Show More
@@ -1,933 +1,932 b'' | |||
|
1 | 1 | # tags.py - read tag info from local repository |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2009 Olivia Mackall <olivia@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 | |
|
14 | 14 | import binascii |
|
15 | 15 | import io |
|
16 | 16 | |
|
17 | 17 | from .node import ( |
|
18 | 18 | bin, |
|
19 | 19 | hex, |
|
20 | 20 | nullrev, |
|
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 | from .utils import stringutil |
|
32 | 32 | |
|
33 | 33 | # Tags computation can be expensive and caches exist to make it fast in |
|
34 | 34 | # the common case. |
|
35 | 35 | # |
|
36 | 36 | # The "hgtagsfnodes1" cache file caches the .hgtags filenode values for |
|
37 | 37 | # each revision in the repository. The file is effectively an array of |
|
38 | 38 | # fixed length records. Read the docs for "hgtagsfnodescache" for technical |
|
39 | 39 | # details. |
|
40 | 40 | # |
|
41 | 41 | # The .hgtags filenode cache grows in proportion to the length of the |
|
42 | 42 | # changelog. The file is truncated when the # changelog is stripped. |
|
43 | 43 | # |
|
44 | 44 | # The purpose of the filenode cache is to avoid the most expensive part |
|
45 | 45 | # of finding global tags, which is looking up the .hgtags filenode in the |
|
46 | 46 | # manifest for each head. This can take dozens or over 100ms for |
|
47 | 47 | # repositories with very large manifests. Multiplied by dozens or even |
|
48 | 48 | # hundreds of heads and there is a significant performance concern. |
|
49 | 49 | # |
|
50 | 50 | # There also exist a separate cache file for each repository filter. |
|
51 | 51 | # These "tags-*" files store information about the history of tags. |
|
52 | 52 | # |
|
53 | 53 | # The tags cache files consists of a cache validation line followed by |
|
54 | 54 | # a history of tags. |
|
55 | 55 | # |
|
56 | 56 | # The cache validation line has the format: |
|
57 | 57 | # |
|
58 | 58 | # <tiprev> <tipnode> [<filteredhash>] |
|
59 | 59 | # |
|
60 | 60 | # <tiprev> is an integer revision and <tipnode> is a 40 character hex |
|
61 | 61 | # node for that changeset. These redundantly identify the repository |
|
62 | 62 | # tip from the time the cache was written. In addition, <filteredhash>, |
|
63 | 63 | # if present, is a 40 character hex hash of the contents of the filtered |
|
64 | 64 | # revisions for this filter. If the set of filtered revs changes, the |
|
65 | 65 | # hash will change and invalidate the cache. |
|
66 | 66 | # |
|
67 | 67 | # The history part of the tags cache consists of lines of the form: |
|
68 | 68 | # |
|
69 | 69 | # <node> <tag> |
|
70 | 70 | # |
|
71 | 71 | # (This format is identical to that of .hgtags files.) |
|
72 | 72 | # |
|
73 | 73 | # <tag> is the tag name and <node> is the 40 character hex changeset |
|
74 | 74 | # the tag is associated with. |
|
75 | 75 | # |
|
76 | 76 | # Tags are written sorted by tag name. |
|
77 | 77 | # |
|
78 | 78 | # Tags associated with multiple changesets have an entry for each changeset. |
|
79 | 79 | # The most recent changeset (in terms of revlog ordering for the head |
|
80 | 80 | # setting it) for each tag is last. |
|
81 | 81 | |
|
82 | 82 | |
|
83 | 83 | def fnoderevs(ui, repo, revs): |
|
84 | 84 | """return the list of '.hgtags' fnodes used in a set revisions |
|
85 | 85 | |
|
86 | 86 | This is returned as list of unique fnodes. We use a list instead of a set |
|
87 | 87 | because order matters when it comes to tags.""" |
|
88 | 88 | unfi = repo.unfiltered() |
|
89 | 89 | tonode = unfi.changelog.node |
|
90 | 90 | nodes = [tonode(r) for r in revs] |
|
91 | 91 | fnodes = _getfnodes(ui, repo, nodes) |
|
92 | 92 | fnodes = _filterfnodes(fnodes, nodes) |
|
93 | 93 | return fnodes |
|
94 | 94 | |
|
95 | 95 | |
|
96 | 96 | def _nulltonone(repo, value): |
|
97 | 97 | """convert nullid to None |
|
98 | 98 | |
|
99 | 99 | For tag value, nullid means "deleted". This small utility function helps |
|
100 | 100 | translating that to None.""" |
|
101 | 101 | if value == repo.nullid: |
|
102 | 102 | return None |
|
103 | 103 | return value |
|
104 | 104 | |
|
105 | 105 | |
|
106 | 106 | def difftags(ui, repo, oldfnodes, newfnodes): |
|
107 | 107 | """list differences between tags expressed in two set of file-nodes |
|
108 | 108 | |
|
109 | 109 | The list contains entries in the form: (tagname, oldvalue, new value). |
|
110 | 110 | None is used to expressed missing value: |
|
111 | 111 | ('foo', None, 'abcd') is a new tag, |
|
112 | 112 | ('bar', 'ef01', None) is a deletion, |
|
113 | 113 | ('baz', 'abcd', 'ef01') is a tag movement. |
|
114 | 114 | """ |
|
115 | 115 | if oldfnodes == newfnodes: |
|
116 | 116 | return [] |
|
117 | 117 | oldtags = _tagsfromfnodes(ui, repo, oldfnodes) |
|
118 | 118 | newtags = _tagsfromfnodes(ui, repo, newfnodes) |
|
119 | 119 | |
|
120 | 120 | # list of (tag, old, new): None means missing |
|
121 | 121 | entries = [] |
|
122 | 122 | for tag, (new, __) in newtags.items(): |
