Show More
@@ -1,3177 +1,3180 b'' | |||
|
1 | 1 | # revlog.py - storage back-end for mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2005-2007 Olivia Mackall <olivia@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | """Storage back-end for Mercurial. |
|
9 | 9 | |
|
10 | 10 | This provides efficient delta storage with O(1) retrieve and append |
|
11 | 11 | and O(changes) merge between branches. |
|
12 | 12 | """ |
|
13 | 13 | |
|
14 | 14 | from __future__ import absolute_import |
|
15 | 15 | |
|
16 | 16 | import binascii |
|
17 | 17 | import collections |
|
18 | 18 | import contextlib |
|
19 | 19 | import errno |
|
20 | 20 | import io |
|
21 | 21 | import os |
|
22 | 22 | import struct |
|
23 | 23 | import zlib |
|
24 | 24 | |
|
25 | 25 | # import stuff from node for others to import from revlog |
|
26 | 26 | from .node import ( |
|
27 | 27 | bin, |
|
28 | 28 | hex, |
|
29 | 29 | nullrev, |
|
30 | 30 | sha1nodeconstants, |
|
31 | 31 | short, |
|
32 | 32 | wdirrev, |
|
33 | 33 | ) |
|
34 | 34 | from .i18n import _ |
|
35 | 35 | from .pycompat import getattr |
|
36 | 36 | from .revlogutils.constants import ( |
|
37 | 37 | ALL_KINDS, |
|
38 | 38 | FLAG_GENERALDELTA, |
|
39 | 39 | FLAG_INLINE_DATA, |
|
40 | 40 | INDEX_HEADER, |
|
41 | 41 | REVLOGV0, |
|
42 | 42 | REVLOGV1, |
|
43 | 43 | REVLOGV1_FLAGS, |
|
44 | 44 | REVLOGV2, |
|
45 | 45 | REVLOGV2_FLAGS, |
|
46 | 46 | REVLOG_DEFAULT_FLAGS, |
|
47 | 47 | REVLOG_DEFAULT_FORMAT, |
|
48 | 48 | REVLOG_DEFAULT_VERSION, |
|
49 | 49 | ) |
|
50 | 50 | from .revlogutils.flagutil import ( |
|
51 | 51 | REVIDX_DEFAULT_FLAGS, |
|
52 | 52 | REVIDX_ELLIPSIS, |
|
53 | 53 | REVIDX_EXTSTORED, |
|
54 | 54 | REVIDX_FLAGS_ORDER, |
|
55 | 55 | REVIDX_HASCOPIESINFO, |
|
56 | 56 | REVIDX_ISCENSORED, |
|
57 | 57 | REVIDX_RAWTEXT_CHANGING_FLAGS, |
|
58 | 58 | ) |
|
59 | 59 | from .thirdparty import attr |
|
60 | 60 | from . import ( |
|
61 | 61 | ancestor, |
|
62 | 62 | dagop, |
|
63 | 63 | error, |
|
64 | 64 | mdiff, |
|
65 | 65 | policy, |
|
66 | 66 | pycompat, |
|
67 | 67 | templatefilters, |
|
68 | 68 | util, |
|
69 | 69 | ) |
|
70 | 70 | from .interfaces import ( |
|
71 | 71 | repository, |
|
72 | 72 | util as interfaceutil, |
|
73 | 73 | ) |
|
74 | 74 | from .revlogutils import ( |
|
75 | 75 | deltas as deltautil, |
|
76 | 76 | flagutil, |
|
77 | 77 | nodemap as nodemaputil, |
|
78 | 78 | revlogv0, |
|
79 | 79 | sidedata as sidedatautil, |
|
80 | 80 | ) |
|
81 | 81 | from .utils import ( |
|
82 | 82 | storageutil, |
|
83 | 83 | stringutil, |
|
84 | 84 | ) |
|
85 | 85 | |
|
86 | 86 | # blanked usage of all the name to prevent pyflakes constraints |
|
87 | 87 | # We need these name available in the module for extensions. |
|
88 | 88 | |
|
89 | 89 | REVLOGV0 |
|
90 | 90 | REVLOGV1 |
|
91 | 91 | REVLOGV2 |
|
92 | 92 | FLAG_INLINE_DATA |
|
93 | 93 | FLAG_GENERALDELTA |
|
94 | 94 | REVLOG_DEFAULT_FLAGS |
|
95 | 95 | REVLOG_DEFAULT_FORMAT |
|
96 | 96 | REVLOG_DEFAULT_VERSION |
|
97 | 97 | REVLOGV1_FLAGS |
|
98 | 98 | REVLOGV2_FLAGS |
|
99 | 99 | REVIDX_ISCENSORED |
|
100 | 100 | REVIDX_ELLIPSIS |
|
101 | 101 | REVIDX_HASCOPIESINFO |
|
102 | 102 | REVIDX_EXTSTORED |
|
103 | 103 | REVIDX_DEFAULT_FLAGS |
|
104 | 104 | REVIDX_FLAGS_ORDER |
|
105 | 105 | REVIDX_RAWTEXT_CHANGING_FLAGS |
|
106 | 106 | |
|
107 | 107 | parsers = policy.importmod('parsers') |
|
108 | 108 | rustancestor = policy.importrust('ancestor') |
|
109 | 109 | rustdagop = policy.importrust('dagop') |
|
110 | 110 | rustrevlog = policy.importrust('revlog') |
|
111 | 111 | |
|
112 | 112 | # Aliased for performance. |
|
113 | 113 | _zlibdecompress = zlib.decompress |
|
114 | 114 | |
|
115 | 115 | # max size of revlog with inline data |
|
116 | 116 | _maxinline = 131072 |
|
117 | 117 | _chunksize = 1048576 |
|
118 | 118 | |
|
119 | 119 | # Flag processors for REVIDX_ELLIPSIS. |
|
120 | 120 | def ellipsisreadprocessor(rl, text): |
|
121 | 121 | return text, False |
|
122 | 122 | |
|
123 | 123 | |
|
124 | 124 | def ellipsiswriteprocessor(rl, text): |
|
125 | 125 | return text, False |
|
126 | 126 | |
|
127 | 127 | |
|
128 | 128 | def ellipsisrawprocessor(rl, text): |
|
129 | 129 | return False |
|
130 | 130 | |
|
131 | 131 | |
|
132 | 132 | ellipsisprocessor = ( |
|
133 | 133 | ellipsisreadprocessor, |
|
134 | 134 | ellipsiswriteprocessor, |
|
135 | 135 | ellipsisrawprocessor, |
|
136 | 136 | ) |
|
137 | 137 | |
|
138 | 138 | |
|
139 | 139 | def offset_type(offset, type): |
|
140 | 140 | if (type & ~flagutil.REVIDX_KNOWN_FLAGS) != 0: |
|
141 | 141 | raise ValueError(b'unknown revlog index flags') |
|
142 | 142 | return int(int(offset) << 16 | type) |
|
143 | 143 | |
|
144 | 144 | |
|
145 | 145 | def _verify_revision(rl, skipflags, state, node): |
|
146 | 146 | """Verify the integrity of the given revlog ``node`` while providing a hook |
|
147 | 147 | point for extensions to influence the operation.""" |
|
148 | 148 | if skipflags: |
|
149 | 149 | state[b'skipread'].add(node) |
|
150 | 150 | else: |
|
151 | 151 | # Side-effect: read content and verify hash. |
|
152 | 152 | rl.revision(node) |
|
153 | 153 | |
|
154 | 154 | |
|
155 | 155 | # True if a fast implementation for persistent-nodemap is available |
|
156 | 156 | # |
|
157 | 157 | # We also consider we have a "fast" implementation in "pure" python because |
|
158 | 158 | # people using pure don't really have performance consideration (and a |
|
159 | 159 | # wheelbarrow of other slowness source) |
|
160 | 160 | HAS_FAST_PERSISTENT_NODEMAP = rustrevlog is not None or util.safehasattr( |
|
161 | 161 | parsers, 'BaseIndexObject' |
|
162 | 162 | ) |
|
163 | 163 | |
|
164 | 164 | |
|
165 | 165 | @attr.s(slots=True, frozen=True) |
|
166 | 166 | class _revisioninfo(object): |
|
167 | 167 | """Information about a revision that allows building its fulltext |
|
168 | 168 | node: expected hash of the revision |
|
169 | 169 | p1, p2: parent revs of the revision |
|
170 | 170 | btext: built text cache consisting of a one-element list |
|
171 | 171 | cachedelta: (baserev, uncompressed_delta) or None |
|
172 | 172 | flags: flags associated to the revision storage |
|
173 | 173 | |
|
174 | 174 | One of btext[0] or cachedelta must be set. |
|
175 | 175 | """ |
|
176 | 176 | |
|
177 | 177 | node = attr.ib() |
|
178 | 178 | p1 = attr.ib() |
|
179 | 179 | p2 = attr.ib() |
|
180 | 180 | btext = attr.ib() |
|
181 | 181 | textlen = attr.ib() |
|
182 | 182 | cachedelta = attr.ib() |
|
183 | 183 | flags = attr.ib() |
|
184 | 184 | |
|
185 | 185 | |
|
186 | 186 | @interfaceutil.implementer(repository.irevisiondelta) |
|
187 | 187 | @attr.s(slots=True) |
|
188 | 188 | class revlogrevisiondelta(object): |
|
189 | 189 | node = attr.ib() |
|
190 | 190 | p1node = attr.ib() |
|
191 | 191 | p2node = attr.ib() |
|
192 | 192 | basenode = attr.ib() |
|
193 | 193 | flags = attr.ib() |
|
194 | 194 | baserevisionsize = attr.ib() |
|
195 | 195 | revision = attr.ib() |
|
196 | 196 | delta = attr.ib() |
|
197 | 197 | sidedata = attr.ib() |
|
198 | 198 | protocol_flags = attr.ib() |
|
199 | 199 | linknode = attr.ib(default=None) |
|
200 | 200 | |
|
201 | 201 | |
|
202 | 202 | @interfaceutil.implementer(repository.iverifyproblem) |
|
203 | 203 | @attr.s(frozen=True) |
|
204 | 204 | class revlogproblem(object): |
|
205 | 205 | warning = attr.ib(default=None) |
|
206 | 206 | error = attr.ib(default=None) |
|
207 | 207 | node = attr.ib(default=None) |
|
208 | 208 | |
|
209 | 209 | |
|
210 | 210 | def parse_index_v1(data, inline): |
|
211 | 211 | # call the C implementation to parse the index data |
|
212 | 212 | index, cache = parsers.parse_index2(data, inline) |
|
213 | 213 | return index, cache |
|
214 | 214 | |
|
215 | 215 | |
|
216 | 216 | def parse_index_v2(data, inline): |
|
217 | 217 | # call the C implementation to parse the index data |
|
218 | 218 | index, cache = parsers.parse_index2(data, inline, revlogv2=True) |
|
219 | 219 | return index, cache |
|
220 | 220 | |
|
221 | 221 | |
|
222 | 222 | if util.safehasattr(parsers, 'parse_index_devel_nodemap'): |
|
223 | 223 | |
|
224 | 224 | def parse_index_v1_nodemap(data, inline): |
|
225 | 225 | index, cache = parsers.parse_index_devel_nodemap(data, inline) |
|
226 | 226 | return index, cache |
|
227 | 227 | |
|
228 | 228 | |
|
229 | 229 | else: |
|
230 | 230 | parse_index_v1_nodemap = None |
|
231 | 231 | |
|
232 | 232 | |
|
233 | 233 | def parse_index_v1_mixed(data, inline): |
|
234 | 234 | index, cache = parse_index_v1(data, inline) |
|
235 | 235 | return rustrevlog.MixedIndex(index), cache |
|
236 | 236 | |
|
237 | 237 | |
|
238 | 238 | # corresponds to uncompressed length of indexformatng (2 gigs, 4-byte |
|
239 | 239 | # signed integer) |
|
240 | 240 | _maxentrysize = 0x7FFFFFFF |
|
241 | 241 | |
|
242 | 242 | |
|
243 | 243 | class revlog(object): |
|
244 | 244 | """ |
|
245 | 245 | the underlying revision storage object |
|
246 | 246 | |
|
247 | 247 | A revlog consists of two parts, an index and the revision data. |
|
248 | 248 | |
|
249 | 249 | The index is a file with a fixed record size containing |
|
250 | 250 | information on each revision, including its nodeid (hash), the |
|
251 | 251 | nodeids of its parents, the position and offset of its data within |
|
252 | 252 | the data file, and the revision it's based on. Finally, each entry |
|
253 | 253 | contains a linkrev entry that can serve as a pointer to external |
|
254 | 254 | data. |
|
255 | 255 | |
|
256 | 256 | The revision data itself is a linear collection of data chunks. |
|
257 | 257 | Each chunk represents a revision and is usually represented as a |
|
258 | 258 | delta against the previous chunk. To bound lookup time, runs of |
|
259 | 259 | deltas are limited to about 2 times the length of the original |
|
260 | 260 | version data. This makes retrieval of a version proportional to |
|
261 | 261 | its size, or O(1) relative to the number of revisions. |
|
262 | 262 | |
|
263 | 263 | Both pieces of the revlog are written to in an append-only |
|
264 | 264 | fashion, which means we never need to rewrite a file to insert or |
|
265 | 265 | remove data, and can use some simple techniques to avoid the need |
|
266 | 266 | for locking while reading. |
|
267 | 267 | |
|
268 | 268 | If checkambig, indexfile is opened with checkambig=True at |
|
269 | 269 | writing, to avoid file stat ambiguity. |
|
270 | 270 | |
|
271 | 271 | If mmaplargeindex is True, and an mmapindexthreshold is set, the |
|
272 | 272 | index will be mmapped rather than read if it is larger than the |
|
273 | 273 | configured threshold. |
|
274 | 274 | |
|
275 | 275 | If censorable is True, the revlog can have censored revisions. |
|
276 | 276 | |
|
277 | 277 | If `upperboundcomp` is not None, this is the expected maximal gain from |
|
278 | 278 | compression for the data content. |
|
279 | 279 | |
|
280 | 280 | `concurrencychecker` is an optional function that receives 3 arguments: a |
|
281 | 281 | file handle, a filename, and an expected position. It should check whether |
|
282 | 282 | the current position in the file handle is valid, and log/warn/fail (by |
|
283 | 283 | raising). |
|
284 | 284 | """ |
|
285 | 285 | |
|
286 | 286 | _flagserrorclass = error.RevlogError |
|
287 | 287 | |
|
288 | 288 | def __init__( |
|
289 | 289 | self, |
|
290 | 290 | opener, |
|
291 | 291 | target, |
|
292 | 292 | radix, |
|
293 | 293 | postfix=None, |
|
294 | 294 | checkambig=False, |
|
295 | 295 | mmaplargeindex=False, |
|
296 | 296 | censorable=False, |
|
297 | 297 | upperboundcomp=None, |
|
298 | 298 | persistentnodemap=False, |
|
299 | 299 | concurrencychecker=None, |
|
300 | 300 | ): |
|
301 | 301 | """ |
|
302 | 302 | create a revlog object |
|
303 | 303 | |
|
304 | 304 | opener is a function that abstracts the file opening operation |
|
305 | 305 | and can be used to implement COW semantics or the like. |
|
306 | 306 | |
|
307 | 307 | `target`: a (KIND, ID) tuple that identify the content stored in |
|
308 | 308 | this revlog. It help the rest of the code to understand what the revlog |
|
309 | 309 | is about without having to resort to heuristic and index filename |
|
310 | 310 | analysis. Note: that this must be reliably be set by normal code, but |
|
311 | 311 | that test, debug, or performance measurement code might not set this to |
|
312 | 312 | accurate value. |
|
313 | 313 | """ |
|
314 | 314 | self.upperboundcomp = upperboundcomp |
|
315 | 315 | |
|
316 | 316 | self.radix = radix |
|
317 | 317 | |
|
318 | 318 | self._indexfile = None |
|
319 | 319 | self._datafile = None |
|
320 | 320 | self._nodemap_file = None |
|
321 | 321 | self.postfix = postfix |
|
322 | 322 | self.opener = opener |
|
323 | 323 | if persistentnodemap: |
|
324 | 324 | self._nodemap_file = nodemaputil.get_nodemap_file(self) |
|
325 | 325 | |
|
326 | 326 | assert target[0] in ALL_KINDS |
|
327 | 327 | assert len(target) == 2 |
|
328 | 328 | self.target = target |
|
329 | 329 | # When True, indexfile is opened with checkambig=True at writing, to |
|
330 | 330 | # avoid file stat ambiguity. |
|
331 | 331 | self._checkambig = checkambig |
|
332 | 332 | self._mmaplargeindex = mmaplargeindex |
|
333 | 333 | self._censorable = censorable |
|
334 | 334 | # 3-tuple of (node, rev, text) for a raw revision. |
|
335 | 335 | self._revisioncache = None |
|
336 | 336 | # Maps rev to chain base rev. |
|
337 | 337 | self._chainbasecache = util.lrucachedict(100) |
|
338 | 338 | # 2-tuple of (offset, data) of raw data from the revlog at an offset. |
|
339 | 339 | self._chunkcache = (0, b'') |
|
340 | 340 | # How much data to read and cache into the raw revlog data cache. |
|
341 | 341 | self._chunkcachesize = 65536 |
|
342 | 342 | self._maxchainlen = None |
|
343 | 343 | self._deltabothparents = True |
|
344 | 344 | self.index = None |
|
345 | 345 | self._nodemap_docket = None |
|
346 | 346 | # Mapping of partial identifiers to full nodes. |
|
347 | 347 | self._pcache = {} |
|
348 | 348 | # Mapping of revision integer to full node. |
|
349 | 349 | self._compengine = b'zlib' |
|
350 | 350 | self._compengineopts = {} |
|
351 | 351 | self._maxdeltachainspan = -1 |
|
352 | 352 | self._withsparseread = False |
|
353 | 353 | self._sparserevlog = False |
|
354 | 354 | self._srdensitythreshold = 0.50 |
|
355 | 355 | self._srmingapsize = 262144 |
|
356 | 356 | |
|
357 | 357 | # Make copy of flag processors so each revlog instance can support |
|
358 | 358 | # custom flags. |
|
359 | 359 | self._flagprocessors = dict(flagutil.flagprocessors) |
|
360 | 360 | |
|
361 | 361 | # 2-tuple of file handles being used for active writing. |
|
362 | 362 | self._writinghandles = None |
|
363 | 363 | |
|
364 | 364 | self._loadindex() |
|
365 | 365 | |
|
366 | 366 | self._concurrencychecker = concurrencychecker |
|
367 | 367 | |
|
368 | 368 | def _init_opts(self): |
|
369 | 369 | """process options (from above/config) to setup associated default revlog mode |
|
370 | 370 | |
|
371 | 371 | These values might be affected when actually reading on disk information. |
|
372 | 372 | |
|
373 | 373 | The relevant values are returned for use in _loadindex(). |
|
374 | 374 | |
|
375 | 375 | * newversionflags: |
|
376 | 376 | version header to use if we need to create a new revlog |
|
377 | 377 | |
|
378 | 378 | * mmapindexthreshold: |
|
379 | 379 | minimal index size for start to use mmap |
|
380 | 380 | |
|
381 | 381 | * force_nodemap: |
|
382 | 382 | force the usage of a "development" version of the nodemap code |
|
383 | 383 | """ |
|
384 | 384 | mmapindexthreshold = None |
|
385 | 385 | opts = self.opener.options |
|
386 | 386 | |
|
387 | 387 | if b'revlogv2' in opts: |
|
388 | 388 | newversionflags = REVLOGV2 | FLAG_INLINE_DATA |
|
389 | 389 | elif b'revlogv1' in opts: |
|
390 | 390 | newversionflags = REVLOGV1 | FLAG_INLINE_DATA |
|
391 | 391 | if b'generaldelta' in opts: |
|
392 | 392 | newversionflags |= FLAG_GENERALDELTA |
|
393 | 393 | elif b'revlogv0' in self.opener.options: |
|
394 | 394 | newversionflags = REVLOGV0 |
|
395 | 395 | else: |
|
396 | 396 | newversionflags = REVLOG_DEFAULT_VERSION |
|
397 | 397 | |
|
398 | 398 | if b'chunkcachesize' in opts: |
|
399 | 399 | self._chunkcachesize = opts[b'chunkcachesize'] |
|
400 | 400 | if b'maxchainlen' in opts: |
|
401 | 401 | self._maxchainlen = opts[b'maxchainlen'] |
|
402 | 402 | if b'deltabothparents' in opts: |
|
403 | 403 | self._deltabothparents = opts[b'deltabothparents'] |
|
404 | 404 | self._lazydelta = bool(opts.get(b'lazydelta', True)) |
|
405 | 405 | self._lazydeltabase = False |
|
406 | 406 | if self._lazydelta: |
|
407 | 407 | self._lazydeltabase = bool(opts.get(b'lazydeltabase', False)) |
|
408 | 408 | if b'compengine' in opts: |
|
409 | 409 | self._compengine = opts[b'compengine'] |
|
410 | 410 | if b'zlib.level' in opts: |
|
411 | 411 | self._compengineopts[b'zlib.level'] = opts[b'zlib.level'] |
|
412 | 412 | if b'zstd.level' in opts: |
|
413 | 413 | self._compengineopts[b'zstd.level'] = opts[b'zstd.level'] |
|
414 | 414 | if b'maxdeltachainspan' in opts: |
|
415 | 415 | self._maxdeltachainspan = opts[b'maxdeltachainspan'] |
|
416 | 416 | if self._mmaplargeindex and b'mmapindexthreshold' in opts: |
|
417 | 417 | mmapindexthreshold = opts[b'mmapindexthreshold'] |
|
418 | 418 | self.hassidedata = bool(opts.get(b'side-data', False)) |
|
419 | 419 | self._sparserevlog = bool(opts.get(b'sparse-revlog', False)) |
|
