Show More
@@ -1,1042 +1,1045 b'' | |||
|
1 | 1 | # branchmap.py - logic to computes, maintain and stores branchmap for local repo |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2005-2007 Olivia Mackall <olivia@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | |
|
9 | 9 | import struct |
|
10 | 10 | |
|
11 | 11 | from .node import ( |
|
12 | 12 | bin, |
|
13 | 13 | hex, |
|
14 | 14 | nullrev, |
|
15 | 15 | ) |
|
16 | 16 | |
|
17 | 17 | from typing import ( |
|
18 | 18 | Any, |
|
19 | 19 | Callable, |
|
20 | 20 | Dict, |
|
21 | 21 | Iterable, |
|
22 | 22 | List, |
|
23 | 23 | Optional, |
|
24 | 24 | Set, |
|
25 | 25 | TYPE_CHECKING, |
|
26 | 26 | Tuple, |
|
27 | 27 | Union, |
|
28 | 28 | ) |
|
29 | 29 | |
|
30 | 30 | from . import ( |
|
31 | 31 | encoding, |
|
32 | 32 | error, |
|
33 | 33 | obsolete, |
|
34 | 34 | scmutil, |
|
35 | 35 | util, |
|
36 | 36 | ) |
|
37 | 37 | |
|
38 | 38 | from .utils import ( |
|
39 | 39 | repoviewutil, |
|
40 | 40 | stringutil, |
|
41 | 41 | ) |
|
42 | 42 | |
|
43 | 43 | if TYPE_CHECKING: |
|
44 | 44 | from . import localrepo |
|
45 | 45 | |
|
46 | 46 | assert [localrepo] |
|
47 | 47 | |
|
48 | 48 | subsettable = repoviewutil.subsettable |
|
49 | 49 | |
|
50 | 50 | calcsize = struct.calcsize |
|
51 | 51 | pack_into = struct.pack_into |
|
52 | 52 | unpack_from = struct.unpack_from |
|
53 | 53 | |
|
54 | 54 | |
|
55 | 55 | class BranchMapCache: |
|
56 | 56 | """mapping of filtered views of repo with their branchcache""" |
|
57 | 57 | |
|
58 | 58 | def __init__(self): |
|
59 | 59 | self._per_filter = {} |
|
60 | 60 | |
|
61 | 61 | def __getitem__(self, repo): |
|
62 | 62 | self.updatecache(repo) |
|
63 | 63 | bcache = self._per_filter[repo.filtername] |
|
64 | 64 | assert bcache._filtername == repo.filtername, ( |
|
65 | 65 | bcache._filtername, |
|
66 | 66 | repo.filtername, |
|
67 | 67 | ) |
|
68 | 68 | return bcache |
|
69 | 69 | |
|
70 | 70 | def update_disk(self, repo): |
|
71 | 71 | """ensure and up-to-date cache is (or will be) written on disk |
|
72 | 72 | |
|
73 | 73 | The cache for this repository view is updated if needed and written on |
|
74 | 74 | disk. |
|
75 | 75 | |
|
76 | 76 | If a transaction is in progress, the writing is schedule to transaction |
|
77 | 77 | close. See the `BranchMapCache.write_delayed` method. |
|
78 | 78 | |
|
79 | 79 | This method exist independently of __getitem__ as it is sometime useful |
|
80 | 80 | to signal that we have no intend to use the data in memory yet. |
|
81 | 81 | """ |
|
82 | 82 | self.updatecache(repo) |
|
83 | 83 | bcache = self._per_filter[repo.filtername] |
|
84 | 84 | assert bcache._filtername == repo.filtername, ( |
|
85 | 85 | bcache._filtername, |
|
86 | 86 | repo.filtername, |
|
87 | 87 | ) |
|
88 | 88 | bcache.write(repo) |
|
89 | 89 | |
|
90 | 90 | def updatecache(self, repo): |
|
91 | 91 | """Update the cache for the given filtered view on a repository""" |
|
92 | 92 | # This can trigger updates for the caches for subsets of the filtered |
|
93 | 93 | # view, e.g. when there is no cache for this filtered view or the cache |
|
94 | 94 | # is stale. |
|
95 | 95 | |
|
96 | 96 | cl = repo.changelog |
|
97 | 97 | filtername = repo.filtername |
|
98 | 98 | bcache = self._per_filter.get(filtername) |
|
99 | 99 | if bcache is None or not bcache.validfor(repo): |
|
100 | 100 | # cache object missing or cache object stale? Read from disk |
|
101 | 101 | bcache = branchcache.fromfile(repo) |
|
102 | 102 | |
|
103 | 103 | revs = [] |
|
104 | 104 | if bcache is None: |
|
105 | 105 | # no (fresh) cache available anymore, perhaps we can re-use |
|
106 | 106 | # the cache for a subset, then extend that to add info on missing |
|
107 | 107 | # revisions. |
|
108 | 108 | subsetname = subsettable.get(filtername) |
|
109 | 109 | if subsetname is not None: |
|
110 | 110 | subset = repo.filtered(subsetname) |
|
111 | 111 | bcache = self[subset].copy(repo) |
|
112 | 112 | extrarevs = subset.changelog.filteredrevs - cl.filteredrevs |
|
113 | 113 | revs.extend(r for r in extrarevs if r <= bcache.tiprev) |
|
114 | 114 | else: |
|
115 | 115 | # nothing to fall back on, start empty. |
|
116 | 116 | bcache = branchcache(repo) |
|
117 | 117 | |
|
118 | 118 | revs.extend(cl.revs(start=bcache.tiprev + 1)) |
|
119 | 119 | if revs: |
|
120 | 120 | bcache.update(repo, revs) |
|
121 | 121 | |
|
122 | 122 | assert bcache.validfor(repo), filtername |
|
123 | 123 | self._per_filter[repo.filtername] = bcache |
|
124 | 124 | |
|
125 | 125 | def replace(self, repo, remotebranchmap): |
|
126 | 126 | """Replace the branchmap cache for a repo with a branch mapping. |
|
127 | 127 | |
|
128 | 128 | This is likely only called during clone with a branch map from a |
|
129 | 129 | remote. |
|
130 | 130 | |
|
131 | 131 | """ |
|
132 | 132 | cl = repo.changelog |
|
133 | 133 | clrev = cl.rev |
|
134 | 134 | clbranchinfo = cl.branchinfo |
|
135 | 135 | rbheads = [] |
|
136 | 136 | closed = set() |
|
137 | 137 | for bheads in remotebranchmap.values(): |
|
138 | 138 | rbheads += bheads |
|
139 | 139 | for h in bheads: |
|
140 | 140 | r = clrev(h) |
|
141 | 141 | b, c = clbranchinfo(r) |
|
142 | 142 | if c: |
|
143 | 143 | closed.add(h) |
|
144 | 144 | |
|
145 | 145 | if rbheads: |
|
146 | 146 | rtiprev = max((int(clrev(node)) for node in rbheads)) |
|
147 | 147 | cache = branchcache( |
|
148 | 148 | repo, |
|
149 | 149 | remotebranchmap, |
|
150 | 150 | repo[rtiprev].node(), |
|
151 | 151 | rtiprev, |
|
152 | 152 | closednodes=closed, |
|
153 | 153 | ) |
|
154 | 154 | |
|
155 | 155 | # Try to stick it as low as possible |
|
156 | 156 | # filter above served are unlikely to be fetch from a clone |
|