|
123 | 123 | new = _nulltonone(repo, new) |
|
124 | 124 | old, __ = oldtags.pop(tag, (None, None)) |
|
125 | 125 | old = _nulltonone(repo, old) |
|
126 | 126 | if old != new: |
|
127 | 127 | entries.append((tag, old, new)) |
|
128 | 128 | # handle deleted tags |
|
129 | 129 | for tag, (old, __) in oldtags.items(): |
|
130 | 130 | old = _nulltonone(repo, old) |
|
131 | 131 | if old is not None: |
|
132 | 132 | entries.append((tag, old, None)) |
|
133 | 133 | entries.sort() |
|
134 | 134 | return entries |
|
135 | 135 | |
|
136 | 136 | |
|
137 | 137 | def writediff(fp, difflist): |
|
138 | 138 | """write tags diff information to a file. |
|
139 | 139 | |
|
140 | 140 | Data are stored with a line based format: |
|
141 | 141 | |
|
142 | 142 | <action> <hex-node> <tag-name>\n |
|
143 | 143 | |
|
144 | 144 | Action are defined as follow: |
|
145 | 145 | -R tag is removed, |
|
146 | 146 | +A tag is added, |
|
147 | 147 | -M tag is moved (old value), |
|
148 | 148 | +M tag is moved (new value), |
|
149 | 149 | |
|
150 | 150 | Example: |
|
151 | 151 | |
|
152 | 152 | +A 875517b4806a848f942811a315a5bce30804ae85 t5 |
|
153 | 153 | |
|
154 | 154 | See documentation of difftags output for details about the input. |
|
155 | 155 | """ |
|
156 | 156 | add = b'+A %s %s\n' |
|
157 | 157 | remove = b'-R %s %s\n' |
|
158 | 158 | updateold = b'-M %s %s\n' |
|
159 | 159 | updatenew = b'+M %s %s\n' |
|
160 | 160 | for tag, old, new in difflist: |
|
161 | 161 | # translate to hex |
|
162 | 162 | if old is not None: |
|
163 | 163 | old = hex(old) |
|
164 | 164 | if new is not None: |
|
165 | 165 | new = hex(new) |
|
166 | 166 | # write to file |
|
167 | 167 | if old is None: |
|
168 | 168 | fp.write(add % (new, tag)) |
|
169 | 169 | elif new is None: |
|
170 | 170 | fp.write(remove % (old, tag)) |
|
171 | 171 | else: |
|
172 | 172 | fp.write(updateold % (old, tag)) |
|
173 | 173 | fp.write(updatenew % (new, tag)) |
|
174 | 174 | |
|
175 | 175 | |
|
176 | 176 | def findglobaltags(ui, repo): |
|
177 | 177 | """Find global tags in a repo: return a tagsmap |
|
178 | 178 | |
|
179 | 179 | tagsmap: tag name to (node, hist) 2-tuples. |
|
180 | 180 | |
|
181 | 181 | The tags cache is read and updated as a side-effect of calling. |
|
182 | 182 | """ |
|
183 | 183 | (heads, tagfnode, valid, cachetags, shouldwrite) = _readtagcache(ui, repo) |
|
184 | 184 | if cachetags is not None: |
|
185 | 185 | assert not shouldwrite |
|
186 | 186 | # XXX is this really 100% correct? are there oddball special |
|
187 | 187 | # cases where a global tag should outrank a local tag but won't, |
|
188 | 188 | # because cachetags does not contain rank info? |
|
189 | 189 | alltags = {} |
|
190 | 190 | _updatetags(cachetags, alltags) |
|
191 | 191 | return alltags |
|
192 | 192 | |
|
193 | has_node = repo.changelog.index.has_node | |
|
193 | 194 | for head in reversed(heads): # oldest to newest |
|
194 | assert repo.changelog.index.has_node( | |
|
195 | head | |
|
196 | ), b"tag cache returned bogus head %s" % short(head) | |
|
195 | assert has_node(head), b"tag cache returned bogus head %s" % short(head) | |
|
197 | 196 | fnodes = _filterfnodes(tagfnode, reversed(heads)) |
|
198 | 197 | alltags = _tagsfromfnodes(ui, repo, fnodes) |
|
199 | 198 | |
|
200 | 199 | # and update the cache (if necessary) |
|
201 | 200 | if shouldwrite: |
|
202 | 201 | _writetagcache(ui, repo, valid, alltags) |
|
203 | 202 | return alltags |
|
204 | 203 | |
|
205 | 204 | |
|
206 | 205 | def _filterfnodes(tagfnode, nodes): |
|
207 | 206 | """return a list of unique fnodes |
|
208 | 207 | |
|
209 | 208 | The order of this list matches the order of "nodes". Preserving this order |
|
210 | 209 | is important as reading tags in different order provides different |
|
211 | 210 | results.""" |
|
212 | 211 | seen = set() # set of fnode |
|
213 | 212 | fnodes = [] |
|
214 | 213 | for no in nodes: # oldest to newest |
|
215 | 214 | fnode = tagfnode.get(no) |
|
216 | 215 | if fnode and fnode not in seen: |
|
217 | 216 | seen.add(fnode) |
|
218 | 217 | fnodes.append(fnode) |
|
219 | 218 | return fnodes |
|
220 | 219 | |
|
221 | 220 | |
|
222 | 221 | def _tagsfromfnodes(ui, repo, fnodes): |
|
223 | 222 | """return a tagsmap from a list of file-node |
|
224 | 223 | |
|
225 | 224 | tagsmap: tag name to (node, hist) 2-tuples. |
|
226 | 225 | |
|
227 | 226 | The order of the list matters.""" |
|
228 | 227 | alltags = {} |
|
229 | 228 | fctx = None |
|
230 | 229 | for fnode in fnodes: |
|
231 | 230 | if fctx is None: |
|
232 | 231 | fctx = repo.filectx(b'.hgtags', fileid=fnode) |
|
233 | 232 | else: |
|
234 | 233 | fctx = fctx.filectx(fnode) |
|
235 | 234 | filetags = _readtags(ui, repo, fctx.data().splitlines(), fctx) |
|
236 | 235 | _updatetags(filetags, alltags) |
|
237 | 236 | return alltags |
|
238 | 237 | |
|
239 | 238 | |
|
240 | 239 | def readlocaltags(ui, repo, alltags, tagtypes): |
|
241 | 240 | '''Read local tags in repo. Update alltags and tagtypes.''' |
|
