Show More
@@ -1,771 +1,785 b'' | |||
|
1 | 1 | # obsutil.py - utility functions for obsolescence |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2017 Boris Feld <boris.feld@octobus.net> |
|
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 re |
|
11 | 11 | |
|
12 | 12 | from . import ( |
|
13 | 13 | phases, |
|
14 | 14 | util |
|
15 | 15 | ) |
|
16 | 16 | |
|
17 | 17 | class marker(object): |
|
18 | 18 | """Wrap obsolete marker raw data""" |
|
19 | 19 | |
|
20 | 20 | def __init__(self, repo, data): |
|
21 | 21 | # the repo argument will be used to create changectx in later version |
|
22 | 22 | self._repo = repo |
|
23 | 23 | self._data = data |
|
24 | 24 | self._decodedmeta = None |
|
25 | 25 | |
|
26 | 26 | def __hash__(self): |
|
27 | 27 | return hash(self._data) |
|
28 | 28 | |
|
29 | 29 | def __eq__(self, other): |
|
30 | 30 | if type(other) != type(self): |
|
31 | 31 | return False |
|
32 | 32 | return self._data == other._data |
|
33 | 33 | |
|
34 | 34 | def precnode(self): |
|
35 | 35 | msg = ("'marker.precnode' is deprecated, " |
|
36 | 36 | "use 'marker.prednode'") |
|
37 | 37 | util.nouideprecwarn(msg, '4.4') |
|
38 | 38 | return self.prednode() |
|
39 | 39 | |
|
40 | 40 | def prednode(self): |
|
41 | 41 | """Predecessor changeset node identifier""" |
|
42 | 42 | return self._data[0] |
|
43 | 43 | |
|
44 | 44 | def succnodes(self): |
|
45 | 45 | """List of successor changesets node identifiers""" |
|
46 | 46 | return self._data[1] |
|
47 | 47 | |
|
48 | 48 | def parentnodes(self): |
|
49 | 49 | """Parents of the predecessors (None if not recorded)""" |
|
50 | 50 | return self._data[5] |
|
51 | 51 | |
|
52 | 52 | def metadata(self): |
|
53 | 53 | """Decoded metadata dictionary""" |
|
54 | 54 | return dict(self._data[3]) |
|
55 | 55 | |
|
56 | 56 | def date(self): |
|
57 | 57 | """Creation date as (unixtime, offset)""" |
|
58 | 58 | return self._data[4] |
|
59 | 59 | |
|
60 | 60 | def flags(self): |
|
61 | 61 | """The flags field of the marker""" |
|
62 | 62 | return self._data[2] |
|
63 | 63 | |
|
64 | 64 | def getmarkers(repo, nodes=None, exclusive=False): |
|
65 | 65 | """returns markers known in a repository |
|
66 | 66 | |
|
67 | 67 | If <nodes> is specified, only markers "relevant" to those nodes are are |
|
68 | 68 | returned""" |
|
69 | 69 | if nodes is None: |
|
70 | 70 | rawmarkers = repo.obsstore |
|
71 | 71 | elif exclusive: |
|
72 | 72 | rawmarkers = exclusivemarkers(repo, nodes) |
|
73 | 73 | else: |
|
74 | 74 | rawmarkers = repo.obsstore.relevantmarkers(nodes) |
|
75 | 75 | |
|
76 | 76 | for markerdata in rawmarkers: |
|
77 | 77 | yield marker(repo, markerdata) |
|
78 | 78 | |
|
79 | 79 | def closestpredecessors(repo, nodeid): |
|
80 | 80 | """yield the list of next predecessors pointing on visible changectx nodes |
|
81 | 81 | |
|
82 | 82 | This function respect the repoview filtering, filtered revision will be |
|
83 | 83 | considered missing. |
|
84 | 84 | """ |
|
85 | 85 | |
|
86 | 86 | precursors = repo.obsstore.predecessors |
|
87 | 87 | stack = [nodeid] |
|
88 | 88 | seen = set(stack) |
|
89 | 89 | |
|
90 | 90 | while stack: |
|
91 | 91 | current = stack.pop() |
|
92 | 92 | currentpreccs = precursors.get(current, ()) |
|
93 | 93 | |
|
94 | 94 | for prec in currentpreccs: |
|
95 | 95 | precnodeid = prec[0] |
|
96 | 96 | |
|
97 | 97 | # Basic cycle protection |
|
98 | 98 | if precnodeid in seen: |
|
99 | 99 | continue |
|
100 | 100 | seen.add(precnodeid) |
|
101 | 101 | |
|
102 | 102 | if precnodeid in repo: |
|
103 | 103 | yield precnodeid |
|
104 | 104 | else: |
|
105 | 105 | stack.append(precnodeid) |
|
106 | 106 | |
|
107 | 107 | def allprecursors(*args, **kwargs): |
|
108 | 108 | """ (DEPRECATED) |
|
109 | 109 | """ |
|
110 | 110 | msg = ("'obsutil.allprecursors' is deprecated, " |
|
111 | 111 | "use 'obsutil.allpredecessors'") |
|
112 | 112 | util.nouideprecwarn(msg, '4.4') |
|
113 | 113 | |
|
114 | 114 | return allpredecessors(*args, **kwargs) |
|
115 | 115 | |
|
116 | 116 | def allpredecessors(obsstore, nodes, ignoreflags=0): |
|
117 | 117 | """Yield node for every precursors of <nodes>. |
|
118 | 118 | |
|
119 | 119 | Some precursors may be unknown locally. |
|
120 | 120 | |
|
121 | 121 | This is a linear yield unsuited to detecting folded changesets. It includes |
|
122 | 122 | initial nodes too.""" |
|
123 | 123 | |
|
124 | 124 | remaining = set(nodes) |
|
125 | 125 | seen = set(remaining) |
