Show More
@@ -1,525 +1,526 b'' | |||
|
1 | 1 | # branchmap.py - logic to computes, maintain and stores branchmap for local repo |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2005-2007 Matt Mackall <mpm@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import struct |
|
11 | 11 | |
|
12 | 12 | from .node import ( |
|
13 | 13 | bin, |
|
14 | 14 | hex, |
|
15 | 15 | nullid, |
|
16 | 16 | nullrev, |
|
17 | 17 | ) |
|
18 | 18 | from . import ( |
|
19 | 19 | encoding, |
|
20 | 20 | error, |
|
21 | 21 | pycompat, |
|
22 | 22 | scmutil, |
|
23 | 23 | util, |
|
24 | 24 | ) |
|
25 | 25 | |
|
26 | 26 | calcsize = struct.calcsize |
|
27 | 27 | pack_into = struct.pack_into |
|
28 | 28 | unpack_from = struct.unpack_from |
|
29 | 29 | |
|
30 | 30 | def _filename(repo): |
|
31 | 31 | """name of a branchcache file for a given repo or repoview""" |
|
32 | 32 | filename = "branch2" |
|
33 | 33 | if repo.filtername: |
|
34 | 34 | filename = '%s-%s' % (filename, repo.filtername) |
|
35 | 35 | return filename |
|
36 | 36 | |
|
37 | 37 | def read(repo): |
|
38 | 38 | try: |
|
39 | 39 | f = repo.cachevfs(_filename(repo)) |
|
40 | 40 | lines = f.read().split('\n') |
|
41 | 41 | f.close() |
|
42 | 42 | except (IOError, OSError): |
|
43 | 43 | return None |
|
44 | 44 | |
|
45 | 45 | try: |
|
46 | 46 | cachekey = lines.pop(0).split(" ", 2) |
|
47 | 47 | last, lrev = cachekey[:2] |
|
48 | 48 | last, lrev = bin(last), int(lrev) |
|
49 | 49 | filteredhash = None |
|
50 | 50 | if len(cachekey) > 2: |
|
51 | 51 | filteredhash = bin(cachekey[2]) |
|
52 | 52 | partial = branchcache(tipnode=last, tiprev=lrev, |
|
53 | 53 | filteredhash=filteredhash) |
|
54 | 54 | if not partial.validfor(repo): |
|
55 | 55 | # invalidate the cache |
|
56 | 56 | raise ValueError(r'tip differs') |
|
57 | 57 | cl = repo.changelog |
|
58 | 58 | for l in lines: |
|
59 | 59 | if not l: |
|
60 | 60 | continue |
|
61 | 61 | node, state, label = l.split(" ", 2) |
|
62 | 62 | if state not in 'oc': |
|
63 | 63 | raise ValueError(r'invalid branch state') |
|
64 | 64 | label = encoding.tolocal(label.strip()) |
|
65 | 65 | node = bin(node) |
|
66 | 66 | if not cl.hasnode(node): |
|
67 | 67 | raise ValueError( |
|
68 | 68 | r'node %s does not exist' % pycompat.sysstr(hex(node))) |
|
69 | 69 | partial.setdefault(label, []).append(node) |
|
70 | 70 | if state == 'c': |
|
71 | 71 | partial._closednodes.add(node) |
|
72 | 72 | except Exception as inst: |
|
73 | 73 | if repo.ui.debugflag: |
|
74 | 74 | msg = 'invalid branchheads cache' |
|
75 | 75 | if repo.filtername is not None: |
|
76 | 76 | msg += ' (%s)' % repo.filtername |
|
77 | 77 | msg += ': %s\n' |
|
78 | 78 | repo.ui.debug(msg % pycompat.bytestr(inst)) |
|
79 | 79 | partial = None |
|
80 | 80 | return partial |
|
81 | 81 | |
|
82 | 82 | ### Nearest subset relation |
|
83 | 83 | # Nearest subset of filter X is a filter Y so that: |
|
84 | 84 | # * Y is included in X, |
|
85 | 85 | # * X - Y is as small as possible. |
|
86 | 86 | # This create and ordering used for branchmap purpose. |
|
87 | 87 | # the ordering may be partial |
|
88 | 88 | subsettable = {None: 'visible', |
|
89 | 89 | 'visible-hidden': 'visible', |
|
90 | 90 | 'visible': 'served', |
|
91 | 91 | 'served': 'immutable', |
|
92 | 92 | 'immutable': 'base'} |
|
93 | 93 | |
|
94 | 94 | def updatecache(repo): |
|
95 | 95 | cl = repo.changelog |
|
96 | 96 | filtername = repo.filtername |
|
97 | 97 | partial = repo._branchcaches.get(filtername) |
|
98 | 98 | |
|
99 | 99 | revs = [] |
|
100 | 100 | if partial is None or not partial.validfor(repo): |
|
101 | 101 | partial = read(repo) |
|
102 | 102 | if partial is None: |
|
103 | 103 | subsetname = subsettable.get(filtername) |
|
104 | 104 | if subsetname is None: |
|
105 | 105 | partial = branchcache() |
|
106 | 106 | else: |
|
107 | 107 | subset = repo.filtered(subsetname) |
|
108 | 108 | partial = subset.branchmap().copy() |
|
109 | 109 | extrarevs = subset.changelog.filteredrevs - cl.filteredrevs |
|
110 | 110 | revs.extend(r for r in extrarevs if r <= partial.tiprev) |
|
111 | 111 | revs.extend(cl.revs(start=partial.tiprev + 1)) |
|
112 | 112 | if revs: |
|
113 | 113 | partial.update(repo, revs) |
|
114 | 114 | partial.write(repo) |
|
115 | 115 | |
|
116 | 116 | assert partial.validfor(repo), filtername |
|
117 | 117 | repo._branchcaches[repo.filtername] = partial |
|
118 | 118 | |
|
119 | 119 | def replacecache(repo, bm): |
|
120 | 120 | """Replace the branchmap cache for a repo with a branch mapping. |
|
121 | 121 | |
|
122 | 122 | This is likely only called during clone with a branch map from a remote. |
|
123 | 123 | """ |
|
124 | 124 | rbheads = [] |
|
125 | 125 | closed = [] |
|
126 | 126 | for bheads in bm.itervalues(): |
|
127 | 127 | rbheads.extend(bheads) |
|
128 | 128 | for h in bheads: |
|
129 | 129 | r = repo.changelog.rev(h) |
|
130 | 130 | b, c = repo.changelog.branchinfo(r) |
|
131 | 131 | if c: |
|
132 | 132 | closed.append(h) |
|
133 | 133 | |
|
134 | 134 | if rbheads: |
|
135 | 135 | rtiprev = max((int(repo.changelog.rev(node)) |
|
136 | 136 | for node in rbheads)) |
|
137 | 137 | cache = branchcache(bm, |
|
138 | 138 | repo[rtiprev].node(), |
|
139 | 139 | rtiprev, |
|
140 | 140 | closednodes=closed) |
|
141 | 141 | |
|
142 | 142 | # Try to stick it as low as possible |
|
143 | 143 | # filter above served are unlikely to be fetch from a clone |
|
144 | 144 | for candidate in ('base', 'immutable', 'served'): |
|
145 | 145 | rview = repo.filtered(candidate) |
|
146 | 146 | if cache.validfor(rview): |
|
147 | 147 | repo._branchcaches[candidate] = cache |
|
148 | 148 | cache.write(rview) |
|
149 | 149 | break |
|
150 | 150 | |
|
151 | 151 | class branchcache(dict): |
|
152 | 152 | """A dict like object that hold branches heads cache. |
|
153 | 153 | |
|
154 | 154 | This cache is used to avoid costly computations to determine all the |
|
155 | 155 | branch heads of a repo. |
|
156 | 156 | |
|
157 | 157 | The cache is serialized on disk in the following format: |
|
158 | 158 | |
|
159 | 159 | <tip hex node> <tip rev number> [optional filtered repo hex hash] |
|
160 | 160 | <branch head hex node> <open/closed state> <branch name> |
|
161 | 161 | <branch head hex node> <open/closed state> <branch name> |
|
162 | 162 | ... |
|
163 | 163 | |
|
164 | 164 | The first line is used to check if the cache is still valid. If the |
|
165 | 165 | branch cache is for a filtered repo view, an optional third hash is |
|
166 | 166 | included that hashes the hashes of all filtered revisions. |
|
167 | 167 | |
|
168 | 168 | The open/closed state is represented by a single letter 'o' or 'c'. |
|
169 | 169 | This field can be used to avoid changelog reads when determining if a |
|
170 | 170 | branch head closes a branch or not. |
|
171 | 171 | """ |
|
172 | 172 | |
|
173 | 173 | def __init__(self, entries=(), tipnode=nullid, tiprev=nullrev, |
|
174 | 174 | filteredhash=None, closednodes=None): |
|
175 | 175 | super(branchcache, self).__init__(entries) |
|
176 | 176 | self.tipnode = tipnode |
|
177 | 177 | self.tiprev = tiprev |
|
178 | 178 | self.filteredhash = filteredhash |
|
179 | 179 | # closednodes is a set of nodes that close their branch. If the branch |
|
180 | 180 | # cache has been updated, it may contain nodes that are no longer |
|
181 | 181 | # heads. |
|
182 | 182 | if closednodes is None: |
|