242 | 241 | try: |
|
243 | 242 | data = repo.vfs.read(b"localtags") |
|
244 | 243 | except FileNotFoundError: |
|
245 | 244 | return |
|
246 | 245 | |
|
247 | 246 | # localtags is in the local encoding; re-encode to UTF-8 on |
|
248 | 247 | # input for consistency with the rest of this module. |
|
249 | 248 | filetags = _readtags( |
|
250 | 249 | ui, repo, data.splitlines(), b"localtags", recode=encoding.fromlocal |
|
251 | 250 | ) |
|
252 | 251 | |
|
253 | 252 | # remove tags pointing to invalid nodes |
|
254 | 253 | cl = repo.changelog |
|
255 | 254 | for t in list(filetags): |
|
256 | 255 | try: |
|
257 | 256 | cl.rev(filetags[t][0]) |
|
258 | 257 | except (LookupError, ValueError): |
|
259 | 258 | del filetags[t] |
|
260 | 259 | |
|
261 | 260 | _updatetags(filetags, alltags, b'local', tagtypes) |
|
262 | 261 | |
|
263 | 262 | |
|
264 | 263 | def _readtaghist(ui, repo, lines, fn, recode=None, calcnodelines=False): |
|
265 | 264 | """Read tag definitions from a file (or any source of lines). |
|
266 | 265 | |
|
267 | 266 | This function returns two sortdicts with similar information: |
|
268 | 267 | |
|
269 | 268 | - the first dict, bintaghist, contains the tag information as expected by |
|
270 | 269 | the _readtags function, i.e. a mapping from tag name to (node, hist): |
|
271 | 270 | - node is the node id from the last line read for that name, |
|
272 | 271 | - hist is the list of node ids previously associated with it (in file |
|
273 | 272 | order). All node ids are binary, not hex. |
|
274 | 273 | |
|
275 | 274 | - the second dict, hextaglines, is a mapping from tag name to a list of |
|
276 | 275 | [hexnode, line number] pairs, ordered from the oldest to the newest node. |
|
277 | 276 | |
|
278 | 277 | When calcnodelines is False the hextaglines dict is not calculated (an |
|
279 | 278 | empty dict is returned). This is done to improve this function's |
|
280 | 279 | performance in cases where the line numbers are not needed. |
|
281 | 280 | """ |
|
282 | 281 | |
|
283 | 282 | bintaghist = util.sortdict() |
|
284 | 283 | hextaglines = util.sortdict() |
|
285 | 284 | count = 0 |
|
286 | 285 | |
|
287 | 286 | def dbg(msg): |
|
288 | 287 | ui.debug(b"%s, line %d: %s\n" % (fn, count, msg)) |
|
289 | 288 | |
|
290 | 289 | for nline, line in enumerate(lines): |
|
291 | 290 | count += 1 |
|
292 | 291 | if not line: |
|
293 | 292 | continue |
|
294 | 293 | try: |
|
295 | 294 | (nodehex, name) = line.split(b" ", 1) |
|
296 | 295 | except ValueError: |
|
297 | 296 | dbg(b"cannot parse entry") |
|
298 | 297 | continue |
|
299 | 298 | name = name.strip() |
|
300 | 299 | if recode: |
|
301 | 300 | name = recode(name) |
|
302 | 301 | try: |
|
303 | 302 | nodebin = bin(nodehex) |
|
304 | 303 | except binascii.Error: |
|
305 | 304 | dbg(b"node '%s' is not well formed" % nodehex) |
|
306 | 305 | continue |
|
307 | 306 | |
|
308 | 307 | # update filetags |
|
309 | 308 | if calcnodelines: |
|
310 | 309 | # map tag name to a list of line numbers |
|
311 | 310 | if name not in hextaglines: |
|
312 | 311 | hextaglines[name] = [] |
|
313 | 312 | hextaglines[name].append([nodehex, nline]) |
|
314 | 313 | continue |
|
315 | 314 | # map tag name to (node, hist) |
|
316 | 315 | if name not in bintaghist: |
|
317 | 316 | bintaghist[name] = [] |
|
318 | 317 | bintaghist[name].append(nodebin) |
|
319 | 318 | return bintaghist, hextaglines |
|
320 | 319 | |
|
321 | 320 | |
|
322 | 321 | def _readtags(ui, repo, lines, fn, recode=None, calcnodelines=False): |
|
323 | 322 | """Read tag definitions from a file (or any source of lines). |
|
324 | 323 | |
|
325 | 324 | Returns a mapping from tag name to (node, hist). |
|
326 | 325 | |
|
327 | 326 | "node" is the node id from the last line read for that name. "hist" |
|
328 | 327 | is the list of node ids previously associated with it (in file order). |
|
329 | 328 | All node ids are binary, not hex. |
|
330 | 329 | """ |
|
331 | 330 | filetags, nodelines = _readtaghist( |
|
332 | 331 | ui, repo, lines, fn, recode=recode, calcnodelines=calcnodelines |
|
333 | 332 | ) |
|
334 | 333 | # util.sortdict().__setitem__ is much slower at replacing then inserting |
|
335 | 334 | # new entries. The difference can matter if there are thousands of tags. |
|
336 | 335 | # Create a new sortdict to avoid the performance penalty. |
|
337 | 336 | newtags = util.sortdict() |
|
338 | 337 | for tag, taghist in filetags.items(): |
|
339 | 338 | newtags[tag] = (taghist[-1], taghist[:-1]) |
|
340 | 339 | return newtags |
|
341 | 340 | |
|
342 | 341 | |
|
343 | 342 | def _updatetags(filetags, alltags, tagtype=None, tagtypes=None): |
|
344 | 343 | """Incorporate the tag info read from one file into dictionnaries |
|
345 | 344 | |
|
346 | 345 | The first one, 'alltags', is a "tagmaps" (see 'findglobaltags' for details). |
|
347 | 346 | |
|
348 | 347 | The second one, 'tagtypes', is optional and will be updated to track the |
|
349 | 348 | "tagtype" of entries in the tagmaps. When set, the 'tagtype' argument also |