|
126 | 126 | while remaining: |
|
127 | 127 | current = remaining.pop() |
|
128 | 128 | yield current |
|
129 | 129 | for mark in obsstore.predecessors.get(current, ()): |
|
130 | 130 | # ignore marker flagged with specified flag |
|
131 | 131 | if mark[2] & ignoreflags: |
|
132 | 132 | continue |
|
133 | 133 | suc = mark[0] |
|
134 | 134 | if suc not in seen: |
|
135 | 135 | seen.add(suc) |
|
136 | 136 | remaining.add(suc) |
|
137 | 137 | |
|
138 | 138 | def allsuccessors(obsstore, nodes, ignoreflags=0): |
|
139 | 139 | """Yield node for every successor of <nodes>. |
|
140 | 140 | |
|
141 | 141 | Some successors may be unknown locally. |
|
142 | 142 | |
|
143 | 143 | This is a linear yield unsuited to detecting split changesets. It includes |
|
144 | 144 | initial nodes too.""" |
|
145 | 145 | remaining = set(nodes) |
|
146 | 146 | seen = set(remaining) |
|
147 | 147 | while remaining: |
|
148 | 148 | current = remaining.pop() |
|
149 | 149 | yield current |
|
150 | 150 | for mark in obsstore.successors.get(current, ()): |
|
151 | 151 | # ignore marker flagged with specified flag |
|
152 | 152 | if mark[2] & ignoreflags: |
|
153 | 153 | continue |
|
154 | 154 | for suc in mark[1]: |
|
155 | 155 | if suc not in seen: |
|
156 | 156 | seen.add(suc) |
|
157 | 157 | remaining.add(suc) |
|
158 | 158 | |
|
159 | 159 | def _filterprunes(markers): |
|
160 | 160 | """return a set with no prune markers""" |
|
161 | 161 | return set(m for m in markers if m[1]) |
|
162 | 162 | |
|
163 | 163 | def exclusivemarkers(repo, nodes): |
|
164 | 164 | """set of markers relevant to "nodes" but no other locally-known nodes |
|
165 | 165 | |
|
166 | 166 | This function compute the set of markers "exclusive" to a locally-known |
|
167 | 167 | node. This means we walk the markers starting from <nodes> until we reach a |
|
168 | 168 | locally-known precursors outside of <nodes>. Element of <nodes> with |
|
169 | 169 | locally-known successors outside of <nodes> are ignored (since their |
|
170 | 170 | precursors markers are also relevant to these successors). |
|
171 | 171 | |
|
172 | 172 | For example: |
|
173 | 173 | |
|
174 | 174 | # (A0 rewritten as A1) |
|
175 | 175 | # |
|
176 | 176 | # A0 <-1- A1 # Marker "1" is exclusive to A1 |
|
177 | 177 | |
|
178 | 178 | or |
|
179 | 179 | |
|
180 | 180 | # (A0 rewritten as AX; AX rewritten as A1; AX is unkown locally) |
|
181 | 181 | # |
|
182 | 182 | # <-1- A0 <-2- AX <-3- A1 # Marker "2,3" are exclusive to A1 |
|
183 | 183 | |
|
184 | 184 | or |
|
185 | 185 | |
|
186 | 186 | # (A0 has unknown precursors, A0 rewritten as A1 and A2 (divergence)) |
|
187 | 187 | # |
|
188 | 188 | # <-2- A1 # Marker "2" is exclusive to A0,A1 |
|
189 | 189 | # / |
|
190 | 190 | # <-1- A0 |
|
191 | 191 | # \ |
|
192 | 192 | # <-3- A2 # Marker "3" is exclusive to A0,A2 |
|
193 | 193 | # |
|
194 | 194 | # in addition: |
|
195 | 195 | # |
|
196 | 196 | # Markers "2,3" are exclusive to A1,A2 |
|
197 | 197 | # Markers "1,2,3" are exclusive to A0,A1,A2 |
|
198 | 198 | |
|
199 | 199 | See test/test-obsolete-bundle-strip.t for more examples. |
|
200 | 200 | |
|
201 | 201 | An example usage is strip. When stripping a changeset, we also want to |
|
202 | 202 | strip the markers exclusive to this changeset. Otherwise we would have |
|
203 | 203 | "dangling"" obsolescence markers from its precursors: Obsolescence markers |
|
204 | 204 | marking a node as obsolete without any successors available locally. |
|
205 | 205 | |
|
206 | 206 | As for relevant markers, the prune markers for children will be followed. |
|
207 | 207 | Of course, they will only be followed if the pruned children is |
|
208 | 208 | locally-known. Since the prune markers are relevant to the pruned node. |
|
209 | 209 | However, while prune markers are considered relevant to the parent of the |
|
210 | 210 | pruned changesets, prune markers for locally-known changeset (with no |
|
211 | 211 | successors) are considered exclusive to the pruned nodes. This allows |
|
212 | 212 | to strip the prune markers (with the rest of the exclusive chain) alongside |
|
213 | 213 | the pruned changesets. |
|
214 | 214 | """ |
|
215 | 215 | # running on a filtered repository would be dangerous as markers could be |
|
216 | 216 | # reported as exclusive when they are relevant for other filtered nodes. |
|
217 | 217 | unfi = repo.unfiltered() |
|
218 | 218 | |
|
