508 | path = b'/'.join([self._storepath, hex(node)]) |
|
509 | 509 | |
|
510 | 510 | self._svfs.write(path, rawtext) |
|
511 | 511 | |
|
512 | 512 | self._indexdata.append( |
|
513 | 513 | { |
|
514 | 514 | b'node': node, |
|
515 | 515 | b'p1': p1, |
|
516 | 516 | b'p2': p2, |
|
517 | 517 | b'linkrev': link, |
|
518 | 518 | b'flags': flags, |
|
519 | 519 | } |
|
520 | 520 | ) |
|
521 | 521 | |
|
522 | 522 | self._reflectindexupdate() |
|
523 | 523 | |
|
524 | 524 | return node |
|
525 | 525 | |
|
526 | 526 | def _reflectindexupdate(self): |
|
527 | 527 | self._refreshindex() |
|
528 | 528 | self._svfs.write( |
|
529 | 529 | self._indexpath, ''.join(cborutil.streamencode(self._indexdata)) |
|
530 | 530 | ) |
|
531 | 531 | |
|
532 | 532 | def addgroup( |
|
533 | 533 | self, |
|
534 | 534 | deltas, |
|
535 | 535 | linkmapper, |
|
536 | 536 | transaction, |
|
537 | 537 | addrevisioncb=None, |
|
538 | 538 | duplicaterevisioncb=None, |
|
539 | 539 | maybemissingparents=False, |
|
540 | 540 | ): |
|
541 | 541 | if maybemissingparents: |
|
542 | 542 | raise error.Abort( |
|
543 | 543 | _('simple store does not support missing parents ' 'write mode') |
|
544 | 544 | ) |
|
545 | 545 | |
|
546 | 546 | empty = True |
|
547 | 547 | |
|
548 | 548 | transaction.addbackup(self._indexpath) |
|
549 | 549 | |
|
550 | 550 | for node, p1, p2, linknode, deltabase, delta, flags in deltas: |
|
551 | 551 | linkrev = linkmapper(linknode) |
|
552 | 552 | flags = flags or revlog.REVIDX_DEFAULT_FLAGS |
|
553 | 553 | |
|
554 | 554 | if node in self._indexbynode: |
|
555 | 555 | if duplicaterevisioncb: |
|
556 | 556 | duplicaterevisioncb(self, self.rev(node)) |
|
557 | 557 | empty = False |
|
558 | 558 | continue |
|
559 | 559 | |
|
560 | 560 | # Need to resolve the fulltext from the delta base. |
|
561 | 561 | if deltabase == self._repo.nullid: |
|
562 | 562 | text = mdiff.patch(b'', delta) |
|
563 | 563 | else: |
|
564 | 564 | text = mdiff.patch(self.revision(deltabase), delta) |
|
565 | 565 | |
|
566 | 566 | rev = self._addrawrevision( |
|
567 | 567 | node, text, transaction, linkrev, p1, p2, flags |
|
568 | 568 | ) |
|
569 | 569 | |
|
570 | 570 | if addrevisioncb: |
|
571 | 571 | addrevisioncb(self, rev) |
|
572 | 572 | empty = False |
|
573 | 573 | return not empty |
|
574 | 574 | |
|
575 | 575 | def _headrevs(self): |
|
576 | 576 | # Assume all revisions are heads by default. |
|
577 | 577 | revishead = {rev: True for rev in self._indexbyrev} |
|
578 | 578 | |
|
579 | 579 | for rev, entry in self._indexbyrev.items(): |
|
580 | 580 | # Unset head flag for all seen parents. |
|
581 | 581 | revishead[self.rev(entry[b'p1'])] = False |
|
582 | 582 | revishead[self.rev(entry[b'p2'])] = False |
|
583 | 583 | |
|
584 | 584 | return [rev for rev, ishead in sorted(revishead.items()) if ishead] |
|
585 | 585 | |
|
586 | 586 | def heads(self, start=None, stop=None): |
|
587 | 587 | # This is copied from revlog.py. |
|
588 | 588 | if start is None and stop is None: |
|
589 | 589 | if not len(self): |
|