|
350 | 349 | needs to be set.""" |
|
351 | 350 | if tagtype is None: |
|
352 | 351 | assert tagtypes is None |
|
353 | 352 | |
|
354 | 353 | for name, nodehist in filetags.items(): |
|
355 | 354 | if name not in alltags: |
|
356 | 355 | alltags[name] = nodehist |
|
357 | 356 | if tagtype is not None: |
|
358 | 357 | tagtypes[name] = tagtype |
|
359 | 358 | continue |
|
360 | 359 | |
|
361 | 360 | # we prefer alltags[name] if: |
|
362 | 361 | # it supersedes us OR |
|
363 | 362 | # mutual supersedes and it has a higher rank |
|
364 | 363 | # otherwise we win because we're tip-most |
|
365 | 364 | anode, ahist = nodehist |
|
366 | 365 | bnode, bhist = alltags[name] |
|
367 | 366 | if ( |
|
368 | 367 | bnode != anode |
|
369 | 368 | and anode in bhist |
|
370 | 369 | and (bnode not in ahist or len(bhist) > len(ahist)) |
|
371 | 370 | ): |
|
372 | 371 | anode = bnode |
|
373 | 372 | elif tagtype is not None: |
|
374 | 373 | tagtypes[name] = tagtype |
|
375 | 374 | ahist.extend([n for n in bhist if n not in ahist]) |
|
376 | 375 | alltags[name] = anode, ahist |
|
377 | 376 | |
|
378 | 377 | |
|
379 | 378 | def _filename(repo): |
|
380 | 379 | """name of a tagcache file for a given repo or repoview""" |
|
381 | 380 | filename = b'tags2' |
|
382 | 381 | if repo.filtername: |
|
383 | 382 | filename = b'%s-%s' % (filename, repo.filtername) |
|
384 | 383 | return filename |
|
385 | 384 | |
|
386 | 385 | |
|
387 | 386 | def _readtagcache(ui, repo): |
|
388 | 387 | """Read the tag cache. |
|
389 | 388 | |
|
390 | 389 | Returns a tuple (heads, fnodes, validinfo, cachetags, shouldwrite). |
|
391 | 390 | |
|
392 | 391 | If the cache is completely up-to-date, "cachetags" is a dict of the |
|
393 | 392 | form returned by _readtags() and "heads", "fnodes", and "validinfo" are |
|
394 | 393 | None and "shouldwrite" is False. |
|
395 | 394 | |
|
396 | 395 | If the cache is not up to date, "cachetags" is None. "heads" is a list |
|
397 | 396 | of all heads currently in the repository, ordered from tip to oldest. |
|
398 | 397 | "validinfo" is a tuple describing cache validation info. This is used |
|
399 | 398 | when writing the tags cache. "fnodes" is a mapping from head to .hgtags |
|
400 | 399 | filenode. "shouldwrite" is True. |
|
401 | 400 | |
|
402 | 401 | If the cache is not up to date, the caller is responsible for reading tag |
|
403 | 402 | info from each returned head. (See findglobaltags().) |
|
404 | 403 | """ |
|
405 | 404 | try: |
|
406 | 405 | cachefile = repo.cachevfs(_filename(repo), b'r') |
|
407 | 406 | # force reading the file for static-http |
|
408 | 407 | cachelines = iter(cachefile) |
|
409 | 408 | except IOError: |
|
410 | 409 | cachefile = None |
|
411 | 410 | |
|
412 | 411 | cacherev = None |
|
413 | 412 | cachenode = None |
|
414 | 413 | cachehash = None |
|
415 | 414 | if cachefile: |
|
416 | 415 | try: |
|
417 | 416 | validline = next(cachelines) |
|
418 | 417 | validline = validline.split() |
|
419 | 418 | cacherev = int(validline[0]) |
|
420 | 419 | cachenode = bin(validline[1]) |
|
421 | 420 | if len(validline) > 2: |
|
422 | 421 | cachehash = bin(validline[2]) |
|
423 | 422 | except Exception: |
|
424 | 423 | # corruption of the cache, just recompute it. |
|
425 | 424 | pass |
|
426 | 425 | |
|
427 | 426 | tipnode = repo.changelog.tip() |
|
428 | 427 | tiprev = len(repo.changelog) - 1 |
|
429 | 428 | |
|
430 | 429 | # Case 1 (common): tip is the same, so nothing has changed. |
|
431 | 430 | # (Unchanged tip trivially means no changesets have been added. |
|
432 | 431 | # But, thanks to localrepository.destroyed(), it also means none |
|
433 | 432 | # have been destroyed by strip or rollback.) |
|
434 | 433 | if ( |
|
435 | 434 | cacherev == tiprev |
|
436 | 435 | and cachenode == tipnode |
|
437 | 436 | and cachehash == scmutil.filteredhash(repo, tiprev) |
|
438 | 437 | ): |
|
439 | 438 | tags = _readtags(ui, repo, cachelines, cachefile.name) |
|
440 | 439 | cachefile.close() |
|
441 | 440 | return (None, None, None, tags, False) |
|
442 | 441 | if cachefile: |
|
443 | 442 | cachefile.close() # ignore rest of file |
|
444 | 443 | |
|
445 | 444 | valid = (tiprev, tipnode, scmutil.filteredhash(repo, tiprev)) |
|
446 | 445 | |
|
447 | 446 | repoheads = repo.heads() |
|
448 | 447 | # Case 2 (uncommon): empty repo; get out quickly and don't bother |
|
449 | 448 | # writing an empty cache. |
|
450 | 449 | if repoheads == [repo.nullid]: |
|
451 | 450 | return ([], {}, valid, {}, False) |
|
452 | 451 | |
|
453 | 452 | # Case 3 (uncommon): cache file missing or empty. |
|
454 | 453 | |
|
455 | 454 | # Case 4 (uncommon): tip rev decreased. This should only happen |
|
456 | 455 | # when we're called from localrepository.destroyed(). Refresh the |
|
457 | 456 | # cache so future invocations will not see disappeared heads in the |
|
458 | 457 | # cache. |
|
459 | 458 | |
|
460 | 459 | # Case 5 (common): tip has changed, so we've added/replaced heads. |
|
461 | 460 | |
|