420 | 420 | withsparseread = bool(opts.get(b'with-sparse-read', False)) |
|
421 | 421 | # sparse-revlog forces sparse-read |
|
422 | 422 | self._withsparseread = self._sparserevlog or withsparseread |
|
423 | 423 | if b'sparse-read-density-threshold' in opts: |
|
424 | 424 | self._srdensitythreshold = opts[b'sparse-read-density-threshold'] |
|
425 | 425 | if b'sparse-read-min-gap-size' in opts: |
|
426 | 426 | self._srmingapsize = opts[b'sparse-read-min-gap-size'] |
|
427 | 427 | if opts.get(b'enableellipsis'): |
|
428 | 428 | self._flagprocessors[REVIDX_ELLIPSIS] = ellipsisprocessor |
|
429 | 429 | |
|
430 | 430 | # revlog v0 doesn't have flag processors |
|
431 | 431 | for flag, processor in pycompat.iteritems( |
|
432 | 432 | opts.get(b'flagprocessors', {}) |
|
433 | 433 | ): |
|
434 | 434 | flagutil.insertflagprocessor(flag, processor, self._flagprocessors) |
|
435 | 435 | |
|
436 | 436 | if self._chunkcachesize <= 0: |
|
437 | 437 | raise error.RevlogError( |
|
438 | 438 | _(b'revlog chunk cache size %r is not greater than 0') |
|
439 | 439 | % self._chunkcachesize |
|
440 | 440 | ) |
|
441 | 441 | elif self._chunkcachesize & (self._chunkcachesize - 1): |
|
442 | 442 | raise error.RevlogError( |
|
443 | 443 | _(b'revlog chunk cache size %r is not a power of 2') |
|
444 | 444 | % self._chunkcachesize |
|
445 | 445 | ) |
|
446 | 446 | force_nodemap = opts.get(b'devel-force-nodemap', False) |
|
447 | 447 | return newversionflags, mmapindexthreshold, force_nodemap |
|
448 | 448 | |
|
449 | def _get_data(self, filepath, mmap_threshold): | |
|
450 | """return a file content with or without mmap | |
|
451 | ||
|
452 | If the file is missing return the empty string""" | |
|
453 | try: | |
|
454 | with self.opener(filepath) as fp: | |
|
455 | if mmap_threshold is not None: | |
|
456 | file_size = self.opener.fstat(fp).st_size | |
|
457 | if file_size >= mmap_threshold: | |
|
458 | # TODO: should .close() to release resources without | |
|
459 | # relying on Python GC | |
|
460 | return util.buffer(util.mmapread(fp)) | |
|
461 | return fp.read() | |
|
462 | except IOError as inst: | |
|
463 | if inst.errno != errno.ENOENT: | |
|
464 | raise | |
|
465 | return b'' | |
|
466 | ||
|
449 | 467 | def _loadindex(self): |
|
450 | 468 | |
|
451 | 469 | newversionflags, mmapindexthreshold, force_nodemap = self._init_opts() |
|
452 | 470 | |
|
453 | 471 | if self.postfix is None: |
|
454 | 472 | index_file = b'%s.i' % self.radix |
|
455 | 473 | data_file = b'%s.d' % self.radix |
|
456 | 474 | elif self.postfix == b'a': |
|
457 | 475 | index_file = b'%s.i.a' % self.radix |
|
458 | 476 | data_file = b'%s.d' % self.radix |
|
459 | 477 | else: |
|
460 | 478 | index_file = b'%s.i.%s' % (self.radix, self.postfix) |
|
461 | 479 | data_file = b'%s.d.%s' % (self.radix, self.postfix) |
|
462 | 480 | |
|
463 | 481 | self._indexfile = index_file |
|
464 | 482 | self._datafile = data_file |
|
465 | 483 | |
|
466 | 484 | indexdata = b'' |
|
467 | 485 | self._initempty = True |
|
468 | try: | |
|
469 | with self._indexfp() as f: | |
|
470 | if ( | |
|
471 | mmapindexthreshold is not None | |
|
472 | and self.opener.fstat(f).st_size >= mmapindexthreshold | |
|
473 | ): | |
|
474 | # TODO: should .close() to release resources without | |
|
475 | # relying on Python GC | |
|
476 | indexdata = util.buffer(util.mmapread(f)) | |
|
477 | else: | |
|
478 | indexdata = f.read() | |
|
486 | indexdata = self._get_data(self._indexfile, mmapindexthreshold) | |
|
479 | 487 |
|
|
480 | 488 |
|
|
481 | 489 |
|
|
482 | 490 |
|
|
483 | 491 |
|
|
484 | except IOError as inst: | |
|
485 | if inst.errno != errno.ENOENT: | |
|
486 | raise | |
|
487 | ||
|
488 | versionflags = newversionflags | |
|
489 | 492 | |
|
490 | 493 | flags = self._format_flags = versionflags & ~0xFFFF |
|
491 | 494 | fmt = self._format_version = versionflags & 0xFFFF |
|
492 | 495 | |
|
493 | 496 | if fmt == REVLOGV0: |
|
494 | 497 | if flags: |
|
495 | 498 | raise error.RevlogError( |
|
496 | 499 | _(b'unknown flags (%#04x) in version %d revlog %s') |
|
497 | 500 | % (flags >> 16, fmt, self.display_id) |
|
498 | 501 | ) |
|
499 | 502 | |
|
500 | 503 | self._inline = False |
|
501 | 504 | self._generaldelta = False |
|
502 | 505 | |
|
503 | 506 | elif fmt == REVLOGV1: |
|
504 | 507 | if flags & ~REVLOGV1_FLAGS: |
|
505 | 508 | raise error.RevlogError( |
|
506 | 509 | _(b'unknown flags (%#04x) in version %d revlog %s') |
|
507 | 510 | % (flags >> 16, fmt, self.display_id) |
|
508 | 511 | ) |
|
509 | 512 | |
|
510 | 513 | self._inline = versionflags & FLAG_INLINE_DATA |
|
511 | 514 | self._generaldelta = versionflags & FLAG_GENERALDELTA |
|
512 | 515 | |
|
513 | 516 | elif fmt == REVLOGV2: |
|
514 | 517 | if flags & ~REVLOGV2_FLAGS: |
|
515 | 518 | raise error.RevlogError( |
|
516 | 519 | _(b'unknown flags (%#04x) in version %d revlog %s') |
|
517 | 520 | % (flags >> 16, fmt, self.display_id) |
|
518 | 521 | ) |
|
519 | 522 | |
|
520 | 523 | # There is a bug in the transaction handling when going from an |
|
521 | 524 | # inline revlog to a separate index and data file. Turn it off until |
|
522 | 525 | # it's fixed, since v2 revlogs sometimes get rewritten on exchange. |
|
523 | 526 | # See issue6485 |
|
524 | 527 | self._inline = False |
|
525 | 528 | # generaldelta implied by version 2 revlogs. |
|
526 | 529 | self._generaldelta = True |
|
527 | 530 | |
|
528 | 531 | else: |
|
529 | 532 | raise error.RevlogError( |
|
530 | 533 | _(b'unknown version (%d) in revlog %s') % (fmt, self.display_id) |
|
531 | 534 | ) |
|
532 | 535 | |
|
533 | 536 | self.nodeconstants = sha1nodeconstants |
|
534 | 537 | self.nullid = self.nodeconstants.nullid |
|
535 | 538 | |
|
536 | 539 | # sparse-revlog can't be on without general-delta (issue6056) |
|
537 | 540 | if not self._generaldelta: |
|
538 | 541 | self._sparserevlog = False |
|
539 | 542 | |
|
540 | 543 | self._storedeltachains = True |
|
541 | 544 | |
|
542 | 545 | devel_nodemap = ( |
|
543 | 546 | self._nodemap_file |
|
544 | 547 | and force_nodemap |
|
545 | 548 | and parse_index_v1_nodemap is not None |
|
546 | 549 | ) |
|
547 | 550 | |
|
548 | 551 | use_rust_index = False |
|
549 | 552 | if rustrevlog is not None: |
|
550 | 553 | if self._nodemap_file is not None: |
|
551 | 554 | use_rust_index = True |
|
552 | 555 | else: |
|
553 | 556 | use_rust_index = self.opener.options.get(b'rust.index') |
|
554 | 557 | |
|
555 | 558 | self._parse_index = parse_index_v1 |
|
556 | 559 | if self._format_version == REVLOGV0: |
|
557 | 560 | self._parse_index = revlogv0.parse_index_v0 |
|
558 | 561 | elif fmt == REVLOGV2: |
|
559 | 562 | self._parse_index = parse_index_v2 |
|
560 | 563 | elif devel_nodemap: |
|
561 | 564 | self._parse_index = parse_index_v1_nodemap |
|
562 | 565 | elif use_rust_index: |
|
563 | 566 | self._parse_index = parse_index_v1_mixed |
|
564 | 567 | try: |
|
565 | 568 | d = self._parse_index(indexdata, self._inline) |
|
566 | 569 | index, _chunkcache = d |
|
567 | 570 | use_nodemap = ( |
|
568 | 571 | not self._inline |
|
569 | 572 | and self._nodemap_file is not None |
|
570 | 573 | and util.safehasattr(index, 'update_nodemap_data') |
|
571 | 574 | ) |
|
572 | 575 | if use_nodemap: |
|
573 | 576 | nodemap_data = nodemaputil.persisted_data(self) |
|
574 | 577 | if nodemap_data is not None: |
|
575 | 578 | docket = nodemap_data[0] |
|
576 | 579 | if ( |
|
577 | 580 | len(d[0]) > docket.tip_rev |
|
578 | 581 | and d[0][docket.tip_rev][7] == docket.tip_node |
|
579 | 582 | ): |
|
580 | 583 | # no changelog tampering |
|
581 | 584 | self._nodemap_docket = docket |
|
582 | 585 | index.update_nodemap_data(*nodemap_data) |
|
583 | 586 | except (ValueError, IndexError): |
|
584 | 587 | raise error.RevlogError( |
|
585 | 588 | _(b"index %s is corrupted") % self.display_id |
|
586 | 589 | ) |
|
587 | 590 | self.index, self._chunkcache = d |
|
588 | 591 | if not self._chunkcache: |
|
589 | 592 | self._chunkclear() |
|
590 | 593 | # revnum -> (chain-length, sum-delta-length) |
|
591 | 594 | self._chaininfocache = util.lrucachedict(500) |
|
592 | 595 | # revlog header -> revlog compressor |
|
593 | 596 | self._decompressors = {} |
|
594 | 597 | |
|
595 | 598 | @util.propertycache |
|
596 | 599 | def revlog_kind(self): |
|
597 | 600 | return self.target[0] |
|
598 | 601 | |
|
599 | 602 | @util.propertycache |
|
600 | 603 | def display_id(self): |
|
601 | 604 | """The public facing "ID" of the revlog that we use in message""" |
|
602 | 605 | # Maybe we should build a user facing representation of |
|
603 | 606 | # revlog.target instead of using `self.radix` |
|
604 | 607 | return self.radix |
|
605 | 608 | |
|
606 | 609 | @util.propertycache |
|
607 | 610 | def _compressor(self): |
|
608 | 611 | engine = util.compengines[self._compengine] |
|
609 | 612 | return engine.revlogcompressor(self._compengineopts) |
|
610 | 613 | |
|
611 | 614 | def _indexfp(self, mode=b'r'): |
|
612 | 615 | """file object for the revlog's index file""" |
|
613 | 616 | args = {'mode': mode} |
|
614 | 617 | if mode != b'r': |
|
615 | 618 | args['checkambig'] = self._checkambig |
|
616 | 619 | if mode == b'w': |
|
617 | 620 | args['atomictemp'] = True |
|
618 | 621 | return self.opener(self._indexfile, **args) |
|
619 | 622 | |
|
620 | 623 | def _datafp(self, mode=b'r'): |
|
621 | 624 | """file object for the revlog's data file""" |
|
622 | 625 | return self.opener(self._datafile, mode=mode) |
|
623 | 626 | |
|
624 | 627 | @contextlib.contextmanager |
|
625 | 628 | def _datareadfp(self, existingfp=None): |
|
626 | 629 | """file object suitable to read data""" |
|
627 | 630 | # Use explicit file handle, if given. |
|
628 | 631 | if existingfp is not None: |
|
629 | 632 | yield existingfp |
|
630 | 633 | |
|
631 | 634 | # Use a file handle being actively used for writes, if available. |
|
632 | 635 | # There is some danger to doing this because reads will seek the |
|
633 | 636 | # file. However, _writeentry() performs a SEEK_END before all writes, |
|
634 | 637 | # so we should be safe. |
|
635 | 638 | elif self._writinghandles: |
|
636 | 639 | if self._inline: |
|
637 | 640 | yield self._writinghandles[0] |
|
638 | 641 | else: |
|
639 | 642 | yield self._writinghandles[1] |
|
640 | 643 | |
|
641 | 644 | # Otherwise open a new file handle. |
|
642 | 645 | else: |
|
643 | 646 | if self._inline: |
|
644 | 647 | func = self._indexfp |
|
645 | 648 | else: |
|
646 | 649 | func = self._datafp |
|
647 | 650 | with func() as fp: |
|
648 | 651 | yield fp |
|
649 | 652 | |
|
650 | 653 | def tiprev(self): |
|
651 | 654 | return len(self.index) - 1 |
|
652 | 655 | |
|
653 | 656 | def tip(self): |
|
654 | 657 | return self.node(self.tiprev()) |
|
655 | 658 | |
|
656 | 659 | def __contains__(self, rev): |
|
657 | 660 | return 0 <= rev < len(self) |
|
658 | 661 | |
|
659 | 662 | def __len__(self): |
|
660 | 663 | return len(self.index) |
|
661 | 664 | |
|
662 | 665 | def __iter__(self): |
|
663 | 666 | return iter(pycompat.xrange(len(self))) |
|
664 | 667 | |
|
665 | 668 | def revs(self, start=0, stop=None): |
|
666 | 669 | """iterate over all rev in this revlog (from start to stop)""" |
|
667 | 670 | return storageutil.iterrevs(len(self), start=start, stop=stop) |
|
668 | 671 | |
|
669 | 672 | @property |
|
670 | 673 | def nodemap(self): |
|
671 | 674 | msg = ( |
|
672 | 675 | b"revlog.nodemap is deprecated, " |
|
673 | 676 | b"use revlog.index.[has_node|rev|get_rev]" |
|
674 | 677 | ) |
|
675 | 678 | util.nouideprecwarn(msg, b'5.3', stacklevel=2) |
|
676 | 679 | return self.index.nodemap |
|
677 | 680 | |
|
678 | 681 | @property |
|
679 | 682 | def _nodecache(self): |
|
680 | 683 | msg = b"revlog._nodecache is deprecated, use revlog.index.nodemap" |
|
681 | 684 | util.nouideprecwarn(msg, b'5.3', stacklevel=2) |
|
682 | 685 | return self.index.nodemap |
|
683 | 686 | |
|
684 | 687 | def hasnode(self, node): |
|
685 | 688 | try: |
|
686 | 689 | self.rev(node) |
|
687 | 690 | return True |
|
688 | 691 | except KeyError: |
|
689 | 692 | return False |
|
690 | 693 | |
|
691 | 694 | def candelta(self, baserev, rev): |
|
692 | 695 | """whether two revisions (baserev, rev) can be delta-ed or not""" |
|
693 | 696 | # Disable delta if either rev requires a content-changing flag |
|
694 | 697 | # processor (ex. LFS). This is because such flag processor can alter |
|
695 | 698 | # the rawtext content that the delta will be based on, and two clients |
|
696 | 699 | # could have a same revlog node with different flags (i.e. different |
|
697 | 700 | # rawtext contents) and the delta could be incompatible. |
|
698 | 701 | if (self.flags(baserev) & REVIDX_RAWTEXT_CHANGING_FLAGS) or ( |
|
699 | 702 | self.flags(rev) & REVIDX_RAWTEXT_CHANGING_FLAGS |
|
700 | 703 | ): |
|
701 | 704 | return False |
|
702 | 705 | return True |
|
703 | 706 | |
|
704 | 707 | def update_caches(self, transaction): |
|
705 | 708 | if self._nodemap_file is not None: |
|
706 | 709 | if transaction is None: |
|
707 | 710 | nodemaputil.update_persistent_nodemap(self) |
|
708 | 711 | else: |
|
709 | 712 | nodemaputil.setup_persistent_nodemap(transaction, self) |
|
710 | 713 | |
|
711 | 714 | def clearcaches(self): |
|
712 | 715 | self._revisioncache = None |
|
713 | 716 | self._chainbasecache.clear() |
|
714 | 717 | self._chunkcache = (0, b'') |
|
715 | 718 | self._pcache = {} |
|
716 | 719 | self._nodemap_docket = None |
|
717 | 720 | self.index.clearcaches() |
|
718 | 721 | # The python code is the one responsible for validating the docket, we |
|
719 | 722 | # end up having to refresh it here. |
|
720 | 723 | use_nodemap = ( |
|
721 | 724 | not self._inline |
|
722 | 725 | and self._nodemap_file is not None |
|
723 | 726 | and util.safehasattr(self.index, 'update_nodemap_data') |
|
724 | 727 | ) |
|
725 | 728 | if use_nodemap: |
|
726 | 729 | nodemap_data = nodemaputil.persisted_data(self) |
|
727 | 730 | if nodemap_data is not None: |
|
728 | 731 | self._nodemap_docket = nodemap_data[0] |
|
729 | 732 | self.index.update_nodemap_data(*nodemap_data) |
|
730 | 733 | |
|
731 | 734 | def rev(self, node): |
|
732 | 735 | try: |
|
733 | 736 | return self.index.rev(node) |
|
734 | 737 | except TypeError: |
|
735 | 738 | raise |
|
736 | 739 | except error.RevlogError: |
|
737 | 740 | # parsers.c radix tree lookup failed |
|
738 | 741 | if ( |
|
739 | 742 | node == self.nodeconstants.wdirid |
|
740 | 743 | or node in self.nodeconstants.wdirfilenodeids |
|
741 | 744 | ): |
|
742 | 745 | raise error.WdirUnsupported |
|
743 | 746 | raise error.LookupError(node, self.display_id, _(b'no node')) |
|
744 | 747 | |
|
745 | 748 | # Accessors for index entries. |
|
746 | 749 | |
|
747 | 750 | # First tuple entry is 8 bytes. First 6 bytes are offset. Last 2 bytes |
|
748 | 751 | # are flags. |
|
749 | 752 | def start(self, rev): |
|
750 | 753 | return int(self.index[rev][0] >> 16) |
|
751 | 754 | |
|
752 | 755 | def flags(self, rev): |
|
753 | 756 | return self.index[rev][0] & 0xFFFF |
|
754 | 757 | |
|
755 | 758 | def length(self, rev): |
|
756 | 759 | return self.index[rev][1] |
|
757 | 760 | |
|
758 | 761 | def sidedata_length(self, rev): |
|
759 | 762 | if not self.hassidedata: |
|
760 | 763 | return 0 |
|
761 | 764 | return self.index[rev][9] |
|
762 | 765 | |
|
763 | 766 | def rawsize(self, rev): |
|
764 | 767 | """return the length of the uncompressed text for a given revision""" |
|
765 | 768 | l = self.index[rev][2] |
|
766 | 769 | if l >= 0: |
|
767 | 770 | return l |
|
768 | 771 | |
|
769 | 772 | t = self.rawdata(rev) |
|
770 | 773 | return len(t) |
|
771 | 774 | |
|
772 | 775 | def size(self, rev): |
|
773 | 776 | """length of non-raw text (processed by a "read" flag processor)""" |
|
774 | 777 | # fast path: if no "read" flag processor could change the content, |
|
775 | 778 | # size is rawsize. note: ELLIPSIS is known to not change the content. |
|
776 | 779 | flags = self.flags(rev) |
|
777 | 780 | if flags & (flagutil.REVIDX_KNOWN_FLAGS ^ REVIDX_ELLIPSIS) == 0: |
|
778 | 781 | return self.rawsize(rev) |
|
779 | 782 | |
|
780 | 783 | return len(self.revision(rev, raw=False)) |
|
781 | 784 | |
|
782 | 785 | def chainbase(self, rev): |
|
783 | 786 | base = self._chainbasecache.get(rev) |
|
784 | 787 | if base is not None: |
|
785 | 788 | return base |
|
786 | 789 | |
|
787 | 790 | index = self.index |
|
788 | 791 | iterrev = rev |
|
789 | 792 | base = index[iterrev][3] |
|
790 | 793 | while base != iterrev: |
|
791 | 794 | iterrev = base |
|
792 | 795 | base = index[iterrev][3] |
|
793 | 796 | |
|
794 | 797 | self._chainbasecache[rev] = base |
|
795 | 798 | return base |
|
796 | 799 | |
|
797 | 800 | def linkrev(self, rev): |
|
798 | 801 | return self.index[rev][4] |
|
799 | 802 | |
|
800 | 803 | def parentrevs(self, rev): |
|
801 | 804 | try: |
|
802 | 805 | entry = self.index[rev] |
|
803 | 806 | except IndexError: |
|
804 | 807 | if rev == wdirrev: |
|
805 | 808 | raise error.WdirUnsupported |
|
806 | 809 | raise |
|
807 | 810 | if entry[5] == nullrev: |
|
808 | 811 | return entry[6], entry[5] |
|
809 | 812 | else: |
|
810 | 813 | return entry[5], entry[6] |
|
811 | 814 | |
|
812 | 815 | # fast parentrevs(rev) where rev isn't filtered |
|
813 | 816 | _uncheckedparentrevs = parentrevs |
|
814 | 817 | |
|
815 | 818 | def node(self, rev): |
|
816 | 819 | try: |
|
817 | 820 | return self.index[rev][7] |
|
818 | 821 | except IndexError: |