219 | 219 | # shortcut to various useful item |
|
220 | 220 | nm = unfi.changelog.nodemap |
|
221 | 221 | precursorsmarkers = unfi.obsstore.predecessors |
|
222 | 222 | successormarkers = unfi.obsstore.successors |
|
223 | 223 | childrenmarkers = unfi.obsstore.children |
|
224 | 224 | |
|
225 | 225 | # exclusive markers (return of the function) |
|
226 | 226 | exclmarkers = set() |
|
227 | 227 | # we need fast membership testing |
|
228 | 228 | nodes = set(nodes) |
|
229 | 229 | # looking for head in the obshistory |
|
230 | 230 | # |
|
231 | 231 | # XXX we are ignoring all issues in regard with cycle for now. |
|
232 | 232 | stack = [n for n in nodes if not _filterprunes(successormarkers.get(n, ()))] |
|
233 | 233 | stack.sort() |
|
234 | 234 | # nodes already stacked |
|
235 | 235 | seennodes = set(stack) |
|
236 | 236 | while stack: |
|
237 | 237 | current = stack.pop() |
|
238 | 238 | # fetch precursors markers |
|
239 | 239 | markers = list(precursorsmarkers.get(current, ())) |
|
240 | 240 | # extend the list with prune markers |
|
241 | 241 | for mark in successormarkers.get(current, ()): |
|
242 | 242 | if not mark[1]: |
|
243 | 243 | markers.append(mark) |
|
244 | 244 | # and markers from children (looking for prune) |
|
245 | 245 | for mark in childrenmarkers.get(current, ()): |
|
246 | 246 | if not mark[1]: |
|
247 | 247 | markers.append(mark) |
|
248 | 248 | # traverse the markers |
|
249 | 249 | for mark in markers: |
|
250 | 250 | if mark in exclmarkers: |
|
251 | 251 | # markers already selected |
|
252 | 252 | continue |
|
253 | 253 | |
|
254 | 254 | # If the markers is about the current node, select it |
|
255 | 255 | # |
|
256 | 256 | # (this delay the addition of markers from children) |
|
257 | 257 | if mark[1] or mark[0] == current: |
|
258 | 258 | exclmarkers.add(mark) |
|
259 | 259 | |
|
260 | 260 | # should we keep traversing through the precursors? |
|
261 | 261 | prec = mark[0] |
|
262 | 262 | |
|
263 | 263 | # nodes in the stack or already processed |
|
264 | 264 | if prec in seennodes: |
|
265 | 265 | continue |
|
266 | 266 | |
|
267 | 267 | # is this a locally known node ? |
|
268 | 268 | known = prec in nm |
|
269 | 269 | # if locally-known and not in the <nodes> set the traversal |
|
270 | 270 | # stop here. |
|
271 | 271 | if known and prec not in nodes: |
|
272 | 272 | continue |
|
273 | 273 | |
|
274 | 274 | # do not keep going if there are unselected markers pointing to this |
|
275 | 275 | # nodes. If we end up traversing these unselected markers later the |
|
276 | 276 | # node will be taken care of at that point. |
|
277 | 277 | precmarkers = _filterprunes(successormarkers.get(prec)) |
|
278 | 278 | if precmarkers.issubset(exclmarkers): |
|
279 | 279 | seennodes.add(prec) |
|
280 | 280 | stack.append(prec) |
|
281 | 281 | |
|
282 | 282 | return exclmarkers |
|
283 | 283 | |
|
284 | 284 | def foreground(repo, nodes): |
|
285 | 285 | """return all nodes in the "foreground" of other node |
|
286 | 286 | |
|
287 | 287 | The foreground of a revision is anything reachable using parent -> children |
|
288 | 288 | or precursor -> successor relation. It is very similar to "descendant" but |
|
289 | 289 | augmented with obsolescence information. |
|
290 | 290 | |
|
291 | 291 | Beware that possible obsolescence cycle may result if complex situation. |
|
292 | 292 | """ |
|
293 | 293 | repo = repo.unfiltered() |
|
294 | 294 | foreground = set(repo.set('%ln::', nodes)) |
|
295 | 295 | if repo.obsstore: |
|
296 | 296 | # We only need this complicated logic if there is obsolescence |
|
297 | 297 | # XXX will probably deserve an optimised revset. |
|
298 | 298 | nm = repo.changelog.nodemap |
|
299 | 299 | plen = -1 |
|
300 | 300 | # compute the whole set of successors or descendants |
|
301 | 301 | while len(foreground) != plen: |
|
302 | 302 | plen = len(foreground) |
|
303 | 303 | succs = set(c.node() for c in foreground) |
|
304 | 304 | mutable = [c.node() for c in foreground if c.mutable()] |
|
305 | 305 | succs.update(allsuccessors(repo.obsstore, mutable)) |
|
306 | 306 | known = (n for n in succs if n in nm) |
|
307 | 307 | foreground = set(repo.set('%ln::', known)) |
|
308 | 308 | return set(c.node() for c in foreground) |
|
309 | 309 | |
|
310 | # logic around storing and using effect flags | |
|
310 | # effectflag field | |
|
311 | # | |
|
312 | # Effect-flag is a 1-byte bit field used to store what changed between a | |