183 | 183 | self._closednodes = set() |
|
184 | 184 | else: |
|
185 | 185 | self._closednodes = closednodes |
|
186 | 186 | |
|
187 | 187 | def validfor(self, repo): |
|
188 | 188 | """Is the cache content valid regarding a repo |
|
189 | 189 | |
|
190 | 190 | - False when cached tipnode is unknown or if we detect a strip. |
|
191 | 191 | - True when cache is up to date or a subset of current repo.""" |
|
192 | 192 | try: |
|
193 | 193 | return ((self.tipnode == repo.changelog.node(self.tiprev)) |
|
194 | 194 | and (self.filteredhash == \ |
|
195 | 195 | scmutil.filteredhash(repo, self.tiprev))) |
|
196 | 196 | except IndexError: |
|
197 | 197 | return False |
|
198 | 198 | |
|
199 | 199 | def _branchtip(self, heads): |
|
200 | 200 | '''Return tuple with last open head in heads and false, |
|
201 | 201 | otherwise return last closed head and true.''' |
|
202 | 202 | tip = heads[-1] |
|
203 | 203 | closed = True |
|
204 | 204 | for h in reversed(heads): |
|
205 | 205 | if h not in self._closednodes: |
|
206 | 206 | tip = h |
|
207 | 207 | closed = False |
|
208 | 208 | break |
|
209 | 209 | return tip, closed |
|
210 | 210 | |
|
211 | 211 | def branchtip(self, branch): |
|
212 | 212 | '''Return the tipmost open head on branch head, otherwise return the |
|
213 | 213 | tipmost closed head on branch. |
|
214 | 214 | Raise KeyError for unknown branch.''' |
|
215 | 215 | return self._branchtip(self[branch])[0] |
|
216 | 216 | |
|
217 | 217 | def iteropen(self, nodes): |
|
218 | 218 | return (n for n in nodes if n not in self._closednodes) |
|
219 | 219 | |
|
220 | 220 | def branchheads(self, branch, closed=False): |
|
221 | 221 | heads = self[branch] |
|
222 | 222 | if not closed: |
|
223 | 223 | heads = list(self.iteropen(heads)) |
|
224 | 224 | return heads |
|
225 | 225 | |
|
226 | 226 | def iterbranches(self): |
|
227 | 227 | for bn, heads in self.iteritems(): |
|
228 | 228 | yield (bn, heads) + self._branchtip(heads) |
|
229 | 229 | |
|
230 | 230 | def copy(self): |
|
231 | 231 | """return an deep copy of the branchcache object""" |
|
232 | 232 | return branchcache(self, self.tipnode, self.tiprev, self.filteredhash, |
|
233 | 233 | self._closednodes) |
|
234 | 234 | |
|
235 | 235 | def write(self, repo): |
|
236 | 236 | try: |
|
237 | 237 | f = repo.cachevfs(_filename(repo), "w", atomictemp=True) |
|
238 | 238 | cachekey = [hex(self.tipnode), '%d' % self.tiprev] |
|
239 | 239 | if self.filteredhash is not None: |
|
240 | 240 | cachekey.append(hex(self.filteredhash)) |
|
241 | 241 | f.write(" ".join(cachekey) + '\n') |
|
242 | 242 | nodecount = 0 |
|
243 | 243 | for label, nodes in sorted(self.iteritems()): |
|
244 | 244 | for node in nodes: |
|
245 | 245 | nodecount += 1 |
|
246 | 246 | if node in self._closednodes: |
|
247 | 247 | state = 'c' |
|
248 | 248 | else: |
|
249 | 249 | state = 'o' |
|
250 | 250 | f.write("%s %s %s\n" % (hex(node), state, |
|
251 | 251 | encoding.fromlocal(label))) |
|
252 | 252 | f.close() |
|
253 | 253 | repo.ui.log('branchcache', |
|
254 | 254 | 'wrote %s branch cache with %d labels and %d nodes\n', |
|
255 | 255 | repo.filtername, len(self), nodecount) |
|
256 | 256 | except (IOError, OSError, error.Abort) as inst: |
|
257 | 257 | # Abort may be raised by read only opener, so log and continue |
|
258 |
repo.ui.debug("couldn't write branch cache: %s\n" % |
|
|
258 | repo.ui.debug("couldn't write branch cache: %s\n" % | |
|
259 | util.forcebytestr(inst)) | |
|
259 | 260 | |
|
260 | 261 | def update(self, repo, revgen): |
|
261 | 262 | """Given a branchhead cache, self, that may have extra nodes or be |
|
262 | 263 | missing heads, and a generator of nodes that are strictly a superset of |
|
263 | 264 | heads missing, this function updates self to be correct. |
|
264 | 265 | """ |
|
265 | 266 | starttime = util.timer() |
|
266 | 267 | cl = repo.changelog |
|
267 | 268 | # collect new branch entries |
|
268 | 269 | newbranches = {} |
|
269 | 270 | getbranchinfo = repo.revbranchcache().branchinfo |
|
270 | 271 | for r in revgen: |
|
271 | 272 | branch, closesbranch = getbranchinfo(r) |
|
272 | 273 | newbranches.setdefault(branch, []).append(r) |
|
273 | 274 | if closesbranch: |
|
274 | 275 | self._closednodes.add(cl.node(r)) |
|
275 | 276 | |
|
276 | 277 | # fetch current topological heads to speed up filtering |
|
277 | 278 | topoheads = set(cl.headrevs()) |
|
278 | 279 | |
|
279 | 280 | # if older branchheads are reachable from new ones, they aren't |
|
280 | 281 | # really branchheads. Note checking parents is insufficient: |
|
281 | 282 | # 1 (branch a) -> 2 (branch b) -> 3 (branch a) |
|
282 | 283 | for branch, newheadrevs in newbranches.iteritems(): |
|
283 | 284 | bheads = self.setdefault(branch, []) |
|
284 | 285 | bheadset = set(cl.rev(node) for node in bheads) |
|
285 | 286 | |
|
286 | 287 | # This have been tested True on all internal usage of this function. |
|
287 | 288 | # run it again in case of doubt |
|
288 | 289 | # assert not (set(bheadrevs) & set(newheadrevs)) |
|
289 | 290 | newheadrevs.sort() |
|
290 | 291 | bheadset.update(newheadrevs) |
|
291 | 292 | |
|
292 | 293 | # This prunes out two kinds of heads - heads that are superseded by |
|
293 | 294 | # a head in newheadrevs, and newheadrevs that are not heads because |
|
294 | 295 | # an existing head is their descendant. |
|
295 | 296 | uncertain = bheadset - topoheads |
|
296 | 297 | if uncertain: |
|
297 | 298 | floorrev = min(uncertain) |
|
298 | 299 | ancestors = set(cl.ancestors(newheadrevs, floorrev)) |
|
299 | 300 | bheadset -= ancestors |
|
300 | 301 | bheadrevs = sorted(bheadset) |
|
301 | 302 | self[branch] = [cl.node(rev) for rev in bheadrevs] |
|
302 | 303 | tiprev = bheadrevs[-1] |
|
303 | 304 | if tiprev > self.tiprev: |
|
304 | 305 | self.tipnode = cl.node(tiprev) |
|
305 | 306 | self.tiprev = tiprev |
|
306 | 307 | |
|
307 | 308 | if not self.validfor(repo): |
|
308 | 309 | # cache key are not valid anymore |
|
309 | 310 | self.tipnode = nullid |
|
310 | 311 | self.tiprev = nullrev |
|
311 | 312 | for heads in self.values(): |
|
312 | 313 | tiprev = max(cl.rev(node) for node in heads) |
|
313 | 314 | if tiprev > self.tiprev: |
|
314 | 315 | self.tipnode = cl.node(tiprev) |
|
315 | 316 | self.tiprev = tiprev |
|
316 | 317 | self.filteredhash = scmutil.filteredhash(repo, self.tiprev) |
|
317 | 318 | |
|
318 | 319 | duration = util.timer() - starttime |
|
319 | 320 | repo.ui.log('branchcache', 'updated %s branch cache in %.4f seconds\n', |
|
320 | 321 | repo.filtername, duration) |
|
321 | 322 | |
|
322 | 323 | # Revision branch info cache |
|
323 | 324 | |
|
324 | 325 | _rbcversion = '-v1' |
|
325 | 326 | _rbcnames = 'rbc-names' + _rbcversion |
|
326 | 327 | _rbcrevs = 'rbc-revs' + _rbcversion |
|
327 | 328 | # [4 byte hash prefix][4 byte branch name number with sign bit indicating open] |
|
328 | 329 | _rbcrecfmt = '>4sI' |
|
329 | 330 | _rbcrecsize = calcsize(_rbcrecfmt) |
|
330 | 331 | _rbcnodelen = 4 |
|
331 | 332 | _rbcbranchidxmask = 0x7fffffff |
|
332 | 333 | _rbccloseflag = 0x80000000 |
|
333 | 334 | |
|
334 | 335 | class revbranchcache(object): |
|
335 | 336 | """Persistent cache, mapping from revision number to branch name and close. |
|
336 | 337 | This is a low level cache, independent of filtering. |
|
337 | 338 | |
|
338 | 339 | Branch names are stored in rbc-names in internal encoding separated by 0. |