157 | 157 | for candidate in (b'base', b'immutable', b'served'): |
|
158 | 158 | rview = repo.filtered(candidate) |
|
159 | 159 | if cache.validfor(rview): |
|
160 | 160 | cache = self._per_filter[candidate] = cache.copy(rview) |
|
161 | 161 | cache.write(rview) |
|
162 | 162 | return |
|
163 | 163 | |
|
164 | 164 | def clear(self): |
|
165 | 165 | self._per_filter.clear() |
|
166 | 166 | |
|
167 | 167 | def write_delayed(self, repo): |
|
168 | 168 | unfi = repo.unfiltered() |
|
169 | 169 | for filtername, cache in self._per_filter.items(): |
|
170 | 170 | if cache._delayed: |
|
171 |
|
|
|
171 | if filtername is None: | |
|
172 | repo = unfi | |
|
173 | else: | |
|
174 | repo = unfi.filtered(filtername) | |
|
172 | 175 | cache.write(repo) |
|
173 | 176 | |
|
174 | 177 | |
|
175 | 178 | def _unknownnode(node): |
|
176 | 179 | """raises ValueError when branchcache found a node which does not exists""" |
|
177 | 180 | raise ValueError('node %s does not exist' % node.hex()) |
|
178 | 181 | |
|
179 | 182 | |
|
180 | 183 | def _branchcachedesc(repo): |
|
181 | 184 | if repo.filtername is not None: |
|
182 | 185 | return b'branch cache (%s)' % repo.filtername |
|
183 | 186 | else: |
|
184 | 187 | return b'branch cache' |
|
185 | 188 | |
|
186 | 189 | |
|
187 | 190 | class _BaseBranchCache: |
|
188 | 191 | """A dict like object that hold branches heads cache. |
|
189 | 192 | |
|
190 | 193 | This cache is used to avoid costly computations to determine all the |
|
191 | 194 | branch heads of a repo. |
|
192 | 195 | |
|
193 | 196 | The cache is serialized on disk in the following format: |
|
194 | 197 | |
|
195 | 198 | <tip hex node> <tip rev number> [optional filtered repo hex hash] |
|
196 | 199 | <branch head hex node> <open/closed state> <branch name> |
|
197 | 200 | <branch head hex node> <open/closed state> <branch name> |
|
198 | 201 | ... |
|
199 | 202 | |
|
200 | 203 | The first line is used to check if the cache is still valid. If the |
|
201 | 204 | branch cache is for a filtered repo view, an optional third hash is |
|
202 | 205 | included that hashes the hashes of all filtered and obsolete revisions. |
|
203 | 206 | |
|
204 | 207 | The open/closed state is represented by a single letter 'o' or 'c'. |
|
205 | 208 | This field can be used to avoid changelog reads when determining if a |
|
206 | 209 | branch head closes a branch or not. |
|
207 | 210 | """ |
|
208 | 211 | |
|
209 | 212 | def __init__( |
|
210 | 213 | self, |
|
211 | 214 | repo: "localrepo.localrepository", |
|
212 | 215 | entries: Union[ |
|
213 | 216 | Dict[bytes, List[bytes]], Iterable[Tuple[bytes, List[bytes]]] |
|
214 | 217 | ] = (), |
|
215 | 218 | closed_nodes: Optional[Set[bytes]] = None, |
|
216 | 219 | ) -> None: |
|
217 | 220 | """hasnode is a function which can be used to verify whether changelog |
|
218 | 221 | has a given node or not. If it's not provided, we assume that every node |
|
219 | 222 | we have exists in changelog""" |
|
220 | 223 | # closednodes is a set of nodes that close their branch. If the branch |
|
221 | 224 | # cache has been updated, it may contain nodes that are no longer |
|
222 | 225 | # heads. |
|
223 | 226 | if closed_nodes is None: |
|
224 | 227 | closed_nodes = set() |
|
225 | 228 | self._closednodes = set(closed_nodes) |
|
226 | 229 | self._entries = dict(entries) |
|
227 | 230 | |
|
228 | 231 | def __iter__(self): |
|
229 | 232 | return iter(self._entries) |
|
230 | 233 | |
|
231 | 234 | def __setitem__(self, key, value): |
|
232 | 235 | self._entries[key] = value |
|
233 | 236 | |
|
234 | 237 | def __getitem__(self, key): |
|
235 | 238 | return self._entries[key] |
|
236 | 239 | |
|
237 | 240 | def __contains__(self, key): |
|
238 | 241 | return key in self._entries |
|
239 | 242 | |
|
240 | 243 | def iteritems(self): |
|
241 | 244 | return self._entries.items() |
|
242 | 245 | |
|
243 | 246 | items = iteritems |
|
244 | 247 | |
|
245 | 248 | def hasbranch(self, label): |
|
246 | 249 | """checks whether a branch of this name exists or not""" |
|
247 | 250 | return label in self._entries |
|
248 | 251 | |
|
249 | 252 | def _branchtip(self, heads): |
|
250 | 253 | """Return tuple with last open head in heads and false, |
|
251 | 254 | otherwise return last closed head and true.""" |
|
252 | 255 | tip = heads[-1] |
|
253 | 256 | closed = True |
|
254 | 257 | for h in reversed(heads): |
|
255 | 258 | if h not in self._closednodes: |
|
256 | 259 | tip = h |
|
257 | 260 | closed = False |
|
258 | 261 | break |
|
259 | 262 | return tip, closed |
|
260 | 263 | |
|
261 | 264 | def branchtip(self, branch): |
|
262 | 265 | """Return the tipmost open head on branch head, otherwise return the |
|
263 | 266 | tipmost closed head on branch. |
|
264 | 267 | Raise KeyError for unknown branch.""" |
|
265 | 268 | return self._branchtip(self[branch])[0] |
|
266 | 269 | |
|
267 | 270 | def iteropen(self, nodes): |
|
268 | 271 | return (n for n in nodes if n not in self._closednodes) |
|
269 | 272 | |
|
270 | 273 | def branchheads(self, branch, closed=False): |
|
271 | 274 | heads = self._entries[branch] |
|
272 | 275 | if not closed: |
|
273 | 276 | heads = list(self.iteropen(heads)) |
|
274 | 277 | return heads |
|
275 | 278 | |
|
276 | 279 | def iterbranches(self): |
|
277 | 280 | for bn, heads in self.items(): |
|
278 | 281 | yield (bn, heads) + self._branchtip(heads) |
|
279 | 282 | |
|
280 | 283 | def iterheads(self): |
|
281 | 284 | """returns all the heads""" |
|
282 | 285 | return self._entries.values() |
|
283 | 286 | |
|
284 | 287 | def update(self, repo, revgen): |
|
285 | 288 | """Given a branchhead cache, self, that may have extra nodes or be |
|