590 | 590 | return [self._repo.nullid] |
|
591 | 591 | return [self.node(r) for r in self._headrevs()] |
|
592 | 592 | |
|
593 | 593 | if start is None: |
|
594 | 594 | start = self._repo.nullid |
|
595 | 595 | if stop is None: |
|
596 | 596 | stop = [] |
|
597 | 597 | stoprevs = {self.rev(n) for n in stop} |
|
598 | 598 | startrev = self.rev(start) |
|
599 | 599 | reachable = {startrev} |
|
600 | 600 | heads = {startrev} |
|
601 | 601 | |
|
602 | 602 | parentrevs = self.parentrevs |
|
603 | 603 | for r in self.revs(start=startrev + 1): |
|
604 | 604 | for p in parentrevs(r): |
|
605 | 605 | if p in reachable: |
|
606 | 606 | if r not in stoprevs: |
|
607 | 607 | reachable.add(r) |
|
608 | 608 | heads.add(r) |
|
609 | 609 | if p in heads and p not in stoprevs: |
|
610 | 610 | heads.remove(p) |
|
611 | 611 | |
|
612 | 612 | return [self.node(r) for r in heads] |
|
613 | 613 | |
|
614 | 614 | def children(self, node): |
|
615 | 615 | validatenode(node) |
|
616 | 616 | |
|
617 | 617 | # This is a copy of revlog.children(). |
|
618 | 618 | c = [] |
|
619 | 619 | p = self.rev(node) |
|
620 | 620 | for r in self.revs(start=p + 1): |
|
621 | 621 | prevs = [pr for pr in self.parentrevs(r) if pr != nullrev] |
|
622 | 622 | if prevs: |
|
623 | 623 | for pr in prevs: |
|
624 | 624 | if pr == p: |
|
625 | 625 | c.append(self.node(r)) |
|
626 | 626 | elif p == nullrev: |
|
627 | 627 | c.append(self.node(r)) |
|
628 | 628 | return c |
|
629 | 629 | |
|
630 | 630 | def getstrippoint(self, minlink): |
|
631 | 631 | return storageutil.resolvestripinfo( |
|
632 | 632 | minlink, |
|
633 | 633 | len(self) - 1, |
|
634 | 634 | self._headrevs(), |
|
635 | 635 | self.linkrev, |
|
636 | 636 | self.parentrevs, |
|
637 | 637 | ) |
|
638 | 638 | |
|
639 | 639 | def strip(self, minlink, transaction): |
|
640 | 640 | if not len(self): |
|
641 | 641 | return |
|
642 | 642 | |
|
643 | 643 | rev, _ignored = self.getstrippoint(minlink) |
|
644 | 644 | if rev == len(self): |
|
645 | 645 | return |
|
646 | 646 | |
|
647 | 647 | # Purge index data starting at the requested revision. |
|
648 | 648 | self._indexdata[rev:] = [] |
|
649 | 649 | self._reflectindexupdate() |
|
650 | 650 | |
|
651 | 651 | |
|
652 | 652 | def issimplestorefile(f, kind, st): |
|
653 | 653 | if kind != stat.S_IFREG: |
|
654 | 654 | return False |
|
655 | 655 | |
|
656 | 656 | if store.isrevlog(f, kind, st): |
|
657 | 657 | return False |
|
658 | 658 | |
|
659 | 659 | # Ignore transaction undo files. |
|
660 | 660 | if f.startswith('undo.'): |
|
661 | 661 | return False |
|
662 | 662 | |
|
663 | 663 | # Otherwise assume it belongs to the simple store. |
|
664 | 664 | return True |
|
665 | 665 | |
|
666 | 666 | |
|
667 | 667 | class simplestore(store.encodedstore): |
|
668 | def datafiles(self): | |
|
668 | def datafiles(self, undecodable=None): | |
|
669 | 669 | for x in super(simplestore, self).datafiles(): |
|
670 | 670 | yield x |
|
671 | 671 | |
|
672 | 672 | # Supplement with non-revlog files. |
|
673 | 673 | extrafiles = self._walk('data', True, filefilter=issimplestorefile) |