|
339 | 340 | rbc-names is append-only, and each branch name is only stored once and will |
|
340 | 341 | thus have a unique index. |
|
341 | 342 | |
|
342 | 343 | The branch info for each revision is stored in rbc-revs as constant size |
|
343 | 344 | records. The whole file is read into memory, but it is only 'parsed' on |
|
344 | 345 | demand. The file is usually append-only but will be truncated if repo |
|
345 | 346 | modification is detected. |
|
346 | 347 | The record for each revision contains the first 4 bytes of the |
|
347 | 348 | corresponding node hash, and the record is only used if it still matches. |
|
348 | 349 | Even a completely trashed rbc-revs fill thus still give the right result |
|
349 | 350 | while converging towards full recovery ... assuming no incorrectly matching |
|
350 | 351 | node hashes. |
|
351 | 352 | The record also contains 4 bytes where 31 bits contains the index of the |
|
352 | 353 | branch and the last bit indicate that it is a branch close commit. |
|
353 | 354 | The usage pattern for rbc-revs is thus somewhat similar to 00changelog.i |
|
354 | 355 | and will grow with it but be 1/8th of its size. |
|
355 | 356 | """ |
|
356 | 357 | |
|
357 | 358 | def __init__(self, repo, readonly=True): |
|
358 | 359 | assert repo.filtername is None |
|
359 | 360 | self._repo = repo |
|
360 | 361 | self._names = [] # branch names in local encoding with static index |
|
361 | 362 | self._rbcrevs = bytearray() |
|
362 | 363 | self._rbcsnameslen = 0 # length of names read at _rbcsnameslen |
|
363 | 364 | try: |
|
364 | 365 | bndata = repo.cachevfs.read(_rbcnames) |
|
365 | 366 | self._rbcsnameslen = len(bndata) # for verification before writing |
|
366 | 367 | if bndata: |
|
367 | 368 | self._names = [encoding.tolocal(bn) |
|
368 | 369 | for bn in bndata.split('\0')] |
|
369 | 370 | except (IOError, OSError): |
|
370 | 371 | if readonly: |
|
371 | 372 | # don't try to use cache - fall back to the slow path |
|
372 | 373 | self.branchinfo = self._branchinfo |
|
373 | 374 | |
|
374 | 375 | if self._names: |
|
375 | 376 | try: |
|
376 | 377 | data = repo.cachevfs.read(_rbcrevs) |
|
377 | 378 | self._rbcrevs[:] = data |
|
378 | 379 | except (IOError, OSError) as inst: |
|
379 | 380 | repo.ui.debug("couldn't read revision branch cache: %s\n" % |
|
380 | 381 | util.forcebytestr(inst)) |
|
381 | 382 | # remember number of good records on disk |
|
382 | 383 | self._rbcrevslen = min(len(self._rbcrevs) // _rbcrecsize, |
|
383 | 384 | len(repo.changelog)) |
|
384 | 385 | if self._rbcrevslen == 0: |
|
385 | 386 | self._names = [] |
|
386 | 387 | self._rbcnamescount = len(self._names) # number of names read at |
|
387 | 388 | # _rbcsnameslen |
|
388 | 389 | self._namesreverse = dict((b, r) for r, b in enumerate(self._names)) |
|
389 | 390 | |
|
390 | 391 | def _clear(self): |
|
391 | 392 | self._rbcsnameslen = 0 |
|
392 | 393 | del self._names[:] |
|
393 | 394 | self._rbcnamescount = 0 |
|
394 | 395 | self._namesreverse.clear() |
|
395 | 396 | self._rbcrevslen = len(self._repo.changelog) |
|
396 | 397 | self._rbcrevs = bytearray(self._rbcrevslen * _rbcrecsize) |
|
397 | 398 | |
|
398 | 399 | def branchinfo(self, rev): |
|
399 | 400 | """Return branch name and close flag for rev, using and updating |
|
400 | 401 | persistent cache.""" |
|
401 | 402 | changelog = self._repo.changelog |
|
402 | 403 | rbcrevidx = rev * _rbcrecsize |
|
403 | 404 | |
|
404 | 405 | # avoid negative index, changelog.read(nullrev) is fast without cache |
|
405 | 406 | if rev == nullrev: |
|
406 | 407 | return changelog.branchinfo(rev) |
|
407 | 408 | |
|
408 | 409 | # if requested rev isn't allocated, grow and cache the rev info |
|
409 | 410 | if len(self._rbcrevs) < rbcrevidx + _rbcrecsize: |
|
410 | 411 | return self._branchinfo(rev) |
|
411 | 412 | |
|
412 | 413 | # fast path: extract data from cache, use it if node is matching |
|
413 | 414 | reponode = changelog.node(rev)[:_rbcnodelen] |
|
414 | 415 | cachenode, branchidx = unpack_from( |
|
415 | 416 | _rbcrecfmt, util.buffer(self._rbcrevs), rbcrevidx) |
|
416 | 417 | close = bool(branchidx & _rbccloseflag) |
|
417 | 418 | if close: |
|
418 | 419 | branchidx &= _rbcbranchidxmask |
|
419 | 420 | if cachenode == '\0\0\0\0': |
|
420 | 421 | pass |
|
421 | 422 | elif cachenode == reponode: |
|
422 | 423 | try: |
|
423 | 424 | return self._names[branchidx], close |
|
424 | 425 | except IndexError: |
|
425 | 426 | # recover from invalid reference to unknown branch |
|
426 | 427 | self._repo.ui.debug("referenced branch names not found" |
|
427 | 428 | " - rebuilding revision branch cache from scratch\n") |
|
428 | 429 | self._clear() |
|
429 | 430 | else: |
|
430 | 431 | # rev/node map has changed, invalidate the cache from here up |
|
431 | 432 | self._repo.ui.debug("history modification detected - truncating " |
|
432 | 433 | "revision branch cache to revision %d\n" % rev) |
|
433 | 434 | truncate = rbcrevidx + _rbcrecsize |
|
434 | 435 | del self._rbcrevs[truncate:] |
|
435 | 436 | self._rbcrevslen = min(self._rbcrevslen, truncate) |
|
436 | 437 | |
|
437 | 438 | # fall back to slow path and make sure it will be written to disk |
|
438 | 439 | return self._branchinfo(rev) |
|
439 | 440 | |
|
440 | 441 | def _branchinfo(self, rev): |
|
441 | 442 | """Retrieve branch info from changelog and update _rbcrevs""" |
|
442 | 443 | changelog = self._repo.changelog |
|
443 | 444 | b, close = changelog.branchinfo(rev) |
|
444 | 445 | if b in self._namesreverse: |
|
445 | 446 | branchidx = self._namesreverse[b] |
|
446 | 447 | else: |
|
447 | 448 | branchidx = len(self._names) |
|
448 | 449 | self._names.append(b) |
|
449 | 450 | self._namesreverse[b] = branchidx |
|
450 | 451 | reponode = changelog.node(rev) |
|
451 | 452 | if close: |
|
452 | 453 | branchidx |= _rbccloseflag |
|
453 | 454 | self._setcachedata(rev, reponode, branchidx) |
|
454 | 455 | return b, close |
|
455 | 456 | |
|
456 | 457 | def _setcachedata(self, rev, node, branchidx): |
|
457 | 458 | """Writes the node's branch data to the in-memory cache data.""" |
|
458 | 459 | if rev == nullrev: |
|
459 | 460 | return |
|
460 | 461 | rbcrevidx = rev * _rbcrecsize |
|
461 | 462 | if len(self._rbcrevs) < rbcrevidx + _rbcrecsize: |
|
462 | 463 | self._rbcrevs.extend('\0' * |
|
463 | 464 | (len(self._repo.changelog) * _rbcrecsize - |
|
464 | 465 | len(self._rbcrevs))) |
|
465 | 466 | pack_into(_rbcrecfmt, self._rbcrevs, rbcrevidx, node, branchidx) |
|
466 | 467 | self._rbcrevslen = min(self._rbcrevslen, rev) |
|
467 | 468 | |
|
468 | 469 | tr = self._repo.currenttransaction() |
|
469 | 470 | if tr: |
|
470 | 471 | tr.addfinalize('write-revbranchcache', self.write) |
|
471 | 472 | |
|
472 | 473 | def write(self, tr=None): |
|
473 | 474 | """Save branch cache if it is dirty.""" |
|
474 | 475 | repo = self._repo |
|
475 | 476 | wlock = None |
|
476 | 477 | step = '' |
|
477 | 478 | try: |
|
478 | 479 | if self._rbcnamescount < len(self._names): |
|
479 | 480 | step = ' names' |
|
480 | 481 | wlock = repo.wlock(wait=False) |
|
481 | 482 | if self._rbcnamescount != 0: |
|
482 | 483 | f = repo.cachevfs.open(_rbcnames, 'ab') |
|
483 | 484 | if f.tell() == self._rbcsnameslen: |
|
484 | 485 | f.write('\0') |
|
485 | 486 | else: |
|
486 | 487 | f.close() |
|
487 | 488 | repo.ui.debug("%s changed - rewriting it\n" % _rbcnames) |
|