|
313 | # changeset and its successor(s). | |
|
314 | # | |
|
315 | # The effect flag is stored in obs-markers metadata while we iterate on the | |
|
316 | # information design. That's why we have the EFFECTFLAGFIELD. If we come up | |
|
317 | # with an incompatible design for effect flag, we can store a new design under | |
|
318 | # another field name so we don't break readers. We plan to extend the existing | |
|
319 | # obsmarkers bit-field when the effect flag design will be stabilized. | |
|
320 | # | |
|
321 | # The effect-flag is placed behind an experimental flag | |
|
322 | # `effect-flags` set to off by default. | |
|
323 | # | |
|
324 | ||
|
311 | 325 | EFFECTFLAGFIELD = "ef1" |
|
312 | 326 | |
|
313 | 327 | DESCCHANGED = 1 << 0 # action changed the description |
|
314 | 328 | METACHANGED = 1 << 1 # action change the meta |
|
315 | 329 | DIFFCHANGED = 1 << 3 # action change diff introduced by the changeset |
|
316 | 330 | PARENTCHANGED = 1 << 2 # action change the parent |
|
317 | 331 | USERCHANGED = 1 << 4 # the user changed |
|
318 | 332 | DATECHANGED = 1 << 5 # the date changed |
|
319 | 333 | BRANCHCHANGED = 1 << 6 # the branch changed |
|
320 | 334 | |
|
321 | 335 | METABLACKLIST = [ |
|
322 | 336 | re.compile('^branch$'), |
|
323 | 337 | re.compile('^.*-source$'), |
|
324 | 338 | re.compile('^.*_source$'), |
|
325 | 339 | re.compile('^source$'), |
|
326 | 340 | ] |
|
327 | 341 | |
|
328 | 342 | def metanotblacklisted(metaitem): |
|
329 | 343 | """ Check that the key of a meta item (extrakey, extravalue) does not |
|
330 | 344 | match at least one of the blacklist pattern |
|
331 | 345 | """ |
|
332 | 346 | metakey = metaitem[0] |
|
333 | 347 | |
|
334 | 348 | return not any(pattern.match(metakey) for pattern in METABLACKLIST) |
|
335 | 349 | |
|
336 | 350 | def _prepare_hunk(hunk): |
|
337 | 351 | """Drop all information but the username and patch""" |
|
338 | 352 | cleanhunk = [] |
|
339 | 353 | for line in hunk.splitlines(): |
|
340 | 354 | if line.startswith(b'# User') or not line.startswith(b'#'): |
|
341 | 355 | if line.startswith(b'@@'): |
|
342 | 356 | line = b'@@\n' |
|
343 | 357 | cleanhunk.append(line) |
|
344 | 358 | return cleanhunk |
|
345 | 359 | |
|
346 | 360 | def _getdifflines(iterdiff): |
|
347 | 361 | """return a cleaned up lines""" |
|
348 | 362 | lines = next(iterdiff, None) |
|
349 | 363 | |
|
350 | 364 | if lines is None: |
|
351 | 365 | return lines |
|
352 | 366 | |
|
353 | 367 | return _prepare_hunk(lines) |
|
354 | 368 | |
|
355 | 369 | def _cmpdiff(leftctx, rightctx): |
|
356 | 370 | """return True if both ctx introduce the "same diff" |
|
357 | 371 | |
|
358 | 372 | This is a first and basic implementation, with many shortcoming. |
|
359 | 373 | """ |
|
360 | 374 | |
|
361 | 375 | # Leftctx or right ctx might be filtered, so we need to use the contexts |
|
362 | 376 | # with an unfiltered repository to safely compute the diff |
|
363 | 377 | leftunfi = leftctx._repo.unfiltered()[leftctx.rev()] |
|
364 | 378 | leftdiff = leftunfi.diff(git=1) |
|
365 | 379 | rightunfi = rightctx._repo.unfiltered()[rightctx.rev()] |
|
366 | 380 | rightdiff = rightunfi.diff(git=1) |
|
367 | 381 | |
|
368 | 382 | left, right = (0, 0) |
|
369 | 383 | while None not in (left, right): |
|
370 | 384 | left = _getdifflines(leftdiff) |
|
371 | 385 | right = _getdifflines(rightdiff) |
|
372 | 386 | |
|
373 | 387 | if left != right: |
|
374 | 388 | return False |
|
375 | 389 | return True |
|
376 | 390 | |
|
377 | 391 | def geteffectflag(relation): |
|
378 | 392 | """ From an obs-marker relation, compute what changed between the |
|
379 | 393 | predecessor and the successor. |
|
380 | 394 | """ |
|
381 | 395 | effects = 0 |
|
382 | 396 | |
|
383 | 397 | source = relation[0] |
|
384 | 398 | |
|
385 | 399 | for changectx in relation[1]: |
|
386 | 400 | # Check if description has changed |
|
387 | 401 | if changectx.description() != source.description(): |
|
388 | 402 | effects |= DESCCHANGED |
|
389 | 403 | |
|
390 | 404 | # Check if user has changed |
|
391 | 405 | if changectx.user() != source.user(): |
|
392 | 406 | effects |= USERCHANGED |
|
393 | 407 | |
|
394 | 408 | # Check if date has changed |
|
395 | 409 | if changectx.date() != source.date(): |
|
396 | 410 | effects |= DATECHANGED |
|
397 | 411 | |
|
398 | 412 | # Check if branch has changed |
|
399 | 413 | if changectx.branch() != source.branch(): |