286 | 289 | missing heads, and a generator of nodes that are strictly a superset of |
|
287 | 290 | heads missing, this function updates self to be correct. |
|
288 | 291 | """ |
|
289 | 292 | starttime = util.timer() |
|
290 | 293 | cl = repo.changelog |
|
291 | 294 | # collect new branch entries |
|
292 | 295 | newbranches = {} |
|
293 | 296 | getbranchinfo = repo.revbranchcache().branchinfo |
|
294 | 297 | max_rev = -1 |
|
295 | 298 | for r in revgen: |
|
296 | 299 | branch, closesbranch = getbranchinfo(r) |
|
297 | 300 | newbranches.setdefault(branch, []).append(r) |
|
298 | 301 | if closesbranch: |
|
299 | 302 | self._closednodes.add(cl.node(r)) |
|
300 | 303 | max_rev = max(max_rev, r) |
|
301 | 304 | if max_rev < 0: |
|
302 | 305 | max_rev = None |
|
303 | 306 | |
|
304 | 307 | # Delay fetching the topological heads until they are needed. |
|
305 | 308 | # A repository without non-continous branches can skip this part. |
|
306 | 309 | topoheads = None |
|
307 | 310 | |
|
308 | 311 | # If a changeset is visible, its parents must be visible too, so |
|
309 | 312 | # use the faster unfiltered parent accessor. |
|
310 | 313 | parentrevs = repo.unfiltered().changelog.parentrevs |
|
311 | 314 | |
|
312 | 315 | # Faster than using ctx.obsolete() |
|
313 | 316 | obsrevs = obsolete.getrevs(repo, b'obsolete') |
|
314 | 317 | |
|
315 | 318 | for branch, newheadrevs in newbranches.items(): |
|
316 | 319 | # For every branch, compute the new branchheads. |
|
317 | 320 | # A branchhead is a revision such that no descendant is on |
|
318 | 321 | # the same branch. |
|
319 | 322 | # |
|
320 | 323 | # The branchheads are computed iteratively in revision order. |
|
321 | 324 | # This ensures topological order, i.e. parents are processed |
|
322 | 325 | # before their children. Ancestors are inclusive here, i.e. |
|
323 | 326 | # any revision is an ancestor of itself. |
|
324 | 327 | # |
|
325 | 328 | # Core observations: |
|
326 | 329 | # - The current revision is always a branchhead for the |
|
327 | 330 | # repository up to that point. |
|
328 | 331 | # - It is the first revision of the branch if and only if |
|
329 | 332 | # there was no branchhead before. In that case, it is the |
|
330 | 333 | # only branchhead as there are no possible ancestors on |
|
331 | 334 | # the same branch. |
|
332 | 335 | # - If a parent is on the same branch, a branchhead can |
|
333 | 336 | # only be an ancestor of that parent, if it is parent |
|
334 | 337 | # itself. Otherwise it would have been removed as ancestor |
|
335 | 338 | # of that parent before. |
|
336 | 339 | # - Therefore, if all parents are on the same branch, they |
|
337 | 340 | # can just be removed from the branchhead set. |
|
338 | 341 | # - If one parent is on the same branch and the other is not |
|
339 | 342 | # and there was exactly one branchhead known, the existing |
|
340 | 343 | # branchhead can only be an ancestor if it is the parent. |
|
341 | 344 | # Otherwise it would have been removed as ancestor of |
|
342 | 345 | # the parent before. The other parent therefore can't have |
|
343 | 346 | # a branchhead as ancestor. |
|
344 | 347 | # - In all other cases, the parents on different branches |
|
345 | 348 | # could have a branchhead as ancestor. Those parents are |
|
346 | 349 | # kept in the "uncertain" set. If all branchheads are also |
|
347 | 350 | # topological heads, they can't have descendants and further |
|
348 | 351 | # checks can be skipped. Otherwise, the ancestors of the |
|
349 | 352 | # "uncertain" set are removed from branchheads. |
|
350 | 353 | # This computation is heavy and avoided if at all possible. |
|
351 | 354 | bheads = self._entries.get(branch, []) |
|
352 | 355 | bheadset = {cl.rev(node) for node in bheads} |
|
353 | 356 | uncertain = set() |
|
354 | 357 | for newrev in sorted(newheadrevs): |
|
355 | 358 | if newrev in obsrevs: |
|
356 | 359 | # We ignore obsolete changesets as they shouldn't be |
|
357 | 360 | # considered heads. |
|
358 | 361 | continue |
|
359 | 362 | |
|
360 | 363 | if not bheadset: |
|
361 | 364 | bheadset.add(newrev) |
|
362 | 365 | continue |
|
363 | 366 | |
|
364 | 367 | parents = [p for p in parentrevs(newrev) if p != nullrev] |
|
365 | 368 | samebranch = set() |
|
366 | 369 | otherbranch = set() |
|
367 | 370 | obsparents = set() |
|
368 | 371 | for p in parents: |
|
369 | 372 | if p in obsrevs: |
|
370 | 373 | # We ignored this obsolete changeset earlier, but now |
|
371 | 374 | # that it has non-ignored children, we need to make |
|
372 | 375 | # sure their ancestors are not considered heads. To |
|
373 | 376 | # achieve that, we will simply treat this obsolete |
|
374 | 377 | # changeset as a parent from other branch. |
|
375 | 378 | obsparents.add(p) |
|
376 | 379 | elif p in bheadset or getbranchinfo(p)[0] == branch: |
|
377 | 380 | samebranch.add(p) |
|
378 | 381 | else: |
|
379 | 382 | otherbranch.add(p) |
|
380 | 383 | if not (len(bheadset) == len(samebranch) == 1): |
|
381 | 384 | uncertain.update(otherbranch) |
|
382 | 385 | uncertain.update(obsparents) |
|
383 | 386 | bheadset.difference_update(samebranch) |
|
384 | 387 | bheadset.add(newrev) |
|
385 | 388 | |
|
386 | 389 | if uncertain: |
|
387 | 390 | if topoheads is None: |
|
388 | 391 | topoheads = set(cl.headrevs()) |
|
389 | 392 | if bheadset - topoheads: |
|
390 | 393 | floorrev = min(bheadset) |
|
391 | 394 | if floorrev <= max(uncertain): |
|
392 | 395 | ancestors = set(cl.ancestors(uncertain, floorrev)) |
|
393 | 396 | bheadset -= ancestors |
|
394 | 397 | if bheadset: |
|
395 | 398 | self[branch] = [cl.node(rev) for rev in sorted(bheadset)] |
|
396 | 399 | |
|
397 | 400 | duration = util.timer() - starttime |
|