488 | 489 | self._rbcnamescount = 0 |
|
489 | 490 | self._rbcrevslen = 0 |
|
490 | 491 | if self._rbcnamescount == 0: |
|
491 | 492 | # before rewriting names, make sure references are removed |
|
492 | 493 | repo.cachevfs.unlinkpath(_rbcrevs, ignoremissing=True) |
|
493 | 494 | f = repo.cachevfs.open(_rbcnames, 'wb') |
|
494 | 495 | f.write('\0'.join(encoding.fromlocal(b) |
|
495 | 496 | for b in self._names[self._rbcnamescount:])) |
|
496 | 497 | self._rbcsnameslen = f.tell() |
|
497 | 498 | f.close() |
|
498 | 499 | self._rbcnamescount = len(self._names) |
|
499 | 500 | |
|
500 | 501 | start = self._rbcrevslen * _rbcrecsize |
|
501 | 502 | if start != len(self._rbcrevs): |
|
502 | 503 | step = '' |
|
503 | 504 | if wlock is None: |
|
504 | 505 | wlock = repo.wlock(wait=False) |
|
505 | 506 | revs = min(len(repo.changelog), |
|
506 | 507 | len(self._rbcrevs) // _rbcrecsize) |
|
507 | 508 | f = repo.cachevfs.open(_rbcrevs, 'ab') |
|
508 | 509 | if f.tell() != start: |
|
509 | 510 | repo.ui.debug("truncating cache/%s to %d\n" |
|
510 | 511 | % (_rbcrevs, start)) |
|
511 | 512 | f.seek(start) |
|
512 | 513 | if f.tell() != start: |
|
513 | 514 | start = 0 |
|
514 | 515 | f.seek(start) |
|
515 | 516 | f.truncate() |
|
516 | 517 | end = revs * _rbcrecsize |
|
517 | 518 | f.write(self._rbcrevs[start:end]) |
|
518 | 519 | f.close() |
|
519 | 520 | self._rbcrevslen = revs |
|
520 | 521 | except (IOError, OSError, error.Abort, error.LockError) as inst: |
|
521 | 522 | repo.ui.debug("couldn't write revision branch cache%s: %s\n" |
|
522 | 523 | % (step, inst)) |
|
523 | 524 | finally: |
|
524 | 525 | if wlock is not None: |
|
525 | 526 | wlock.release() |
@@ -1,788 +1,788 b'' | |||
|
1 | 1 | # tags.py - read tag info from local repository |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2009 Matt Mackall <mpm@selenic.com> |
|
4 | 4 | # Copyright 2009 Greg Ward <greg@gerg.ca> |
|
5 | 5 | # |
|
6 | 6 | # This software may be used and distributed according to the terms of the |
|
7 | 7 | # GNU General Public License version 2 or any later version. |
|
8 | 8 | |
|
9 | 9 | # Currently this module only deals with reading and caching tags. |
|
10 | 10 | # Eventually, it could take care of updating (adding/removing/moving) |
|
11 | 11 | # tags too. |
|
12 | 12 | |
|
13 | 13 | from __future__ import absolute_import |
|
14 | 14 | |
|
15 | 15 | import errno |
|
16 | 16 | |
|
17 | 17 | from .node import ( |
|
18 | 18 | bin, |
|
19 | 19 | hex, |
|
20 | 20 | nullid, |
|
21 | 21 | short, |
|
22 | 22 | ) |
|
23 | 23 | from .i18n import _ |
|
24 | 24 | from . import ( |
|
25 | 25 | encoding, |
|
26 | 26 | error, |
|
27 | 27 | match as matchmod, |
|
28 | 28 | scmutil, |
|
29 | 29 | util, |
|
30 | 30 | ) |
|
31 | 31 | |
|
32 | 32 | # Tags computation can be expensive and caches exist to make it fast in |
|
33 | 33 | # the common case. |
|
34 | 34 | # |
|
35 | 35 | # The "hgtagsfnodes1" cache file caches the .hgtags filenode values for |
|
36 | 36 | # each revision in the repository. The file is effectively an array of |
|
37 | 37 | # fixed length records. Read the docs for "hgtagsfnodescache" for technical |
|
38 | 38 | # details. |
|
39 | 39 | # |
|
40 | 40 | # The .hgtags filenode cache grows in proportion to the length of the |
|
41 | 41 | # changelog. The file is truncated when the # changelog is stripped. |
|
42 | 42 | # |
|
43 | 43 | # The purpose of the filenode cache is to avoid the most expensive part |
|
44 | 44 | # of finding global tags, which is looking up the .hgtags filenode in the |
|
45 | 45 | # manifest for each head. This can take dozens or over 100ms for |
|
46 | 46 | # repositories with very large manifests. Multiplied by dozens or even |
|
47 | 47 | # hundreds of heads and there is a significant performance concern. |
|
48 | 48 | # |
|
49 | 49 | # There also exist a separate cache file for each repository filter. |
|
50 | 50 | # These "tags-*" files store information about the history of tags. |
|
51 | 51 | # |
|
52 | 52 | # The tags cache files consists of a cache validation line followed by |
|
53 | 53 | # a history of tags. |
|
54 | 54 | # |
|
55 | 55 | # The cache validation line has the format: |
|
56 | 56 | # |
|
57 | 57 | # <tiprev> <tipnode> [<filteredhash>] |
|
58 | 58 | # |
|
59 | 59 | # <tiprev> is an integer revision and <tipnode> is a 40 character hex |
|
60 | 60 | # node for that changeset. These redundantly identify the repository |
|
61 | 61 | # tip from the time the cache was written. In addition, <filteredhash>, |
|
62 | 62 | # if present, is a 40 character hex hash of the contents of the filtered |
|
63 | 63 | # revisions for this filter. If the set of filtered revs changes, the |
|
64 | 64 | # hash will change and invalidate the cache. |
|
65 | 65 | # |
|
66 | 66 | # The history part of the tags cache consists of lines of the form: |
|
67 | 67 | # |
|
68 | 68 | # <node> <tag> |
|
69 | 69 | # |
|
70 | 70 | # (This format is identical to that of .hgtags files.) |
|
71 | 71 | # |
|
72 | 72 | # <tag> is the tag name and <node> is the 40 character hex changeset |
|
73 | 73 | # the tag is associated with. |
|
74 | 74 | # |
|
75 | 75 | # Tags are written sorted by tag name. |
|
76 | 76 | # |
|
77 | 77 | # Tags associated with multiple changesets have an entry for each changeset. |
|
78 | 78 | # The most recent changeset (in terms of revlog ordering for the head |
|
79 | 79 | # setting it) for each tag is last. |
|
80 | 80 | |
|
81 | 81 | def fnoderevs(ui, repo, revs): |
|
82 | 82 | """return the list of '.hgtags' fnodes used in a set revisions |
|
83 | 83 | |
|
84 | 84 | This is returned as list of unique fnodes. We use a list instead of a set |
|
85 | 85 | because order matters when it comes to tags.""" |
|
86 | 86 | unfi = repo.unfiltered() |
|
87 | 87 | tonode = unfi.changelog.node |
|
88 | 88 | nodes = [tonode(r) for r in revs] |
|
89 | 89 | fnodes = _getfnodes(ui, repo, nodes[::-1]) # reversed help the cache |
|
90 | 90 | fnodes = _filterfnodes(fnodes, nodes) |
|
91 | 91 | return fnodes |
|
92 | 92 | |
|
93 | 93 | def _nulltonone(value): |
|
94 | 94 | """convert nullid to None |
|
95 | 95 | |
|
96 | 96 | For tag value, nullid means "deleted". This small utility function helps |
|
97 | 97 | translating that to None.""" |
|
98 | 98 | if value == nullid: |
|
99 | 99 | return None |
|
100 | 100 | return value |
|
101 | 101 | |
|
102 | 102 | def difftags(ui, repo, oldfnodes, newfnodes): |
|
103 | 103 | """list differences between tags expressed in two set of file-nodes |
|
104 | 104 | |
|
105 | 105 | The list contains entries in the form: (tagname, oldvalue, new value). |
|
106 | 106 | None is used to expressed missing value: |
|
107 | 107 | ('foo', None, 'abcd') is a new tag, |
|
108 | 108 | ('bar', 'ef01', None) is a deletion, |
|
109 | 109 | ('baz', 'abcd', 'ef01') is a tag movement. |
|
110 | 110 | """ |
|
111 | 111 | if oldfnodes == newfnodes: |
|
112 | 112 | return [] |
|
113 | 113 | oldtags = _tagsfromfnodes(ui, repo, oldfnodes) |
|
114 | 114 | newtags = _tagsfromfnodes(ui, repo, newfnodes) |
|
115 | 115 | |
|
116 | 116 | # list of (tag, old, new): None means missing |
|
117 | 117 | entries = [] |
|
118 | 118 | for tag, (new, __) in newtags.items(): |
|
119 | 119 | new = _nulltonone(new) |
|
120 | 120 | old, __ = oldtags.pop(tag, (None, None)) |
|
121 | 121 | old = _nulltonone(old) |
|
122 | 122 | if old != new: |