|
819 | 822 | if rev == wdirrev: |
|
820 | 823 | raise error.WdirUnsupported |
|
821 | 824 | raise |
|
822 | 825 | |
|
823 | 826 | # Derived from index values. |
|
824 | 827 | |
|
825 | 828 | def end(self, rev): |
|
826 | 829 | return self.start(rev) + self.length(rev) |
|
827 | 830 | |
|
828 | 831 | def parents(self, node): |
|
829 | 832 | i = self.index |
|
830 | 833 | d = i[self.rev(node)] |
|
831 | 834 | # inline node() to avoid function call overhead |
|
832 | 835 | if d[5] == self.nullid: |
|
833 | 836 | return i[d[6]][7], i[d[5]][7] |
|
834 | 837 | else: |
|
835 | 838 | return i[d[5]][7], i[d[6]][7] |
|
836 | 839 | |
|
837 | 840 | def chainlen(self, rev): |
|
838 | 841 | return self._chaininfo(rev)[0] |
|
839 | 842 | |
|
840 | 843 | def _chaininfo(self, rev): |
|
841 | 844 | chaininfocache = self._chaininfocache |
|
842 | 845 | if rev in chaininfocache: |
|
843 | 846 | return chaininfocache[rev] |
|
844 | 847 | index = self.index |
|
845 | 848 | generaldelta = self._generaldelta |
|
846 | 849 | iterrev = rev |
|
847 | 850 | e = index[iterrev] |
|
848 | 851 | clen = 0 |
|
849 | 852 | compresseddeltalen = 0 |
|
850 | 853 | while iterrev != e[3]: |
|
851 | 854 | clen += 1 |
|
852 | 855 | compresseddeltalen += e[1] |
|
853 | 856 | if generaldelta: |
|
854 | 857 | iterrev = e[3] |
|
855 | 858 | else: |
|
856 | 859 | iterrev -= 1 |
|
857 | 860 | if iterrev in chaininfocache: |
|
858 | 861 | t = chaininfocache[iterrev] |
|
859 | 862 | clen += t[0] |
|
860 | 863 | compresseddeltalen += t[1] |
|
861 | 864 | break |
|
862 | 865 | e = index[iterrev] |
|
863 | 866 | else: |
|
864 | 867 | # Add text length of base since decompressing that also takes |
|
865 | 868 | # work. For cache hits the length is already included. |
|
866 | 869 | compresseddeltalen += e[1] |
|
867 | 870 | r = (clen, compresseddeltalen) |
|
868 | 871 | chaininfocache[rev] = r |
|
869 | 872 | return r |
|
870 | 873 | |
|
871 | 874 | def _deltachain(self, rev, stoprev=None): |
|
872 | 875 | """Obtain the delta chain for a revision. |
|
873 | 876 | |
|
874 | 877 | ``stoprev`` specifies a revision to stop at. If not specified, we |
|
875 | 878 | stop at the base of the chain. |
|
876 | 879 | |
|
877 | 880 | Returns a 2-tuple of (chain, stopped) where ``chain`` is a list of |
|
878 | 881 | revs in ascending order and ``stopped`` is a bool indicating whether |
|
879 | 882 | ``stoprev`` was hit. |
|
880 | 883 | """ |
|
881 | 884 | # Try C implementation. |
|
882 | 885 | try: |
|
883 | 886 | return self.index.deltachain(rev, stoprev, self._generaldelta) |
|
884 | 887 | except AttributeError: |
|
885 | 888 | pass |
|
886 | 889 | |
|
887 | 890 | chain = [] |
|
888 | 891 | |
|
889 | 892 | # Alias to prevent attribute lookup in tight loop. |
|
890 | 893 | index = self.index |
|
891 | 894 | generaldelta = self._generaldelta |
|
892 | 895 | |
|
893 | 896 | iterrev = rev |
|
894 | 897 | e = index[iterrev] |
|
895 | 898 | while iterrev != e[3] and iterrev != stoprev: |
|
896 | 899 | chain.append(iterrev) |
|
897 | 900 | if generaldelta: |
|
898 | 901 | iterrev = e[3] |
|
899 | 902 | else: |
|
900 | 903 | iterrev -= 1 |
|
901 | 904 | e = index[iterrev] |
|
902 | 905 | |
|
903 | 906 | if iterrev == stoprev: |
|
904 | 907 | stopped = True |
|
905 | 908 | else: |
|
906 | 909 | chain.append(iterrev) |
|
907 | 910 | stopped = False |
|
908 | 911 | |
|
909 | 912 | chain.reverse() |
|
910 | 913 | return chain, stopped |
|
911 | 914 | |
|
912 | 915 | def ancestors(self, revs, stoprev=0, inclusive=False): |
|
913 | 916 | """Generate the ancestors of 'revs' in reverse revision order. |
|
914 | 917 | Does not generate revs lower than stoprev. |
|
915 | 918 | |
|
916 | 919 | See the documentation for ancestor.lazyancestors for more details.""" |
|
917 | 920 | |
|
918 | 921 | # first, make sure start revisions aren't filtered |
|
919 | 922 | revs = list(revs) |
|
920 | 923 | checkrev = self.node |
|
921 | 924 | for r in revs: |
|
922 | 925 | checkrev(r) |
|
923 | 926 | # and we're sure ancestors aren't filtered as well |
|
924 | 927 | |
|
925 | 928 | if rustancestor is not None: |
|
926 | 929 | lazyancestors = rustancestor.LazyAncestors |
|
927 | 930 | arg = self.index |
|
928 | 931 | else: |
|
929 | 932 | lazyancestors = ancestor.lazyancestors |
|
930 | 933 | arg = self._uncheckedparentrevs |
|
931 | 934 | return lazyancestors(arg, revs, stoprev=stoprev, inclusive=inclusive) |
|
932 | 935 | |
|
933 | 936 | def descendants(self, revs): |
|
934 | 937 | return dagop.descendantrevs(revs, self.revs, self.parentrevs) |
|
935 | 938 | |
|
936 | 939 | def findcommonmissing(self, common=None, heads=None): |
|
937 | 940 | """Return a tuple of the ancestors of common and the ancestors of heads |
|
938 | 941 | that are not ancestors of common. In revset terminology, we return the |
|
939 | 942 | tuple: |
|
940 | 943 | |
|
941 | 944 | ::common, (::heads) - (::common) |
|
942 | 945 | |
|
943 | 946 | The list is sorted by revision number, meaning it is |
|
944 | 947 | topologically sorted. |
|
945 | 948 | |
|
946 | 949 | 'heads' and 'common' are both lists of node IDs. If heads is |
|
947 | 950 | not supplied, uses all of the revlog's heads. If common is not |
|
948 | 951 | supplied, uses nullid.""" |
|
949 | 952 | if common is None: |
|
950 | 953 | common = [self.nullid] |
|
951 | 954 | if heads is None: |
|
952 | 955 | heads = self.heads() |
|
953 | 956 | |
|
954 | 957 | common = [self.rev(n) for n in common] |
|
955 | 958 | heads = [self.rev(n) for n in heads] |
|
956 | 959 | |
|
957 | 960 | # we want the ancestors, but inclusive |
|
958 | 961 | class lazyset(object): |
|
959 | 962 | def __init__(self, lazyvalues): |
|
960 | 963 | self.addedvalues = set() |
|
961 | 964 | self.lazyvalues = lazyvalues |
|
962 | 965 | |
|
963 | 966 | def __contains__(self, value): |
|
964 | 967 | return value in self.addedvalues or value in self.lazyvalues |
|
965 | 968 | |
|
966 | 969 | def __iter__(self): |
|
967 | 970 | added = self.addedvalues |
|
968 | 971 | for r in added: |
|
969 | 972 | yield r |
|
970 | 973 | for r in self.lazyvalues: |
|
971 | 974 | if not r in added: |
|
972 | 975 | yield r |
|
973 | 976 | |
|
974 | 977 | def add(self, value): |
|
975 | 978 | self.addedvalues.add(value) |
|
976 | 979 | |
|
977 | 980 | def update(self, values): |
|
978 | 981 | self.addedvalues.update(values) |
|
979 | 982 | |
|
980 | 983 | has = lazyset(self.ancestors(common)) |
|
981 | 984 | has.add(nullrev) |
|
982 | 985 | has.update(common) |
|
983 | 986 | |
|
984 | 987 | # take all ancestors from heads that aren't in has |
|
985 | 988 | missing = set() |
|
986 | 989 | visit = collections.deque(r for r in heads if r not in has) |
|
987 | 990 | while visit: |
|
988 | 991 | r = visit.popleft() |
|
989 | 992 | if r in missing: |
|
990 | 993 | continue |
|
991 | 994 | else: |
|
992 | 995 | missing.add(r) |
|
993 | 996 | for p in self.parentrevs(r): |
|
994 | 997 | if p not in has: |
|
995 | 998 | visit.append(p) |
|
996 | 999 | missing = list(missing) |
|
997 | 1000 | missing.sort() |
|
998 | 1001 | return has, [self.node(miss) for miss in missing] |
|
999 | 1002 | |
|
1000 | 1003 | def incrementalmissingrevs(self, common=None): |
|
1001 | 1004 | """Return an object that can be used to incrementally compute the |
|
1002 | 1005 | revision numbers of the ancestors of arbitrary sets that are not |
|
1003 | 1006 | ancestors of common. This is an ancestor.incrementalmissingancestors |
|
1004 | 1007 | object. |
|
1005 | 1008 | |
|
1006 | 1009 | 'common' is a list of revision numbers. If common is not supplied, uses |
|
1007 | 1010 | nullrev. |
|
1008 | 1011 | """ |
|
1009 | 1012 | if common is None: |
|
1010 | 1013 | common = [nullrev] |
|
1011 | 1014 | |
|
1012 | 1015 | if rustancestor is not None: |
|
1013 | 1016 | return rustancestor.MissingAncestors(self.index, common) |
|
1014 | 1017 | return ancestor.incrementalmissingancestors(self.parentrevs, common) |
|
1015 | 1018 | |
|
1016 | 1019 | def findmissingrevs(self, common=None, heads=None): |
|
1017 | 1020 | """Return the revision numbers of the ancestors of heads that |
|
1018 | 1021 | are not ancestors of common. |
|
1019 | 1022 | |
|
1020 | 1023 | More specifically, return a list of revision numbers corresponding to |
|
1021 | 1024 | nodes N such that every N satisfies the following constraints: |
|
1022 | 1025 | |
|
1023 | 1026 | 1. N is an ancestor of some node in 'heads' |
|
1024 | 1027 | 2. N is not an ancestor of any node in 'common' |
|
1025 | 1028 | |
|
1026 | 1029 | The list is sorted by revision number, meaning it is |
|
1027 | 1030 | topologically sorted. |
|
1028 | 1031 | |
|
1029 | 1032 | 'heads' and 'common' are both lists of revision numbers. If heads is |
|
1030 | 1033 | not supplied, uses all of the revlog's heads. If common is not |
|
1031 | 1034 | supplied, uses nullid.""" |
|
1032 | 1035 | if common is None: |
|
1033 | 1036 | common = [nullrev] |
|
1034 | 1037 | if heads is None: |
|
1035 | 1038 | heads = self.headrevs() |
|
1036 | 1039 | |
|
1037 | 1040 | inc = self.incrementalmissingrevs(common=common) |
|
1038 | 1041 | return inc.missingancestors(heads) |
|
1039 | 1042 | |
|
1040 | 1043 | def findmissing(self, common=None, heads=None): |
|
1041 | 1044 | """Return the ancestors of heads that are not ancestors of common. |
|
1042 | 1045 | |
|
1043 | 1046 | More specifically, return a list of nodes N such that every N |
|
1044 | 1047 | satisfies the following constraints: |
|
1045 | 1048 | |
|
1046 | 1049 | 1. N is an ancestor of some node in 'heads' |
|
1047 | 1050 | 2. N is not an ancestor of any node in 'common' |
|
1048 | 1051 | |
|
1049 | 1052 | The list is sorted by revision number, meaning it is |
|
1050 | 1053 | topologically sorted. |
|
1051 | 1054 | |
|
1052 | 1055 | 'heads' and 'common' are both lists of node IDs. If heads is |
|
1053 | 1056 | not supplied, uses all of the revlog's heads. If common is not |
|
1054 | 1057 | supplied, uses nullid.""" |
|
1055 | 1058 | if common is None: |
|
1056 | 1059 | common = [self.nullid] |
|
1057 | 1060 | if heads is None: |
|
1058 | 1061 | heads = self.heads() |
|
1059 | 1062 | |
|
1060 | 1063 | common = [self.rev(n) for n in common] |
|
1061 | 1064 | heads = [self.rev(n) for n in heads] |
|
1062 | 1065 | |
|
1063 | 1066 | inc = self.incrementalmissingrevs(common=common) |
|
1064 | 1067 | return [self.node(r) for r in inc.missingancestors(heads)] |
|
1065 | 1068 | |
|
1066 | 1069 | def nodesbetween(self, roots=None, heads=None): |
|
1067 | 1070 | """Return a topological path from 'roots' to 'heads'. |
|
1068 | 1071 | |
|
1069 | 1072 | Return a tuple (nodes, outroots, outheads) where 'nodes' is a |
|
1070 | 1073 | topologically sorted list of all nodes N that satisfy both of |
|
1071 | 1074 | these constraints: |
|
1072 | 1075 | |
|
1073 | 1076 | 1. N is a descendant of some node in 'roots' |
|
1074 | 1077 | 2. N is an ancestor of some node in 'heads' |
|
1075 | 1078 | |
|
1076 | 1079 | Every node is considered to be both a descendant and an ancestor |
|
1077 | 1080 | of itself, so every reachable node in 'roots' and 'heads' will be |
|
1078 | 1081 | included in 'nodes'. |
|
1079 | 1082 | |
|
1080 | 1083 | 'outroots' is the list of reachable nodes in 'roots', i.e., the |
|
1081 | 1084 | subset of 'roots' that is returned in 'nodes'. Likewise, |
|
1082 | 1085 | 'outheads' is the subset of 'heads' that is also in 'nodes'. |
|
1083 | 1086 | |
|
1084 | 1087 | 'roots' and 'heads' are both lists of node IDs. If 'roots' is |
|
1085 | 1088 | unspecified, uses nullid as the only root. If 'heads' is |
|
1086 | 1089 | unspecified, uses list of all of the revlog's heads.""" |
|
1087 | 1090 | nonodes = ([], [], []) |
|
1088 | 1091 | if roots is not None: |
|
1089 | 1092 | roots = list(roots) |
|
1090 | 1093 | if not roots: |
|
1091 | 1094 | return nonodes |
|
1092 | 1095 | lowestrev = min([self.rev(n) for n in roots]) |
|
1093 | 1096 | else: |
|
1094 | 1097 | roots = [self.nullid] # Everybody's a descendant of nullid |
|
1095 | 1098 | lowestrev = nullrev |
|
1096 | 1099 | if (lowestrev == nullrev) and (heads is None): |
|
1097 | 1100 | # We want _all_ the nodes! |
|
1098 | 1101 | return ( |
|
1099 | 1102 | [self.node(r) for r in self], |
|
1100 | 1103 | [self.nullid], |
|
1101 | 1104 | list(self.heads()), |
|
1102 | 1105 | ) |
|
1103 | 1106 | if heads is None: |
|
1104 | 1107 | # All nodes are ancestors, so the latest ancestor is the last |
|
1105 | 1108 | # node. |
|
1106 | 1109 | highestrev = len(self) - 1 |
|
1107 | 1110 | # Set ancestors to None to signal that every node is an ancestor. |
|
1108 | 1111 | ancestors = None |
|
1109 | 1112 | # Set heads to an empty dictionary for later discovery of heads |
|
1110 | 1113 | heads = {} |
|
1111 | 1114 | else: |
|
1112 | 1115 | heads = list(heads) |
|
1113 | 1116 | if not heads: |
|
1114 | 1117 | return nonodes |
|
1115 | 1118 | ancestors = set() |
|
1116 | 1119 | # Turn heads into a dictionary so we can remove 'fake' heads. |
|
1117 | 1120 | # Also, later we will be using it to filter out the heads we can't |
|
1118 | 1121 | # find from roots. |
|
1119 | 1122 | heads = dict.fromkeys(heads, False) |
|
1120 | 1123 | # Start at the top and keep marking parents until we're done. |
|
1121 | 1124 | nodestotag = set(heads) |
|
1122 | 1125 | # Remember where the top was so we can use it as a limit later. |
|
1123 | 1126 | highestrev = max([self.rev(n) for n in nodestotag]) |
|
1124 | 1127 | while nodestotag: |
|
1125 | 1128 | # grab a node to tag |
|
1126 | 1129 | n = nodestotag.pop() |
|
1127 | 1130 | # Never tag nullid |
|
1128 | 1131 | if n == self.nullid: |
|
1129 | 1132 | continue |
|
1130 | 1133 | # A node's revision number represents its place in a |
|
1131 | 1134 | # topologically sorted list of nodes. |
|
1132 | 1135 | r = self.rev(n) |
|
1133 | 1136 | if r >= lowestrev: |
|
1134 | 1137 | if n not in ancestors: |
|
1135 | 1138 | # If we are possibly a descendant of one of the roots |
|
1136 | 1139 | # and we haven't already been marked as an ancestor |
|
1137 | 1140 | ancestors.add(n) # Mark as ancestor |
|
1138 | 1141 | # Add non-nullid parents to list of nodes to tag. |
|
1139 | 1142 | nodestotag.update( |
|
1140 | 1143 | [p for p in self.parents(n) if p != self.nullid] |
|
1141 | 1144 | ) |
|
1142 | 1145 | elif n in heads: # We've seen it before, is it a fake head? |
|
1143 | 1146 | # So it is, real heads should not be the ancestors of |
|
1144 | 1147 | # any other heads. |
|
1145 | 1148 | heads.pop(n) |
|
1146 | 1149 | if not ancestors: |
|
1147 | 1150 | return nonodes |
|
1148 | 1151 | # Now that we have our set of ancestors, we want to remove any |
|
1149 | 1152 | # roots that are not ancestors. |
|
1150 | 1153 | |
|
1151 | 1154 | # If one of the roots was nullid, everything is included anyway. |
|
1152 | 1155 | if lowestrev > nullrev: |
|
1153 | 1156 | # But, since we weren't, let's recompute the lowest rev to not |
|
1154 | 1157 | # include roots that aren't ancestors. |
|
1155 | 1158 | |
|
1156 | 1159 | # Filter out roots that aren't ancestors of heads |
|
1157 | 1160 | roots = [root for root in roots if root in ancestors] |
|
1158 | 1161 | # Recompute the lowest revision |
|
1159 | 1162 | if roots: |
|
1160 | 1163 | lowestrev = min([self.rev(root) for root in roots]) |
|
1161 | 1164 | else: |
|
1162 | 1165 | # No more roots? Return empty list |
|
1163 | 1166 | return nonodes |
|
1164 | 1167 | else: |
|
1165 | 1168 | # We are descending from nullid, and don't need to care about |
|
1166 | 1169 | # any other roots. |
|
1167 | 1170 | lowestrev = nullrev |
|
1168 | 1171 | roots = [self.nullid] |
|
1169 | 1172 | # Transform our roots list into a set. |
|
1170 | 1173 | descendants = set(roots) |
|
1171 | 1174 | # Also, keep the original roots so we can filter out roots that aren't |
|
1172 | 1175 | # 'real' roots (i.e. are descended from other roots). |
|
1173 | 1176 | roots = descendants.copy() |
|
1174 | 1177 | # Our topologically sorted list of output nodes. |
|
1175 | 1178 | orderedout = [] |
|
1176 | 1179 | # Don't start at nullid since we don't want nullid in our output list, |
|
1177 | 1180 | # and if nullid shows up in descendants, empty parents will look like |
|
1178 | 1181 | # they're descendants. |
|
1179 | 1182 | for r in self.revs(start=max(lowestrev, 0), stop=highestrev + 1): |
|
1180 | 1183 | n = self.node(r) |
|
1181 | 1184 | isdescendant = False |
|
1182 | 1185 | if lowestrev == nullrev: # Everybody is a descendant of nullid |
|
1183 | 1186 | isdescendant = True |
|
1184 | 1187 | elif n in descendants: |
|
1185 | 1188 | # n is already a descendant |
|
1186 | 1189 | isdescendant = True |
|
1187 | 1190 | # This check only needs to be done here because all the roots |
|
1188 | 1191 | # will start being marked is descendants before the loop. |
|
1189 | 1192 | if n in roots: |
|
1190 | 1193 | # If n was a root, check if it's a 'real' root. |
|
1191 | 1194 | p = tuple(self.parents(n)) |
|
1192 | 1195 | # If any of its parents are descendants, it's not a root. |
|
1193 | 1196 | if (p[0] in descendants) or (p[1] in descendants): |
|
1194 | 1197 | roots.remove(n) |
|
1195 | 1198 | else: |
|
1196 | 1199 | p = tuple(self.parents(n)) |
|
1197 | 1200 | # A node is a descendant if either of its parents are |
|
1198 | 1201 | # descendants. (We seeded the dependents list with the roots |
|
1199 | 1202 | # up there, remember?) |
|