|
400 | 414 | effects |= BRANCHCHANGED |
|
401 | 415 | |
|
402 | 416 | # Check if at least one of the parent has changed |
|
403 | 417 | if changectx.parents() != source.parents(): |
|
404 | 418 | effects |= PARENTCHANGED |
|
405 | 419 | |
|
406 | 420 | # Check if other meta has changed |
|
407 | 421 | changeextra = changectx.extra().items() |
|
408 | 422 | ctxmeta = filter(metanotblacklisted, changeextra) |
|
409 | 423 | |
|
410 | 424 | sourceextra = source.extra().items() |
|
411 | 425 | srcmeta = filter(metanotblacklisted, sourceextra) |
|
412 | 426 | |
|
413 | 427 | if ctxmeta != srcmeta: |
|
414 | 428 | effects |= METACHANGED |
|
415 | 429 | |
|
416 | 430 | # Check if the diff has changed |
|
417 | 431 | if not _cmpdiff(source, changectx): |
|
418 | 432 | effects |= DIFFCHANGED |
|
419 | 433 | |
|
420 | 434 | return effects |
|
421 | 435 | |
|
422 | 436 | def getobsoleted(repo, tr): |
|
423 | 437 | """return the set of pre-existing revisions obsoleted by a transaction""" |
|
424 | 438 | torev = repo.unfiltered().changelog.nodemap.get |
|
425 | 439 | phase = repo._phasecache.phase |
|
426 | 440 | succsmarkers = repo.obsstore.successors.get |
|
427 | 441 | public = phases.public |
|
428 | 442 | addedmarkers = tr.changes.get('obsmarkers') |
|
429 | 443 | addedrevs = tr.changes.get('revs') |
|
430 | 444 | seenrevs = set(addedrevs) |
|
431 | 445 | obsoleted = set() |
|
432 | 446 | for mark in addedmarkers: |
|
433 | 447 | node = mark[0] |
|
434 | 448 | rev = torev(node) |
|
435 | 449 | if rev is None or rev in seenrevs: |
|
436 | 450 | continue |
|
437 | 451 | seenrevs.add(rev) |
|
438 | 452 | if phase(repo, rev) == public: |
|
439 | 453 | continue |
|
440 | 454 | if set(succsmarkers(node) or []).issubset(addedmarkers): |
|
441 | 455 | obsoleted.add(rev) |
|
442 | 456 | return obsoleted |
|
443 | 457 | |
|
444 | 458 | class _succs(list): |
|
445 | 459 | """small class to represent a successors with some metadata about it""" |
|
446 | 460 | |
|
447 | 461 | def __init__(self, *args, **kwargs): |
|
448 | 462 | super(_succs, self).__init__(*args, **kwargs) |
|
449 | 463 | self.markers = set() |
|
450 | 464 | |
|
451 | 465 | def copy(self): |
|
452 | 466 | new = _succs(self) |
|
453 | 467 | new.markers = self.markers.copy() |
|
454 | 468 | return new |
|
455 | 469 | |
|
456 | 470 | @util.propertycache |
|
457 | 471 | def _set(self): |
|
458 | 472 | # immutable |
|
459 | 473 | return set(self) |
|
460 | 474 | |
|
461 | 475 | def canmerge(self, other): |
|
462 | 476 | return self._set.issubset(other._set) |
|
463 | 477 | |
|
464 | 478 | def successorssets(repo, initialnode, closest=False, cache=None): |
|
465 | 479 | """Return set of all latest successors of initial nodes |
|
466 | 480 | |
|
467 | 481 | The successors set of a changeset A are the group of revisions that succeed |
|
468 | 482 | A. It succeeds A as a consistent whole, each revision being only a partial |
|
469 | 483 | replacement. By default, the successors set contains non-obsolete |
|
470 | 484 | changesets only, walking the obsolescence graph until reaching a leaf. If |
|
471 | 485 | 'closest' is set to True, closest successors-sets are return (the |
|
472 | 486 | obsolescence walk stops on known changesets). |
|
473 | 487 | |
|
474 | 488 | This function returns the full list of successor sets which is why it |
|
475 | 489 | returns a list of tuples and not just a single tuple. Each tuple is a valid |
|
476 | 490 | successors set. Note that (A,) may be a valid successors set for changeset A |
|
477 | 491 | (see below). |
|
478 | 492 | |
|
479 | 493 | In most cases, a changeset A will have a single element (e.g. the changeset |
|
480 | 494 | A is replaced by A') in its successors set. Though, it is also common for a |
|
481 | 495 | changeset A to have no elements in its successor set (e.g. the changeset |
|
482 | 496 | has been pruned). Therefore, the returned list of successors sets will be |
|
483 | 497 | [(A',)] or [], respectively. |
|
484 | 498 | |
|
485 | 499 | When a changeset A is split into A' and B', however, it will result in a |
|
486 | 500 | successors set containing more than a single element, i.e. [(A',B')]. |
|
487 | 501 | Divergent changesets will result in multiple successors sets, i.e. [(A',), |
|
488 | 502 | (A'')]. |
|
489 | 503 | |
|
490 | 504 | If a changeset A is not obsolete, then it will conceptually have no |
|