462 | 461 | # As it happens, the code to handle cases 3, 4, 5 is the same. |
|
463 | 462 | |
|
464 | 463 | # N.B. in case 4 (nodes destroyed), "new head" really means "newly |
|
465 | 464 | # exposed". |
|
466 | 465 | if not len(repo.file(b'.hgtags')): |
|
467 | 466 | # No tags have ever been committed, so we can avoid a |
|
468 | 467 | # potentially expensive search. |
|
469 | 468 | return ([], {}, valid, None, True) |
|
470 | 469 | |
|
471 | 470 | # Now we have to lookup the .hgtags filenode for every new head. |
|
472 | 471 | # This is the most expensive part of finding tags, so performance |
|
473 | 472 | # depends primarily on the size of newheads. Worst case: no cache |
|
474 | 473 | # file, so newheads == repoheads. |
|
475 | 474 | # Reversed order helps the cache ('repoheads' is in descending order) |
|
476 | 475 | cachefnode = _getfnodes(ui, repo, reversed(repoheads)) |
|
477 | 476 | |
|
478 | 477 | # Caller has to iterate over all heads, but can use the filenodes in |
|
479 | 478 | # cachefnode to get to each .hgtags revision quickly. |
|
480 | 479 | return (repoheads, cachefnode, valid, None, True) |
|
481 | 480 | |
|
482 | 481 | |
|
483 | 482 | def _getfnodes(ui, repo, nodes): |
|
484 | 483 | """return .hgtags fnodes for a list of changeset nodes |
|
485 | 484 | |
|
486 | 485 | Return value is a {node: fnode} mapping. There will be no entry for nodes |
|
487 | 486 | without a '.hgtags' file. |
|
488 | 487 | """ |
|
489 | 488 | starttime = util.timer() |
|
490 | 489 | fnodescache = hgtagsfnodescache(repo.unfiltered()) |
|
491 | 490 | cachefnode = {} |
|
492 | 491 | validated_fnodes = set() |
|
493 | 492 | unknown_entries = set() |
|
494 | 493 | |
|
495 | 494 | flog = None |
|
496 | 495 | for node in nodes: |
|
497 | 496 | fnode = fnodescache.getfnode(node) |
|
498 | 497 | if fnode != repo.nullid: |
|
499 | 498 | if fnode not in validated_fnodes: |
|
500 | 499 | if flog is None: |
|
501 | 500 | flog = repo.file(b'.hgtags') |
|
502 | 501 | if flog.hasnode(fnode): |
|
503 | 502 | validated_fnodes.add(fnode) |
|
504 | 503 | else: |
|
505 | 504 | unknown_entries.add(node) |
|
506 | 505 | cachefnode[node] = fnode |
|
507 | 506 | |
|
508 | 507 | if unknown_entries: |
|
509 | 508 | fixed_nodemap = fnodescache.refresh_invalid_nodes(unknown_entries) |
|
510 | 509 | for node, fnode in fixed_nodemap.items(): |
|
511 | 510 | if fnode != repo.nullid: |
|
512 | 511 | cachefnode[node] = fnode |
|
513 | 512 | |
|
514 | 513 | fnodescache.write() |
|
515 | 514 | |
|
516 | 515 | duration = util.timer() - starttime |
|
517 | 516 | ui.log( |
|
518 | 517 | b'tagscache', |
|
519 | 518 | b'%d/%d cache hits/lookups in %0.4f seconds\n', |
|
520 | 519 | fnodescache.hitcount, |
|
521 | 520 | fnodescache.lookupcount, |
|
522 | 521 | duration, |
|
523 | 522 | ) |
|
524 | 523 | return cachefnode |
|
525 | 524 | |
|
526 | 525 | |
|
527 | 526 | def _writetagcache(ui, repo, valid, cachetags): |
|
528 | 527 | filename = _filename(repo) |
|
529 | 528 | try: |
|
530 | 529 | cachefile = repo.cachevfs(filename, b'w', atomictemp=True) |
|
531 | 530 | except (OSError, IOError): |
|
532 | 531 | return |
|
533 | 532 | |
|
534 | 533 | ui.log( |
|
535 | 534 | b'tagscache', |
|
536 | 535 | b'writing .hg/cache/%s with %d tags\n', |
|
537 | 536 | filename, |
|
538 | 537 | len(cachetags), |
|
539 | 538 | ) |
|
540 | 539 | |
|
541 | 540 | if valid[2]: |
|
542 | 541 | cachefile.write( |
|
543 | 542 | b'%d %s %s\n' % (valid[0], hex(valid[1]), hex(valid[2])) |
|
544 | 543 | ) |
|
545 | 544 | else: |
|
546 | 545 | cachefile.write(b'%d %s\n' % (valid[0], hex(valid[1]))) |
|
547 | 546 | |
|
548 | 547 | # Tag names in the cache are in UTF-8 -- which is the whole reason |
|
549 | 548 | # we keep them in UTF-8 throughout this module. If we converted |
|
550 | 549 | # them local encoding on input, we would lose info writing them to |
|
551 | 550 | # the cache. |
|
552 | 551 | for (name, (node, hist)) in sorted(cachetags.items()): |
|
553 | 552 | for n in hist: |
|
554 | 553 | cachefile.write(b"%s %s\n" % (hex(n), name)) |
|
555 | 554 | cachefile.write(b"%s %s\n" % (hex(node), name)) |
|
556 | 555 | |
|
557 | 556 | try: |
|
558 | 557 | cachefile.close() |
|
559 | 558 | except (OSError, IOError): |
|
560 | 559 | pass |
|
561 | 560 | |
|
562 | 561 | |
|
563 | 562 | def tag(repo, names, node, message, local, user, date, editor=False): |
|
564 | 563 | """tag a revision with one or more symbolic names. |
|
565 | 564 | |
|
566 | 565 | names is a list of strings or, when adding a single tag, names may be a |
|
567 | 566 | string. |
|
568 | 567 | |
|
569 | 568 | if local is True, the tags are stored in a per-repository file. |
|
570 | 569 | otherwise, they are stored in the .hgtags file, and a new |
|
571 | 570 | changeset is committed with the change. |
|
572 | 571 | |
|
573 | 572 | keyword arguments: |
|
574 | 573 | |
|
575 | 574 | local: whether to store tags in non-version-controlled file |
|
576 | 575 | (default False) |
|
577 | 576 | |
|