|
123 | 123 | entries.append((tag, old, new)) |
|
124 | 124 | # handle deleted tags |
|
125 | 125 | for tag, (old, __) in oldtags.items(): |
|
126 | 126 | old = _nulltonone(old) |
|
127 | 127 | if old is not None: |
|
128 | 128 | entries.append((tag, old, None)) |
|
129 | 129 | entries.sort() |
|
130 | 130 | return entries |
|
131 | 131 | |
|
132 | 132 | def writediff(fp, difflist): |
|
133 | 133 | """write tags diff information to a file. |
|
134 | 134 | |
|
135 | 135 | Data are stored with a line based format: |
|
136 | 136 | |
|
137 | 137 | <action> <hex-node> <tag-name>\n |
|
138 | 138 | |
|
139 | 139 | Action are defined as follow: |
|
140 | 140 | -R tag is removed, |
|
141 | 141 | +A tag is added, |
|
142 | 142 | -M tag is moved (old value), |
|
143 | 143 | +M tag is moved (new value), |
|
144 | 144 | |
|
145 | 145 | Example: |
|
146 | 146 | |
|
147 | 147 | +A 875517b4806a848f942811a315a5bce30804ae85 t5 |
|
148 | 148 | |
|
149 | 149 | See documentation of difftags output for details about the input. |
|
150 | 150 | """ |
|
151 | 151 | add = '+A %s %s\n' |
|
152 | 152 | remove = '-R %s %s\n' |
|
153 | 153 | updateold = '-M %s %s\n' |
|
154 | 154 | updatenew = '+M %s %s\n' |
|
155 | 155 | for tag, old, new in difflist: |
|
156 | 156 | # translate to hex |
|
157 | 157 | if old is not None: |
|
158 | 158 | old = hex(old) |
|
159 | 159 | if new is not None: |
|
160 | 160 | new = hex(new) |
|
161 | 161 | # write to file |
|
162 | 162 | if old is None: |
|
163 | 163 | fp.write(add % (new, tag)) |
|
164 | 164 | elif new is None: |
|
165 | 165 | fp.write(remove % (old, tag)) |
|
166 | 166 | else: |
|
167 | 167 | fp.write(updateold % (old, tag)) |
|
168 | 168 | fp.write(updatenew % (new, tag)) |
|
169 | 169 | |
|
170 | 170 | def findglobaltags(ui, repo): |
|
171 | 171 | '''Find global tags in a repo: return a tagsmap |
|
172 | 172 | |
|
173 | 173 | tagsmap: tag name to (node, hist) 2-tuples. |
|
174 | 174 | |
|
175 | 175 | The tags cache is read and updated as a side-effect of calling. |
|
176 | 176 | ''' |
|
177 | 177 | (heads, tagfnode, valid, cachetags, shouldwrite) = _readtagcache(ui, repo) |
|
178 | 178 | if cachetags is not None: |
|
179 | 179 | assert not shouldwrite |
|
180 | 180 | # XXX is this really 100% correct? are there oddball special |
|
181 | 181 | # cases where a global tag should outrank a local tag but won't, |
|
182 | 182 | # because cachetags does not contain rank info? |
|
183 | 183 | alltags = {} |
|
184 | 184 | _updatetags(cachetags, alltags) |
|
185 | 185 | return alltags |
|
186 | 186 | |
|
187 | 187 | for head in reversed(heads): # oldest to newest |
|
188 | 188 | assert head in repo.changelog.nodemap, \ |
|
189 | 189 | "tag cache returned bogus head %s" % short(head) |
|
190 | 190 | fnodes = _filterfnodes(tagfnode, reversed(heads)) |
|
191 | 191 | alltags = _tagsfromfnodes(ui, repo, fnodes) |
|
192 | 192 | |
|
193 | 193 | # and update the cache (if necessary) |
|
194 | 194 | if shouldwrite: |
|
195 | 195 | _writetagcache(ui, repo, valid, alltags) |
|
196 | 196 | return alltags |
|
197 | 197 | |
|
198 | 198 | def _filterfnodes(tagfnode, nodes): |
|
199 | 199 | """return a list of unique fnodes |
|
200 | 200 | |
|
201 | 201 | The order of this list matches the order of "nodes". Preserving this order |
|
202 | 202 | is important as reading tags in different order provides different |
|
203 | 203 | results.""" |
|
204 | 204 | seen = set() # set of fnode |
|
205 | 205 | fnodes = [] |
|
206 | 206 | for no in nodes: # oldest to newest |
|
207 | 207 | fnode = tagfnode.get(no) |
|
208 | 208 | if fnode and fnode not in seen: |
|
209 | 209 | seen.add(fnode) |
|
210 | 210 | fnodes.append(fnode) |
|
211 | 211 | return fnodes |
|
212 | 212 | |
|
213 | 213 | def _tagsfromfnodes(ui, repo, fnodes): |
|
214 | 214 | """return a tagsmap from a list of file-node |
|
215 | 215 | |
|
216 | 216 | tagsmap: tag name to (node, hist) 2-tuples. |
|
217 | 217 | |
|
218 | 218 | The order of the list matters.""" |
|
219 | 219 | alltags = {} |
|
220 | 220 | fctx = None |
|
221 | 221 | for fnode in fnodes: |
|
222 | 222 | if fctx is None: |
|
223 | 223 | fctx = repo.filectx('.hgtags', fileid=fnode) |
|
224 | 224 | else: |
|
225 | 225 | fctx = fctx.filectx(fnode) |
|
226 | 226 | filetags = _readtags(ui, repo, fctx.data().splitlines(), fctx) |
|
227 | 227 | _updatetags(filetags, alltags) |
|
228 | 228 | return alltags |
|
229 | 229 | |
|
230 | 230 | def readlocaltags(ui, repo, alltags, tagtypes): |
|
231 | 231 | '''Read local tags in repo. Update alltags and tagtypes.''' |
|
232 | 232 | try: |
|
233 | 233 | data = repo.vfs.read("localtags") |
|
234 | 234 | except IOError as inst: |
|
235 | 235 | if inst.errno != errno.ENOENT: |
|
236 | 236 | raise |
|
237 | 237 | return |
|
238 | 238 | |
|
239 | 239 | # localtags is in the local encoding; re-encode to UTF-8 on |
|
240 | 240 | # input for consistency with the rest of this module. |
|
241 | 241 | filetags = _readtags( |
|
242 | 242 | ui, repo, data.splitlines(), "localtags", |
|
243 | 243 | recode=encoding.fromlocal) |
|
244 | 244 | |
|
245 | 245 | # remove tags pointing to invalid nodes |
|
246 | 246 | cl = repo.changelog |
|
247 | 247 | for t in list(filetags): |
|
248 | 248 | try: |
|
249 | 249 | cl.rev(filetags[t][0]) |
|
250 | 250 | except (LookupError, ValueError): |
|
251 | 251 | del filetags[t] |
|
252 | 252 | |
|
253 | 253 | _updatetags(filetags, alltags, 'local', tagtypes) |
|
254 | 254 | |
|
255 | 255 | def _readtaghist(ui, repo, lines, fn, recode=None, calcnodelines=False): |
|
256 | 256 | '''Read tag definitions from a file (or any source of lines). |
|
257 | 257 | |
|
258 | 258 | This function returns two sortdicts with similar information: |
|
259 | 259 | |
|
260 | 260 | - the first dict, bintaghist, contains the tag information as expected by |
|
261 | 261 | the _readtags function, i.e. a mapping from tag name to (node, hist): |
|
262 | 262 | - node is the node id from the last line read for that name, |
|
263 | 263 | - hist is the list of node ids previously associated with it (in file |
|
264 | 264 | order). All node ids are binary, not hex. |
|
265 | 265 | |
|
266 | 266 | - the second dict, hextaglines, is a mapping from tag name to a list of |
|
267 | 267 | [hexnode, line number] pairs, ordered from the oldest to the newest node. |
|
268 | 268 | |
|
269 | 269 | When calcnodelines is False the hextaglines dict is not calculated (an |
|
270 | 270 | empty dict is returned). This is done to improve this function's |
|
271 | 271 | performance in cases where the line numbers are not needed. |
|
272 | 272 | ''' |
|
273 | 273 | |
|
274 | 274 | bintaghist = util.sortdict() |
|
275 | 275 | hextaglines = util.sortdict() |
|
276 | 276 | count = 0 |
|
277 | 277 | |
|
278 | 278 | def dbg(msg): |
|
279 | 279 | ui.debug("%s, line %s: %s\n" % (fn, count, msg)) |
|
280 | 280 | |
|
281 | 281 | for nline, line in enumerate(lines): |
|
282 | 282 | count += 1 |
|
283 | 283 | if not line: |
|
284 | 284 | continue |
|
285 | 285 | try: |
|
286 | 286 | (nodehex, name) = line.split(" ", 1) |
|
287 | 287 | except ValueError: |
|
288 | 288 | dbg("cannot parse entry") |
|
289 | 289 | continue |
|
290 | 290 | name = name.strip() |
|
291 | 291 | if recode: |
|
292 | 292 | name = recode(name) |
|
293 | 293 | try: |
|
294 | 294 | nodebin = bin(nodehex) |
|
295 | 295 | except TypeError: |