491 | 505 | successors set. To distinguish this from a pruned changeset, the successor |
|
492 | 506 | set will contain itself only, i.e. [(A,)]. |
|
493 | 507 | |
|
494 | 508 | Finally, final successors unknown locally are considered to be pruned |
|
495 | 509 | (pruned: obsoleted without any successors). (Final: successors not affected |
|
496 | 510 | by markers). |
|
497 | 511 | |
|
498 | 512 | The 'closest' mode respect the repoview filtering. For example, without |
|
499 | 513 | filter it will stop at the first locally known changeset, with 'visible' |
|
500 | 514 | filter it will stop on visible changesets). |
|
501 | 515 | |
|
502 | 516 | The optional `cache` parameter is a dictionary that may contains |
|
503 | 517 | precomputed successors sets. It is meant to reuse the computation of a |
|
504 | 518 | previous call to `successorssets` when multiple calls are made at the same |
|
505 | 519 | time. The cache dictionary is updated in place. The caller is responsible |
|
506 | 520 | for its life span. Code that makes multiple calls to `successorssets` |
|
507 | 521 | *should* use this cache mechanism or risk a performance hit. |
|
508 | 522 | |
|
509 | 523 | Since results are different depending of the 'closest' most, the same cache |
|
510 | 524 | cannot be reused for both mode. |
|
511 | 525 | """ |
|
512 | 526 | |
|
513 | 527 | succmarkers = repo.obsstore.successors |
|
514 | 528 | |
|
515 | 529 | # Stack of nodes we search successors sets for |
|
516 | 530 | toproceed = [initialnode] |
|
517 | 531 | # set version of above list for fast loop detection |
|
518 | 532 | # element added to "toproceed" must be added here |
|
519 | 533 | stackedset = set(toproceed) |
|
520 | 534 | if cache is None: |
|
521 | 535 | cache = {} |
|
522 | 536 | |
|
523 | 537 | # This while loop is the flattened version of a recursive search for |
|
524 | 538 | # successors sets |
|
525 | 539 | # |
|
526 | 540 | # def successorssets(x): |
|
527 | 541 | # successors = directsuccessors(x) |
|
528 | 542 | # ss = [[]] |
|
529 | 543 | # for succ in directsuccessors(x): |
|
530 | 544 | # # product as in itertools cartesian product |
|
531 | 545 | # ss = product(ss, successorssets(succ)) |
|
532 | 546 | # return ss |
|
533 | 547 | # |
|
534 | 548 | # But we can not use plain recursive calls here: |
|
535 | 549 | # - that would blow the python call stack |
|
536 | 550 | # - obsolescence markers may have cycles, we need to handle them. |
|
537 | 551 | # |
|
538 | 552 | # The `toproceed` list act as our call stack. Every node we search |
|
539 | 553 | # successors set for are stacked there. |
|
540 | 554 | # |
|
541 | 555 | # The `stackedset` is set version of this stack used to check if a node is |
|
542 | 556 | # already stacked. This check is used to detect cycles and prevent infinite |
|
543 | 557 | # loop. |
|
544 | 558 | # |
|
545 | 559 | # successors set of all nodes are stored in the `cache` dictionary. |
|
546 | 560 | # |
|
547 | 561 | # After this while loop ends we use the cache to return the successors sets |
|
548 | 562 | # for the node requested by the caller. |
|
549 | 563 | while toproceed: |
|
550 | 564 | # Every iteration tries to compute the successors sets of the topmost |
|
551 | 565 | # node of the stack: CURRENT. |
|
552 | 566 | # |
|
553 | 567 | # There are four possible outcomes: |
|
554 | 568 | # |
|
555 | 569 | # 1) We already know the successors sets of CURRENT: |
|
556 | 570 | # -> mission accomplished, pop it from the stack. |
|
557 | 571 | # 2) Stop the walk: |
|
558 | 572 | # default case: Node is not obsolete |
|
559 | 573 | # closest case: Node is known at this repo filter level |
|
560 | 574 | # -> the node is its own successors sets. Add it to the cache. |
|
561 | 575 | # 3) We do not know successors set of direct successors of CURRENT: |
|
562 | 576 | # -> We add those successors to the stack. |
|
563 | 577 | # 4) We know successors sets of all direct successors of CURRENT: |
|
564 | 578 | # -> We can compute CURRENT successors set and add it to the |
|
565 | 579 | # cache. |
|
566 | 580 | # |
|
567 | 581 | current = toproceed[-1] |
|
568 | 582 | |
|
569 | 583 | # case 2 condition is a bit hairy because of closest, |
|
570 | 584 | # we compute it on its own |
|
571 | 585 | case2condition = ((current not in succmarkers) |
|
572 | 586 | or (closest and current != initialnode |
|
573 | 587 | and current in repo)) |
|
574 | 588 | |
|