398 | 401 | repo.ui.log( |
|
399 | 402 | b'branchcache', |
|
400 | 403 | b'updated %s in %.4f seconds\n', |
|
401 | 404 | _branchcachedesc(repo), |
|
402 | 405 | duration, |
|
403 | 406 | ) |
|
404 | 407 | return max_rev |
|
405 | 408 | |
|
406 | 409 | |
|
407 | 410 | class branchcache(_BaseBranchCache): |
|
408 | 411 | """Branchmap info for a local repo or repoview""" |
|
409 | 412 | |
|
410 | 413 | _base_filename = b"branch2" |
|
411 | 414 | |
|
412 | 415 | def __init__( |
|
413 | 416 | self, |
|
414 | 417 | repo: "localrepo.localrepository", |
|
415 | 418 | entries: Union[ |
|
416 | 419 | Dict[bytes, List[bytes]], Iterable[Tuple[bytes, List[bytes]]] |
|
417 | 420 | ] = (), |
|
418 | 421 | tipnode: Optional[bytes] = None, |
|
419 | 422 | tiprev: Optional[int] = nullrev, |
|
420 | 423 | filteredhash: Optional[bytes] = None, |
|
421 | 424 | closednodes: Optional[Set[bytes]] = None, |
|
422 | 425 | hasnode: Optional[Callable[[bytes], bool]] = None, |
|
423 | 426 | verify_node: bool = False, |
|
424 | 427 | ) -> None: |
|
425 | 428 | """hasnode is a function which can be used to verify whether changelog |
|
426 | 429 | has a given node or not. If it's not provided, we assume that every node |
|
427 | 430 | we have exists in changelog""" |
|
428 | 431 | self._filtername = repo.filtername |
|
429 | 432 | self._delayed = False |
|
430 | 433 | if tipnode is None: |
|
431 | 434 | self.tipnode = repo.nullid |
|
432 | 435 | else: |
|
433 | 436 | self.tipnode = tipnode |
|
434 | 437 | self.tiprev = tiprev |
|
435 | 438 | self.filteredhash = filteredhash |
|
436 | 439 | |
|
437 | 440 | super().__init__(repo=repo, entries=entries, closed_nodes=closednodes) |
|
438 | 441 | # closednodes is a set of nodes that close their branch. If the branch |
|
439 | 442 | # cache has been updated, it may contain nodes that are no longer |
|
440 | 443 | # heads. |
|
441 | 444 | |
|
442 | 445 | # Do we need to verify branch at all ? |
|
443 | 446 | self._verify_node = verify_node |
|
444 | 447 | # branches for which nodes are verified |
|
445 | 448 | self._verifiedbranches = set() |
|
446 | 449 | self._hasnode = None |
|
447 | 450 | if self._verify_node: |
|
448 | 451 | self._hasnode = repo.changelog.hasnode |
|
449 | 452 | |
|
450 | 453 | def validfor(self, repo): |
|
451 | 454 | """check that cache contents are valid for (a subset of) this repo |
|
452 | 455 | |
|
453 | 456 | - False when the order of changesets changed or if we detect a strip. |
|
454 | 457 | - True when cache is up-to-date for the current repo or its subset.""" |
|
455 | 458 | try: |
|
456 | 459 | node = repo.changelog.node(self.tiprev) |
|
457 | 460 | except IndexError: |
|
458 | 461 | # changesets were stripped and now we don't even have enough to |
|
459 | 462 | # find tiprev |
|
460 | 463 | return False |
|
461 | 464 | if self.tipnode != node: |
|
462 | 465 | # tiprev doesn't correspond to tipnode: repo was stripped, or this |
|
463 | 466 | # repo has a different order of changesets |
|
464 | 467 | return False |
|
465 | 468 | tiphash = scmutil.filteredhash(repo, self.tiprev, needobsolete=True) |
|
466 | 469 | # hashes don't match if this repo view has a different set of filtered |
|
467 | 470 | # revisions (e.g. due to phase changes) or obsolete revisions (e.g. |
|
468 | 471 | # history was rewritten) |
|
469 | 472 | return self.filteredhash == tiphash |
|
470 | 473 | |
|
471 | 474 | @classmethod |
|
472 | 475 | def fromfile(cls, repo): |
|
473 | 476 | f = None |
|
474 | 477 | try: |
|
475 | 478 | f = repo.cachevfs(cls._filename(repo)) |
|
476 | 479 | lineiter = iter(f) |
|
477 | 480 | init_kwargs = cls._load_header(repo, lineiter) |
|
478 | 481 | bcache = cls( |
|
479 | 482 | repo, |
|
480 | 483 | verify_node=True, |
|
481 | 484 | **init_kwargs, |
|
482 | 485 | ) |
|
483 | 486 | if not bcache.validfor(repo): |
|
484 | 487 | # invalidate the cache |
|
485 | 488 | raise ValueError('tip differs') |
|
486 | 489 | bcache._load_heads(repo, lineiter) |
|
487 | 490 | except (IOError, OSError): |
|
488 | 491 | return None |
|
489 | 492 | |
|
490 | 493 | except Exception as inst: |
|
491 | 494 | if repo.ui.debugflag: |
|
492 | 495 | msg = b'invalid %s: %s\n' |
|
493 | 496 | msg %= ( |
|
494 | 497 | _branchcachedesc(repo), |
|
495 | 498 | stringutil.forcebytestr(inst), |
|
496 | 499 | ) |
|
497 | 500 | repo.ui.debug(msg) |
|
498 | 501 | bcache = None |
|
499 | 502 | |
|
500 | 503 | finally: |
|
501 | 504 | if f: |
|
502 | 505 | f.close() |
|
503 | 506 | |
|
504 | 507 | return bcache |
|
505 | 508 | |
|
506 | 509 | @classmethod |
|
507 | 510 | def _load_header(cls, repo, lineiter) -> "dict[str, Any]": |
|
508 | 511 | """parse the head of a branchmap file |
|
509 | 512 | |
|
510 | 513 | return parameters to pass to a newly created class instance. |
|
511 | 514 | """ |
|
512 | 515 | cachekey = next(lineiter).rstrip(b'\n').split(b" ", 2) |
|
513 | 516 | last, lrev = cachekey[:2] |
|
514 | 517 | last, lrev = bin(last), int(lrev) |
|
515 | 518 | filteredhash = None |
|
516 | 519 | if len(cachekey) > 2: |
|
517 | 520 | filteredhash = bin(cachekey[2]) |
|
518 | 521 | return { |
|
519 | 522 | "tipnode": last, |
|
520 | 523 | "tiprev": lrev, |
|
521 | 524 | "filteredhash": filteredhash, |
|
522 | 525 | } |
|
523 | 526 | |
|
524 | 527 | def _load_heads(self, repo, lineiter): |
|
525 | 528 | """fully loads the branchcache by reading from the file using the line |
|
526 | 529 | iterator passed""" |
|
527 | 530 | for line in lineiter: |
|
528 | 531 | line = line.rstrip(b'\n') |
|
529 | 532 | if not line: |
|
530 | 533 | continue |
|
531 | 534 | node, state, label = line.split(b" ", 2) |
|
532 | 535 | if state not in b'oc': |
|