578 | 577 | message: commit message to use if committing |
|
579 | 578 | |
|
580 | 579 | user: name of user to use if committing |
|
581 | 580 | |
|
582 | 581 | date: date tuple to use if committing""" |
|
583 | 582 | |
|
584 | 583 | if not local: |
|
585 | 584 | m = matchmod.exact([b'.hgtags']) |
|
586 | 585 | st = repo.status(match=m, unknown=True, ignored=True) |
|
587 | 586 | if any( |
|
588 | 587 | ( |
|
589 | 588 | st.modified, |
|
590 | 589 | st.added, |
|
591 | 590 | st.removed, |
|
592 | 591 | st.deleted, |
|
593 | 592 | st.unknown, |
|
594 | 593 | st.ignored, |
|
595 | 594 | ) |
|
596 | 595 | ): |
|
597 | 596 | raise error.Abort( |
|
598 | 597 | _(b'working copy of .hgtags is changed'), |
|
599 | 598 | hint=_(b'please commit .hgtags manually'), |
|
600 | 599 | ) |
|
601 | 600 | |
|
602 | 601 | with repo.wlock(): |
|
603 | 602 | repo.tags() # instantiate the cache |
|
604 | 603 | _tag(repo, names, node, message, local, user, date, editor=editor) |
|
605 | 604 | |
|
606 | 605 | |
|
607 | 606 | def _tag( |
|
608 | 607 | repo, names, node, message, local, user, date, extra=None, editor=False |
|
609 | 608 | ): |
|
610 | 609 | if isinstance(names, bytes): |
|
611 | 610 | names = (names,) |
|
612 | 611 | |
|
613 | 612 | branches = repo.branchmap() |
|
614 | 613 | for name in names: |
|
615 | 614 | repo.hook(b'pretag', throw=True, node=hex(node), tag=name, local=local) |
|
616 | 615 | if name in branches: |
|
617 | 616 | repo.ui.warn( |
|
618 | 617 | _(b"warning: tag %s conflicts with existing branch name\n") |
|
619 | 618 | % name |
|
620 | 619 | ) |
|
621 | 620 | |
|
622 | 621 | def writetags(fp, names, munge, prevtags): |
|
623 | 622 | fp.seek(0, io.SEEK_END) |
|
624 | 623 | if prevtags and not prevtags.endswith(b'\n'): |
|
625 | 624 | fp.write(b'\n') |
|
626 | 625 | for name in names: |
|
627 | 626 | if munge: |
|
628 | 627 | m = munge(name) |
|
629 | 628 | else: |
|
630 | 629 | m = name |
|
631 | 630 | |
|
632 | 631 | if repo._tagscache.tagtypes and name in repo._tagscache.tagtypes: |
|
633 | 632 | old = repo.tags().get(name, repo.nullid) |
|
634 | 633 | fp.write(b'%s %s\n' % (hex(old), m)) |
|
635 | 634 | fp.write(b'%s %s\n' % (hex(node), m)) |
|
636 | 635 | fp.close() |
|
637 | 636 | |
|
638 | 637 | prevtags = b'' |
|
639 | 638 | if local: |
|
640 | 639 | try: |
|
641 | 640 | fp = repo.vfs(b'localtags', b'r+') |
|
642 | 641 | except IOError: |
|
643 | 642 | fp = repo.vfs(b'localtags', b'a') |
|
644 | 643 | else: |
|
645 | 644 | prevtags = fp.read() |
|
646 | 645 | |
|
647 | 646 | # local tags are stored in the current charset |
|
648 | 647 | writetags(fp, names, None, prevtags) |
|
649 | 648 | for name in names: |
|
650 | 649 | repo.hook(b'tag', node=hex(node), tag=name, local=local) |
|
651 | 650 | return |
|
652 | 651 | |
|
653 | 652 | try: |
|
654 | 653 | fp = repo.wvfs(b'.hgtags', b'rb+') |
|
655 | 654 | except FileNotFoundError: |
|
656 | 655 | fp = repo.wvfs(b'.hgtags', b'ab') |
|
657 | 656 | else: |
|
658 | 657 | prevtags = fp.read() |
|
659 | 658 | |
|
660 | 659 | # committed tags are stored in UTF-8 |
|
661 | 660 | writetags(fp, names, encoding.fromlocal, prevtags) |
|
662 | 661 | |
|
663 | 662 | fp.close() |
|
664 | 663 | |
|
665 | 664 | repo.invalidatecaches() |
|
666 | 665 | |
|
667 | 666 | with repo.dirstate.changing_files(repo): |
|
668 | 667 | if b'.hgtags' not in repo.dirstate: |
|
669 | 668 | repo[None].add([b'.hgtags']) |
|
670 | 669 | |
|
671 | 670 | m = matchmod.exact([b'.hgtags']) |
|
672 | 671 | tagnode = repo.commit( |
|
673 | 672 | message, user, date, extra=extra, match=m, editor=editor |
|
674 | 673 | ) |
|
675 | 674 | |
|
676 | 675 | for name in names: |
|
677 | 676 | repo.hook(b'tag', node=hex(node), tag=name, local=local) |
|
678 | 677 | |
|
679 | 678 | return tagnode |
|
680 | 679 | |
|
681 | 680 | |
|
682 | 681 | _fnodescachefile = b'hgtagsfnodes1' |
|
683 | 682 | _fnodesrecsize = 4 + 20 # changeset fragment + filenode |
|
684 | 683 | _fnodesmissingrec = b'\xff' * 24 |
|
685 | 684 | |
|
686 | 685 | |
|
687 | 686 | class hgtagsfnodescache: |
|
688 | 687 | """Persistent cache mapping revisions to .hgtags filenodes. |
|
689 | 688 | |
|
690 | 689 | The cache is an array of records. Each item in the array corresponds to |
|
691 | 690 | a changelog revision. Values in the array contain the first 4 bytes of |
|
692 | 691 | the node hash and the 20 bytes .hgtags filenode for that revision. |
|
693 | 692 | |
|
694 | 693 | The first 4 bytes are present as a form of verification. Repository |
|
695 | 694 | stripping and rewriting may change the node at a numeric revision in the |
|
696 | 695 | changelog. The changeset fragment serves as a verifier to detect |
|
697 | 696 | rewriting. This logic is shared with the rev branch cache (see |
|
698 | 697 | branchmap.py). |
|
699 | 698 | |
|
700 | 699 | The instance holds in memory the full cache content but entries are |
|
701 | 700 | only parsed on read. |
|
702 | 701 | |
|
703 | 702 | Instances behave like lists. ``c[i]`` works where i is a rev or |