1200 | 1203 | if (p[0] in descendants) or (p[1] in descendants): |
|
1201 | 1204 | descendants.add(n) |
|
1202 | 1205 | isdescendant = True |
|
1203 | 1206 | if isdescendant and ((ancestors is None) or (n in ancestors)): |
|
1204 | 1207 | # Only include nodes that are both descendants and ancestors. |
|
1205 | 1208 | orderedout.append(n) |
|
1206 | 1209 | if (ancestors is not None) and (n in heads): |
|
1207 | 1210 | # We're trying to figure out which heads are reachable |
|
1208 | 1211 | # from roots. |
|
1209 | 1212 | # Mark this head as having been reached |
|
1210 | 1213 | heads[n] = True |
|
1211 | 1214 | elif ancestors is None: |
|
1212 | 1215 | # Otherwise, we're trying to discover the heads. |
|
1213 | 1216 | # Assume this is a head because if it isn't, the next step |
|
1214 | 1217 | # will eventually remove it. |
|
1215 | 1218 | heads[n] = True |
|
1216 | 1219 | # But, obviously its parents aren't. |
|
1217 | 1220 | for p in self.parents(n): |
|
1218 | 1221 | heads.pop(p, None) |
|
1219 | 1222 | heads = [head for head, flag in pycompat.iteritems(heads) if flag] |
|
1220 | 1223 | roots = list(roots) |
|
1221 | 1224 | assert orderedout |
|
1222 | 1225 | assert roots |
|
1223 | 1226 | assert heads |
|
1224 | 1227 | return (orderedout, roots, heads) |
|
1225 | 1228 | |
|
1226 | 1229 | def headrevs(self, revs=None): |
|
1227 | 1230 | if revs is None: |
|
1228 | 1231 | try: |
|
1229 | 1232 | return self.index.headrevs() |
|
1230 | 1233 | except AttributeError: |
|
1231 | 1234 | return self._headrevs() |
|
1232 | 1235 | if rustdagop is not None: |
|
1233 | 1236 | return rustdagop.headrevs(self.index, revs) |
|
1234 | 1237 | return dagop.headrevs(revs, self._uncheckedparentrevs) |
|
1235 | 1238 | |
|
1236 | 1239 | def computephases(self, roots): |
|
1237 | 1240 | return self.index.computephasesmapsets(roots) |
|
1238 | 1241 | |
|
1239 | 1242 | def _headrevs(self): |
|
1240 | 1243 | count = len(self) |
|
1241 | 1244 | if not count: |
|
1242 | 1245 | return [nullrev] |
|
1243 | 1246 | # we won't iter over filtered rev so nobody is a head at start |
|
1244 | 1247 | ishead = [0] * (count + 1) |
|
1245 | 1248 | index = self.index |
|
1246 | 1249 | for r in self: |
|
1247 | 1250 | ishead[r] = 1 # I may be an head |
|
1248 | 1251 | e = index[r] |
|
1249 | 1252 | ishead[e[5]] = ishead[e[6]] = 0 # my parent are not |
|
1250 | 1253 | return [r for r, val in enumerate(ishead) if val] |
|
1251 | 1254 | |
|
1252 | 1255 | def heads(self, start=None, stop=None): |
|
1253 | 1256 | """return the list of all nodes that have no children |
|
1254 | 1257 | |
|
1255 | 1258 | if start is specified, only heads that are descendants of |
|
1256 | 1259 | start will be returned |
|
1257 | 1260 | if stop is specified, it will consider all the revs from stop |
|
1258 | 1261 | as if they had no children |
|
1259 | 1262 | """ |
|
1260 | 1263 | if start is None and stop is None: |
|
1261 | 1264 | if not len(self): |
|
1262 | 1265 | return [self.nullid] |
|
1263 | 1266 | return [self.node(r) for r in self.headrevs()] |
|
1264 | 1267 | |
|
1265 | 1268 | if start is None: |
|
1266 | 1269 | start = nullrev |
|
1267 | 1270 | else: |
|
1268 | 1271 | start = self.rev(start) |
|
1269 | 1272 | |
|
1270 | 1273 | stoprevs = {self.rev(n) for n in stop or []} |
|
1271 | 1274 | |
|
1272 | 1275 | revs = dagop.headrevssubset( |
|
1273 | 1276 | self.revs, self.parentrevs, startrev=start, stoprevs=stoprevs |
|
1274 | 1277 | ) |
|
1275 | 1278 | |
|
1276 | 1279 | return [self.node(rev) for rev in revs] |
|
1277 | 1280 | |
|
1278 | 1281 | def children(self, node): |
|
1279 | 1282 | """find the children of a given node""" |
|
1280 | 1283 | c = [] |
|
1281 | 1284 | p = self.rev(node) |
|
1282 | 1285 | for r in self.revs(start=p + 1): |
|
1283 | 1286 | prevs = [pr for pr in self.parentrevs(r) if pr != nullrev] |
|
1284 | 1287 | if prevs: |
|
1285 | 1288 | for pr in prevs: |
|
1286 | 1289 | if pr == p: |
|
1287 | 1290 | c.append(self.node(r)) |
|
1288 | 1291 | elif p == nullrev: |
|
1289 | 1292 | c.append(self.node(r)) |
|
1290 | 1293 | return c |
|
1291 | 1294 | |
|
1292 | 1295 | def commonancestorsheads(self, a, b): |
|
1293 | 1296 | """calculate all the heads of the common ancestors of nodes a and b""" |
|
1294 | 1297 | a, b = self.rev(a), self.rev(b) |
|
1295 | 1298 | ancs = self._commonancestorsheads(a, b) |
|
1296 | 1299 | return pycompat.maplist(self.node, ancs) |
|
1297 | 1300 | |
|
1298 | 1301 | def _commonancestorsheads(self, *revs): |
|
1299 | 1302 | """calculate all the heads of the common ancestors of revs""" |
|
1300 | 1303 | try: |
|
1301 | 1304 | ancs = self.index.commonancestorsheads(*revs) |
|
1302 | 1305 | except (AttributeError, OverflowError): # C implementation failed |
|
1303 | 1306 | ancs = ancestor.commonancestorsheads(self.parentrevs, *revs) |
|
1304 | 1307 | return ancs |
|
1305 | 1308 | |
|
1306 | 1309 | def isancestor(self, a, b): |
|
1307 | 1310 | """return True if node a is an ancestor of node b |
|
1308 | 1311 | |
|
1309 | 1312 | A revision is considered an ancestor of itself.""" |
|
1310 | 1313 | a, b = self.rev(a), self.rev(b) |
|
1311 | 1314 | return self.isancestorrev(a, b) |
|
1312 | 1315 | |
|
1313 | 1316 | def isancestorrev(self, a, b): |
|
1314 | 1317 | """return True if revision a is an ancestor of revision b |
|
1315 | 1318 | |
|
1316 | 1319 | A revision is considered an ancestor of itself. |
|
1317 | 1320 | |
|
1318 | 1321 | The implementation of this is trivial but the use of |
|
1319 | 1322 | reachableroots is not.""" |
|
1320 | 1323 | if a == nullrev: |
|
1321 | 1324 | return True |
|
1322 | 1325 | elif a == b: |
|
1323 | 1326 | return True |
|
1324 | 1327 | elif a > b: |
|
1325 | 1328 | return False |
|
1326 | 1329 | return bool(self.reachableroots(a, [b], [a], includepath=False)) |
|
1327 | 1330 | |
|
1328 | 1331 | def reachableroots(self, minroot, heads, roots, includepath=False): |
|
1329 | 1332 | """return (heads(::(<roots> and <roots>::<heads>))) |
|
1330 | 1333 | |
|
1331 | 1334 | If includepath is True, return (<roots>::<heads>).""" |
|
1332 | 1335 | try: |
|
1333 | 1336 | return self.index.reachableroots2( |
|
1334 | 1337 | minroot, heads, roots, includepath |
|
1335 | 1338 | ) |
|
1336 | 1339 | except AttributeError: |
|
1337 | 1340 | return dagop._reachablerootspure( |
|
1338 | 1341 | self.parentrevs, minroot, roots, heads, includepath |
|
1339 | 1342 | ) |
|
1340 | 1343 | |
|
1341 | 1344 | def ancestor(self, a, b): |
|
1342 | 1345 | """calculate the "best" common ancestor of nodes a and b""" |
|
1343 | 1346 | |
|
1344 | 1347 | a, b = self.rev(a), self.rev(b) |
|
1345 | 1348 | try: |
|
1346 | 1349 | ancs = self.index.ancestors(a, b) |
|
1347 | 1350 | except (AttributeError, OverflowError): |
|
1348 | 1351 | ancs = ancestor.ancestors(self.parentrevs, a, b) |
|
1349 | 1352 | if ancs: |
|
1350 | 1353 | # choose a consistent winner when there's a tie |
|
1351 | 1354 | return min(map(self.node, ancs)) |
|
1352 | 1355 | return self.nullid |
|
1353 | 1356 | |
|
1354 | 1357 | def _match(self, id): |
|
1355 | 1358 | if isinstance(id, int): |
|
1356 | 1359 | # rev |
|
1357 | 1360 | return self.node(id) |
|
1358 | 1361 | if len(id) == self.nodeconstants.nodelen: |
|
1359 | 1362 | # possibly a binary node |
|
1360 | 1363 | # odds of a binary node being all hex in ASCII are 1 in 10**25 |
|
1361 | 1364 | try: |
|
1362 | 1365 | node = id |
|
1363 | 1366 | self.rev(node) # quick search the index |
|
1364 | 1367 | return node |
|
1365 | 1368 | except error.LookupError: |
|
1366 | 1369 | pass # may be partial hex id |
|
1367 | 1370 | try: |
|
1368 | 1371 | # str(rev) |
|
1369 | 1372 | rev = int(id) |
|
1370 | 1373 | if b"%d" % rev != id: |
|
1371 | 1374 | raise ValueError |
|
1372 | 1375 | if rev < 0: |
|
1373 | 1376 | rev = len(self) + rev |
|
1374 | 1377 | if rev < 0 or rev >= len(self): |
|
1375 | 1378 | raise ValueError |
|
1376 | 1379 | return self.node(rev) |
|
1377 | 1380 | except (ValueError, OverflowError): |
|
1378 | 1381 | pass |
|
1379 | 1382 | if len(id) == 2 * self.nodeconstants.nodelen: |
|
1380 | 1383 | try: |
|
1381 | 1384 | # a full hex nodeid? |
|
1382 | 1385 | node = bin(id) |
|
1383 | 1386 | self.rev(node) |
|
1384 | 1387 | return node |
|
1385 | 1388 | except (TypeError, error.LookupError): |
|
1386 | 1389 | pass |
|
1387 | 1390 | |
|
1388 | 1391 | def _partialmatch(self, id): |
|
1389 | 1392 | # we don't care wdirfilenodeids as they should be always full hash |
|
1390 | 1393 | maybewdir = self.nodeconstants.wdirhex.startswith(id) |
|
1391 | 1394 | try: |
|
1392 | 1395 | partial = self.index.partialmatch(id) |
|
1393 | 1396 | if partial and self.hasnode(partial): |
|
1394 | 1397 | if maybewdir: |
|
1395 | 1398 | # single 'ff...' match in radix tree, ambiguous with wdir |
|
1396 | 1399 | raise error.RevlogError |
|
1397 | 1400 | return partial |
|
1398 | 1401 | if maybewdir: |
|
1399 | 1402 | # no 'ff...' match in radix tree, wdir identified |
|
1400 | 1403 | raise error.WdirUnsupported |
|
1401 | 1404 | return None |
|
1402 | 1405 | except error.RevlogError: |
|
1403 | 1406 | # parsers.c radix tree lookup gave multiple matches |
|
1404 | 1407 | # fast path: for unfiltered changelog, radix tree is accurate |
|
1405 | 1408 | if not getattr(self, 'filteredrevs', None): |
|
1406 | 1409 | raise error.AmbiguousPrefixLookupError( |
|
1407 | 1410 | id, self.display_id, _(b'ambiguous identifier') |
|
1408 | 1411 | ) |
|
1409 | 1412 | # fall through to slow path that filters hidden revisions |
|
1410 | 1413 | except (AttributeError, ValueError): |
|
1411 | 1414 | # we are pure python, or key was too short to search radix tree |
|
1412 | 1415 | pass |
|
1413 | 1416 | |
|
1414 | 1417 | if id in self._pcache: |
|
1415 | 1418 | return self._pcache[id] |
|
1416 | 1419 | |
|
1417 | 1420 | if len(id) <= 40: |
|
1418 | 1421 | try: |
|
1419 | 1422 | # hex(node)[:...] |
|
1420 | 1423 | l = len(id) // 2 # grab an even number of digits |
|
1421 | 1424 | prefix = bin(id[: l * 2]) |
|
1422 | 1425 | nl = [e[7] for e in self.index if e[7].startswith(prefix)] |
|
1423 | 1426 | nl = [ |
|
1424 | 1427 | n for n in nl if hex(n).startswith(id) and self.hasnode(n) |
|
1425 | 1428 | ] |
|
1426 | 1429 | if self.nodeconstants.nullhex.startswith(id): |
|
1427 | 1430 | nl.append(self.nullid) |
|
1428 | 1431 | if len(nl) > 0: |
|
1429 | 1432 | if len(nl) == 1 and not maybewdir: |
|
1430 | 1433 | self._pcache[id] = nl[0] |
|
1431 | 1434 | return nl[0] |
|
1432 | 1435 | raise error.AmbiguousPrefixLookupError( |
|
1433 | 1436 | id, self.display_id, _(b'ambiguous identifier') |
|
1434 | 1437 | ) |
|
1435 | 1438 | if maybewdir: |
|
1436 | 1439 | raise error.WdirUnsupported |
|
1437 | 1440 | return None |
|
1438 | 1441 | except TypeError: |
|
1439 | 1442 | pass |
|
1440 | 1443 | |
|
1441 | 1444 | def lookup(self, id): |
|
1442 | 1445 | """locate a node based on: |
|
1443 | 1446 | - revision number or str(revision number) |
|
1444 | 1447 | - nodeid or subset of hex nodeid |
|
1445 | 1448 | """ |
|
1446 | 1449 | n = self._match(id) |
|
1447 | 1450 | if n is not None: |
|
1448 | 1451 | return n |
|
1449 | 1452 | n = self._partialmatch(id) |
|
1450 | 1453 | if n: |
|
1451 | 1454 | return n |
|
1452 | 1455 | |
|
1453 | 1456 | raise error.LookupError(id, self.display_id, _(b'no match found')) |
|
1454 | 1457 | |
|
1455 | 1458 | def shortest(self, node, minlength=1): |
|
1456 | 1459 | """Find the shortest unambiguous prefix that matches node.""" |
|
1457 | 1460 | |
|
1458 | 1461 | def isvalid(prefix): |
|
1459 | 1462 | try: |
|
1460 | 1463 | matchednode = self._partialmatch(prefix) |
|
1461 | 1464 | except error.AmbiguousPrefixLookupError: |
|
1462 | 1465 | return False |
|
1463 | 1466 | except error.WdirUnsupported: |
|
1464 | 1467 | # single 'ff...' match |
|
1465 | 1468 | return True |
|
1466 | 1469 | if matchednode is None: |
|
1467 | 1470 | raise error.LookupError(node, self.display_id, _(b'no node')) |
|
1468 | 1471 | return True |
|
1469 | 1472 | |
|
1470 | 1473 | def maybewdir(prefix): |
|
1471 | 1474 | return all(c == b'f' for c in pycompat.iterbytestr(prefix)) |
|
1472 | 1475 | |
|
1473 | 1476 | hexnode = hex(node) |
|
1474 | 1477 | |
|
1475 | 1478 | def disambiguate(hexnode, minlength): |
|
1476 | 1479 | """Disambiguate against wdirid.""" |
|
1477 | 1480 | for length in range(minlength, len(hexnode) + 1): |
|
1478 | 1481 | prefix = hexnode[:length] |
|
1479 | 1482 | if not maybewdir(prefix): |
|
1480 | 1483 | return prefix |
|
1481 | 1484 | |
|
1482 | 1485 | if not getattr(self, 'filteredrevs', None): |
|
1483 | 1486 | try: |
|
1484 | 1487 | length = max(self.index.shortest(node), minlength) |
|
1485 | 1488 | return disambiguate(hexnode, length) |
|
1486 | 1489 | except error.RevlogError: |
|
1487 | 1490 | if node != self.nodeconstants.wdirid: |
|
1488 | 1491 | raise error.LookupError( |
|
1489 | 1492 | node, self.display_id, _(b'no node') |
|
1490 | 1493 | ) |
|
1491 | 1494 | except AttributeError: |
|
1492 | 1495 | # Fall through to pure code |
|
1493 | 1496 | pass |
|
1494 | 1497 | |
|
1495 | 1498 | if node == self.nodeconstants.wdirid: |
|
1496 | 1499 | for length in range(minlength, len(hexnode) + 1): |
|
1497 | 1500 | prefix = hexnode[:length] |
|
1498 | 1501 | if isvalid(prefix): |
|
1499 | 1502 | return prefix |
|
1500 | 1503 | |
|
1501 | 1504 | for length in range(minlength, len(hexnode) + 1): |
|
1502 | 1505 | prefix = hexnode[:length] |
|
1503 | 1506 | if isvalid(prefix): |
|
1504 | 1507 | return disambiguate(hexnode, length) |
|
1505 | 1508 | |
|
1506 | 1509 | def cmp(self, node, text): |
|
1507 | 1510 | """compare text with a given file revision |
|
1508 | 1511 | |
|
1509 | 1512 | returns True if text is different than what is stored. |
|
1510 | 1513 | """ |
|
1511 | 1514 | p1, p2 = self.parents(node) |
|
1512 | 1515 | return storageutil.hashrevisionsha1(text, p1, p2) != node |
|
1513 | 1516 | |
|
1514 | 1517 | def _cachesegment(self, offset, data): |
|
1515 | 1518 | """Add a segment to the revlog cache. |
|
1516 | 1519 | |
|
1517 | 1520 | Accepts an absolute offset and the data that is at that location. |
|
1518 | 1521 | """ |
|
1519 | 1522 | o, d = self._chunkcache |
|
1520 | 1523 | # try to add to existing cache |
|
1521 | 1524 | if o + len(d) == offset and len(d) + len(data) < _chunksize: |
|
1522 | 1525 | self._chunkcache = o, d + data |
|
1523 | 1526 | else: |
|
1524 | 1527 | self._chunkcache = offset, data |
|
1525 | 1528 | |
|
1526 | 1529 | def _readsegment(self, offset, length, df=None): |
|
1527 | 1530 | """Load a segment of raw data from the revlog. |
|
1528 | 1531 | |
|
1529 | 1532 | Accepts an absolute offset, length to read, and an optional existing |
|
1530 | 1533 | file handle to read from. |
|
1531 | 1534 | |
|
1532 | 1535 | If an existing file handle is passed, it will be seeked and the |
|
1533 | 1536 | original seek position will NOT be restored. |
|
1534 | 1537 | |
|
1535 | 1538 | Returns a str or buffer of raw byte data. |
|
1536 | 1539 | |
|
1537 | 1540 | Raises if the requested number of bytes could not be read. |
|
1538 | 1541 | """ |
|
1539 | 1542 | # Cache data both forward and backward around the requested |
|
1540 | 1543 | # data, in a fixed size window. This helps speed up operations |
|
1541 | 1544 | # involving reading the revlog backwards. |
|
1542 | 1545 | cachesize = self._chunkcachesize |
|
1543 | 1546 | realoffset = offset & ~(cachesize - 1) |
|
1544 | 1547 | reallength = ( |
|
1545 | 1548 | (offset + length + cachesize) & ~(cachesize - 1) |
|
1546 | 1549 | ) - realoffset |
|
1547 | 1550 | with self._datareadfp(df) as df: |
|
1548 | 1551 | df.seek(realoffset) |
|
1549 | 1552 | d = df.read(reallength) |
|
1550 | 1553 | |
|
1551 | 1554 | self._cachesegment(realoffset, d) |
|
1552 | 1555 | if offset != realoffset or reallength != length: |
|
1553 | 1556 | startoffset = offset - realoffset |
|
1554 | 1557 | if len(d) - startoffset < length: |
|
1555 | 1558 | raise error.RevlogError( |
|
1556 | 1559 | _( |
|
1557 | 1560 | b'partial read of revlog %s; expected %d bytes from ' |
|
1558 | 1561 | b'offset %d, got %d' |
|
1559 | 1562 | ) |
|
1560 | 1563 | % ( |
|
1561 | 1564 | self._indexfile if self._inline else self._datafile, |
|
1562 | 1565 | length, |
|
1563 | 1566 | offset, |
|
1564 | 1567 | len(d) - startoffset, |
|
1565 | 1568 | ) |
|
1566 | 1569 | ) |
|
1567 | 1570 | |
|
1568 | 1571 | return util.buffer(d, startoffset, length) |
|
1569 | 1572 | |
|
1570 | 1573 | if len(d) < length: |
|
1571 | 1574 | raise error.RevlogError( |
|
1572 | 1575 | _( |
|
1573 | 1576 | b'partial read of revlog %s; expected %d bytes from offset ' |
|
1574 | 1577 | b'%d, got %d' |
|
1575 | 1578 | ) |
|
1576 | 1579 | % ( |
|
1577 | 1580 | self._indexfile if self._inline else self._datafile, |
|
1578 | 1581 | length, |
|
1579 | 1582 | offset, |
|
1580 | 1583 | len(d), |
|
1581 | 1584 | ) |
|
1582 | 1585 | ) |
|
1583 | 1586 | |
|
1584 | 1587 | return d |
|
1585 | 1588 | |
|
1586 | 1589 | def _getsegment(self, offset, length, df=None): |
|
1587 | 1590 | """Obtain a segment of raw data from the revlog. |
|
1588 | 1591 | |
|
1589 | 1592 | Accepts an absolute offset, length of bytes to obtain, and an |
|
1590 | 1593 | optional file handle to the already-opened revlog. If the file |
|
1591 | 1594 | handle is used, it's original seek position will not be preserved. |
|
1592 | 1595 | |
|
1593 | 1596 | Requests for data may be returned from a cache. |