533 | 536 | raise ValueError('invalid branch state') |
|
534 | 537 | label = encoding.tolocal(label.strip()) |
|
535 | 538 | node = bin(node) |
|
536 | 539 | self._entries.setdefault(label, []).append(node) |
|
537 | 540 | if state == b'c': |
|
538 | 541 | self._closednodes.add(node) |
|
539 | 542 | |
|
540 | 543 | @classmethod |
|
541 | 544 | def _filename(cls, repo): |
|
542 | 545 | """name of a branchcache file for a given repo or repoview""" |
|
543 | 546 | filename = cls._base_filename |
|
544 | 547 | if repo.filtername: |
|
545 | 548 | filename = b'%s-%s' % (filename, repo.filtername) |
|
546 | 549 | return filename |
|
547 | 550 | |
|
548 | 551 | def copy(self, repo): |
|
549 | 552 | """return a deep copy of the branchcache object""" |
|
550 | 553 | other = type(self)( |
|
551 | 554 | repo=repo, |
|
552 | 555 | # we always do a shally copy of self._entries, and the values is |
|
553 | 556 | # always replaced, so no need to deepcopy until the above remains |
|
554 | 557 | # true. |
|
555 | 558 | entries=self._entries, |
|
556 | 559 | tipnode=self.tipnode, |
|
557 | 560 | tiprev=self.tiprev, |
|
558 | 561 | filteredhash=self.filteredhash, |
|
559 | 562 | closednodes=set(self._closednodes), |
|
560 | 563 | verify_node=self._verify_node, |
|
561 | 564 | ) |
|
562 | 565 | # we copy will likely schedule a write anyway, but that does not seems |
|
563 | 566 | # to hurt to overschedule |
|
564 | 567 | other._delayed = self._delayed |
|
565 | 568 | # also copy information about the current verification state |
|
566 | 569 | other._verifiedbranches = set(self._verifiedbranches) |
|
567 | 570 | return other |
|
568 | 571 | |
|
569 | 572 | def write(self, repo): |
|
570 | 573 | assert self._filtername == repo.filtername, ( |
|
571 | 574 | self._filtername, |
|
572 | 575 | repo.filtername, |
|
573 | 576 | ) |
|
574 | 577 | tr = repo.currenttransaction() |
|
575 | 578 | if not getattr(tr, 'finalized', True): |
|
576 | 579 | # Avoid premature writing. |
|
577 | 580 | # |
|
578 | 581 | # (The cache warming setup by localrepo will update the file later.) |
|
579 | 582 | self._delayed = True |
|
580 | 583 | return |
|
581 | 584 | try: |
|
582 | 585 | filename = self._filename(repo) |
|
583 | 586 | with repo.cachevfs(filename, b"w", atomictemp=True) as f: |
|
584 | 587 | self._write_header(f) |
|
585 | 588 | nodecount = self._write_heads(f) |
|
586 | 589 | repo.ui.log( |
|
587 | 590 | b'branchcache', |
|
588 | 591 | b'wrote %s with %d labels and %d nodes\n', |
|
589 | 592 | _branchcachedesc(repo), |
|
590 | 593 | len(self._entries), |
|
591 | 594 | nodecount, |
|
592 | 595 | ) |
|
593 | 596 | self._delayed = False |
|
594 | 597 | except (IOError, OSError, error.Abort) as inst: |
|
595 | 598 | # Abort may be raised by read only opener, so log and continue |
|
596 | 599 | repo.ui.debug( |
|
597 | 600 | b"couldn't write branch cache: %s\n" |
|
598 | 601 | % stringutil.forcebytestr(inst) |
|
599 | 602 | ) |
|
600 | 603 | |
|
601 | 604 | def _write_header(self, fp) -> None: |
|
602 | 605 | """write the branch cache header to a file""" |
|
603 | 606 | cachekey = [hex(self.tipnode), b'%d' % self.tiprev] |
|
604 | 607 | if self.filteredhash is not None: |
|
605 | 608 | cachekey.append(hex(self.filteredhash)) |
|
606 | 609 | fp.write(b" ".join(cachekey) + b'\n') |
|
607 | 610 | |
|
608 | 611 | def _write_heads(self, fp) -> int: |
|
609 | 612 | """write list of heads to a file |
|
610 | 613 | |
|
611 | 614 | Return the number of heads written.""" |
|
612 | 615 | nodecount = 0 |
|
613 | 616 | for label, nodes in sorted(self._entries.items()): |
|
614 | 617 | label = encoding.fromlocal(label) |
|
615 | 618 | for node in nodes: |
|
616 | 619 | nodecount += 1 |
|
617 | 620 | if node in self._closednodes: |
|
618 | 621 | state = b'c' |
|
619 | 622 | else: |
|
620 | 623 | state = b'o' |
|
621 | 624 | fp.write(b"%s %s %s\n" % (hex(node), state, label)) |
|
622 | 625 | return nodecount |
|
623 | 626 | |
|
624 | 627 | def _verifybranch(self, branch): |
|
625 | 628 | """verify head nodes for the given branch.""" |
|
626 | 629 | if not self._verify_node: |
|
627 | 630 | return |
|
628 | 631 | if branch not in self._entries or branch in self._verifiedbranches: |
|
629 | 632 | return |
|
630 | 633 | assert self._hasnode is not None |
|
631 | 634 | for n in self._entries[branch]: |
|
632 | 635 | if not self._hasnode(n): |
|
633 | 636 | _unknownnode(n) |
|
634 | 637 | |
|
635 | 638 | self._verifiedbranches.add(branch) |
|
636 | 639 | |
|
637 | 640 | def _verifyall(self): |
|
638 | 641 | """verifies nodes of all the branches""" |
|
639 | 642 | for b in self._entries.keys(): |
|
640 | 643 | if b not in self._verifiedbranches: |
|
641 | 644 | self._verifybranch(b) |
|
642 | 645 | |
|
643 | 646 | def __getitem__(self, key): |
|
644 | 647 | self._verifybranch(key) |
|
645 | 648 | return super().__getitem__(key) |
|
646 | 649 | |
|
647 | 650 | def __contains__(self, key): |
|
648 | 651 | self._verifybranch(key) |
|
649 | 652 | return super().__contains__(key) |
|
650 | 653 | |
|
651 | 654 | def iteritems(self): |
|
652 | 655 | self._verifyall() |
|
653 | 656 | return super().iteritems() |
|
654 | 657 | |
|
655 | 658 | items = iteritems |
|
656 | 659 | |
|
657 | 660 | def iterheads(self): |
|
658 | 661 | """returns all the heads""" |
|
659 | 662 | self._verifyall() |
|
660 | 663 | return super().iterheads() |
|
661 | 664 | |
|
662 | 665 | def hasbranch(self, label): |
|
663 | 666 | """checks whether a branch of this name exists or not""" |
|
664 | 667 | self._verifybranch(label) |
|
665 | 668 | return super().hasbranch(label) |
|
666 | 669 | |
|
667 | 670 | def branchheads(self, branch, closed=False): |
|
668 | 671 | self._verifybranch(branch) |