|
704 | 703 | changeset node. Missing indexes are populated automatically on access. |
|
705 | 704 | """ |
|
706 | 705 | |
|
707 | 706 | def __init__(self, repo): |
|
708 | 707 | assert repo.filtername is None |
|
709 | 708 | |
|
710 | 709 | self._repo = repo |
|
711 | 710 | |
|
712 | 711 | # Only for reporting purposes. |
|
713 | 712 | self.lookupcount = 0 |
|
714 | 713 | self.hitcount = 0 |
|
715 | 714 | |
|
716 | 715 | try: |
|
717 | 716 | data = repo.cachevfs.read(_fnodescachefile) |
|
718 | 717 | except (OSError, IOError): |
|
719 | 718 | data = b"" |
|
720 | 719 | self._raw = bytearray(data) |
|
721 | 720 | |
|
722 | 721 | # The end state of self._raw is an array that is of the exact length |
|
723 | 722 | # required to hold a record for every revision in the repository. |
|
724 | 723 | # We truncate or extend the array as necessary. self._dirtyoffset is |
|
725 | 724 | # defined to be the start offset at which we need to write the output |
|
726 | 725 | # file. This offset is also adjusted when new entries are calculated |
|
727 | 726 | # for array members. |
|
728 | 727 | cllen = len(repo.changelog) |
|
729 | 728 | wantedlen = cllen * _fnodesrecsize |
|
730 | 729 | rawlen = len(self._raw) |
|
731 | 730 | |
|
732 | 731 | self._dirtyoffset = None |
|
733 | 732 | |
|
734 | 733 | rawlentokeep = min( |
|
735 | 734 | wantedlen, (rawlen // _fnodesrecsize) * _fnodesrecsize |
|
736 | 735 | ) |
|
737 | 736 | if rawlen > rawlentokeep: |
|
738 | 737 | # There's no easy way to truncate array instances. This seems |
|
739 | 738 | # slightly less evil than copying a potentially large array slice. |
|
740 | 739 | for i in range(rawlen - rawlentokeep): |
|
741 | 740 | self._raw.pop() |
|
742 | 741 | rawlen = len(self._raw) |
|
743 | 742 | self._dirtyoffset = rawlen |
|
744 | 743 | if rawlen < wantedlen: |
|
745 | 744 | if self._dirtyoffset is None: |
|
746 | 745 | self._dirtyoffset = rawlen |
|
747 | 746 | # TODO: zero fill entire record, because it's invalid not missing? |
|
748 | 747 | self._raw.extend(b'\xff' * (wantedlen - rawlen)) |
|
749 | 748 | |
|
750 | 749 | def getfnode(self, node, computemissing=True): |
|
751 | 750 | """Obtain the filenode of the .hgtags file at a specified revision. |
|
752 | 751 | |
|
753 | 752 | If the value is in the cache, the entry will be validated and returned. |
|
754 | 753 | Otherwise, the filenode will be computed and returned unless |
|
755 | 754 | "computemissing" is False. In that case, None will be returned if |
|
756 | 755 | the entry is missing or False if the entry is invalid without |
|
757 | 756 | any potentially expensive computation being performed. |
|
758 | 757 | |
|
759 | 758 | If an .hgtags does not exist at the specified revision, nullid is |
|
760 | 759 | returned. |
|
761 | 760 | """ |
|
762 | 761 | if node == self._repo.nullid: |
|
763 | 762 | return node |
|
764 | 763 | |
|
765 | 764 | rev = self._repo.changelog.rev(node) |
|
766 | 765 | |
|
767 | 766 | self.lookupcount += 1 |
|
768 | 767 | |
|
769 | 768 | offset = rev * _fnodesrecsize |
|
770 | 769 | record = b'%s' % self._raw[offset : offset + _fnodesrecsize] |
|
771 | 770 | properprefix = node[0:4] |
|
772 | 771 | |
|
773 | 772 | # Validate and return existing entry. |
|
774 | 773 | if record != _fnodesmissingrec and len(record) == _fnodesrecsize: |
|
775 | 774 | fileprefix = record[0:4] |
|
776 | 775 | |
|
777 | 776 | if fileprefix == properprefix: |
|
778 | 777 | self.hitcount += 1 |
|
779 | 778 | return record[4:] |
|
780 | 779 | |
|
781 | 780 | # Fall through. |
|
782 | 781 | |
|
783 | 782 | # If we get here, the entry is either missing or invalid. |
|
784 | 783 | |
|
785 | 784 | if not computemissing: |
|
786 | 785 | if record != _fnodesmissingrec: |
|
787 | 786 | return False |
|
788 | 787 | return None |
|
789 | 788 | |
|
790 | 789 | fnode = self._computefnode(node) |
|
791 | 790 | self._writeentry(offset, properprefix, fnode) |
|
792 | 791 | return fnode |
|
793 | 792 | |
|
794 | 793 | def _computefnode(self, node): |
|
795 | 794 | """Finds the tag filenode for a node which is missing or invalid |
|
796 | 795 | in cache""" |
|
797 | 796 | ctx = self._repo[node] |
|
798 | 797 | rev = ctx.rev() |
|
799 | 798 | fnode = None |
|
800 | 799 | cl = self._repo.changelog |
|
801 | 800 | p1rev, p2rev = cl._uncheckedparentrevs(rev) |
|
802 | 801 | p1node = cl.node(p1rev) |
|
803 | 802 | p1fnode = self.getfnode(p1node, computemissing=False) |
|
804 | 803 | if p2rev != nullrev: |
|
805 | 804 | # There is some no-merge changeset where p1 is null and p2 is set |
|
806 | 805 | # Processing them as merge is just slower, but still gives a good |
|
807 | 806 | # result. |
|
808 | 807 | p2node = cl.node(p2rev) |
|
809 | 808 | p2fnode = self.getfnode(p2node, computemissing=False) |
|
810 | 809 | if p1fnode != p2fnode: |
|
811 | 810 | # we cannot rely on readfast because we don't know against what |
|
812 | 811 | # parent the readfast delta is computed |
|