|
1594 | 1597 | |
|
1595 | 1598 | Returns a str or a buffer instance of raw byte data. |
|
1596 | 1599 | """ |
|
1597 | 1600 | o, d = self._chunkcache |
|
1598 | 1601 | l = len(d) |
|
1599 | 1602 | |
|
1600 | 1603 | # is it in the cache? |
|
1601 | 1604 | cachestart = offset - o |
|
1602 | 1605 | cacheend = cachestart + length |
|
1603 | 1606 | if cachestart >= 0 and cacheend <= l: |
|
1604 | 1607 | if cachestart == 0 and cacheend == l: |
|
1605 | 1608 | return d # avoid a copy |
|
1606 | 1609 | return util.buffer(d, cachestart, cacheend - cachestart) |
|
1607 | 1610 | |
|
1608 | 1611 | return self._readsegment(offset, length, df=df) |
|
1609 | 1612 | |
|
1610 | 1613 | def _getsegmentforrevs(self, startrev, endrev, df=None): |
|
1611 | 1614 | """Obtain a segment of raw data corresponding to a range of revisions. |
|
1612 | 1615 | |
|
1613 | 1616 | Accepts the start and end revisions and an optional already-open |
|
1614 | 1617 | file handle to be used for reading. If the file handle is read, its |
|
1615 | 1618 | seek position will not be preserved. |
|
1616 | 1619 | |
|
1617 | 1620 | Requests for data may be satisfied by a cache. |
|
1618 | 1621 | |
|
1619 | 1622 | Returns a 2-tuple of (offset, data) for the requested range of |
|
1620 | 1623 | revisions. Offset is the integer offset from the beginning of the |
|
1621 | 1624 | revlog and data is a str or buffer of the raw byte data. |
|
1622 | 1625 | |
|
1623 | 1626 | Callers will need to call ``self.start(rev)`` and ``self.length(rev)`` |
|
1624 | 1627 | to determine where each revision's data begins and ends. |
|
1625 | 1628 | """ |
|
1626 | 1629 | # Inlined self.start(startrev) & self.end(endrev) for perf reasons |
|
1627 | 1630 | # (functions are expensive). |
|
1628 | 1631 | index = self.index |
|
1629 | 1632 | istart = index[startrev] |
|
1630 | 1633 | start = int(istart[0] >> 16) |
|
1631 | 1634 | if startrev == endrev: |
|
1632 | 1635 | end = start + istart[1] |
|
1633 | 1636 | else: |
|
1634 | 1637 | iend = index[endrev] |
|
1635 | 1638 | end = int(iend[0] >> 16) + iend[1] |
|
1636 | 1639 | |
|
1637 | 1640 | if self._inline: |
|
1638 | 1641 | start += (startrev + 1) * self.index.entry_size |
|
1639 | 1642 | end += (endrev + 1) * self.index.entry_size |
|
1640 | 1643 | length = end - start |
|
1641 | 1644 | |
|
1642 | 1645 | return start, self._getsegment(start, length, df=df) |
|
1643 | 1646 | |
|
1644 | 1647 | def _chunk(self, rev, df=None): |
|
1645 | 1648 | """Obtain a single decompressed chunk for a revision. |
|
1646 | 1649 | |
|
1647 | 1650 | Accepts an integer revision and an optional already-open file handle |
|
1648 | 1651 | to be used for reading. If used, the seek position of the file will not |
|
1649 | 1652 | be preserved. |
|
1650 | 1653 | |
|
1651 | 1654 | Returns a str holding uncompressed data for the requested revision. |
|
1652 | 1655 | """ |
|
1653 | 1656 | return self.decompress(self._getsegmentforrevs(rev, rev, df=df)[1]) |
|
1654 | 1657 | |
|
1655 | 1658 | def _chunks(self, revs, df=None, targetsize=None): |
|
1656 | 1659 | """Obtain decompressed chunks for the specified revisions. |
|
1657 | 1660 | |
|
1658 | 1661 | Accepts an iterable of numeric revisions that are assumed to be in |
|
1659 | 1662 | ascending order. Also accepts an optional already-open file handle |
|
1660 | 1663 | to be used for reading. If used, the seek position of the file will |
|
1661 | 1664 | not be preserved. |
|
1662 | 1665 | |
|
1663 | 1666 | This function is similar to calling ``self._chunk()`` multiple times, |
|
1664 | 1667 | but is faster. |
|
1665 | 1668 | |
|
1666 | 1669 | Returns a list with decompressed data for each requested revision. |
|
1667 | 1670 | """ |
|
1668 | 1671 | if not revs: |
|
1669 | 1672 | return [] |
|
1670 | 1673 | start = self.start |
|
1671 | 1674 | length = self.length |
|
1672 | 1675 | inline = self._inline |
|
1673 | 1676 | iosize = self.index.entry_size |
|
1674 | 1677 | buffer = util.buffer |
|
1675 | 1678 | |
|
1676 | 1679 | l = [] |
|
1677 | 1680 | ladd = l.append |
|
1678 | 1681 | |
|
1679 | 1682 | if not self._withsparseread: |
|
1680 | 1683 | slicedchunks = (revs,) |
|
1681 | 1684 | else: |
|
1682 | 1685 | slicedchunks = deltautil.slicechunk( |
|
1683 | 1686 | self, revs, targetsize=targetsize |
|
1684 | 1687 | ) |
|
1685 | 1688 | |
|
1686 | 1689 | for revschunk in slicedchunks: |
|
1687 | 1690 | firstrev = revschunk[0] |
|
1688 | 1691 | # Skip trailing revisions with empty diff |
|
1689 | 1692 | for lastrev in revschunk[::-1]: |
|
1690 | 1693 | if length(lastrev) != 0: |
|
1691 | 1694 | break |
|
1692 | 1695 | |
|
1693 | 1696 | try: |
|
1694 | 1697 | offset, data = self._getsegmentforrevs(firstrev, lastrev, df=df) |
|
1695 | 1698 | except OverflowError: |
|
1696 | 1699 | # issue4215 - we can't cache a run of chunks greater than |
|
1697 | 1700 | # 2G on Windows |
|
1698 | 1701 | return [self._chunk(rev, df=df) for rev in revschunk] |
|
1699 | 1702 | |
|
1700 | 1703 | decomp = self.decompress |
|
1701 | 1704 | for rev in revschunk: |
|
1702 | 1705 | chunkstart = start(rev) |
|
1703 | 1706 | if inline: |
|
1704 | 1707 | chunkstart += (rev + 1) * iosize |
|
1705 | 1708 | chunklength = length(rev) |
|
1706 | 1709 | ladd(decomp(buffer(data, chunkstart - offset, chunklength))) |
|
1707 | 1710 | |
|
1708 | 1711 | return l |
|
1709 | 1712 | |
|
1710 | 1713 | def _chunkclear(self): |
|
1711 | 1714 | """Clear the raw chunk cache.""" |
|
1712 | 1715 | self._chunkcache = (0, b'') |
|
1713 | 1716 | |
|
1714 | 1717 | def deltaparent(self, rev): |
|
1715 | 1718 | """return deltaparent of the given revision""" |
|
1716 | 1719 | base = self.index[rev][3] |
|
1717 | 1720 | if base == rev: |
|
1718 | 1721 | return nullrev |
|
1719 | 1722 | elif self._generaldelta: |
|
1720 | 1723 | return base |
|
1721 | 1724 | else: |
|
1722 | 1725 | return rev - 1 |
|
1723 | 1726 | |
|
1724 | 1727 | def issnapshot(self, rev): |
|
1725 | 1728 | """tells whether rev is a snapshot""" |
|
1726 | 1729 | if not self._sparserevlog: |
|
1727 | 1730 | return self.deltaparent(rev) == nullrev |
|
1728 | 1731 | elif util.safehasattr(self.index, b'issnapshot'): |
|
1729 | 1732 | # directly assign the method to cache the testing and access |
|
1730 | 1733 | self.issnapshot = self.index.issnapshot |
|
1731 | 1734 | return self.issnapshot(rev) |
|
1732 | 1735 | if rev == nullrev: |
|
1733 | 1736 | return True |
|
1734 | 1737 | entry = self.index[rev] |
|
1735 | 1738 | base = entry[3] |
|
1736 | 1739 | if base == rev: |
|
1737 | 1740 | return True |
|
1738 | 1741 | if base == nullrev: |
|
1739 | 1742 | return True |
|
1740 | 1743 | p1 = entry[5] |
|
1741 | 1744 | p2 = entry[6] |
|
1742 | 1745 | if base == p1 or base == p2: |
|
1743 | 1746 | return False |
|
1744 | 1747 | return self.issnapshot(base) |
|
1745 | 1748 | |
|
1746 | 1749 | def snapshotdepth(self, rev): |
|
1747 | 1750 | """number of snapshot in the chain before this one""" |
|
1748 | 1751 | if not self.issnapshot(rev): |
|
1749 | 1752 | raise error.ProgrammingError(b'revision %d not a snapshot') |
|
1750 | 1753 | return len(self._deltachain(rev)[0]) - 1 |
|
1751 | 1754 | |
|
1752 | 1755 | def revdiff(self, rev1, rev2): |
|
1753 | 1756 | """return or calculate a delta between two revisions |
|
1754 | 1757 | |
|
1755 | 1758 | The delta calculated is in binary form and is intended to be written to |
|
1756 | 1759 | revlog data directly. So this function needs raw revision data. |
|
1757 | 1760 | """ |
|
1758 | 1761 | if rev1 != nullrev and self.deltaparent(rev2) == rev1: |
|
1759 | 1762 | return bytes(self._chunk(rev2)) |
|
1760 | 1763 | |
|
1761 | 1764 | return mdiff.textdiff(self.rawdata(rev1), self.rawdata(rev2)) |
|
1762 | 1765 | |
|
1763 | 1766 | def _processflags(self, text, flags, operation, raw=False): |
|
1764 | 1767 | """deprecated entry point to access flag processors""" |
|
1765 | 1768 | msg = b'_processflag(...) use the specialized variant' |
|
1766 | 1769 | util.nouideprecwarn(msg, b'5.2', stacklevel=2) |
|
1767 | 1770 | if raw: |
|
1768 | 1771 | return text, flagutil.processflagsraw(self, text, flags) |
|
1769 | 1772 | elif operation == b'read': |
|
1770 | 1773 | return flagutil.processflagsread(self, text, flags) |
|
1771 | 1774 | else: # write operation |
|
1772 | 1775 | return flagutil.processflagswrite(self, text, flags) |
|
1773 | 1776 | |
|
1774 | 1777 | def revision(self, nodeorrev, _df=None, raw=False): |
|
1775 | 1778 | """return an uncompressed revision of a given node or revision |
|
1776 | 1779 | number. |
|
1777 | 1780 | |
|
1778 | 1781 | _df - an existing file handle to read from. (internal-only) |
|
1779 | 1782 | raw - an optional argument specifying if the revision data is to be |
|
1780 | 1783 | treated as raw data when applying flag transforms. 'raw' should be set |
|
1781 | 1784 | to True when generating changegroups or in debug commands. |
|
1782 | 1785 | """ |
|
1783 | 1786 | if raw: |
|
1784 | 1787 | msg = ( |
|
1785 | 1788 | b'revlog.revision(..., raw=True) is deprecated, ' |
|
1786 | 1789 | b'use revlog.rawdata(...)' |
|
1787 | 1790 | ) |
|
1788 | 1791 | util.nouideprecwarn(msg, b'5.2', stacklevel=2) |
|
1789 | 1792 | return self._revisiondata(nodeorrev, _df, raw=raw)[0] |
|
1790 | 1793 | |
|
1791 | 1794 | def sidedata(self, nodeorrev, _df=None): |
|
1792 | 1795 | """a map of extra data related to the changeset but not part of the hash |
|
1793 | 1796 | |
|
1794 | 1797 | This function currently return a dictionary. However, more advanced |
|
1795 | 1798 | mapping object will likely be used in the future for a more |
|
1796 | 1799 | efficient/lazy code. |
|
1797 | 1800 | """ |
|
1798 | 1801 | return self._revisiondata(nodeorrev, _df)[1] |
|
1799 | 1802 | |
|
1800 | 1803 | def _revisiondata(self, nodeorrev, _df=None, raw=False): |
|
1801 | 1804 | # deal with <nodeorrev> argument type |
|
1802 | 1805 | if isinstance(nodeorrev, int): |
|
1803 | 1806 | rev = nodeorrev |
|
1804 | 1807 | node = self.node(rev) |
|
1805 | 1808 | else: |
|
1806 | 1809 | node = nodeorrev |
|
1807 | 1810 | rev = None |
|
1808 | 1811 | |
|
1809 | 1812 | # fast path the special `nullid` rev |
|
1810 | 1813 | if node == self.nullid: |
|
1811 | 1814 | return b"", {} |
|
1812 | 1815 | |
|
1813 | 1816 | # ``rawtext`` is the text as stored inside the revlog. Might be the |
|
1814 | 1817 | # revision or might need to be processed to retrieve the revision. |
|
1815 | 1818 | rev, rawtext, validated = self._rawtext(node, rev, _df=_df) |
|
1816 | 1819 | |
|
1817 | 1820 | if self.hassidedata: |
|
1818 | 1821 | if rev is None: |
|
1819 | 1822 | rev = self.rev(node) |
|
1820 | 1823 | sidedata = self._sidedata(rev) |
|
1821 | 1824 | else: |
|
1822 | 1825 | sidedata = {} |
|
1823 | 1826 | |
|
1824 | 1827 | if raw and validated: |
|
1825 | 1828 | # if we don't want to process the raw text and that raw |
|
1826 | 1829 | # text is cached, we can exit early. |
|
1827 | 1830 | return rawtext, sidedata |
|
1828 | 1831 | if rev is None: |
|
1829 | 1832 | rev = self.rev(node) |
|
1830 | 1833 | # the revlog's flag for this revision |
|
1831 | 1834 | # (usually alter its state or content) |
|
1832 | 1835 | flags = self.flags(rev) |
|
1833 | 1836 | |
|
1834 | 1837 | if validated and flags == REVIDX_DEFAULT_FLAGS: |
|
1835 | 1838 | # no extra flags set, no flag processor runs, text = rawtext |
|
1836 | 1839 | return rawtext, sidedata |
|
1837 | 1840 | |
|
1838 | 1841 | if raw: |
|
1839 | 1842 | validatehash = flagutil.processflagsraw(self, rawtext, flags) |
|
1840 | 1843 | text = rawtext |
|
1841 | 1844 | else: |
|
1842 | 1845 | r = flagutil.processflagsread(self, rawtext, flags) |
|
1843 | 1846 | text, validatehash = r |
|
1844 | 1847 | if validatehash: |
|
1845 | 1848 | self.checkhash(text, node, rev=rev) |
|
1846 | 1849 | if not validated: |
|
1847 | 1850 | self._revisioncache = (node, rev, rawtext) |
|
1848 | 1851 | |
|
1849 | 1852 | return text, sidedata |
|
1850 | 1853 | |
|
1851 | 1854 | def _rawtext(self, node, rev, _df=None): |
|
1852 | 1855 | """return the possibly unvalidated rawtext for a revision |
|
1853 | 1856 | |
|
1854 | 1857 | returns (rev, rawtext, validated) |
|
1855 | 1858 | """ |
|
1856 | 1859 | |
|
1857 | 1860 | # revision in the cache (could be useful to apply delta) |
|
1858 | 1861 | cachedrev = None |
|
1859 | 1862 | # An intermediate text to apply deltas to |
|
1860 | 1863 | basetext = None |
|
1861 | 1864 | |
|
1862 | 1865 | # Check if we have the entry in cache |
|
1863 | 1866 | # The cache entry looks like (node, rev, rawtext) |
|
1864 | 1867 | if self._revisioncache: |
|
1865 | 1868 | if self._revisioncache[0] == node: |
|
1866 | 1869 | return (rev, self._revisioncache[2], True) |
|
1867 | 1870 | cachedrev = self._revisioncache[1] |
|
1868 | 1871 | |
|
1869 | 1872 | if rev is None: |
|
1870 | 1873 | rev = self.rev(node) |
|
1871 | 1874 | |
|
1872 | 1875 | chain, stopped = self._deltachain(rev, stoprev=cachedrev) |
|
1873 | 1876 | if stopped: |
|
1874 | 1877 | basetext = self._revisioncache[2] |
|
1875 | 1878 | |
|
1876 | 1879 | # drop cache to save memory, the caller is expected to |
|
1877 | 1880 | # update self._revisioncache after validating the text |
|
1878 | 1881 | self._revisioncache = None |
|
1879 | 1882 | |
|
1880 | 1883 | targetsize = None |
|
1881 | 1884 | rawsize = self.index[rev][2] |
|
1882 | 1885 | if 0 <= rawsize: |
|
1883 | 1886 | targetsize = 4 * rawsize |
|
1884 | 1887 | |
|
1885 | 1888 | bins = self._chunks(chain, df=_df, targetsize=targetsize) |
|
1886 | 1889 | if basetext is None: |
|
1887 | 1890 | basetext = bytes(bins[0]) |
|
1888 | 1891 | bins = bins[1:] |
|
1889 | 1892 | |
|
1890 | 1893 | rawtext = mdiff.patches(basetext, bins) |
|
1891 | 1894 | del basetext # let us have a chance to free memory early |
|
1892 | 1895 | return (rev, rawtext, False) |
|
1893 | 1896 | |
|
1894 | 1897 | def _sidedata(self, rev): |
|
1895 | 1898 | """Return the sidedata for a given revision number.""" |
|
1896 | 1899 | index_entry = self.index[rev] |
|
1897 | 1900 | sidedata_offset = index_entry[8] |
|
1898 | 1901 | sidedata_size = index_entry[9] |
|
1899 | 1902 | |
|
1900 | 1903 | if self._inline: |
|
1901 | 1904 | sidedata_offset += self.index.entry_size * (1 + rev) |
|
1902 | 1905 | if sidedata_size == 0: |
|
1903 | 1906 | return {} |
|
1904 | 1907 | |
|
1905 | 1908 | segment = self._getsegment(sidedata_offset, sidedata_size) |
|
1906 | 1909 | sidedata = sidedatautil.deserialize_sidedata(segment) |
|
1907 | 1910 | return sidedata |
|
1908 | 1911 | |
|
1909 | 1912 | def rawdata(self, nodeorrev, _df=None): |
|
1910 | 1913 | """return an uncompressed raw data of a given node or revision number. |
|
1911 | 1914 | |
|
1912 | 1915 | _df - an existing file handle to read from. (internal-only) |
|
1913 | 1916 | """ |
|
1914 | 1917 | return self._revisiondata(nodeorrev, _df, raw=True)[0] |
|
1915 | 1918 | |
|
1916 | 1919 | def hash(self, text, p1, p2): |
|
1917 | 1920 | """Compute a node hash. |
|
1918 | 1921 | |
|
1919 | 1922 | Available as a function so that subclasses can replace the hash |
|
1920 | 1923 | as needed. |
|
1921 | 1924 | """ |
|
1922 | 1925 | return storageutil.hashrevisionsha1(text, p1, p2) |
|
1923 | 1926 | |
|
1924 | 1927 | def checkhash(self, text, node, p1=None, p2=None, rev=None): |
|
1925 | 1928 | """Check node hash integrity. |
|
1926 | 1929 | |
|
1927 | 1930 | Available as a function so that subclasses can extend hash mismatch |
|
1928 | 1931 | behaviors as needed. |
|
1929 | 1932 | """ |
|
1930 | 1933 | try: |
|
1931 | 1934 | if p1 is None and p2 is None: |
|
1932 | 1935 | p1, p2 = self.parents(node) |
|
1933 | 1936 | if node != self.hash(text, p1, p2): |
|
1934 | 1937 | # Clear the revision cache on hash failure. The revision cache |
|
1935 | 1938 | # only stores the raw revision and clearing the cache does have |
|
1936 | 1939 | # the side-effect that we won't have a cache hit when the raw |
|
1937 | 1940 | # revision data is accessed. But this case should be rare and |
|
1938 | 1941 | # it is extra work to teach the cache about the hash |
|
1939 | 1942 | # verification state. |
|
1940 | 1943 | if self._revisioncache and self._revisioncache[0] == node: |
|
1941 | 1944 | self._revisioncache = None |
|
1942 | 1945 | |
|
1943 | 1946 | revornode = rev |
|
1944 | 1947 | if revornode is None: |
|
1945 | 1948 | revornode = templatefilters.short(hex(node)) |
|
1946 | 1949 | raise error.RevlogError( |
|
1947 | 1950 | _(b"integrity check failed on %s:%s") |
|
1948 | 1951 | % (self.display_id, pycompat.bytestr(revornode)) |
|
1949 | 1952 | ) |
|
1950 | 1953 | except error.RevlogError: |
|
1951 | 1954 | if self._censorable and storageutil.iscensoredtext(text): |
|
1952 | 1955 | raise error.CensoredNodeError(self.display_id, node, text) |
|
1953 | 1956 | raise |
|
1954 | 1957 | |
|
1955 | 1958 | def _enforceinlinesize(self, tr, fp=None): |
|
1956 | 1959 | """Check if the revlog is too big for inline and convert if so. |
|
1957 | 1960 | |
|
1958 | 1961 | This should be called after revisions are added to the revlog. If the |
|
1959 | 1962 | revlog has grown too large to be an inline revlog, it will convert it |
|
1960 | 1963 | to use multiple index and data files. |
|
1961 | 1964 | """ |
|
1962 | 1965 | tiprev = len(self) - 1 |
|
1963 | 1966 | total_size = self.start(tiprev) + self.length(tiprev) |
|
1964 | 1967 | if not self._inline or total_size < _maxinline: |