575 | 589 | if current in cache: |
|
576 | 590 | # case (1): We already know the successors sets |
|
577 | 591 | stackedset.remove(toproceed.pop()) |
|
578 | 592 | elif case2condition: |
|
579 | 593 | # case (2): end of walk. |
|
580 | 594 | if current in repo: |
|
581 | 595 | # We have a valid successors. |
|
582 | 596 | cache[current] = [_succs((current,))] |
|
583 | 597 | else: |
|
584 | 598 | # Final obsolete version is unknown locally. |
|
585 | 599 | # Do not count that as a valid successors |
|
586 | 600 | cache[current] = [] |
|
587 | 601 | else: |
|
588 | 602 | # cases (3) and (4) |
|
589 | 603 | # |
|
590 | 604 | # We proceed in two phases. Phase 1 aims to distinguish case (3) |
|
591 | 605 | # from case (4): |
|
592 | 606 | # |
|
593 | 607 | # For each direct successors of CURRENT, we check whether its |
|
594 | 608 | # successors sets are known. If they are not, we stack the |
|
595 | 609 | # unknown node and proceed to the next iteration of the while |
|
596 | 610 | # loop. (case 3) |
|
597 | 611 | # |
|
598 | 612 | # During this step, we may detect obsolescence cycles: a node |
|
599 | 613 | # with unknown successors sets but already in the call stack. |
|
600 | 614 | # In such a situation, we arbitrary set the successors sets of |
|
601 | 615 | # the node to nothing (node pruned) to break the cycle. |
|
602 | 616 | # |
|
603 | 617 | # If no break was encountered we proceed to phase 2. |
|
604 | 618 | # |
|
605 | 619 | # Phase 2 computes successors sets of CURRENT (case 4); see details |
|
606 | 620 | # in phase 2 itself. |
|
607 | 621 | # |
|
608 | 622 | # Note the two levels of iteration in each phase. |
|
609 | 623 | # - The first one handles obsolescence markers using CURRENT as |
|
610 | 624 | # precursor (successors markers of CURRENT). |
|
611 | 625 | # |
|
612 | 626 | # Having multiple entry here means divergence. |
|
613 | 627 | # |
|
614 | 628 | # - The second one handles successors defined in each marker. |
|
615 | 629 | # |
|
616 | 630 | # Having none means pruned node, multiple successors means split, |
|
617 | 631 | # single successors are standard replacement. |
|
618 | 632 | # |
|
619 | 633 | for mark in sorted(succmarkers[current]): |
|
620 | 634 | for suc in mark[1]: |
|
621 | 635 | if suc not in cache: |
|
622 | 636 | if suc in stackedset: |
|
623 | 637 | # cycle breaking |
|
624 | 638 | cache[suc] = [] |
|
625 | 639 | else: |
|
626 | 640 | # case (3) If we have not computed successors sets |
|
627 | 641 | # of one of those successors we add it to the |
|
628 | 642 | # `toproceed` stack and stop all work for this |
|
629 | 643 | # iteration. |
|
630 | 644 | toproceed.append(suc) |
|
631 | 645 | stackedset.add(suc) |
|
632 | 646 | break |
|
633 | 647 | else: |
|
634 | 648 | continue |
|
635 | 649 | break |
|
636 | 650 | else: |
|
637 | 651 | # case (4): we know all successors sets of all direct |
|
638 | 652 | # successors |
|
639 | 653 | # |
|
640 | 654 | # Successors set contributed by each marker depends on the |
|
641 | 655 | # successors sets of all its "successors" node. |
|
642 | 656 | # |
|
643 | 657 | # Each different marker is a divergence in the obsolescence |
|
644 | 658 | # history. It contributes successors sets distinct from other |
|
645 | 659 | # markers. |
|
646 | 660 | # |
|
647 | 661 | # Within a marker, a successor may have divergent successors |
|
648 | 662 | # sets. In such a case, the marker will contribute multiple |
|
649 | 663 | # divergent successors sets. If multiple successors have |
|
650 | 664 | # divergent successors sets, a Cartesian product is used. |
|
651 | 665 | # |
|
652 | 666 | # At the end we post-process successors sets to remove |
|
653 | 667 | # duplicated entry and successors set that are strict subset of |
|
654 | 668 | # another one. |
|
655 | 669 | succssets = [] |
|
656 | 670 | for mark in sorted(succmarkers[current]): |
|
657 | 671 | # successors sets contributed by this marker |
|
658 | 672 | base = _succs() |
|
659 | 673 | base.markers.add(mark) |
|
660 | 674 | markss = [base] |
|
661 | 675 | for suc in mark[1]: |
|
662 | 676 | # cardinal product with previous successors |
|
663 | 677 | productresult = [] |
|
664 | 678 | for prefix in markss: |
|
665 | 679 | for suffix in cache[suc]: |
|
666 | 680 | newss = prefix.copy() |
|
667 | 681 | newss.markers.update(suffix.markers) |
|
668 | 682 | for part in suffix: |
|
669 | 683 | # do not duplicated entry in successors set |