|
296 | 296 | dbg("node '%s' is not well formed" % nodehex) |
|
297 | 297 | continue |
|
298 | 298 | |
|
299 | 299 | # update filetags |
|
300 | 300 | if calcnodelines: |
|
301 | 301 | # map tag name to a list of line numbers |
|
302 | 302 | if name not in hextaglines: |
|
303 | 303 | hextaglines[name] = [] |
|
304 | 304 | hextaglines[name].append([nodehex, nline]) |
|
305 | 305 | continue |
|
306 | 306 | # map tag name to (node, hist) |
|
307 | 307 | if name not in bintaghist: |
|
308 | 308 | bintaghist[name] = [] |
|
309 | 309 | bintaghist[name].append(nodebin) |
|
310 | 310 | return bintaghist, hextaglines |
|
311 | 311 | |
|
312 | 312 | def _readtags(ui, repo, lines, fn, recode=None, calcnodelines=False): |
|
313 | 313 | '''Read tag definitions from a file (or any source of lines). |
|
314 | 314 | |
|
315 | 315 | Returns a mapping from tag name to (node, hist). |
|
316 | 316 | |
|
317 | 317 | "node" is the node id from the last line read for that name. "hist" |
|
318 | 318 | is the list of node ids previously associated with it (in file order). |
|
319 | 319 | All node ids are binary, not hex. |
|
320 | 320 | ''' |
|
321 | 321 | filetags, nodelines = _readtaghist(ui, repo, lines, fn, recode=recode, |
|
322 | 322 | calcnodelines=calcnodelines) |
|
323 | 323 | # util.sortdict().__setitem__ is much slower at replacing then inserting |
|
324 | 324 | # new entries. The difference can matter if there are thousands of tags. |
|
325 | 325 | # Create a new sortdict to avoid the performance penalty. |
|
326 | 326 | newtags = util.sortdict() |
|
327 | 327 | for tag, taghist in filetags.items(): |
|
328 | 328 | newtags[tag] = (taghist[-1], taghist[:-1]) |
|
329 | 329 | return newtags |
|
330 | 330 | |
|
331 | 331 | def _updatetags(filetags, alltags, tagtype=None, tagtypes=None): |
|
332 | 332 | """Incorporate the tag info read from one file into dictionnaries |
|
333 | 333 | |
|
334 | 334 | The first one, 'alltags', is a "tagmaps" (see 'findglobaltags' for details). |
|
335 | 335 | |
|
336 | 336 | The second one, 'tagtypes', is optional and will be updated to track the |
|
337 | 337 | "tagtype" of entries in the tagmaps. When set, the 'tagtype' argument also |
|
338 | 338 | needs to be set.""" |
|
339 | 339 | if tagtype is None: |
|
340 | 340 | assert tagtypes is None |
|
341 | 341 | |
|
342 | 342 | for name, nodehist in filetags.iteritems(): |
|
343 | 343 | if name not in alltags: |
|
344 | 344 | alltags[name] = nodehist |
|
345 | 345 | if tagtype is not None: |
|
346 | 346 | tagtypes[name] = tagtype |
|
347 | 347 | continue |
|
348 | 348 | |
|
349 | 349 | # we prefer alltags[name] if: |
|
350 | 350 | # it supersedes us OR |
|
351 | 351 | # mutual supersedes and it has a higher rank |
|
352 | 352 | # otherwise we win because we're tip-most |
|
353 | 353 | anode, ahist = nodehist |
|
354 | 354 | bnode, bhist = alltags[name] |
|
355 | 355 | if (bnode != anode and anode in bhist and |
|
356 | 356 | (bnode not in ahist or len(bhist) > len(ahist))): |
|
357 | 357 | anode = bnode |
|
358 | 358 | elif tagtype is not None: |
|
359 | 359 | tagtypes[name] = tagtype |
|
360 | 360 | ahist.extend([n for n in bhist if n not in ahist]) |
|
361 | 361 | alltags[name] = anode, ahist |
|
362 | 362 | |
|
363 | 363 | def _filename(repo): |
|
364 | 364 | """name of a tagcache file for a given repo or repoview""" |
|
365 | 365 | filename = 'tags2' |
|
366 | 366 | if repo.filtername: |
|
367 | 367 | filename = '%s-%s' % (filename, repo.filtername) |
|
368 | 368 | return filename |
|
369 | 369 | |
|
370 | 370 | def _readtagcache(ui, repo): |
|
371 | 371 | '''Read the tag cache. |
|
372 | 372 | |
|
373 | 373 | Returns a tuple (heads, fnodes, validinfo, cachetags, shouldwrite). |
|
374 | 374 | |
|
375 | 375 | If the cache is completely up-to-date, "cachetags" is a dict of the |
|
376 | 376 | form returned by _readtags() and "heads", "fnodes", and "validinfo" are |
|
377 | 377 | None and "shouldwrite" is False. |
|
378 | 378 | |
|
379 | 379 | If the cache is not up to date, "cachetags" is None. "heads" is a list |
|
380 | 380 | of all heads currently in the repository, ordered from tip to oldest. |
|
381 | 381 | "validinfo" is a tuple describing cache validation info. This is used |
|
382 | 382 | when writing the tags cache. "fnodes" is a mapping from head to .hgtags |
|
383 | 383 | filenode. "shouldwrite" is True. |
|
384 | 384 | |
|
385 | 385 | If the cache is not up to date, the caller is responsible for reading tag |
|
386 | 386 | info from each returned head. (See findglobaltags().) |
|
387 | 387 | ''' |
|
388 | 388 | try: |
|
389 | 389 | cachefile = repo.cachevfs(_filename(repo), 'r') |
|
390 | 390 | # force reading the file for static-http |
|
391 | 391 | cachelines = iter(cachefile) |
|
392 | 392 | except IOError: |
|
393 | 393 | cachefile = None |
|
394 | 394 | |
|
395 | 395 | cacherev = None |
|
396 | 396 | cachenode = None |
|
397 | 397 | cachehash = None |
|
398 | 398 | if cachefile: |
|
399 | 399 | try: |
|
400 | 400 | validline = next(cachelines) |
|
401 | 401 | validline = validline.split() |
|
402 | 402 | cacherev = int(validline[0]) |
|
403 | 403 | cachenode = bin(validline[1]) |
|
404 | 404 | if len(validline) > 2: |
|
405 | 405 | cachehash = bin(validline[2]) |
|
406 | 406 | except Exception: |
|
407 | 407 | # corruption of the cache, just recompute it. |
|
408 | 408 | pass |
|
409 | 409 | |
|
410 | 410 | tipnode = repo.changelog.tip() |
|
411 | 411 | tiprev = len(repo.changelog) - 1 |
|
412 | 412 | |
|
413 | 413 | # Case 1 (common): tip is the same, so nothing has changed. |
|
414 | 414 | # (Unchanged tip trivially means no changesets have been added. |
|
415 | 415 | # But, thanks to localrepository.destroyed(), it also means none |
|
416 | 416 | # have been destroyed by strip or rollback.) |
|
417 | 417 | if (cacherev == tiprev |
|
418 | 418 | and cachenode == tipnode |
|
419 | 419 | and cachehash == scmutil.filteredhash(repo, tiprev)): |
|
420 | 420 | tags = _readtags(ui, repo, cachelines, cachefile.name) |
|
421 | 421 | cachefile.close() |
|
422 | 422 | return (None, None, None, tags, False) |
|
423 | 423 | if cachefile: |
|
424 | 424 | cachefile.close() # ignore rest of file |
|
425 | 425 | |
|
426 | 426 | valid = (tiprev, tipnode, scmutil.filteredhash(repo, tiprev)) |
|
427 | 427 | |
|
428 | 428 | repoheads = repo.heads() |
|
429 | 429 | # Case 2 (uncommon): empty repo; get out quickly and don't bother |
|
430 | 430 | # writing an empty cache. |
|
431 | 431 | if repoheads == [nullid]: |
|
432 | 432 | return ([], {}, valid, {}, False) |
|
433 | 433 | |
|
434 | 434 | # Case 3 (uncommon): cache file missing or empty. |
|
435 | 435 | |
|
436 | 436 | # Case 4 (uncommon): tip rev decreased. This should only happen |
|
437 | 437 | # when we're called from localrepository.destroyed(). Refresh the |
|
438 | 438 | # cache so future invocations will not see disappeared heads in the |
|
439 | 439 | # cache. |
|
440 | 440 | |
|
441 | 441 | # Case 5 (common): tip has changed, so we've added/replaced heads. |
|
442 | 442 | |
|
443 | 443 | # As it happens, the code to handle cases 3, 4, 5 is the same. |
|
444 | 444 | |
|
445 | 445 | # N.B. in case 4 (nodes destroyed), "new head" really means "newly |
|
446 | 446 | # exposed". |
|
447 | 447 | if not len(repo.file('.hgtags')): |
|
448 | 448 | # No tags have ever been committed, so we can avoid a |