813 | 812 | p1fnode = None |
|
814 | 813 | if p1fnode: |
|
815 | 814 | mctx = ctx.manifestctx() |
|
816 | 815 | fnode = mctx.readfast().get(b'.hgtags') |
|
817 | 816 | if fnode is None: |
|
818 | 817 | fnode = p1fnode |
|
819 | 818 | if fnode is None: |
|
820 | 819 | # Populate missing entry. |
|
821 | 820 | try: |
|
822 | 821 | fnode = ctx.filenode(b'.hgtags') |
|
823 | 822 | except error.LookupError: |
|
824 | 823 | # No .hgtags file on this revision. |
|
825 | 824 | fnode = self._repo.nullid |
|
826 | 825 | return fnode |
|
827 | 826 | |
|
828 | 827 | def setfnode(self, node, fnode): |
|
829 | 828 | """Set the .hgtags filenode for a given changeset.""" |
|
830 | 829 | assert len(fnode) == 20 |
|
831 | 830 | ctx = self._repo[node] |
|
832 | 831 | |
|
833 | 832 | # Do a lookup first to avoid writing if nothing has changed. |
|
834 | 833 | if self.getfnode(ctx.node(), computemissing=False) == fnode: |
|
835 | 834 | return |
|
836 | 835 | |
|
837 | 836 | self._writeentry(ctx.rev() * _fnodesrecsize, node[0:4], fnode) |
|
838 | 837 | |
|
839 | 838 | def refresh_invalid_nodes(self, nodes): |
|
840 | 839 | """recomputes file nodes for a given set of nodes which has unknown |
|
841 | 840 | filenodes for them in the cache |
|
842 | 841 | Also updates the in-memory cache with the correct filenode. |
|
843 | 842 | Caller needs to take care about calling `.write()` so that updates are |
|
844 | 843 | persisted. |
|
845 | 844 | Returns a map {node: recomputed fnode} |
|
846 | 845 | """ |
|
847 | 846 | fixed_nodemap = {} |
|
848 | 847 | for node in nodes: |
|
849 | 848 | fnode = self._computefnode(node) |
|
850 | 849 | fixed_nodemap[node] = fnode |
|
851 | 850 | self.setfnode(node, fnode) |
|
852 | 851 | return fixed_nodemap |
|
853 | 852 | |
|
854 | 853 | def _writeentry(self, offset, prefix, fnode): |
|
855 | 854 | # Slices on array instances only accept other array. |
|
856 | 855 | entry = bytearray(prefix + fnode) |
|
857 | 856 | self._raw[offset : offset + _fnodesrecsize] = entry |
|
858 | 857 | # self._dirtyoffset could be None. |
|
859 | 858 | self._dirtyoffset = min(self._dirtyoffset or 0, offset or 0) |
|
860 | 859 | |
|
861 | 860 | def write(self): |
|
862 | 861 | """Perform all necessary writes to cache file. |
|
863 | 862 | |
|
864 | 863 | This may no-op if no writes are needed or if a write lock could |
|
865 | 864 | not be obtained. |
|
866 | 865 | """ |
|
867 | 866 | if self._dirtyoffset is None: |
|
868 | 867 | return |
|
869 | 868 | |
|
870 | 869 | data = self._raw[self._dirtyoffset :] |
|
871 | 870 | if not data: |
|
872 | 871 | return |
|
873 | 872 | |
|
874 | 873 | repo = self._repo |
|
875 | 874 | |
|
876 | 875 | try: |
|
877 | 876 | lock = repo.lock(wait=False) |
|
878 | 877 | except error.LockError: |
|
879 | 878 | repo.ui.log( |
|
880 | 879 | b'tagscache', |
|
881 | 880 | b'not writing .hg/cache/%s because ' |
|
882 | 881 | b'lock cannot be acquired\n' % _fnodescachefile, |
|
883 | 882 | ) |
|
884 | 883 | return |
|
885 | 884 | |
|
886 | 885 | try: |
|
887 | 886 | f = repo.cachevfs.open(_fnodescachefile, b'ab') |
|
888 | 887 | try: |
|
889 | 888 | # if the file has been truncated |
|
890 | 889 | actualoffset = f.tell() |
|
891 | 890 | if actualoffset < self._dirtyoffset: |
|
892 | 891 | self._dirtyoffset = actualoffset |
|
893 | 892 | data = self._raw[self._dirtyoffset :] |
|
894 | 893 | f.seek(self._dirtyoffset) |
|
895 | 894 | f.truncate() |
|
896 | 895 | repo.ui.log( |
|
897 | 896 | b'tagscache', |
|
898 | 897 | b'writing %d bytes to cache/%s\n' |
|
899 | 898 | % (len(data), _fnodescachefile), |
|
900 | 899 | ) |
|
901 | 900 | f.write(data) |
|
902 | 901 | self._dirtyoffset = None |
|
903 | 902 | finally: |
|
904 | 903 | f.close() |
|
905 | 904 | except (IOError, OSError) as inst: |
|
906 | 905 | repo.ui.log( |
|
907 | 906 | b'tagscache', |
|
908 | 907 | b"couldn't write cache/%s: %s\n" |
|
909 | 908 | % (_fnodescachefile, stringutil.forcebytestr(inst)), |
|
910 | 909 | ) |
|
911 | 910 | finally: |
|
912 | 911 | lock.release() |
|
913 | 912 | |
|
914 | 913 | |
|
915 | 914 | def clear_cache_on_disk(repo): |
|
916 | 915 | """function used by the perf extension to "tags" cache""" |
|
917 | 916 | repo.cachevfs.tryunlink(_filename(repo)) |
|
918 | 917 | |
|
919 | 918 | |
|
920 | 919 | def clear_cache_fnodes(repo): |
|
921 | 920 | """function used by the perf extension to clear "file node cache""" |
|
922 | 921 | repo.cachevfs.tryunlink(_filename(repo)) |
|
923 | 922 | |
|
924 | 923 | |
|
925 | 924 | def forget_fnodes(repo, revs): |
|
926 | 925 | """function used by the perf extension to prune some entries from the fnodes |
|
927 | 926 | cache""" |
|
928 | 927 | missing_1 = b'\xff' * 4 |
|
929 | 928 | missing_2 = b'\xff' * 20 |
|
930 | 929 | cache = hgtagsfnodescache(repo.unfiltered()) |
|
931 | 930 | for r in revs: |
|
932 | 931 | cache._writeentry(r * _fnodesrecsize, missing_1, missing_2) |
|
933 | 932 | cache.write() |
General Comments 0
You need to be logged in to leave comments.
Login now