|
670 | 684 | # first entry wins. |
|
671 | 685 | if part not in newss: |
|
672 | 686 | newss.append(part) |
|
673 | 687 | productresult.append(newss) |
|
674 | 688 | markss = productresult |
|
675 | 689 | succssets.extend(markss) |
|
676 | 690 | # remove duplicated and subset |
|
677 | 691 | seen = [] |
|
678 | 692 | final = [] |
|
679 | 693 | candidates = sorted((s for s in succssets if s), |
|
680 | 694 | key=len, reverse=True) |
|
681 | 695 | for cand in candidates: |
|
682 | 696 | for seensuccs in seen: |
|
683 | 697 | if cand.canmerge(seensuccs): |
|
684 | 698 | seensuccs.markers.update(cand.markers) |
|
685 | 699 | break |
|
686 | 700 | else: |
|
687 | 701 | final.append(cand) |
|
688 | 702 | seen.append(cand) |
|
689 | 703 | final.reverse() # put small successors set first |
|
690 | 704 | cache[current] = final |
|
691 | 705 | return cache[initialnode] |
|
692 | 706 | |
|
693 | 707 | def successorsandmarkers(repo, ctx): |
|
694 | 708 | """compute the raw data needed for computing obsfate |
|
695 | 709 | Returns a list of dict, one dict per successors set |
|
696 | 710 | """ |
|
697 | 711 | if not ctx.obsolete(): |
|
698 | 712 | return None |
|
699 | 713 | |
|
700 | 714 | ssets = successorssets(repo, ctx.node(), closest=True) |
|
701 | 715 | |
|
702 | 716 | # closestsuccessors returns an empty list for pruned revisions, remap it |
|
703 | 717 | # into a list containing an empty list for future processing |
|
704 | 718 | if ssets == []: |
|
705 | 719 | ssets = [[]] |
|
706 | 720 | |
|
707 | 721 | # Try to recover pruned markers |
|
708 | 722 | succsmap = repo.obsstore.successors |
|
709 | 723 | fullsuccessorsets = [] # successor set + markers |
|
710 | 724 | for sset in ssets: |
|
711 | 725 | if sset: |
|
712 | 726 | fullsuccessorsets.append(sset) |
|
713 | 727 | else: |
|
714 | 728 | # successorsset return an empty set() when ctx or one of its |
|
715 | 729 | # successors is pruned. |
|
716 | 730 | # In this case, walk the obs-markers tree again starting with ctx |
|
717 | 731 | # and find the relevant pruning obs-makers, the ones without |
|
718 | 732 | # successors. |
|
719 | 733 | # Having these markers allow us to compute some information about |
|
720 | 734 | # its fate, like who pruned this changeset and when. |
|
721 | 735 | |
|
722 | 736 | # XXX we do not catch all prune markers (eg rewritten then pruned) |
|
723 | 737 | # (fix me later) |
|
724 | 738 | foundany = False |
|
725 | 739 | for mark in succsmap.get(ctx.node(), ()): |
|
726 | 740 | if not mark[1]: |
|
727 | 741 | foundany = True |
|
728 | 742 | sset = _succs() |
|
729 | 743 | sset.markers.add(mark) |
|
730 | 744 | fullsuccessorsets.append(sset) |
|
731 | 745 | if not foundany: |
|
732 | 746 | fullsuccessorsets.append(_succs()) |
|
733 | 747 | |
|
734 | 748 | values = [] |
|
735 | 749 | for sset in fullsuccessorsets: |
|
736 | 750 | values.append({'successors': sset, 'markers': sset.markers}) |
|
737 | 751 | |
|
738 | 752 | return values |
|
739 | 753 | |
|
740 | 754 | def successorsetverb(successorset): |
|
741 | 755 | """ Return the verb summarizing the successorset |
|
742 | 756 | """ |
|
743 | 757 | if not successorset: |
|
744 | 758 | verb = 'pruned' |
|
745 | 759 | elif len(successorset) == 1: |
|
746 | 760 | verb = 'rewritten' |
|
747 | 761 | else: |
|
748 | 762 | verb = 'split' |
|
749 | 763 | return verb |
|
750 | 764 | |
|
751 | 765 | def markersdates(markers): |
|
752 | 766 | """returns the list of dates for a list of markers |
|
753 | 767 | """ |
|
754 | 768 | return [m[4] for m in markers] |
|
755 | 769 | |
|
756 | 770 | def markersusers(markers): |
|
757 | 771 | """ Returns a sorted list of markers users without duplicates |
|
758 | 772 | """ |
|
759 | 773 | markersmeta = [dict(m[3]) for m in markers] |
|
760 | 774 | users = set(meta.get('user') for meta in markersmeta if meta.get('user')) |
|
761 | 775 | |
|
762 | 776 | return sorted(users) |
|
763 | 777 | |
|
764 | 778 | def markersoperations(markers): |
|
765 | 779 | """ Returns a sorted list of markers operations without duplicates |
|
766 | 780 | """ |
|
767 | 781 | markersmeta = [dict(m[3]) for m in markers] |
|
768 | 782 | operations = set(meta.get('operation') for meta in markersmeta |
|
769 | 783 | if meta.get('operation')) |
|
770 | 784 | |
|
771 | 785 | return sorted(operations) |
General Comments 0
You need to be logged in to leave comments.
Login now