|
1965 | 1968 | return |
|
1966 | 1969 | |
|
1967 | 1970 | troffset = tr.findoffset(self._indexfile) |
|
1968 | 1971 | if troffset is None: |
|
1969 | 1972 | raise error.RevlogError( |
|
1970 | 1973 | _(b"%s not found in the transaction") % self._indexfile |
|
1971 | 1974 | ) |
|
1972 | 1975 | trindex = 0 |
|
1973 | 1976 | tr.add(self._datafile, 0) |
|
1974 | 1977 | |
|
1975 | 1978 | if fp: |
|
1976 | 1979 | fp.flush() |
|
1977 | 1980 | fp.close() |
|
1978 | 1981 | # We can't use the cached file handle after close(). So prevent |
|
1979 | 1982 | # its usage. |
|
1980 | 1983 | self._writinghandles = None |
|
1981 | 1984 | |
|
1982 | 1985 | with self._indexfp(b'r') as ifh, self._datafp(b'w') as dfh: |
|
1983 | 1986 | for r in self: |
|
1984 | 1987 | dfh.write(self._getsegmentforrevs(r, r, df=ifh)[1]) |
|
1985 | 1988 | if troffset <= self.start(r): |
|
1986 | 1989 | trindex = r |
|
1987 | 1990 | |
|
1988 | 1991 | with self._indexfp(b'w') as fp: |
|
1989 | 1992 | self._format_flags &= ~FLAG_INLINE_DATA |
|
1990 | 1993 | self._inline = False |
|
1991 | 1994 | for i in self: |
|
1992 | 1995 | e = self.index.entry_binary(i) |
|
1993 | 1996 | if i == 0: |
|
1994 | 1997 | header = self._format_flags | self._format_version |
|
1995 | 1998 | header = self.index.pack_header(header) |
|
1996 | 1999 | e = header + e |
|
1997 | 2000 | fp.write(e) |
|
1998 | 2001 | |
|
1999 | 2002 | # the temp file replace the real index when we exit the context |
|
2000 | 2003 | # manager |
|
2001 | 2004 | |
|
2002 | 2005 | tr.replace(self._indexfile, trindex * self.index.entry_size) |
|
2003 | 2006 | nodemaputil.setup_persistent_nodemap(tr, self) |
|
2004 | 2007 | self._chunkclear() |
|
2005 | 2008 | |
|
2006 | 2009 | def _nodeduplicatecallback(self, transaction, node): |
|
2007 | 2010 | """called when trying to add a node already stored.""" |
|
2008 | 2011 | |
|
2009 | 2012 | def addrevision( |
|
2010 | 2013 | self, |
|
2011 | 2014 | text, |
|
2012 | 2015 | transaction, |
|
2013 | 2016 | link, |
|
2014 | 2017 | p1, |
|
2015 | 2018 | p2, |
|
2016 | 2019 | cachedelta=None, |
|
2017 | 2020 | node=None, |
|
2018 | 2021 | flags=REVIDX_DEFAULT_FLAGS, |
|
2019 | 2022 | deltacomputer=None, |
|
2020 | 2023 | sidedata=None, |
|
2021 | 2024 | ): |
|
2022 | 2025 | """add a revision to the log |
|
2023 | 2026 | |
|
2024 | 2027 | text - the revision data to add |
|
2025 | 2028 | transaction - the transaction object used for rollback |
|
2026 | 2029 | link - the linkrev data to add |
|
2027 | 2030 | p1, p2 - the parent nodeids of the revision |
|
2028 | 2031 | cachedelta - an optional precomputed delta |
|
2029 | 2032 | node - nodeid of revision; typically node is not specified, and it is |
|
2030 | 2033 | computed by default as hash(text, p1, p2), however subclasses might |
|
2031 | 2034 | use different hashing method (and override checkhash() in such case) |
|
2032 | 2035 | flags - the known flags to set on the revision |
|
2033 | 2036 | deltacomputer - an optional deltacomputer instance shared between |
|
2034 | 2037 | multiple calls |
|
2035 | 2038 | """ |
|
2036 | 2039 | if link == nullrev: |
|
2037 | 2040 | raise error.RevlogError( |
|
2038 | 2041 | _(b"attempted to add linkrev -1 to %s") % self.display_id |
|
2039 | 2042 | ) |
|
2040 | 2043 | |
|
2041 | 2044 | if sidedata is None: |
|
2042 | 2045 | sidedata = {} |
|
2043 | 2046 | elif sidedata and not self.hassidedata: |
|
2044 | 2047 | raise error.ProgrammingError( |
|
2045 | 2048 | _(b"trying to add sidedata to a revlog who don't support them") |
|
2046 | 2049 | ) |
|
2047 | 2050 | |
|
2048 | 2051 | if flags: |
|
2049 | 2052 | node = node or self.hash(text, p1, p2) |
|
2050 | 2053 | |
|
2051 | 2054 | rawtext, validatehash = flagutil.processflagswrite(self, text, flags) |
|
2052 | 2055 | |
|
2053 | 2056 | # If the flag processor modifies the revision data, ignore any provided |
|
2054 | 2057 | # cachedelta. |
|
2055 | 2058 | if rawtext != text: |
|
2056 | 2059 | cachedelta = None |
|
2057 | 2060 | |
|
2058 | 2061 | if len(rawtext) > _maxentrysize: |
|
2059 | 2062 | raise error.RevlogError( |
|
2060 | 2063 | _( |
|
2061 | 2064 | b"%s: size of %d bytes exceeds maximum revlog storage of 2GiB" |
|
2062 | 2065 | ) |
|
2063 | 2066 | % (self.display_id, len(rawtext)) |
|
2064 | 2067 | ) |
|
2065 | 2068 | |
|
2066 | 2069 | node = node or self.hash(rawtext, p1, p2) |
|
2067 | 2070 | rev = self.index.get_rev(node) |
|
2068 | 2071 | if rev is not None: |
|
2069 | 2072 | return rev |
|
2070 | 2073 | |
|
2071 | 2074 | if validatehash: |
|
2072 | 2075 | self.checkhash(rawtext, node, p1=p1, p2=p2) |
|
2073 | 2076 | |
|
2074 | 2077 | return self.addrawrevision( |
|
2075 | 2078 | rawtext, |
|
2076 | 2079 | transaction, |
|
2077 | 2080 | link, |
|
2078 | 2081 | p1, |
|
2079 | 2082 | p2, |
|
2080 | 2083 | node, |
|
2081 | 2084 | flags, |
|
2082 | 2085 | cachedelta=cachedelta, |
|
2083 | 2086 | deltacomputer=deltacomputer, |
|
2084 | 2087 | sidedata=sidedata, |
|
2085 | 2088 | ) |
|
2086 | 2089 | |
|
2087 | 2090 | def addrawrevision( |
|
2088 | 2091 | self, |
|
2089 | 2092 | rawtext, |
|
2090 | 2093 | transaction, |
|
2091 | 2094 | link, |
|
2092 | 2095 | p1, |
|
2093 | 2096 | p2, |
|
2094 | 2097 | node, |
|
2095 | 2098 | flags, |
|
2096 | 2099 | cachedelta=None, |
|
2097 | 2100 | deltacomputer=None, |
|
2098 | 2101 | sidedata=None, |
|
2099 | 2102 | ): |
|
2100 | 2103 | """add a raw revision with known flags, node and parents |
|
2101 | 2104 | useful when reusing a revision not stored in this revlog (ex: received |
|
2102 | 2105 | over wire, or read from an external bundle). |
|
2103 | 2106 | """ |
|
2104 | 2107 | dfh = None |
|
2105 | 2108 | if not self._inline: |
|
2106 | 2109 | dfh = self._datafp(b"a+") |
|
2107 | 2110 | ifh = self._indexfp(b"a+") |
|
2108 | 2111 | try: |
|
2109 | 2112 | return self._addrevision( |
|
2110 | 2113 | node, |
|
2111 | 2114 | rawtext, |
|
2112 | 2115 | transaction, |
|
2113 | 2116 | link, |
|
2114 | 2117 | p1, |
|
2115 | 2118 | p2, |
|
2116 | 2119 | flags, |
|
2117 | 2120 | cachedelta, |
|
2118 | 2121 | ifh, |
|
2119 | 2122 | dfh, |
|
2120 | 2123 | deltacomputer=deltacomputer, |
|
2121 | 2124 | sidedata=sidedata, |
|
2122 | 2125 | ) |
|
2123 | 2126 | finally: |
|
2124 | 2127 | if dfh: |
|
2125 | 2128 | dfh.close() |
|
2126 | 2129 | ifh.close() |
|
2127 | 2130 | |
|
2128 | 2131 | def compress(self, data): |
|
2129 | 2132 | """Generate a possibly-compressed representation of data.""" |
|
2130 | 2133 | if not data: |
|
2131 | 2134 | return b'', data |
|
2132 | 2135 | |
|
2133 | 2136 | compressed = self._compressor.compress(data) |
|
2134 | 2137 | |
|
2135 | 2138 | if compressed: |
|
2136 | 2139 | # The revlog compressor added the header in the returned data. |
|
2137 | 2140 | return b'', compressed |
|
2138 | 2141 | |
|
2139 | 2142 | if data[0:1] == b'\0': |
|
2140 | 2143 | return b'', data |
|
2141 | 2144 | return b'u', data |
|
2142 | 2145 | |
|
2143 | 2146 | def decompress(self, data): |
|
2144 | 2147 | """Decompress a revlog chunk. |
|
2145 | 2148 | |
|
2146 | 2149 | The chunk is expected to begin with a header identifying the |
|
2147 | 2150 | format type so it can be routed to an appropriate decompressor. |
|
2148 | 2151 | """ |
|
2149 | 2152 | if not data: |
|
2150 | 2153 | return data |
|
2151 | 2154 | |
|
2152 | 2155 | # Revlogs are read much more frequently than they are written and many |
|
2153 | 2156 | # chunks only take microseconds to decompress, so performance is |
|
2154 | 2157 | # important here. |
|
2155 | 2158 | # |
|
2156 | 2159 | # We can make a few assumptions about revlogs: |
|
2157 | 2160 | # |
|
2158 | 2161 | # 1) the majority of chunks will be compressed (as opposed to inline |
|
2159 | 2162 | # raw data). |
|
2160 | 2163 | # 2) decompressing *any* data will likely by at least 10x slower than |
|
2161 | 2164 | # returning raw inline data. |
|
2162 | 2165 | # 3) we want to prioritize common and officially supported compression |
|
2163 | 2166 | # engines |
|
2164 | 2167 | # |
|
2165 | 2168 | # It follows that we want to optimize for "decompress compressed data |
|
2166 | 2169 | # when encoded with common and officially supported compression engines" |
|
2167 | 2170 | # case over "raw data" and "data encoded by less common or non-official |
|
2168 | 2171 | # compression engines." That is why we have the inline lookup first |
|
2169 | 2172 | # followed by the compengines lookup. |
|
2170 | 2173 | # |
|
2171 | 2174 | # According to `hg perfrevlogchunks`, this is ~0.5% faster for zlib |
|
2172 | 2175 | # compressed chunks. And this matters for changelog and manifest reads. |
|
2173 | 2176 | t = data[0:1] |
|
2174 | 2177 | |
|
2175 | 2178 | if t == b'x': |
|
2176 | 2179 | try: |
|
2177 | 2180 | return _zlibdecompress(data) |
|
2178 | 2181 | except zlib.error as e: |
|
2179 | 2182 | raise error.RevlogError( |
|
2180 | 2183 | _(b'revlog decompress error: %s') |
|
2181 | 2184 | % stringutil.forcebytestr(e) |
|
2182 | 2185 | ) |
|
2183 | 2186 | # '\0' is more common than 'u' so it goes first. |
|
2184 | 2187 | elif t == b'\0': |
|
2185 | 2188 | return data |
|
2186 | 2189 | elif t == b'u': |
|
2187 | 2190 | return util.buffer(data, 1) |
|
2188 | 2191 | |
|
2189 | 2192 | try: |
|
2190 | 2193 | compressor = self._decompressors[t] |
|
2191 | 2194 | except KeyError: |
|
2192 | 2195 | try: |
|
2193 | 2196 | engine = util.compengines.forrevlogheader(t) |
|
2194 | 2197 | compressor = engine.revlogcompressor(self._compengineopts) |
|
2195 | 2198 | self._decompressors[t] = compressor |
|
2196 | 2199 | except KeyError: |
|
2197 | 2200 | raise error.RevlogError( |
|
2198 | 2201 | _(b'unknown compression type %s') % binascii.hexlify(t) |
|
2199 | 2202 | ) |
|
2200 | 2203 | |
|
2201 | 2204 | return compressor.decompress(data) |
|
2202 | 2205 | |
|
2203 | 2206 | def _addrevision( |
|
2204 | 2207 | self, |
|
2205 | 2208 | node, |
|
2206 | 2209 | rawtext, |
|
2207 | 2210 | transaction, |
|
2208 | 2211 | link, |
|
2209 | 2212 | p1, |
|
2210 | 2213 | p2, |
|
2211 | 2214 | flags, |
|
2212 | 2215 | cachedelta, |
|
2213 | 2216 | ifh, |
|
2214 | 2217 | dfh, |
|
2215 | 2218 | alwayscache=False, |
|
2216 | 2219 | deltacomputer=None, |
|
2217 | 2220 | sidedata=None, |
|
2218 | 2221 | ): |
|
2219 | 2222 | """internal function to add revisions to the log |
|
2220 | 2223 | |
|
2221 | 2224 | see addrevision for argument descriptions. |
|
2222 | 2225 | |
|
2223 | 2226 | note: "addrevision" takes non-raw text, "_addrevision" takes raw text. |
|
2224 | 2227 | |
|
2225 | 2228 | if "deltacomputer" is not provided or None, a defaultdeltacomputer will |
|
2226 | 2229 | be used. |
|
2227 | 2230 | |
|
2228 | 2231 | invariants: |
|
2229 | 2232 | - rawtext is optional (can be None); if not set, cachedelta must be set. |
|
2230 | 2233 | if both are set, they must correspond to each other. |
|
2231 | 2234 | """ |
|
2232 | 2235 | if node == self.nullid: |
|
2233 | 2236 | raise error.RevlogError( |
|
2234 | 2237 | _(b"%s: attempt to add null revision") % self.display_id |
|
2235 | 2238 | ) |
|
2236 | 2239 | if ( |
|
2237 | 2240 | node == self.nodeconstants.wdirid |
|
2238 | 2241 | or node in self.nodeconstants.wdirfilenodeids |
|
2239 | 2242 | ): |
|
2240 | 2243 | raise error.RevlogError( |
|
2241 | 2244 | _(b"%s: attempt to add wdir revision") % self.display_id |
|
2242 | 2245 | ) |
|
2243 | 2246 | |
|
2244 | 2247 | if self._inline: |
|
2245 | 2248 | fh = ifh |
|
2246 | 2249 | else: |
|
2247 | 2250 | fh = dfh |
|
2248 | 2251 | |
|
2249 | 2252 | btext = [rawtext] |
|
2250 | 2253 | |
|
2251 | 2254 | curr = len(self) |
|
2252 | 2255 | prev = curr - 1 |
|
2253 | 2256 | |
|
2254 | 2257 | offset = self._get_data_offset(prev) |
|
2255 | 2258 | |
|
2256 | 2259 | if self._concurrencychecker: |
|
2257 | 2260 | if self._inline: |
|
2258 | 2261 | # offset is "as if" it were in the .d file, so we need to add on |
|
2259 | 2262 | # the size of the entry metadata. |
|
2260 | 2263 | self._concurrencychecker( |
|
2261 | 2264 | ifh, self._indexfile, offset + curr * self.index.entry_size |
|
2262 | 2265 | ) |
|
2263 | 2266 | else: |
|
2264 | 2267 | # Entries in the .i are a consistent size. |
|
2265 | 2268 | self._concurrencychecker( |
|
2266 | 2269 | ifh, self._indexfile, curr * self.index.entry_size |
|
2267 | 2270 | ) |
|
2268 | 2271 | self._concurrencychecker(dfh, self._datafile, offset) |
|
2269 | 2272 | |
|
2270 | 2273 | p1r, p2r = self.rev(p1), self.rev(p2) |
|
2271 | 2274 | |
|
2272 | 2275 | # full versions are inserted when the needed deltas |
|
2273 | 2276 | # become comparable to the uncompressed text |
|
2274 | 2277 | if rawtext is None: |
|
2275 | 2278 | # need rawtext size, before changed by flag processors, which is |
|
2276 | 2279 | # the non-raw size. use revlog explicitly to avoid filelog's extra |
|
2277 | 2280 | # logic that might remove metadata size. |
|
2278 | 2281 | textlen = mdiff.patchedsize( |
|
2279 | 2282 | revlog.size(self, cachedelta[0]), cachedelta[1] |
|
2280 | 2283 | ) |
|
2281 | 2284 | else: |
|
2282 | 2285 | textlen = len(rawtext) |
|
2283 | 2286 | |
|
2284 | 2287 | if deltacomputer is None: |
|
2285 | 2288 | deltacomputer = deltautil.deltacomputer(self) |
|
2286 | 2289 | |
|
2287 | 2290 | revinfo = _revisioninfo(node, p1, p2, btext, textlen, cachedelta, flags) |
|
2288 | 2291 | |
|
2289 | 2292 | deltainfo = deltacomputer.finddeltainfo(revinfo, fh) |
|
2290 | 2293 | |
|
2291 | 2294 | if sidedata and self.hassidedata: |
|
2292 | 2295 | serialized_sidedata = sidedatautil.serialize_sidedata(sidedata) |
|
2293 | 2296 | sidedata_offset = offset + deltainfo.deltalen |
|
2294 | 2297 | else: |
|
2295 | 2298 | serialized_sidedata = b"" |
|
2296 | 2299 | # Don't store the offset if the sidedata is empty, that way |
|
2297 | 2300 | # we can easily detect empty sidedata and they will be no different |
|
2298 | 2301 | # than ones we manually add. |
|
2299 | 2302 | sidedata_offset = 0 |
|
2300 | 2303 | |
|
2301 | 2304 | e = ( |
|
2302 | 2305 | offset_type(offset, flags), |
|
2303 | 2306 | deltainfo.deltalen, |
|
2304 | 2307 | textlen, |
|
2305 | 2308 | deltainfo.base, |
|
2306 | 2309 | link, |
|
2307 | 2310 | p1r, |
|
2308 | 2311 | p2r, |
|
2309 | 2312 | node, |
|
2310 | 2313 | sidedata_offset, |
|
2311 | 2314 | len(serialized_sidedata), |
|
2312 | 2315 | ) |
|
2313 | 2316 | |
|
2314 | 2317 | self.index.append(e) |
|
2315 | 2318 | entry = self.index.entry_binary(curr) |
|
2316 | 2319 | if curr == 0: |
|
2317 | 2320 | header = self._format_flags | self._format_version |
|
2318 | 2321 | header = self.index.pack_header(header) |
|
2319 | 2322 | entry = header + entry |
|
2320 | 2323 | self._writeentry( |
|
2321 | 2324 | transaction, |
|
2322 | 2325 | ifh, |
|
2323 | 2326 | dfh, |
|
2324 | 2327 | entry, |
|
2325 | 2328 | deltainfo.data, |
|
2326 | 2329 | link, |
|
2327 | 2330 | offset, |
|
2328 | 2331 | serialized_sidedata, |
|
2329 | 2332 | ) |
|
2330 | 2333 | |
|
2331 | 2334 | rawtext = btext[0] |
|
2332 | 2335 | |
|
2333 | 2336 | if alwayscache and rawtext is None: |
|
2334 | 2337 | rawtext = deltacomputer.buildtext(revinfo, fh) |
|
2335 | 2338 | |
|
2336 | 2339 | if type(rawtext) == bytes: # only accept immutable objects |
|
2337 | 2340 | self._revisioncache = (node, curr, rawtext) |
|
2338 | 2341 | self._chainbasecache[curr] = deltainfo.chainbase |
|
2339 | 2342 | return curr |
|
2340 | 2343 | |
|
2341 | 2344 | def _get_data_offset(self, prev): |
|
2342 | 2345 | """Returns the current offset in the (in-transaction) data file. |
|
2343 | 2346 | Versions < 2 of the revlog can get this 0(1), revlog v2 needs a docket |
|
2344 | 2347 | file to store that information: since sidedata can be rewritten to the |
|
2345 | 2348 | end of the data file within a transaction, you can have cases where, for |
|
2346 | 2349 | example, rev `n` does not have sidedata while rev `n - 1` does, leading |
|
2347 | 2350 | to `n - 1`'s sidedata being written after `n`'s data. |
|
2348 | 2351 | |
|
2349 | 2352 | TODO cache this in a docket file before getting out of experimental.""" |
|
2350 | 2353 | if self._format_version != REVLOGV2: |
|
2351 | 2354 | return self.end(prev) |
|
2352 | 2355 | |
|
2353 | 2356 | offset = 0 |
|
2354 | 2357 | for rev, entry in enumerate(self.index): |
|
2355 | 2358 | sidedata_end = entry[8] + entry[9] |
|
2356 | 2359 | # Sidedata for a previous rev has potentially been written after |
|
2357 | 2360 | # this rev's end, so take the max. |
|
2358 | 2361 | offset = max(self.end(rev), offset, sidedata_end) |
|
2359 | 2362 | return offset |
|
2360 | 2363 | |
|
2361 | 2364 | def _writeentry( |
|
2362 | 2365 | self, transaction, ifh, dfh, entry, data, link, offset, sidedata |
|
2363 | 2366 | ): |
|
2364 | 2367 | # Files opened in a+ mode have inconsistent behavior on various |
|
2365 | 2368 | # platforms. Windows requires that a file positioning call be made |
|
2366 | 2369 | # when the file handle transitions between reads and writes. See |
|
2367 | 2370 | # 3686fa2b8eee and the mixedfilemodewrapper in windows.py. On other |
|
2368 | 2371 | # platforms, Python or the platform itself can be buggy. Some versions |
|
2369 | 2372 | # of Solaris have been observed to not append at the end of the file |
|
2370 | 2373 | # if the file was seeked to before the end. See issue4943 for more. |
|
2371 | 2374 | # |
|
2372 | 2375 | # We work around this issue by inserting a seek() before writing. |
|
