Show More
@@ -1,766 +1,768 b'' | |||
|
1 | 1 | # changegroup.py - Mercurial changegroup manipulation functions |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2006 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 | import weakref |
|
9 | 9 | from i18n import _ |
|
10 | 10 | from node import nullrev, nullid, hex, short |
|
11 | 11 | import mdiff, util, dagutil |
|
12 | 12 | import struct, os, bz2, zlib, tempfile |
|
13 | 13 | import discovery, error, phases, branchmap |
|
14 | 14 | |
|
15 | 15 | _CHANGEGROUPV1_DELTA_HEADER = "20s20s20s20s" |
|
16 | 16 | |
|
17 | 17 | def readexactly(stream, n): |
|
18 | 18 | '''read n bytes from stream.read and abort if less was available''' |
|
19 | 19 | s = stream.read(n) |
|
20 | 20 | if len(s) < n: |
|
21 | 21 | raise util.Abort(_("stream ended unexpectedly" |
|
22 | 22 | " (got %d bytes, expected %d)") |
|
23 | 23 | % (len(s), n)) |
|
24 | 24 | return s |
|
25 | 25 | |
|
26 | 26 | def getchunk(stream): |
|
27 | 27 | """return the next chunk from stream as a string""" |
|
28 | 28 | d = readexactly(stream, 4) |
|
29 | 29 | l = struct.unpack(">l", d)[0] |
|
30 | 30 | if l <= 4: |
|
31 | 31 | if l: |
|
32 | 32 | raise util.Abort(_("invalid chunk length %d") % l) |
|
33 | 33 | return "" |
|
34 | 34 | return readexactly(stream, l - 4) |
|
35 | 35 | |
|
36 | 36 | def chunkheader(length): |
|
37 | 37 | """return a changegroup chunk header (string)""" |
|
38 | 38 | return struct.pack(">l", length + 4) |
|
39 | 39 | |
|
40 | 40 | def closechunk(): |
|
41 | 41 | """return a changegroup chunk header (string) for a zero-length chunk""" |
|
42 | 42 | return struct.pack(">l", 0) |
|
43 | 43 | |
|
44 | 44 | class nocompress(object): |
|
45 | 45 | def compress(self, x): |
|
46 | 46 | return x |
|
47 | 47 | def flush(self): |
|
48 | 48 | return "" |
|
49 | 49 | |
|
50 | 50 | bundletypes = { |
|
51 | 51 | "": ("", nocompress), # only when using unbundle on ssh and old http servers |
|
52 | 52 | # since the unification ssh accepts a header but there |
|
53 | 53 | # is no capability signaling it. |
|
54 | 54 | "HG10UN": ("HG10UN", nocompress), |
|
55 | 55 | "HG10BZ": ("HG10", lambda: bz2.BZ2Compressor()), |
|
56 | 56 | "HG10GZ": ("HG10GZ", lambda: zlib.compressobj()), |
|
57 | 57 | } |
|
58 | 58 | |
|
59 | 59 | # hgweb uses this list to communicate its preferred type |
|
60 | 60 | bundlepriority = ['HG10GZ', 'HG10BZ', 'HG10UN'] |
|
61 | 61 | |
|
62 | 62 | def writebundle(cg, filename, bundletype, vfs=None): |
|
63 | 63 | """Write a bundle file and return its filename. |
|
64 | 64 | |
|
65 | 65 | Existing files will not be overwritten. |
|
66 | 66 | If no filename is specified, a temporary file is created. |
|
67 | 67 | bz2 compression can be turned off. |
|
68 | 68 | The bundle file will be deleted in case of errors. |
|
69 | 69 | """ |
|
70 | 70 | |
|
71 | 71 | fh = None |
|
72 | 72 | cleanup = None |
|
73 | 73 | try: |
|
74 | 74 | if filename: |
|
75 | 75 | if vfs: |
|
76 | 76 | fh = vfs.open(filename, "wb") |
|
77 | 77 | else: |
|
78 | 78 | fh = open(filename, "wb") |
|
79 | 79 | else: |
|
80 | 80 | fd, filename = tempfile.mkstemp(prefix="hg-bundle-", suffix=".hg") |
|
81 | 81 | fh = os.fdopen(fd, "wb") |
|
82 | 82 | cleanup = filename |
|
83 | 83 | |
|
84 | 84 | header, compressor = bundletypes[bundletype] |
|
85 | 85 | fh.write(header) |
|
86 | 86 | z = compressor() |
|
87 | 87 | |
|
88 | 88 | # parse the changegroup data, otherwise we will block |
|
89 | 89 | # in case of sshrepo because we don't know the end of the stream |
|
90 | 90 | |
|
91 | 91 | # an empty chunkgroup is the end of the changegroup |
|
92 | 92 | # a changegroup has at least 2 chunkgroups (changelog and manifest). |
|
93 | 93 | # after that, an empty chunkgroup is the end of the changegroup |
|
94 | 94 | for chunk in cg.getchunks(): |
|
95 | 95 | fh.write(z.compress(chunk)) |
|
96 | 96 | fh.write(z.flush()) |
|
97 | 97 | cleanup = None |
|
98 | 98 | return filename |
|
99 | 99 | finally: |
|
100 | 100 | if fh is not None: |
|
101 | 101 | fh.close() |
|
102 | 102 | if cleanup is not None: |
|
103 | 103 | if filename and vfs: |
|
104 | 104 | vfs.unlink(cleanup) |
|
105 | 105 | else: |
|
106 | 106 | os.unlink(cleanup) |
|
107 | 107 | |
|
108 | 108 | def decompressor(fh, alg): |
|