|
669 | 672 | return super().branchheads(branch, closed=closed) |
|
670 | 673 | |
|
671 | 674 | def update(self, repo, revgen): |
|
672 | 675 | assert self._filtername == repo.filtername, ( |
|
673 | 676 | self._filtername, |
|
674 | 677 | repo.filtername, |
|
675 | 678 | ) |
|
676 | 679 | cl = repo.changelog |
|
677 | 680 | max_rev = super().update(repo, revgen) |
|
678 | 681 | # new tip revision which we found after iterating items from new |
|
679 | 682 | # branches |
|
680 | 683 | if max_rev is not None and max_rev > self.tiprev: |
|
681 | 684 | self.tiprev = max_rev |
|
682 | 685 | self.tipnode = cl.node(max_rev) |
|
683 | 686 | |
|
684 | 687 | if not self.validfor(repo): |
|
685 | 688 | # old cache key is now invalid for the repo, but we've just updated |
|
686 | 689 | # the cache and we assume it's valid, so let's make the cache key |
|
687 | 690 | # valid as well by recomputing it from the cached data |
|
688 | 691 | self.tipnode = repo.nullid |
|
689 | 692 | self.tiprev = nullrev |
|
690 | 693 | for heads in self.iterheads(): |
|
691 | 694 | if not heads: |
|
692 | 695 | # all revisions on a branch are obsolete |
|
693 | 696 | continue |
|
694 | 697 | # note: tiprev is not necessarily the tip revision of repo, |
|
695 | 698 | # because the tip could be obsolete (i.e. not a head) |
|
696 | 699 | tiprev = max(cl.rev(node) for node in heads) |
|
697 | 700 | if tiprev > self.tiprev: |
|
698 | 701 | self.tipnode = cl.node(tiprev) |
|
699 | 702 | self.tiprev = tiprev |
|
700 | 703 | self.filteredhash = scmutil.filteredhash( |
|
701 | 704 | repo, self.tiprev, needobsolete=True |
|
702 | 705 | ) |
|
703 | 706 | |
|
704 | 707 | self.write(repo) |
|
705 | 708 | |
|
706 | 709 | |
|
707 | 710 | class remotebranchcache(_BaseBranchCache): |
|
708 | 711 | """Branchmap info for a remote connection, should not write locally""" |
|
709 | 712 | |
|
710 | 713 | def __init__( |
|
711 | 714 | self, |
|
712 | 715 | repo: "localrepo.localrepository", |
|
713 | 716 | entries: Union[ |
|
714 | 717 | Dict[bytes, List[bytes]], Iterable[Tuple[bytes, List[bytes]]] |
|
715 | 718 | ] = (), |
|
716 | 719 | closednodes: Optional[Set[bytes]] = None, |
|
717 | 720 | ) -> None: |
|
718 | 721 | super().__init__(repo=repo, entries=entries, closed_nodes=closednodes) |
|
719 | 722 | |
|
720 | 723 | |
|
721 | 724 | # Revision branch info cache |
|
722 | 725 | |
|
723 | 726 | _rbcversion = b'-v1' |
|
724 | 727 | _rbcnames = b'rbc-names' + _rbcversion |
|
725 | 728 | _rbcrevs = b'rbc-revs' + _rbcversion |
|
726 | 729 | # [4 byte hash prefix][4 byte branch name number with sign bit indicating open] |
|
727 | 730 | _rbcrecfmt = b'>4sI' |
|
728 | 731 | _rbcrecsize = calcsize(_rbcrecfmt) |
|
729 | 732 | _rbcmininc = 64 * _rbcrecsize |
|
730 | 733 | _rbcnodelen = 4 |
|
731 | 734 | _rbcbranchidxmask = 0x7FFFFFFF |
|
732 | 735 | _rbccloseflag = 0x80000000 |
|
733 | 736 | |
|
734 | 737 | |
|
735 | 738 | class rbcrevs: |
|
736 | 739 | """a byte string consisting of an immutable prefix followed by a mutable suffix""" |
|
737 | 740 | |
|
738 | 741 | def __init__(self, revs): |
|
739 | 742 | self._prefix = revs |
|
740 | 743 | self._rest = bytearray() |
|
741 | 744 | |
|
742 | 745 | def __len__(self): |
|
743 | 746 | return len(self._prefix) + len(self._rest) |
|
744 | 747 | |
|
745 | 748 | def unpack_record(self, rbcrevidx): |
|
746 | 749 | if rbcrevidx < len(self._prefix): |
|
747 | 750 | return unpack_from(_rbcrecfmt, util.buffer(self._prefix), rbcrevidx) |
|
748 | 751 | else: |
|
749 | 752 | return unpack_from( |
|
750 | 753 | _rbcrecfmt, |
|
751 | 754 | util.buffer(self._rest), |
|
752 | 755 | rbcrevidx - len(self._prefix), |
|
753 | 756 | ) |
|
754 | 757 | |
|
755 | 758 | def make_mutable(self): |
|
756 | 759 | if len(self._prefix) > 0: |
|
757 | 760 | entirety = bytearray() |
|
758 | 761 | entirety[:] = self._prefix |
|
759 | 762 | entirety.extend(self._rest) |
|
760 | 763 | self._rest = entirety |
|
761 | 764 | self._prefix = bytearray() |
|
762 | 765 | |
|
763 | 766 | def truncate(self, pos): |
|
764 | 767 | self.make_mutable() |
|
765 | 768 | del self._rest[pos:] |
|
766 | 769 | |
|
767 | 770 | def pack_into(self, rbcrevidx, node, branchidx): |
|
768 | 771 | if rbcrevidx < len(self._prefix): |
|
769 | 772 | self.make_mutable() |
|
770 | 773 | buf = self._rest |
|
771 | 774 | start_offset = rbcrevidx - len(self._prefix) |
|
772 | 775 | end_offset = start_offset + _rbcrecsize |
|
773 | 776 | |
|
774 | 777 | if len(self._rest) < end_offset: |
|
775 | 778 | # bytearray doesn't allocate extra space at least in Python 3.7. |
|
776 | 779 | # When multiple changesets are added in a row, precise resize would |
|
777 | 780 | # result in quadratic complexity. Overallocate to compensate by |
|
778 | 781 | # using the classic doubling technique for dynamic arrays instead. |
|
779 | 782 | # If there was a gap in the map before, less space will be reserved. |
|
780 | 783 | self._rest.extend(b'\0' * end_offset) |
|
781 | 784 | return pack_into( |
|
782 | 785 | _rbcrecfmt, |
|
783 | 786 | buf, |
|
784 | 787 | start_offset, |
|
785 | 788 | node, |
|
786 | 789 | branchidx, |
|
787 | 790 | ) |
|
788 | 791 | |
|
789 | 792 | def extend(self, extension): |
|
790 | 793 | return self._rest.extend(extension) |
|
791 | 794 | |
|
792 | 795 | def slice(self, begin, end): |
|
793 | 796 | if begin < len(self._prefix): |
|
794 | 797 | acc = bytearray() |
|
795 | 798 | acc[:] = self._prefix[begin:end] |
|
796 | 799 | acc.extend( |
|
797 | 800 | self._rest[begin - len(self._prefix) : end - len(self._prefix)] |
|
798 | 801 | ) |
|
799 | 802 | return acc |
|
800 | 803 | return self._rest[begin - len(self._prefix) : end - len(self._prefix)] |