2373 | 2376 | # Note: This is likely not necessary on Python 3. However, because |
|
2374 | 2377 | # the file handle is reused for reads and may be seeked there, we need |
|
2375 | 2378 | # to be careful before changing this. |
|
2376 | 2379 | ifh.seek(0, os.SEEK_END) |
|
2377 | 2380 | if dfh: |
|
2378 | 2381 | dfh.seek(0, os.SEEK_END) |
|
2379 | 2382 | |
|
2380 | 2383 | curr = len(self) - 1 |
|
2381 | 2384 | if not self._inline: |
|
2382 | 2385 | transaction.add(self._datafile, offset) |
|
2383 | 2386 | transaction.add(self._indexfile, curr * len(entry)) |
|
2384 | 2387 | if data[0]: |
|
2385 | 2388 | dfh.write(data[0]) |
|
2386 | 2389 | dfh.write(data[1]) |
|
2387 | 2390 | if sidedata: |
|
2388 | 2391 | dfh.write(sidedata) |
|
2389 | 2392 | ifh.write(entry) |
|
2390 | 2393 | else: |
|
2391 | 2394 | offset += curr * self.index.entry_size |
|
2392 | 2395 | transaction.add(self._indexfile, offset) |
|
2393 | 2396 | ifh.write(entry) |
|
2394 | 2397 | ifh.write(data[0]) |
|
2395 | 2398 | ifh.write(data[1]) |
|
2396 | 2399 | if sidedata: |
|
2397 | 2400 | ifh.write(sidedata) |
|
2398 | 2401 | self._enforceinlinesize(transaction, ifh) |
|
2399 | 2402 | nodemaputil.setup_persistent_nodemap(transaction, self) |
|
2400 | 2403 | |
|
2401 | 2404 | def addgroup( |
|
2402 | 2405 | self, |
|
2403 | 2406 | deltas, |
|
2404 | 2407 | linkmapper, |
|
2405 | 2408 | transaction, |
|
2406 | 2409 | alwayscache=False, |
|
2407 | 2410 | addrevisioncb=None, |
|
2408 | 2411 | duplicaterevisioncb=None, |
|
2409 | 2412 | ): |
|
2410 | 2413 | """ |
|
2411 | 2414 | add a delta group |
|
2412 | 2415 | |
|
2413 | 2416 | given a set of deltas, add them to the revision log. the |
|
2414 | 2417 | first delta is against its parent, which should be in our |
|
2415 | 2418 | log, the rest are against the previous delta. |
|
2416 | 2419 | |
|
2417 | 2420 | If ``addrevisioncb`` is defined, it will be called with arguments of |
|
2418 | 2421 | this revlog and the node that was added. |
|
2419 | 2422 | """ |
|
2420 | 2423 | |
|
2421 | 2424 | if self._writinghandles: |
|
2422 | 2425 | raise error.ProgrammingError(b'cannot nest addgroup() calls') |
|
2423 | 2426 | |
|
2424 | 2427 | r = len(self) |
|
2425 | 2428 | end = 0 |
|
2426 | 2429 | if r: |
|
2427 | 2430 | end = self.end(r - 1) |
|
2428 | 2431 | ifh = self._indexfp(b"a+") |
|
2429 | 2432 | isize = r * self.index.entry_size |
|
2430 | 2433 | if self._inline: |
|
2431 | 2434 | transaction.add(self._indexfile, end + isize) |
|
2432 | 2435 | dfh = None |
|
2433 | 2436 | else: |
|
2434 | 2437 | transaction.add(self._indexfile, isize) |
|
2435 | 2438 | transaction.add(self._datafile, end) |
|
2436 | 2439 | dfh = self._datafp(b"a+") |
|
2437 | 2440 | |
|
2438 | 2441 | self._writinghandles = (ifh, dfh) |
|
2439 | 2442 | empty = True |
|
2440 | 2443 | |
|
2441 | 2444 | try: |
|
2442 | 2445 | deltacomputer = deltautil.deltacomputer(self) |
|
2443 | 2446 | # loop through our set of deltas |
|
2444 | 2447 | for data in deltas: |
|
2445 | 2448 | node, p1, p2, linknode, deltabase, delta, flags, sidedata = data |
|
2446 | 2449 | link = linkmapper(linknode) |
|
2447 | 2450 | flags = flags or REVIDX_DEFAULT_FLAGS |
|
2448 | 2451 | |
|
2449 | 2452 | rev = self.index.get_rev(node) |
|
2450 | 2453 | if rev is not None: |
|
2451 | 2454 | # this can happen if two branches make the same change |
|
2452 | 2455 | self._nodeduplicatecallback(transaction, rev) |
|
2453 | 2456 | if duplicaterevisioncb: |
|
2454 | 2457 | duplicaterevisioncb(self, rev) |
|
2455 | 2458 | empty = False |
|
2456 | 2459 | continue |
|
2457 | 2460 | |
|
2458 | 2461 | for p in (p1, p2): |
|
2459 | 2462 | if not self.index.has_node(p): |
|
2460 | 2463 | raise error.LookupError( |
|
2461 | 2464 | p, self.radix, _(b'unknown parent') |
|
2462 | 2465 | ) |
|
2463 | 2466 | |
|
2464 | 2467 | if not self.index.has_node(deltabase): |
|
2465 | 2468 | raise error.LookupError( |
|
2466 | 2469 | deltabase, self.display_id, _(b'unknown delta base') |
|
2467 | 2470 | ) |
|
2468 | 2471 | |
|
2469 | 2472 | baserev = self.rev(deltabase) |
|
2470 | 2473 | |
|
2471 | 2474 | if baserev != nullrev and self.iscensored(baserev): |
|
2472 | 2475 | # if base is censored, delta must be full replacement in a |
|
2473 | 2476 | # single patch operation |
|
2474 | 2477 | hlen = struct.calcsize(b">lll") |
|
2475 | 2478 | oldlen = self.rawsize(baserev) |
|
2476 | 2479 | newlen = len(delta) - hlen |
|
2477 | 2480 | if delta[:hlen] != mdiff.replacediffheader(oldlen, newlen): |
|
2478 | 2481 | raise error.CensoredBaseError( |
|
2479 | 2482 | self.display_id, self.node(baserev) |
|
2480 | 2483 | ) |
|
2481 | 2484 | |
|
2482 | 2485 | if not flags and self._peek_iscensored(baserev, delta): |
|
2483 | 2486 | flags |= REVIDX_ISCENSORED |
|
2484 | 2487 | |
|
2485 | 2488 | # We assume consumers of addrevisioncb will want to retrieve |
|
2486 | 2489 | # the added revision, which will require a call to |
|
2487 | 2490 | # revision(). revision() will fast path if there is a cache |
|
2488 | 2491 | # hit. So, we tell _addrevision() to always cache in this case. |
|
2489 | 2492 | # We're only using addgroup() in the context of changegroup |
|
2490 | 2493 | # generation so the revision data can always be handled as raw |
|
2491 | 2494 | # by the flagprocessor. |
|
2492 | 2495 | rev = self._addrevision( |
|
2493 | 2496 | node, |
|
2494 | 2497 | None, |
|
2495 | 2498 | transaction, |
|
2496 | 2499 | link, |
|
2497 | 2500 | p1, |
|
2498 | 2501 | p2, |
|
2499 | 2502 | flags, |
|
2500 | 2503 | (baserev, delta), |
|
2501 | 2504 | ifh, |
|
2502 | 2505 | dfh, |
|
2503 | 2506 | alwayscache=alwayscache, |
|
2504 | 2507 | deltacomputer=deltacomputer, |
|
2505 | 2508 | sidedata=sidedata, |
|
2506 | 2509 | ) |
|
2507 | 2510 | |
|
2508 | 2511 | if addrevisioncb: |
|
2509 | 2512 | addrevisioncb(self, rev) |
|
2510 | 2513 | empty = False |
|
2511 | 2514 | |
|
2512 | 2515 | if not dfh and not self._inline: |
|
2513 | 2516 | # addrevision switched from inline to conventional |
|
2514 | 2517 | # reopen the index |
|
2515 | 2518 | ifh.close() |
|
2516 | 2519 | dfh = self._datafp(b"a+") |
|
2517 | 2520 | ifh = self._indexfp(b"a+") |
|
2518 | 2521 | self._writinghandles = (ifh, dfh) |
|
2519 | 2522 | finally: |
|
2520 | 2523 | self._writinghandles = None |
|
2521 | 2524 | |
|
2522 | 2525 | if dfh: |
|
2523 | 2526 | dfh.close() |
|
2524 | 2527 | ifh.close() |
|
2525 | 2528 | return not empty |
|
2526 | 2529 | |
|
2527 | 2530 | def iscensored(self, rev): |
|
2528 | 2531 | """Check if a file revision is censored.""" |
|
2529 | 2532 | if not self._censorable: |
|
2530 | 2533 | return False |
|
2531 | 2534 | |
|
2532 | 2535 | return self.flags(rev) & REVIDX_ISCENSORED |
|
2533 | 2536 | |
|
2534 | 2537 | def _peek_iscensored(self, baserev, delta): |
|
2535 | 2538 | """Quickly check if a delta produces a censored revision.""" |
|
2536 | 2539 | if not self._censorable: |
|
2537 | 2540 | return False |
|
2538 | 2541 | |
|
2539 | 2542 | return storageutil.deltaiscensored(delta, baserev, self.rawsize) |
|
2540 | 2543 | |
|
2541 | 2544 | def getstrippoint(self, minlink): |
|
2542 | 2545 | """find the minimum rev that must be stripped to strip the linkrev |
|
2543 | 2546 | |
|
2544 | 2547 | Returns a tuple containing the minimum rev and a set of all revs that |
|
2545 | 2548 | have linkrevs that will be broken by this strip. |
|
2546 | 2549 | """ |
|
2547 | 2550 | return storageutil.resolvestripinfo( |
|
2548 | 2551 | minlink, |
|
2549 | 2552 | len(self) - 1, |
|
2550 | 2553 | self.headrevs(), |
|
2551 | 2554 | self.linkrev, |
|
2552 | 2555 | self.parentrevs, |
|
2553 | 2556 | ) |
|
2554 | 2557 | |
|
2555 | 2558 | def strip(self, minlink, transaction): |
|
2556 | 2559 | """truncate the revlog on the first revision with a linkrev >= minlink |
|
2557 | 2560 | |
|
2558 | 2561 | This function is called when we're stripping revision minlink and |
|
2559 | 2562 | its descendants from the repository. |
|
2560 | 2563 | |
|
2561 | 2564 | We have to remove all revisions with linkrev >= minlink, because |
|
2562 | 2565 | the equivalent changelog revisions will be renumbered after the |
|
2563 | 2566 | strip. |
|
2564 | 2567 | |
|
2565 | 2568 | So we truncate the revlog on the first of these revisions, and |
|
2566 | 2569 | trust that the caller has saved the revisions that shouldn't be |
|
2567 | 2570 | removed and that it'll re-add them after this truncation. |
|
2568 | 2571 | """ |
|
2569 | 2572 | if len(self) == 0: |
|
2570 | 2573 | return |
|
2571 | 2574 | |
|
2572 | 2575 | rev, _ = self.getstrippoint(minlink) |
|
2573 | 2576 | if rev == len(self): |
|
2574 | 2577 | return |
|
2575 | 2578 | |
|
2576 | 2579 | # first truncate the files on disk |
|
2577 | 2580 | end = self.start(rev) |
|
2578 | 2581 | if not self._inline: |
|
2579 | 2582 | transaction.add(self._datafile, end) |
|
2580 | 2583 | end = rev * self.index.entry_size |
|
2581 | 2584 | else: |
|
2582 | 2585 | end += rev * self.index.entry_size |
|
2583 | 2586 | |
|
2584 | 2587 | transaction.add(self._indexfile, end) |
|
2585 | 2588 | |
|
2586 | 2589 | # then reset internal state in memory to forget those revisions |
|
2587 | 2590 | self._revisioncache = None |
|
2588 | 2591 | self._chaininfocache = util.lrucachedict(500) |
|
2589 | 2592 | self._chunkclear() |
|
2590 | 2593 | |
|
2591 | 2594 | del self.index[rev:-1] |
|
2592 | 2595 | |
|
2593 | 2596 | def checksize(self): |
|
2594 | 2597 | """Check size of index and data files |
|
2595 | 2598 | |
|
2596 | 2599 | return a (dd, di) tuple. |
|
2597 | 2600 | - dd: extra bytes for the "data" file |
|
2598 | 2601 | - di: extra bytes for the "index" file |
|
2599 | 2602 | |
|
2600 | 2603 | A healthy revlog will return (0, 0). |
|
2601 | 2604 | """ |
|
2602 | 2605 | expected = 0 |
|
2603 | 2606 | if len(self): |
|
2604 | 2607 | expected = max(0, self.end(len(self) - 1)) |
|
2605 | 2608 | |
|
2606 | 2609 | try: |
|
2607 | 2610 | with self._datafp() as f: |
|
2608 | 2611 | f.seek(0, io.SEEK_END) |
|
2609 | 2612 | actual = f.tell() |
|
2610 | 2613 | dd = actual - expected |
|
2611 | 2614 | except IOError as inst: |
|
2612 | 2615 | if inst.errno != errno.ENOENT: |
|
2613 | 2616 | raise |
|
2614 | 2617 | dd = 0 |
|
2615 | 2618 | |
|
2616 | 2619 | try: |
|
2617 | 2620 | f = self.opener(self._indexfile) |
|
2618 | 2621 | f.seek(0, io.SEEK_END) |
|
2619 | 2622 | actual = f.tell() |
|
2620 | 2623 | f.close() |
|
2621 | 2624 | s = self.index.entry_size |
|
2622 | 2625 | i = max(0, actual // s) |
|
2623 | 2626 | di = actual - (i * s) |
|
2624 | 2627 | if self._inline: |
|
2625 | 2628 | databytes = 0 |
|
2626 | 2629 | for r in self: |
|
2627 | 2630 | databytes += max(0, self.length(r)) |
|
2628 | 2631 | dd = 0 |
|
2629 | 2632 | di = actual - len(self) * s - databytes |
|
2630 | 2633 | except IOError as inst: |
|
2631 | 2634 | if inst.errno != errno.ENOENT: |
|
2632 | 2635 | raise |
|
2633 | 2636 | di = 0 |
|
2634 | 2637 | |
|
2635 | 2638 | return (dd, di) |
|
2636 | 2639 | |
|
2637 | 2640 | def files(self): |
|
2638 | 2641 | res = [self._indexfile] |
|
2639 | 2642 | if not self._inline: |
|
2640 | 2643 | res.append(self._datafile) |
|
2641 | 2644 | return res |
|
2642 | 2645 | |
|
2643 | 2646 | def emitrevisions( |
|
2644 | 2647 | self, |
|
2645 | 2648 | nodes, |
|
2646 | 2649 | nodesorder=None, |
|
2647 | 2650 | revisiondata=False, |
|
2648 | 2651 | assumehaveparentrevisions=False, |
|
2649 | 2652 | deltamode=repository.CG_DELTAMODE_STD, |
|
2650 | 2653 | sidedata_helpers=None, |
|
2651 | 2654 | ): |
|
2652 | 2655 | if nodesorder not in (b'nodes', b'storage', b'linear', None): |
|
2653 | 2656 | raise error.ProgrammingError( |
|
2654 | 2657 | b'unhandled value for nodesorder: %s' % nodesorder |
|
2655 | 2658 | ) |
|
2656 | 2659 | |
|
2657 | 2660 | if nodesorder is None and not self._generaldelta: |
|
2658 | 2661 | nodesorder = b'storage' |
|
2659 | 2662 | |
|
2660 | 2663 | if ( |
|
2661 | 2664 | not self._storedeltachains |
|
2662 | 2665 | and deltamode != repository.CG_DELTAMODE_PREV |
|
2663 | 2666 | ): |
|
2664 | 2667 | deltamode = repository.CG_DELTAMODE_FULL |
|
2665 | 2668 | |
|
2666 | 2669 | return storageutil.emitrevisions( |
|
2667 | 2670 | self, |
|
2668 | 2671 | nodes, |
|
2669 | 2672 | nodesorder, |
|
2670 | 2673 | revlogrevisiondelta, |
|
2671 | 2674 | deltaparentfn=self.deltaparent, |
|
2672 | 2675 | candeltafn=self.candelta, |
|
2673 | 2676 | rawsizefn=self.rawsize, |
|
2674 | 2677 | revdifffn=self.revdiff, |
|
2675 | 2678 | flagsfn=self.flags, |
|
2676 | 2679 | deltamode=deltamode, |
|
2677 | 2680 | revisiondata=revisiondata, |
|
2678 | 2681 | assumehaveparentrevisions=assumehaveparentrevisions, |
|
2679 | 2682 | sidedata_helpers=sidedata_helpers, |
|
2680 | 2683 | ) |
|
2681 | 2684 | |
|
2682 | 2685 | DELTAREUSEALWAYS = b'always' |
|
2683 | 2686 | DELTAREUSESAMEREVS = b'samerevs' |
|
2684 | 2687 | DELTAREUSENEVER = b'never' |
|
2685 | 2688 | |
|
2686 | 2689 | DELTAREUSEFULLADD = b'fulladd' |
|
2687 | 2690 | |
|
2688 | 2691 | DELTAREUSEALL = {b'always', b'samerevs', b'never', b'fulladd'} |
|
2689 | 2692 | |
|
2690 | 2693 | def clone( |
|
2691 | 2694 | self, |
|
2692 | 2695 | tr, |
|
2693 | 2696 | destrevlog, |
|
2694 | 2697 | addrevisioncb=None, |
|
2695 | 2698 | deltareuse=DELTAREUSESAMEREVS, |
|
2696 | 2699 | forcedeltabothparents=None, |
|
2697 | 2700 | sidedata_helpers=None, |
|
2698 | 2701 | ): |
|
2699 | 2702 | """Copy this revlog to another, possibly with format changes. |
|
2700 | 2703 | |
|
2701 | 2704 | The destination revlog will contain the same revisions and nodes. |
|
2702 | 2705 | However, it may not be bit-for-bit identical due to e.g. delta encoding |
|
2703 | 2706 | differences. |
|
2704 | 2707 | |
|
2705 | 2708 | The ``deltareuse`` argument control how deltas from the existing revlog |
|
2706 | 2709 | are preserved in the destination revlog. The argument can have the |
|
2707 | 2710 | following values: |
|
2708 | 2711 | |
|
2709 | 2712 | DELTAREUSEALWAYS |
|
2710 | 2713 | Deltas will always be reused (if possible), even if the destination |
|
2711 | 2714 | revlog would not select the same revisions for the delta. This is the |
|
2712 | 2715 | fastest mode of operation. |
|
2713 | 2716 | DELTAREUSESAMEREVS |
|
2714 | 2717 | Deltas will be reused if the destination revlog would pick the same |
|
2715 | 2718 | revisions for the delta. This mode strikes a balance between speed |
|
2716 | 2719 | and optimization. |
|
2717 | 2720 | DELTAREUSENEVER |
|
2718 | 2721 | Deltas will never be reused. This is the slowest mode of execution. |
|
2719 | 2722 | This mode can be used to recompute deltas (e.g. if the diff/delta |
|
2720 | 2723 | algorithm changes). |
|
2721 | 2724 | DELTAREUSEFULLADD |
|
2722 | 2725 | Revision will be re-added as if their were new content. This is |
|
2723 | 2726 | slower than DELTAREUSEALWAYS but allow more mechanism to kicks in. |
|
2724 | 2727 | eg: large file detection and handling. |
|
2725 | 2728 | |
|
2726 | 2729 | Delta computation can be slow, so the choice of delta reuse policy can |
|
2727 | 2730 | significantly affect run time. |
|
2728 | 2731 | |
|
2729 | 2732 | The default policy (``DELTAREUSESAMEREVS``) strikes a balance between |
|
2730 | 2733 | two extremes. Deltas will be reused if they are appropriate. But if the |
|
2731 | 2734 | delta could choose a better revision, it will do so. This means if you |
|
2732 | 2735 | are converting a non-generaldelta revlog to a generaldelta revlog, |
|
2733 | 2736 | deltas will be recomputed if the delta's parent isn't a parent of the |
|
2734 | 2737 | revision. |
|
2735 | 2738 | |
|
2736 | 2739 | In addition to the delta policy, the ``forcedeltabothparents`` |
|
2737 | 2740 | argument controls whether to force compute deltas against both parents |
|
2738 | 2741 | for merges. By default, the current default is used. |
|
2739 | 2742 | |
|
2740 | 2743 | See `revlogutil.sidedata.get_sidedata_helpers` for the doc on |
|
2741 | 2744 | `sidedata_helpers`. |
|
2742 | 2745 | """ |
|
2743 | 2746 | if deltareuse not in self.DELTAREUSEALL: |
|
2744 | 2747 | raise ValueError( |
|
2745 | 2748 | _(b'value for deltareuse invalid: %s') % deltareuse |
|
2746 | 2749 | ) |
|
2747 | 2750 | |
|
2748 | 2751 | if len(destrevlog): |
|
2749 | 2752 | raise ValueError(_(b'destination revlog is not empty')) |
|
2750 | 2753 | |
|
2751 | 2754 | if getattr(self, 'filteredrevs', None): |
|
2752 | 2755 | raise ValueError(_(b'source revlog has filtered revisions')) |
|
2753 | 2756 | if getattr(destrevlog, 'filteredrevs', None): |
|
2754 | 2757 | raise ValueError(_(b'destination revlog has filtered revisions')) |
|
2755 | 2758 | |
|
2756 | 2759 | # lazydelta and lazydeltabase controls whether to reuse a cached delta, |
|
2757 | 2760 | # if possible. |
|
2758 | 2761 | oldlazydelta = destrevlog._lazydelta |
|
2759 | 2762 | oldlazydeltabase = destrevlog._lazydeltabase |
|
2760 | 2763 | oldamd = destrevlog._deltabothparents |
|
2761 | 2764 | |
|
2762 | 2765 | try: |
|
2763 | 2766 | if deltareuse == self.DELTAREUSEALWAYS: |
|
2764 | 2767 | destrevlog._lazydeltabase = True |
|
2765 | 2768 | destrevlog._lazydelta = True |
|
2766 | 2769 | elif deltareuse == self.DELTAREUSESAMEREVS: |
|
2767 | 2770 | destrevlog._lazydeltabase = False |
|
2768 | 2771 | destrevlog._lazydelta = True |
|
2769 | 2772 | elif deltareuse == self.DELTAREUSENEVER: |
|
2770 | 2773 | destrevlog._lazydeltabase = False |
|