|
674 | 674 | |
|
675 |
for |
|
|
675 | for f1, size in extrafiles: | |
|
676 | 676 | try: |
|
677 |
|
|
|
677 | f2 = store.decodefilename(f1) | |
|
678 | 678 | except KeyError: |
|
679 |
|
|
|
679 | if undecodable is None: | |
|
680 | raise error.StorageError(b'undecodable revlog name %s' % f1) | |
|
681 | else: | |
|
682 | undecodable.append(f1) | |
|
683 | continue | |
|
680 | 684 | |
|
681 |
yield |
|
|
685 | yield f2, size | |
|
682 | 686 | |
|
683 | 687 | |
|
684 | 688 | def reposetup(ui, repo): |
|
685 | 689 | if not repo.local(): |
|
686 | 690 | return |
|
687 | 691 | |
|
688 | 692 | if isinstance(repo, bundlerepo.bundlerepository): |
|
689 | 693 | raise error.Abort(_('cannot use simple store with bundlerepo')) |
|
690 | 694 | |
|
691 | 695 | class simplestorerepo(repo.__class__): |
|
692 | 696 | def file(self, f): |
|
693 | 697 | return filestorage(repo, self.svfs, f) |
|
694 | 698 | |
|
695 | 699 | repo.__class__ = simplestorerepo |
|
696 | 700 | |
|
697 | 701 | |
|
698 | 702 | def featuresetup(ui, supported): |
|
699 | 703 | supported.add(REQUIREMENT) |
|
700 | 704 | |
|
701 | 705 | |
|
702 | 706 | def newreporequirements(orig, ui, createopts): |
|
703 | 707 | """Modifies default requirements for new repos to use the simple store.""" |
|
704 | 708 | requirements = orig(ui, createopts) |
|
705 | 709 | |
|
706 | 710 | # These requirements are only used to affect creation of the store |
|
707 | 711 | # object. We have our own store. So we can remove them. |
|
708 | 712 | # TODO do this once we feel like taking the test hit. |
|
709 | 713 | # if 'fncache' in requirements: |
|
710 | 714 | # requirements.remove('fncache') |
|
711 | 715 | # if 'dotencode' in requirements: |
|
712 | 716 | # requirements.remove('dotencode') |
|
713 | 717 | |
|
714 | 718 | requirements.add(REQUIREMENT) |
|
715 | 719 | |
|
716 | 720 | return requirements |
|
717 | 721 | |
|
718 | 722 | |
|
719 | 723 | def makestore(orig, requirements, path, vfstype): |
|
720 | 724 | if REQUIREMENT not in requirements: |
|
721 | 725 | return orig(requirements, path, vfstype) |
|
722 | 726 | |
|
723 | 727 | return simplestore(path, vfstype) |
|
724 | 728 | |
|
725 | 729 | |
|
726 | 730 | def verifierinit(orig, self, *args, **kwargs): |
|
727 | 731 | orig(self, *args, **kwargs) |
|
728 | 732 | |
|
729 | 733 | # We don't care that files in the store don't align with what is |
|
730 | 734 | # advertised. So suppress these warnings. |
|
731 | 735 | self.warnorphanstorefiles = False |
|
732 | 736 | |
|
733 | 737 | |
|
734 | 738 | def extsetup(ui): |
|
735 | 739 | localrepo.featuresetupfuncs.add(featuresetup) |
|
736 | 740 | |
|
737 | 741 | extensions.wrapfunction( |
|
738 | 742 | localrepo, 'newreporequirements', newreporequirements |
|
739 | 743 | ) |
|
740 | 744 | extensions.wrapfunction(localrepo, 'makestore', makestore) |
|
741 | 745 | extensions.wrapfunction(verify.verifier, '__init__', verifierinit) |
General Comments 0
You need to be logged in to leave comments.
Login now