|
449 | 449 | # potentially expensive search. |
|
450 | 450 | return ([], {}, valid, None, True) |
|
451 | 451 | |
|
452 | 452 | |
|
453 | 453 | # Now we have to lookup the .hgtags filenode for every new head. |
|
454 | 454 | # This is the most expensive part of finding tags, so performance |
|
455 | 455 | # depends primarily on the size of newheads. Worst case: no cache |
|
456 | 456 | # file, so newheads == repoheads. |
|
457 | 457 | cachefnode = _getfnodes(ui, repo, repoheads) |
|
458 | 458 | |
|
459 | 459 | # Caller has to iterate over all heads, but can use the filenodes in |
|
460 | 460 | # cachefnode to get to each .hgtags revision quickly. |
|
461 | 461 | return (repoheads, cachefnode, valid, None, True) |
|
462 | 462 | |
|
463 | 463 | def _getfnodes(ui, repo, nodes): |
|
464 | 464 | """return .hgtags fnodes for a list of changeset nodes |
|
465 | 465 | |
|
466 | 466 | Return value is a {node: fnode} mapping. There will be no entry for nodes |
|
467 | 467 | without a '.hgtags' file. |
|
468 | 468 | """ |
|
469 | 469 | starttime = util.timer() |
|
470 | 470 | fnodescache = hgtagsfnodescache(repo.unfiltered()) |
|
471 | 471 | cachefnode = {} |
|
472 | 472 | for node in reversed(nodes): |
|
473 | 473 | fnode = fnodescache.getfnode(node) |
|
474 | 474 | if fnode != nullid: |
|
475 | 475 | cachefnode[node] = fnode |
|
476 | 476 | |
|
477 | 477 | fnodescache.write() |
|
478 | 478 | |
|
479 | 479 | duration = util.timer() - starttime |
|
480 | 480 | ui.log('tagscache', |
|
481 | 481 | '%d/%d cache hits/lookups in %0.4f ' |
|
482 | 482 | 'seconds\n', |
|
483 | 483 | fnodescache.hitcount, fnodescache.lookupcount, duration) |
|
484 | 484 | return cachefnode |
|
485 | 485 | |
|
486 | 486 | def _writetagcache(ui, repo, valid, cachetags): |
|
487 | 487 | filename = _filename(repo) |
|
488 | 488 | try: |
|
489 | 489 | cachefile = repo.cachevfs(filename, 'w', atomictemp=True) |
|
490 | 490 | except (OSError, IOError): |
|
491 | 491 | return |
|
492 | 492 | |
|
493 | 493 | ui.log('tagscache', 'writing .hg/cache/%s with %d tags\n', |
|
494 | 494 | filename, len(cachetags)) |
|
495 | 495 | |
|
496 | 496 | if valid[2]: |
|
497 | 497 | cachefile.write('%d %s %s\n' % (valid[0], hex(valid[1]), hex(valid[2]))) |
|
498 | 498 | else: |
|
499 | 499 | cachefile.write('%d %s\n' % (valid[0], hex(valid[1]))) |
|
500 | 500 | |
|
501 | 501 | # Tag names in the cache are in UTF-8 -- which is the whole reason |
|
502 | 502 | # we keep them in UTF-8 throughout this module. If we converted |
|
503 | 503 | # them local encoding on input, we would lose info writing them to |
|
504 | 504 | # the cache. |
|
505 | 505 | for (name, (node, hist)) in sorted(cachetags.iteritems()): |
|
506 | 506 | for n in hist: |
|
507 | 507 | cachefile.write("%s %s\n" % (hex(n), name)) |
|
508 | 508 | cachefile.write("%s %s\n" % (hex(node), name)) |
|
509 | 509 | |
|
510 | 510 | try: |
|
511 | 511 | cachefile.close() |
|
512 | 512 | except (OSError, IOError): |
|
513 | 513 | pass |
|
514 | 514 | |
|
515 | 515 | def tag(repo, names, node, message, local, user, date, editor=False): |
|
516 | 516 | '''tag a revision with one or more symbolic names. |
|
517 | 517 | |
|
518 | 518 | names is a list of strings or, when adding a single tag, names may be a |
|
519 | 519 | string. |
|
520 | 520 | |
|
521 | 521 | if local is True, the tags are stored in a per-repository file. |
|
522 | 522 | otherwise, they are stored in the .hgtags file, and a new |
|
523 | 523 | changeset is committed with the change. |
|
524 | 524 | |
|
525 | 525 | keyword arguments: |
|
526 | 526 | |
|
527 | 527 | local: whether to store tags in non-version-controlled file |
|
528 | 528 | (default False) |
|
529 | 529 | |
|
530 | 530 | message: commit message to use if committing |
|
531 | 531 | |
|
532 | 532 | user: name of user to use if committing |
|
533 | 533 | |
|
534 | 534 | date: date tuple to use if committing''' |
|
535 | 535 | |
|
536 | 536 | if not local: |
|
537 | 537 | m = matchmod.exact(repo.root, '', ['.hgtags']) |
|
538 | 538 | if any(repo.status(match=m, unknown=True, ignored=True)): |
|
539 | 539 | raise error.Abort(_('working copy of .hgtags is changed'), |
|
540 | 540 | hint=_('please commit .hgtags manually')) |
|
541 | 541 | |
|
542 | 542 | with repo.wlock(): |
|
543 | 543 | repo.tags() # instantiate the cache |
|
544 | 544 | _tag(repo, names, node, message, local, user, date, |
|
545 | 545 | editor=editor) |
|
546 | 546 | |
|
547 | 547 | def _tag(repo, names, node, message, local, user, date, extra=None, |
|
548 | 548 | editor=False): |
|
549 | 549 | if isinstance(names, str): |
|
550 | 550 | names = (names,) |
|
551 | 551 | |
|
552 | 552 | branches = repo.branchmap() |
|
553 | 553 | for name in names: |
|
554 | 554 | repo.hook('pretag', throw=True, node=hex(node), tag=name, |
|
555 | 555 | local=local) |
|
556 | 556 | if name in branches: |
|
557 | 557 | repo.ui.warn(_("warning: tag %s conflicts with existing" |
|
558 | 558 | " branch name\n") % name) |
|
559 | 559 | |
|
560 | 560 | def writetags(fp, names, munge, prevtags): |
|
561 | 561 | fp.seek(0, 2) |
|
562 | 562 | if prevtags and prevtags[-1] != '\n': |
|
563 | 563 | fp.write('\n') |
|
564 | 564 | for name in names: |
|
565 | 565 | if munge: |
|
566 | 566 | m = munge(name) |
|
567 | 567 | else: |
|
568 | 568 | m = name |
|
569 | 569 | |
|
570 | 570 | if (repo._tagscache.tagtypes and |
|
571 | 571 | name in repo._tagscache.tagtypes): |
|
572 | 572 | old = repo.tags().get(name, nullid) |
|
573 | 573 | fp.write('%s %s\n' % (hex(old), m)) |
|
574 | 574 | fp.write('%s %s\n' % (hex(node), m)) |
|
575 | 575 | fp.close() |
|
576 | 576 | |
|
577 | 577 | prevtags = '' |
|
578 | 578 | if local: |
|
579 | 579 | try: |
|
580 | 580 | fp = repo.vfs('localtags', 'r+') |
|
581 | 581 | except IOError: |
|
582 | 582 | fp = repo.vfs('localtags', 'a') |
|
583 | 583 | else: |
|
584 | 584 | prevtags = fp.read() |
|
585 | 585 | |
|
586 | 586 | # local tags are stored in the current charset |
|
587 | 587 | writetags(fp, names, None, prevtags) |
|
588 | 588 | for name in names: |
|
589 | 589 | repo.hook('tag', node=hex(node), tag=name, local=local) |
|
590 | 590 | return |
|
591 | 591 | |
|
592 | 592 | try: |
|
593 | 593 | fp = repo.wvfs('.hgtags', 'rb+') |
|
594 | 594 | except IOError as e: |
|
595 | 595 | if e.errno != errno.ENOENT: |
|
596 | 596 | raise |
|
597 | 597 | fp = repo.wvfs('.hgtags', 'ab') |
|
598 | 598 | else: |
|
599 | 599 | prevtags = fp.read() |
|
600 | 600 | |
|
601 | 601 | # committed tags are stored in UTF-8 |
|
602 | 602 | writetags(fp, names, encoding.fromlocal, prevtags) |
|
603 | 603 | |
|
604 | 604 | fp.close() |
|
605 | 605 | |
|
606 | 606 | repo.invalidatecaches() |
|
607 | 607 | |
|
608 | 608 | if '.hgtags' not in repo.dirstate: |
|
609 | 609 | repo[None].add(['.hgtags']) |
|
610 | 610 | |
|
611 | 611 | m = matchmod.exact(repo.root, '', ['.hgtags']) |
|
612 | 612 | tagnode = repo.commit(message, user, date, extra=extra, match=m, |
|
613 | 613 | editor=editor) |
|
614 | 614 | |
|
615 | 615 | for name in names: |
|
616 | 616 | repo.hook('tag', node=hex(node), tag=name, local=local) |
|
617 | 617 | |
|
618 | 618 | return tagnode |
|
619 | 619 | |
|
620 | 620 | _fnodescachefile = 'hgtagsfnodes1' |
|
621 | 621 | _fnodesrecsize = 4 + 20 # changeset fragment + filenode |
|
622 | 622 | _fnodesmissingrec = '\xff' * 24 |
|