|
801 | 804 | |
|
802 | 805 | |
|
803 | 806 | class revbranchcache: |
|
804 | 807 | """Persistent cache, mapping from revision number to branch name and close. |
|
805 | 808 | This is a low level cache, independent of filtering. |
|
806 | 809 | |
|
807 | 810 | Branch names are stored in rbc-names in internal encoding separated by 0. |
|
808 | 811 | rbc-names is append-only, and each branch name is only stored once and will |
|
809 | 812 | thus have a unique index. |
|
810 | 813 | |
|
811 | 814 | The branch info for each revision is stored in rbc-revs as constant size |
|
812 | 815 | records. The whole file is read into memory, but it is only 'parsed' on |
|
813 | 816 | demand. The file is usually append-only but will be truncated if repo |
|
814 | 817 | modification is detected. |
|
815 | 818 | The record for each revision contains the first 4 bytes of the |
|
816 | 819 | corresponding node hash, and the record is only used if it still matches. |
|
817 | 820 | Even a completely trashed rbc-revs fill thus still give the right result |
|
818 | 821 | while converging towards full recovery ... assuming no incorrectly matching |
|
819 | 822 | node hashes. |
|
820 | 823 | The record also contains 4 bytes where 31 bits contains the index of the |
|
821 | 824 | branch and the last bit indicate that it is a branch close commit. |
|
822 | 825 | The usage pattern for rbc-revs is thus somewhat similar to 00changelog.i |
|
823 | 826 | and will grow with it but be 1/8th of its size. |
|
824 | 827 | """ |
|
825 | 828 | |
|
826 | 829 | def __init__(self, repo, readonly=True): |
|
827 | 830 | assert repo.filtername is None |
|
828 | 831 | self._repo = repo |
|
829 | 832 | self._names = [] # branch names in local encoding with static index |
|
830 | 833 | self._rbcrevs = rbcrevs(bytearray()) |
|
831 | 834 | self._rbcsnameslen = 0 # length of names read at _rbcsnameslen |
|
832 | 835 | try: |
|
833 | 836 | bndata = repo.cachevfs.read(_rbcnames) |
|
834 | 837 | self._rbcsnameslen = len(bndata) # for verification before writing |
|
835 | 838 | if bndata: |
|
836 | 839 | self._names = [ |
|
837 | 840 | encoding.tolocal(bn) for bn in bndata.split(b'\0') |
|
838 | 841 | ] |
|
839 | 842 | except (IOError, OSError): |
|
840 | 843 | if readonly: |
|
841 | 844 | # don't try to use cache - fall back to the slow path |
|
842 | 845 | self.branchinfo = self._branchinfo |
|
843 | 846 | |
|
844 | 847 | if self._names: |
|
845 | 848 | try: |
|
846 | 849 | if repo.ui.configbool(b'format', b'mmap-revbranchcache'): |
|
847 | 850 | with repo.cachevfs(_rbcrevs) as fp: |
|
848 | 851 | data = util.buffer(util.mmapread(fp)) |
|
849 | 852 | else: |
|
850 | 853 | data = repo.cachevfs.read(_rbcrevs) |
|
851 | 854 | self._rbcrevs = rbcrevs(data) |
|
852 | 855 | except (IOError, OSError) as inst: |
|
853 | 856 | repo.ui.debug( |
|
854 | 857 | b"couldn't read revision branch cache: %s\n" |
|
855 | 858 | % stringutil.forcebytestr(inst) |
|
856 | 859 | ) |
|
857 | 860 | # remember number of good records on disk |
|
858 | 861 | self._rbcrevslen = min( |
|
859 | 862 | len(self._rbcrevs) // _rbcrecsize, len(repo.changelog) |
|
860 | 863 | ) |
|
861 | 864 | if self._rbcrevslen == 0: |
|
862 | 865 | self._names = [] |
|
863 | 866 | self._rbcnamescount = len(self._names) # number of names read at |
|
864 | 867 | # _rbcsnameslen |
|
865 | 868 | |
|
866 | 869 | def _clear(self): |
|
867 | 870 | self._rbcsnameslen = 0 |
|
868 | 871 | del self._names[:] |
|
869 | 872 | self._rbcnamescount = 0 |
|
870 | 873 | self._rbcrevslen = len(self._repo.changelog) |
|
871 | 874 | self._rbcrevs = rbcrevs(bytearray(self._rbcrevslen * _rbcrecsize)) |
|
872 | 875 | util.clearcachedproperty(self, b'_namesreverse') |
|
873 | 876 | |
|
874 | 877 | @util.propertycache |
|
875 | 878 | def _namesreverse(self): |
|
876 | 879 | return {b: r for r, b in enumerate(self._names)} |
|
877 | 880 | |
|
878 | 881 | def branchinfo(self, rev): |
|
879 | 882 | """Return branch name and close flag for rev, using and updating |
|
880 | 883 | persistent cache.""" |
|
881 | 884 | changelog = self._repo.changelog |
|
882 | 885 | rbcrevidx = rev * _rbcrecsize |
|
883 | 886 | |
|
884 | 887 | # avoid negative index, changelog.read(nullrev) is fast without cache |
|
885 | 888 | if rev == nullrev: |
|
886 | 889 | return changelog.branchinfo(rev) |
|
887 | 890 | |
|
888 | 891 | # if requested rev isn't allocated, grow and cache the rev info |
|
889 | 892 | if len(self._rbcrevs) < rbcrevidx + _rbcrecsize: |
|
890 | 893 | return self._branchinfo(rev) |
|
891 | 894 | |
|
892 | 895 | # fast path: extract data from cache, use it if node is matching |
|
893 | 896 | reponode = changelog.node(rev)[:_rbcnodelen] |
|
894 | 897 | cachenode, branchidx = self._rbcrevs.unpack_record(rbcrevidx) |
|
895 | 898 | close = bool(branchidx & _rbccloseflag) |
|
896 | 899 | if close: |
|
897 | 900 | branchidx &= _rbcbranchidxmask |
|
898 | 901 | if cachenode == b'\0\0\0\0': |
|
899 | 902 | pass |
|
900 | 903 | elif cachenode == reponode: |
|
901 | 904 | try: |
|
902 | 905 | return self._names[branchidx], close |
|
903 | 906 | except IndexError: |
|
904 | 907 | # recover from invalid reference to unknown branch |
|
905 | 908 | self._repo.ui.debug( |
|
906 | 909 | b"referenced branch names not found" |
|
907 | 910 | b" - rebuilding revision branch cache from scratch\n" |
|
908 | 911 | ) |
|
909 | 912 | self._clear() |
|
910 | 913 | else: |
|
911 | 914 | # rev/node map has changed, invalidate the cache from here up |
|
912 | 915 | self._repo.ui.debug( |
|
913 | 916 | b"history modification detected - truncating " |
|
914 | 917 | b"revision branch cache to revision %d\n" % rev |
|