2771 | 2774 | destrevlog._lazydelta = False |
|
2772 | 2775 | |
|
2773 | 2776 | destrevlog._deltabothparents = forcedeltabothparents or oldamd |
|
2774 | 2777 | |
|
2775 | 2778 | self._clone( |
|
2776 | 2779 | tr, |
|
2777 | 2780 | destrevlog, |
|
2778 | 2781 | addrevisioncb, |
|
2779 | 2782 | deltareuse, |
|
2780 | 2783 | forcedeltabothparents, |
|
2781 | 2784 | sidedata_helpers, |
|
2782 | 2785 | ) |
|
2783 | 2786 | |
|
2784 | 2787 | finally: |
|
2785 | 2788 | destrevlog._lazydelta = oldlazydelta |
|
2786 | 2789 | destrevlog._lazydeltabase = oldlazydeltabase |
|
2787 | 2790 | destrevlog._deltabothparents = oldamd |
|
2788 | 2791 | |
|
2789 | 2792 | def _clone( |
|
2790 | 2793 | self, |
|
2791 | 2794 | tr, |
|
2792 | 2795 | destrevlog, |
|
2793 | 2796 | addrevisioncb, |
|
2794 | 2797 | deltareuse, |
|
2795 | 2798 | forcedeltabothparents, |
|
2796 | 2799 | sidedata_helpers, |
|
2797 | 2800 | ): |
|
2798 | 2801 | """perform the core duty of `revlog.clone` after parameter processing""" |
|
2799 | 2802 | deltacomputer = deltautil.deltacomputer(destrevlog) |
|
2800 | 2803 | index = self.index |
|
2801 | 2804 | for rev in self: |
|
2802 | 2805 | entry = index[rev] |
|
2803 | 2806 | |
|
2804 | 2807 | # Some classes override linkrev to take filtered revs into |
|
2805 | 2808 | # account. Use raw entry from index. |
|
2806 | 2809 | flags = entry[0] & 0xFFFF |
|
2807 | 2810 | linkrev = entry[4] |
|
2808 | 2811 | p1 = index[entry[5]][7] |
|
2809 | 2812 | p2 = index[entry[6]][7] |
|
2810 | 2813 | node = entry[7] |
|
2811 | 2814 | |
|
2812 | 2815 | # (Possibly) reuse the delta from the revlog if allowed and |
|
2813 | 2816 | # the revlog chunk is a delta. |
|
2814 | 2817 | cachedelta = None |
|
2815 | 2818 | rawtext = None |
|
2816 | 2819 | if deltareuse == self.DELTAREUSEFULLADD: |
|
2817 | 2820 | text, sidedata = self._revisiondata(rev) |
|
2818 | 2821 | |
|
2819 | 2822 | if sidedata_helpers is not None: |
|
2820 | 2823 | (sidedata, new_flags) = sidedatautil.run_sidedata_helpers( |
|
2821 | 2824 | self, sidedata_helpers, sidedata, rev |
|
2822 | 2825 | ) |
|
2823 | 2826 | flags = flags | new_flags[0] & ~new_flags[1] |
|
2824 | 2827 | |
|
2825 | 2828 | destrevlog.addrevision( |
|
2826 | 2829 | text, |
|
2827 | 2830 | tr, |
|
2828 | 2831 | linkrev, |
|
2829 | 2832 | p1, |
|
2830 | 2833 | p2, |
|
2831 | 2834 | cachedelta=cachedelta, |
|
2832 | 2835 | node=node, |
|
2833 | 2836 | flags=flags, |
|
2834 | 2837 | deltacomputer=deltacomputer, |
|
2835 | 2838 | sidedata=sidedata, |
|
2836 | 2839 | ) |
|
2837 | 2840 | else: |
|
2838 | 2841 | if destrevlog._lazydelta: |
|
2839 | 2842 | dp = self.deltaparent(rev) |
|
2840 | 2843 | if dp != nullrev: |
|
2841 | 2844 | cachedelta = (dp, bytes(self._chunk(rev))) |
|
2842 | 2845 | |
|
2843 | 2846 | sidedata = None |
|
2844 | 2847 | if not cachedelta: |
|
2845 | 2848 | rawtext, sidedata = self._revisiondata(rev) |
|
2846 | 2849 | if sidedata is None: |
|
2847 | 2850 | sidedata = self.sidedata(rev) |
|
2848 | 2851 | |
|
2849 | 2852 | if sidedata_helpers is not None: |
|
2850 | 2853 | (sidedata, new_flags) = sidedatautil.run_sidedata_helpers( |
|
2851 | 2854 | self, sidedata_helpers, sidedata, rev |
|
2852 | 2855 | ) |
|
2853 | 2856 | flags = flags | new_flags[0] & ~new_flags[1] |
|
2854 | 2857 | |
|
2855 | 2858 | ifh = destrevlog.opener( |
|
2856 | 2859 | destrevlog._indexfile, b'a+', checkambig=False |
|
2857 | 2860 | ) |
|
2858 | 2861 | dfh = None |
|
2859 | 2862 | if not destrevlog._inline: |
|
2860 | 2863 | dfh = destrevlog.opener(destrevlog._datafile, b'a+') |
|
2861 | 2864 | try: |
|
2862 | 2865 | destrevlog._addrevision( |
|
2863 | 2866 | node, |
|
2864 | 2867 | rawtext, |
|
2865 | 2868 | tr, |
|
2866 | 2869 | linkrev, |
|
2867 | 2870 | p1, |
|
2868 | 2871 | p2, |
|
2869 | 2872 | flags, |
|
2870 | 2873 | cachedelta, |
|
2871 | 2874 | ifh, |
|
2872 | 2875 | dfh, |
|
2873 | 2876 | deltacomputer=deltacomputer, |
|
2874 | 2877 | sidedata=sidedata, |
|
2875 | 2878 | ) |
|
2876 | 2879 | finally: |
|
2877 | 2880 | if dfh: |
|
2878 | 2881 | dfh.close() |
|
2879 | 2882 | ifh.close() |
|
2880 | 2883 | |
|
2881 | 2884 | if addrevisioncb: |
|
2882 | 2885 | addrevisioncb(self, rev, node) |
|
2883 | 2886 | |
|
2884 | 2887 | def censorrevision(self, tr, censornode, tombstone=b''): |
|
2885 | 2888 | if self._format_version == REVLOGV0: |
|
2886 | 2889 | raise error.RevlogError( |
|
2887 | 2890 | _(b'cannot censor with version %d revlogs') |
|
2888 | 2891 | % self._format_version |
|
2889 | 2892 | ) |
|
2890 | 2893 | |
|
2891 | 2894 | censorrev = self.rev(censornode) |
|
2892 | 2895 | tombstone = storageutil.packmeta({b'censored': tombstone}, b'') |
|
2893 | 2896 | |
|
2894 | 2897 | if len(tombstone) > self.rawsize(censorrev): |
|
2895 | 2898 | raise error.Abort( |
|
2896 | 2899 | _(b'censor tombstone must be no longer than censored data') |
|
2897 | 2900 | ) |
|
2898 | 2901 | |
|
2899 | 2902 | # Rewriting the revlog in place is hard. Our strategy for censoring is |
|
2900 | 2903 | # to create a new revlog, copy all revisions to it, then replace the |
|
2901 | 2904 | # revlogs on transaction close. |
|
2902 | 2905 | # |
|
2903 | 2906 | # This is a bit dangerous. We could easily have a mismatch of state. |
|
2904 | 2907 | newrl = revlog( |
|
2905 | 2908 | self.opener, |
|
2906 | 2909 | target=self.target, |
|
2907 | 2910 | radix=self.radix, |
|
2908 | 2911 | postfix=b'tmpcensored', |
|
2909 | 2912 | censorable=True, |
|
2910 | 2913 | ) |
|
2911 | 2914 | newrl._format_version = self._format_version |
|
2912 | 2915 | newrl._format_flags = self._format_flags |
|
2913 | 2916 | newrl._generaldelta = self._generaldelta |
|
2914 | 2917 | newrl._parse_index = self._parse_index |
|
2915 | 2918 | |
|
2916 | 2919 | for rev in self.revs(): |
|
2917 | 2920 | node = self.node(rev) |
|
2918 | 2921 | p1, p2 = self.parents(node) |
|
2919 | 2922 | |
|
2920 | 2923 | if rev == censorrev: |
|
2921 | 2924 | newrl.addrawrevision( |
|
2922 | 2925 | tombstone, |
|
2923 | 2926 | tr, |
|
2924 | 2927 | self.linkrev(censorrev), |
|
2925 | 2928 | p1, |
|
2926 | 2929 | p2, |
|
2927 | 2930 | censornode, |
|
2928 | 2931 | REVIDX_ISCENSORED, |
|
2929 | 2932 | ) |
|
2930 | 2933 | |
|
2931 | 2934 | if newrl.deltaparent(rev) != nullrev: |
|
2932 | 2935 | raise error.Abort( |
|
2933 | 2936 | _( |
|
2934 | 2937 | b'censored revision stored as delta; ' |
|
2935 | 2938 | b'cannot censor' |
|
2936 | 2939 | ), |
|
2937 | 2940 | hint=_( |
|
2938 | 2941 | b'censoring of revlogs is not ' |
|
2939 | 2942 | b'fully implemented; please report ' |
|
2940 | 2943 | b'this bug' |
|
2941 | 2944 | ), |
|
2942 | 2945 | ) |
|
2943 | 2946 | continue |
|
2944 | 2947 | |
|
2945 | 2948 | if self.iscensored(rev): |
|
2946 | 2949 | if self.deltaparent(rev) != nullrev: |
|
2947 | 2950 | raise error.Abort( |
|
2948 | 2951 | _( |
|
2949 | 2952 | b'cannot censor due to censored ' |
|
2950 | 2953 | b'revision having delta stored' |
|
2951 | 2954 | ) |
|
2952 | 2955 | ) |
|
2953 | 2956 | rawtext = self._chunk(rev) |
|
2954 | 2957 | else: |
|
2955 | 2958 | rawtext = self.rawdata(rev) |
|
2956 | 2959 | |
|
2957 | 2960 | newrl.addrawrevision( |
|
2958 | 2961 | rawtext, tr, self.linkrev(rev), p1, p2, node, self.flags(rev) |
|
2959 | 2962 | ) |
|
2960 | 2963 | |
|
2961 | 2964 | tr.addbackup(self._indexfile, location=b'store') |
|
2962 | 2965 | if not self._inline: |
|
2963 | 2966 | tr.addbackup(self._datafile, location=b'store') |
|
2964 | 2967 | |
|
2965 | 2968 | self.opener.rename(newrl._indexfile, self._indexfile) |
|
2966 | 2969 | if not self._inline: |
|
2967 | 2970 | self.opener.rename(newrl._datafile, self._datafile) |
|
2968 | 2971 | |
|
2969 | 2972 | self.clearcaches() |
|
2970 | 2973 | self._loadindex() |
|
2971 | 2974 | |
|
2972 | 2975 | def verifyintegrity(self, state): |
|
2973 | 2976 | """Verifies the integrity of the revlog. |
|
2974 | 2977 | |
|
2975 | 2978 | Yields ``revlogproblem`` instances describing problems that are |
|
2976 | 2979 | found. |
|
2977 | 2980 | """ |
|
2978 | 2981 | dd, di = self.checksize() |
|
2979 | 2982 | if dd: |
|
2980 | 2983 | yield revlogproblem(error=_(b'data length off by %d bytes') % dd) |
|
2981 | 2984 | if di: |
|
2982 | 2985 | yield revlogproblem(error=_(b'index contains %d extra bytes') % di) |
|
2983 | 2986 | |
|
2984 | 2987 | version = self._format_version |
|
2985 | 2988 | |
|
2986 | 2989 | # The verifier tells us what version revlog we should be. |
|
2987 | 2990 | if version != state[b'expectedversion']: |
|
2988 | 2991 | yield revlogproblem( |
|
2989 | 2992 | warning=_(b"warning: '%s' uses revlog format %d; expected %d") |
|
2990 | 2993 | % (self.display_id, version, state[b'expectedversion']) |
|
2991 | 2994 | ) |
|
2992 | 2995 | |
|
2993 | 2996 | state[b'skipread'] = set() |
|
2994 | 2997 | state[b'safe_renamed'] = set() |
|
2995 | 2998 | |
|
2996 | 2999 | for rev in self: |
|
2997 | 3000 | node = self.node(rev) |
|
2998 | 3001 | |
|
2999 | 3002 | # Verify contents. 4 cases to care about: |
|
3000 | 3003 | # |
|
3001 | 3004 | # common: the most common case |
|
3002 | 3005 | # rename: with a rename |
|
3003 | 3006 | # meta: file content starts with b'\1\n', the metadata |
|
3004 | 3007 | # header defined in filelog.py, but without a rename |
|
3005 | 3008 | # ext: content stored externally |
|
3006 | 3009 | # |
|
3007 | 3010 | # More formally, their differences are shown below: |
|
3008 | 3011 | # |
|
3009 | 3012 | # | common | rename | meta | ext |
|
3010 | 3013 | # ------------------------------------------------------- |
|
3011 | 3014 | # flags() | 0 | 0 | 0 | not 0 |
|
3012 | 3015 | # renamed() | False | True | False | ? |
|
3013 | 3016 | # rawtext[0:2]=='\1\n'| False | True | True | ? |
|
3014 | 3017 | # |
|
3015 | 3018 | # "rawtext" means the raw text stored in revlog data, which |
|
3016 | 3019 | # could be retrieved by "rawdata(rev)". "text" |
|
3017 | 3020 | # mentioned below is "revision(rev)". |
|
3018 | 3021 | # |
|
3019 | 3022 | # There are 3 different lengths stored physically: |
|
3020 | 3023 | # 1. L1: rawsize, stored in revlog index |
|
3021 | 3024 | # 2. L2: len(rawtext), stored in revlog data |
|
3022 | 3025 | # 3. L3: len(text), stored in revlog data if flags==0, or |
|
3023 | 3026 | # possibly somewhere else if flags!=0 |
|
3024 | 3027 | # |
|
3025 | 3028 | # L1 should be equal to L2. L3 could be different from them. |
|
3026 | 3029 | # "text" may or may not affect commit hash depending on flag |
|
3027 | 3030 | # processors (see flagutil.addflagprocessor). |
|
3028 | 3031 | # |
|
3029 | 3032 | # | common | rename | meta | ext |
|
3030 | 3033 | # ------------------------------------------------- |
|
3031 | 3034 | # rawsize() | L1 | L1 | L1 | L1 |
|
3032 | 3035 | # size() | L1 | L2-LM | L1(*) | L1 (?) |
|
3033 | 3036 | # len(rawtext) | L2 | L2 | L2 | L2 |
|
3034 | 3037 | # len(text) | L2 | L2 | L2 | L3 |
|
3035 | 3038 | # len(read()) | L2 | L2-LM | L2-LM | L3 (?) |
|
3036 | 3039 | # |
|
3037 | 3040 | # LM: length of metadata, depending on rawtext |
|
3038 | 3041 | # (*): not ideal, see comment in filelog.size |
|
3039 | 3042 | # (?): could be "- len(meta)" if the resolved content has |
|
3040 | 3043 | # rename metadata |
|
3041 | 3044 | # |
|
3042 | 3045 | # Checks needed to be done: |
|
3043 | 3046 | # 1. length check: L1 == L2, in all cases. |
|
3044 | 3047 | # 2. hash check: depending on flag processor, we may need to |
|
3045 | 3048 | # use either "text" (external), or "rawtext" (in revlog). |
|
3046 | 3049 | |
|
3047 | 3050 | try: |
|
3048 | 3051 | skipflags = state.get(b'skipflags', 0) |
|
3049 | 3052 | if skipflags: |
|
3050 | 3053 | skipflags &= self.flags(rev) |
|
3051 | 3054 | |
|
3052 | 3055 | _verify_revision(self, skipflags, state, node) |
|
3053 | 3056 | |
|
3054 | 3057 | l1 = self.rawsize(rev) |
|
3055 | 3058 | l2 = len(self.rawdata(node)) |
|
3056 | 3059 | |
|
3057 | 3060 | if l1 != l2: |
|
3058 | 3061 | yield revlogproblem( |
|
3059 | 3062 | error=_(b'unpacked size is %d, %d expected') % (l2, l1), |
|
3060 | 3063 | node=node, |
|
3061 | 3064 | ) |
|
3062 | 3065 | |
|
3063 | 3066 | except error.CensoredNodeError: |
|
3064 | 3067 | if state[b'erroroncensored']: |
|
3065 | 3068 | yield revlogproblem( |
|
3066 | 3069 | error=_(b'censored file data'), node=node |
|
3067 | 3070 | ) |
|
3068 | 3071 | state[b'skipread'].add(node) |
|
3069 | 3072 | except Exception as e: |
|
3070 | 3073 | yield revlogproblem( |
|
3071 | 3074 | error=_(b'unpacking %s: %s') |
|
3072 | 3075 | % (short(node), stringutil.forcebytestr(e)), |
|
3073 | 3076 | node=node, |
|
3074 | 3077 | ) |
|
3075 | 3078 | state[b'skipread'].add(node) |
|
3076 | 3079 | |
|
3077 | 3080 | def storageinfo( |
|
3078 | 3081 | self, |
|
3079 | 3082 | exclusivefiles=False, |
|
3080 | 3083 | sharedfiles=False, |
|
3081 | 3084 | revisionscount=False, |
|
3082 | 3085 | trackedsize=False, |
|
3083 | 3086 | storedsize=False, |
|
3084 | 3087 | ): |
|
3085 | 3088 | d = {} |
|
3086 | 3089 | |
|
3087 | 3090 | if exclusivefiles: |
|
3088 | 3091 | d[b'exclusivefiles'] = [(self.opener, self._indexfile)] |
|
3089 | 3092 | if not self._inline: |
|
3090 | 3093 | d[b'exclusivefiles'].append((self.opener, self._datafile)) |
|
3091 | 3094 | |
|
3092 | 3095 | if sharedfiles: |
|
3093 | 3096 | d[b'sharedfiles'] = [] |
|
3094 | 3097 | |
|
3095 | 3098 | if revisionscount: |
|
3096 | 3099 | d[b'revisionscount'] = len(self) |
|
3097 | 3100 | |
|
3098 | 3101 | if trackedsize: |
|
3099 | 3102 | d[b'trackedsize'] = sum(map(self.rawsize, iter(self))) |
|
3100 | 3103 | |
|
3101 | 3104 | if storedsize: |
|
3102 | 3105 | d[b'storedsize'] = sum( |
|
3103 | 3106 | self.opener.stat(path).st_size for path in self.files() |
|
3104 | 3107 | ) |
|
3105 | 3108 | |
|
3106 | 3109 | return d |
|
3107 | 3110 | |
|
3108 | 3111 | def rewrite_sidedata(self, helpers, startrev, endrev): |
|
3109 | 3112 | if not self.hassidedata: |
|
3110 | 3113 | return |
|
3111 | 3114 | # inline are not yet supported because they suffer from an issue when |
|
3112 | 3115 | # rewriting them (since it's not an append-only operation). |
|
3113 | 3116 | # See issue6485. |
|
3114 | 3117 | assert not self._inline |
|
3115 | 3118 | if not helpers[1] and not helpers[2]: |
|
3116 | 3119 | # Nothing to generate or remove |
|
3117 | 3120 | return |
|
3118 | 3121 | |
|
3119 | 3122 | # changelog implement some "delayed" writing mechanism that assume that |
|
3120 | 3123 | # all index data is writen in append mode and is therefor incompatible |
|
3121 | 3124 | # with the seeked write done in this method. The use of such "delayed" |
|
3122 | 3125 | # writing will soon be removed for revlog version that support side |
|
3123 | 3126 | # data, so for now, we only keep this simple assert to highlight the |
|
3124 | 3127 | # situation. |
|
3125 | 3128 | delayed = getattr(self, '_delayed', False) |
|
3126 | 3129 | diverted = getattr(self, '_divert', False) |
|
3127 | 3130 | if delayed and not diverted: |
|
3128 | 3131 | msg = "cannot rewrite_sidedata of a delayed revlog" |
|
3129 | 3132 | raise error.ProgrammingError(msg) |
|
3130 | 3133 | |
|
3131 | 3134 | new_entries = [] |
|
3132 | 3135 | # append the new sidedata |
|
3133 | 3136 | with self._datafp(b'a+') as fp: |
|
3134 | 3137 | # Maybe this bug still exists, see revlog._writeentry |
|
3135 | 3138 | fp.seek(0, os.SEEK_END) |
|
3136 | 3139 | current_offset = fp.tell() |
|
3137 | 3140 | for rev in range(startrev, endrev + 1): |
|
3138 | 3141 | entry = self.index[rev] |
|
3139 | 3142 | new_sidedata, flags = sidedatautil.run_sidedata_helpers( |
|
3140 | 3143 | store=self, |
|
3141 | 3144 | sidedata_helpers=helpers, |
|
3142 | 3145 | sidedata={}, |
|
3143 | 3146 | rev=rev, |
|
3144 | 3147 | ) |
|
3145 | 3148 | |
|
3146 | 3149 | serialized_sidedata = sidedatautil.serialize_sidedata( |
|
3147 | 3150 | new_sidedata |
|
3148 | 3151 | ) |
|
3149 | 3152 | if entry[8] != 0 or entry[9] != 0: |
|
3150 | 3153 | # rewriting entries that already have sidedata is not |
|
3151 | 3154 | # supported yet, because it introduces garbage data in the |
|
3152 | 3155 | # revlog. |
|
3153 | 3156 | msg = b"Rewriting existing sidedata is not supported yet" |
|
3154 | 3157 | raise error.Abort(msg) |
|
3155 | 3158 | |
|
3156 | 3159 | # Apply (potential) flags to add and to remove after running |
|
3157 | 3160 | # the sidedata helpers |
|
3158 | 3161 | new_offset_flags = entry[0] | flags[0] & ~flags[1] |
|
3159 | 3162 | entry = (new_offset_flags,) + entry[1:8] |
|
3160 | 3163 | entry += (current_offset, len(serialized_sidedata)) |
|
3161 | 3164 | |
|
3162 | 3165 | fp.write(serialized_sidedata) |
|
3163 | 3166 | new_entries.append(entry) |
|
3164 | 3167 | current_offset += len(serialized_sidedata) |
|
3165 | 3168 | |
|
3166 | 3169 | # rewrite the new index entries |
|
3167 | 3170 | with self._indexfp(b'r+') as fp: |
|
3168 | 3171 | fp.seek(startrev * self.index.entry_size) |
|
3169 | 3172 | for i, e in enumerate(new_entries): |
|
3170 | 3173 | rev = startrev + i |
|
3171 | 3174 | self.index.replace_sidedata_info(rev, e[8], e[9], e[0]) |
|
3172 | 3175 | packed = self.index.entry_binary(rev) |
|
3173 | 3176 | if rev == 0: |
|
3174 | 3177 | header = self._format_flags | self._format_version |
|
3175 | 3178 | header = self.index.pack_header(header) |
|
3176 | 3179 | packed = header + packed |
|
3177 | 3180 | fp.write(packed) |
General Comments 0
You need to be logged in to leave comments.
Login now