623 | 623 | |
|
624 | 624 | class hgtagsfnodescache(object): |
|
625 | 625 | """Persistent cache mapping revisions to .hgtags filenodes. |
|
626 | 626 | |
|
627 | 627 | The cache is an array of records. Each item in the array corresponds to |
|
628 | 628 | a changelog revision. Values in the array contain the first 4 bytes of |
|
629 | 629 | the node hash and the 20 bytes .hgtags filenode for that revision. |
|
630 | 630 | |
|
631 | 631 | The first 4 bytes are present as a form of verification. Repository |
|
632 | 632 | stripping and rewriting may change the node at a numeric revision in the |
|
633 | 633 | changelog. The changeset fragment serves as a verifier to detect |
|
634 | 634 | rewriting. This logic is shared with the rev branch cache (see |
|
635 | 635 | branchmap.py). |
|
636 | 636 | |
|
637 | 637 | The instance holds in memory the full cache content but entries are |
|
638 | 638 | only parsed on read. |
|
639 | 639 | |
|
640 | 640 | Instances behave like lists. ``c[i]`` works where i is a rev or |
|
641 | 641 | changeset node. Missing indexes are populated automatically on access. |
|
642 | 642 | """ |
|
643 | 643 | def __init__(self, repo): |
|
644 | 644 | assert repo.filtername is None |
|
645 | 645 | |
|
646 | 646 | self._repo = repo |
|
647 | 647 | |
|
648 | 648 | # Only for reporting purposes. |
|
649 | 649 | self.lookupcount = 0 |
|
650 | 650 | self.hitcount = 0 |
|
651 | 651 | |
|
652 | 652 | |
|
653 | 653 | try: |
|
654 | 654 | data = repo.cachevfs.read(_fnodescachefile) |
|
655 | 655 | except (OSError, IOError): |
|
656 | 656 | data = "" |
|
657 | 657 | self._raw = bytearray(data) |
|
658 | 658 | |
|
659 | 659 | # The end state of self._raw is an array that is of the exact length |
|
660 | 660 | # required to hold a record for every revision in the repository. |
|
661 | 661 | # We truncate or extend the array as necessary. self._dirtyoffset is |
|
662 | 662 | # defined to be the start offset at which we need to write the output |
|
663 | 663 | # file. This offset is also adjusted when new entries are calculated |
|
664 | 664 | # for array members. |
|
665 | 665 | cllen = len(repo.changelog) |
|
666 | 666 | wantedlen = cllen * _fnodesrecsize |
|
667 | 667 | rawlen = len(self._raw) |
|
668 | 668 | |
|
669 | 669 | self._dirtyoffset = None |
|
670 | 670 | |
|
671 | 671 | if rawlen < wantedlen: |
|
672 | 672 | self._dirtyoffset = rawlen |
|
673 | 673 | self._raw.extend('\xff' * (wantedlen - rawlen)) |
|
674 | 674 | elif rawlen > wantedlen: |
|
675 | 675 | # There's no easy way to truncate array instances. This seems |
|
676 | 676 | # slightly less evil than copying a potentially large array slice. |
|
677 | 677 | for i in range(rawlen - wantedlen): |
|
678 | 678 | self._raw.pop() |
|
679 | 679 | self._dirtyoffset = len(self._raw) |
|
680 | 680 | |
|
681 | 681 | def getfnode(self, node, computemissing=True): |
|
682 | 682 | """Obtain the filenode of the .hgtags file at a specified revision. |
|
683 | 683 | |
|
684 | 684 | If the value is in the cache, the entry will be validated and returned. |
|
685 | 685 | Otherwise, the filenode will be computed and returned unless |
|
686 | 686 | "computemissing" is False, in which case None will be returned without |
|
687 | 687 | any potentially expensive computation being performed. |
|
688 | 688 | |
|
689 | 689 | If an .hgtags does not exist at the specified revision, nullid is |
|
690 | 690 | returned. |
|
691 | 691 | """ |
|
692 | 692 | ctx = self._repo[node] |
|
693 | 693 | rev = ctx.rev() |
|
694 | 694 | |
|
695 | 695 | self.lookupcount += 1 |
|
696 | 696 | |
|
697 | 697 | offset = rev * _fnodesrecsize |
|
698 | 698 | record = '%s' % self._raw[offset:offset + _fnodesrecsize] |
|
699 | 699 | properprefix = node[0:4] |
|
700 | 700 | |
|
701 | 701 | # Validate and return existing entry. |
|
702 | 702 | if record != _fnodesmissingrec: |
|
703 | 703 | fileprefix = record[0:4] |
|
704 | 704 | |
|
705 | 705 | if fileprefix == properprefix: |
|
706 | 706 | self.hitcount += 1 |
|
707 | 707 | return record[4:] |
|
708 | 708 | |
|
709 | 709 | # Fall through. |
|
710 | 710 | |
|
711 | 711 | # If we get here, the entry is either missing or invalid. |
|
712 | 712 | |
|
713 | 713 | if not computemissing: |
|
714 | 714 | return None |
|
715 | 715 | |
|
716 | 716 | # Populate missing entry. |
|
717 | 717 | try: |
|
718 | 718 | fnode = ctx.filenode('.hgtags') |
|
719 | 719 | except error.LookupError: |
|
720 | 720 | # No .hgtags file on this revision. |
|
721 | 721 | fnode = nullid |
|
722 | 722 | |
|
723 | 723 | self._writeentry(offset, properprefix, fnode) |
|
724 | 724 | return fnode |
|
725 | 725 | |
|
726 | 726 | def setfnode(self, node, fnode): |
|
727 | 727 | """Set the .hgtags filenode for a given changeset.""" |
|
728 | 728 | assert len(fnode) == 20 |
|
729 | 729 | ctx = self._repo[node] |
|
730 | 730 | |
|
731 | 731 | # Do a lookup first to avoid writing if nothing has changed. |
|
732 | 732 | if self.getfnode(ctx.node(), computemissing=False) == fnode: |
|
733 | 733 | return |
|
734 | 734 | |
|
735 | 735 | self._writeentry(ctx.rev() * _fnodesrecsize, node[0:4], fnode) |
|
736 | 736 | |
|
737 | 737 | def _writeentry(self, offset, prefix, fnode): |
|
738 | 738 | # Slices on array instances only accept other array. |
|
739 | 739 | entry = bytearray(prefix + fnode) |
|
740 | 740 | self._raw[offset:offset + _fnodesrecsize] = entry |
|
741 | 741 | # self._dirtyoffset could be None. |
|
742 | 742 | self._dirtyoffset = min(self._dirtyoffset or 0, offset or 0) |
|
743 | 743 | |
|
744 | 744 | def write(self): |
|
745 | 745 | """Perform all necessary writes to cache file. |
|
746 | 746 | |
|
747 | 747 | This may no-op if no writes are needed or if a write lock could |
|
748 | 748 | not be obtained. |
|
749 | 749 | """ |
|
750 | 750 | if self._dirtyoffset is None: |
|
751 | 751 | return |
|
752 | 752 | |
|
753 | 753 | data = self._raw[self._dirtyoffset:] |
|
754 | 754 | if not data: |
|
755 | 755 | return |
|
756 | 756 | |
|
757 | 757 | repo = self._repo |
|
758 | 758 | |
|
759 | 759 | try: |
|
760 | 760 | lock = repo.wlock(wait=False) |
|
761 | 761 | except error.LockError: |
|
762 | 762 | repo.ui.log('tagscache', 'not writing .hg/cache/%s because ' |
|
763 | 763 | 'lock cannot be acquired\n' % (_fnodescachefile)) |
|
764 | 764 | return |
|
765 | 765 | |
|
766 | 766 | try: |
|
767 | 767 | f = repo.cachevfs.open(_fnodescachefile, 'ab') |
|
768 | 768 | try: |
|
769 | 769 | # if the file has been truncated |
|
770 | 770 | actualoffset = f.tell() |
|
771 | 771 | if actualoffset < self._dirtyoffset: |
|
772 | 772 | self._dirtyoffset = actualoffset |
|
773 | 773 | data = self._raw[self._dirtyoffset:] |
|
774 | 774 | f.seek(self._dirtyoffset) |
|
775 | 775 | f.truncate() |
|
776 | 776 | repo.ui.log('tagscache', |
|
777 | 777 | 'writing %d bytes to cache/%s\n' % ( |
|
778 | 778 | len(data), _fnodescachefile)) |
|
779 | 779 | f.write(data) |
|
780 | 780 | self._dirtyoffset = None |
|
781 | 781 | finally: |
|
782 | 782 | f.close() |
|
783 | 783 | except (IOError, OSError) as inst: |
|
784 | 784 | repo.ui.log('tagscache', |
|
785 | 785 | "couldn't write cache/%s: %s\n" % ( |
|
786 | _fnodescachefile, inst)) | |
|
786 | _fnodescachefile, util.forcebytestr(inst))) | |
|
787 | 787 | finally: |
|
788 | 788 | lock.release() |
General Comments 0
You need to be logged in to leave comments.
Login now