915 | 918 | ) |
|
916 | 919 | truncate = rbcrevidx + _rbcrecsize |
|
917 | 920 | self._rbcrevs.truncate(truncate) |
|
918 | 921 | self._rbcrevslen = min(self._rbcrevslen, truncate) |
|
919 | 922 | |
|
920 | 923 | # fall back to slow path and make sure it will be written to disk |
|
921 | 924 | return self._branchinfo(rev) |
|
922 | 925 | |
|
923 | 926 | def _branchinfo(self, rev): |
|
924 | 927 | """Retrieve branch info from changelog and update _rbcrevs""" |
|
925 | 928 | changelog = self._repo.changelog |
|
926 | 929 | b, close = changelog.branchinfo(rev) |
|
927 | 930 | if b in self._namesreverse: |
|
928 | 931 | branchidx = self._namesreverse[b] |
|
929 | 932 | else: |
|
930 | 933 | branchidx = len(self._names) |
|
931 | 934 | self._names.append(b) |
|
932 | 935 | self._namesreverse[b] = branchidx |
|
933 | 936 | reponode = changelog.node(rev) |
|
934 | 937 | if close: |
|
935 | 938 | branchidx |= _rbccloseflag |
|
936 | 939 | self._setcachedata(rev, reponode, branchidx) |
|
937 | 940 | return b, close |
|
938 | 941 | |
|
939 | 942 | def setdata(self, rev, changelogrevision): |
|
940 | 943 | """add new data information to the cache""" |
|
941 | 944 | branch, close = changelogrevision.branchinfo |
|
942 | 945 | |
|
943 | 946 | if branch in self._namesreverse: |
|
944 | 947 | branchidx = self._namesreverse[branch] |
|
945 | 948 | else: |
|
946 | 949 | branchidx = len(self._names) |
|
947 | 950 | self._names.append(branch) |
|
948 | 951 | self._namesreverse[branch] = branchidx |
|
949 | 952 | if close: |
|
950 | 953 | branchidx |= _rbccloseflag |
|
951 | 954 | self._setcachedata(rev, self._repo.changelog.node(rev), branchidx) |
|
952 | 955 | # If no cache data were readable (non exists, bad permission, etc) |
|
953 | 956 | # the cache was bypassing itself by setting: |
|
954 | 957 | # |
|
955 | 958 | # self.branchinfo = self._branchinfo |
|
956 | 959 | # |
|
957 | 960 | # Since we now have data in the cache, we need to drop this bypassing. |
|
958 | 961 | if 'branchinfo' in vars(self): |
|
959 | 962 | del self.branchinfo |
|
960 | 963 | |
|
961 | 964 | def _setcachedata(self, rev, node, branchidx): |
|
962 | 965 | """Writes the node's branch data to the in-memory cache data.""" |
|
963 | 966 | if rev == nullrev: |
|
964 | 967 | return |
|
965 | 968 | rbcrevidx = rev * _rbcrecsize |
|
966 | 969 | self._rbcrevs.pack_into(rbcrevidx, node, branchidx) |
|
967 | 970 | self._rbcrevslen = min(self._rbcrevslen, rev) |
|
968 | 971 | |
|
969 | 972 | tr = self._repo.currenttransaction() |
|
970 | 973 | if tr: |
|
971 | 974 | tr.addfinalize(b'write-revbranchcache', self.write) |
|
972 | 975 | |
|
973 | 976 | def write(self, tr=None): |
|
974 | 977 | """Save branch cache if it is dirty.""" |
|
975 | 978 | repo = self._repo |
|
976 | 979 | wlock = None |
|
977 | 980 | step = b'' |
|
978 | 981 | try: |
|
979 | 982 | # write the new names |
|
980 | 983 | if self._rbcnamescount < len(self._names): |
|
981 | 984 | wlock = repo.wlock(wait=False) |
|
982 | 985 | step = b' names' |
|
983 | 986 | self._writenames(repo) |
|
984 | 987 | |
|
985 | 988 | # write the new revs |
|
986 | 989 | start = self._rbcrevslen * _rbcrecsize |
|
987 | 990 | if start != len(self._rbcrevs): |
|
988 | 991 | step = b'' |
|
989 | 992 | if wlock is None: |
|
990 | 993 | wlock = repo.wlock(wait=False) |
|
991 | 994 | self._writerevs(repo, start) |
|
992 | 995 | |
|
993 | 996 | except (IOError, OSError, error.Abort, error.LockError) as inst: |
|
994 | 997 | repo.ui.debug( |
|
995 | 998 | b"couldn't write revision branch cache%s: %s\n" |
|
996 | 999 | % (step, stringutil.forcebytestr(inst)) |
|
997 | 1000 | ) |
|
998 | 1001 | finally: |
|
999 | 1002 | if wlock is not None: |
|
1000 | 1003 | wlock.release() |
|
1001 | 1004 | |
|
1002 | 1005 | def _writenames(self, repo): |
|
1003 | 1006 | """write the new branch names to revbranchcache""" |
|
1004 | 1007 | if self._rbcnamescount != 0: |
|
1005 | 1008 | f = repo.cachevfs.open(_rbcnames, b'ab') |
|
1006 | 1009 | if f.tell() == self._rbcsnameslen: |
|
1007 | 1010 | f.write(b'\0') |
|
1008 | 1011 | else: |
|
1009 | 1012 | f.close() |
|
1010 | 1013 | repo.ui.debug(b"%s changed - rewriting it\n" % _rbcnames) |
|
1011 | 1014 | self._rbcnamescount = 0 |
|
1012 | 1015 | self._rbcrevslen = 0 |
|
1013 | 1016 | if self._rbcnamescount == 0: |
|
1014 | 1017 | # before rewriting names, make sure references are removed |
|
1015 | 1018 | repo.cachevfs.unlinkpath(_rbcrevs, ignoremissing=True) |
|
1016 | 1019 | f = repo.cachevfs.open(_rbcnames, b'wb') |
|
1017 | 1020 | f.write( |
|
1018 | 1021 | b'\0'.join( |
|
1019 | 1022 | encoding.fromlocal(b) |
|
1020 | 1023 | for b in self._names[self._rbcnamescount :] |
|
1021 | 1024 | ) |
|
1022 | 1025 | ) |
|
1023 | 1026 | self._rbcsnameslen = f.tell() |
|
1024 | 1027 | f.close() |
|
1025 | 1028 | self._rbcnamescount = len(self._names) |
|
1026 | 1029 | |
|
1027 | 1030 | def _writerevs(self, repo, start): |
|
1028 | 1031 | """write the new revs to revbranchcache""" |
|
1029 | 1032 | revs = min(len(repo.changelog), len(self._rbcrevs) // _rbcrecsize) |
|
1030 | 1033 | with repo.cachevfs.open(_rbcrevs, b'ab') as f: |
|
1031 | 1034 | if f.tell() != start: |
|
1032 | 1035 | repo.ui.debug( |
|
1033 | 1036 | b"truncating cache/%s to %d\n" % (_rbcrevs, start) |
|
1034 | 1037 | ) |
|
1035 | 1038 | f.seek(start) |
|
1036 | 1039 | if f.tell() != start: |
|
1037 | 1040 | start = 0 |
|
1038 | 1041 | f.seek(start) |
|
1039 | 1042 | f.truncate() |
|
1040 | 1043 | end = revs * _rbcrecsize |
|
1041 | 1044 | f.write(self._rbcrevs.slice(start, end)) |
|
1042 | 1045 | self._rbcrevslen = revs |
General Comments 0
You need to be logged in to leave comments.
Login now