109 | 109 | if alg == 'UN': |
|
110 | 110 | return fh |
|
111 | 111 | elif alg == 'GZ': |
|
112 | 112 | def generator(f): |
|
113 | 113 | zd = zlib.decompressobj() |
|
114 | 114 | for chunk in util.filechunkiter(f): |
|
115 | 115 | yield zd.decompress(chunk) |
|
116 | 116 | elif alg == 'BZ': |
|
117 | 117 | def generator(f): |
|
118 | 118 | zd = bz2.BZ2Decompressor() |
|
119 | 119 | zd.decompress("BZ") |
|
120 | 120 | for chunk in util.filechunkiter(f, 4096): |
|
121 | 121 | yield zd.decompress(chunk) |
|
122 | 122 | else: |
|
123 | 123 | raise util.Abort("unknown bundle compression '%s'" % alg) |
|
124 | 124 | return util.chunkbuffer(generator(fh)) |
|
125 | 125 | |
|
126 | 126 | class cg1unpacker(object): |
|
127 | 127 | deltaheader = _CHANGEGROUPV1_DELTA_HEADER |
|
128 | 128 | deltaheadersize = struct.calcsize(deltaheader) |
|
129 | 129 | def __init__(self, fh, alg): |
|
130 | 130 | self._stream = decompressor(fh, alg) |
|
131 | 131 | self._type = alg |
|
132 | 132 | self.callback = None |
|
133 | 133 | def compressed(self): |
|
134 | 134 | return self._type != 'UN' |
|
135 | 135 | def read(self, l): |
|
136 | 136 | return self._stream.read(l) |
|
137 | 137 | def seek(self, pos): |
|
138 | 138 | return self._stream.seek(pos) |
|
139 | 139 | def tell(self): |
|
140 | 140 | return self._stream.tell() |
|
141 | 141 | def close(self): |
|
142 | 142 | return self._stream.close() |
|
143 | 143 | |
|
144 | 144 | def chunklength(self): |
|
145 | 145 | d = readexactly(self._stream, 4) |
|
146 | 146 | l = struct.unpack(">l", d)[0] |
|
147 | 147 | if l <= 4: |
|
148 | 148 | if l: |
|
149 | 149 | raise util.Abort(_("invalid chunk length %d") % l) |
|
150 | 150 | return 0 |
|
151 | 151 | if self.callback: |
|
152 | 152 | self.callback() |
|
153 | 153 | return l - 4 |
|
154 | 154 | |
|
155 | 155 | def changelogheader(self): |
|
156 | 156 | """v10 does not have a changelog header chunk""" |
|
157 | 157 | return {} |
|
158 | 158 | |
|
159 | 159 | def manifestheader(self): |
|
160 | 160 | """v10 does not have a manifest header chunk""" |
|
161 | 161 | return {} |
|
162 | 162 | |
|
163 | 163 | def filelogheader(self): |
|
164 | 164 | """return the header of the filelogs chunk, v10 only has the filename""" |
|
165 | 165 | l = self.chunklength() |
|
166 | 166 | if not l: |
|
167 | 167 | return {} |
|
168 | 168 | fname = readexactly(self._stream, l) |
|
169 | 169 | return {'filename': fname} |
|
170 | 170 | |
|
171 | 171 | def _deltaheader(self, headertuple, prevnode): |
|
172 | 172 | node, p1, p2, cs = headertuple |
|
173 | 173 | if prevnode is None: |
|
174 | 174 | deltabase = p1 |
|
175 | 175 | else: |
|
176 | 176 | deltabase = prevnode |
|
177 | 177 | return node, p1, p2, deltabase, cs |
|
178 | 178 | |
|
179 | 179 | def deltachunk(self, prevnode): |
|
180 | 180 | l = self.chunklength() |
|
181 | 181 | if not l: |
|
182 | 182 | return {} |
|
183 | 183 | headerdata = readexactly(self._stream, self.deltaheadersize) |
|
184 | 184 | header = struct.unpack(self.deltaheader, headerdata) |
|
185 | 185 | delta = readexactly(self._stream, l - self.deltaheadersize) |
|
186 | 186 | node, p1, p2, deltabase, cs = self._deltaheader(header, prevnode) |
|
187 | 187 | return {'node': node, 'p1': p1, 'p2': p2, 'cs': cs, |
|
188 | 188 | 'deltabase': deltabase, 'delta': delta} |
|
189 | 189 | |
|
190 | 190 | def getchunks(self): |
|
191 | 191 | """returns all the chunks contains in the bundle |
|
192 | 192 | |
|
193 | 193 | Used when you need to forward the binary stream to a file or another |
|
194 | 194 | network API. To do so, it parse the changegroup data, otherwise it will |
|
195 | 195 | block in case of sshrepo because it don't know the end of the stream. |
|
196 | 196 | """ |
|
197 | 197 | # an empty chunkgroup is the end of the changegroup |
|
198 | 198 | # a changegroup has at least 2 chunkgroups (changelog and manifest). |
|
199 | 199 | # after that, an empty chunkgroup is the end of the changegroup |
|
200 | 200 | empty = False |
|
201 | 201 | count = 0 |
|
202 | 202 | while not empty or count <= 2: |
|
203 | 203 | empty = True |
|
204 | 204 | count += 1 |
|
205 | 205 | while True: |
|
206 | 206 | chunk = getchunk(self) |
|
207 | 207 | if not chunk: |
|
208 | 208 | break |
|
209 | 209 | empty = False |
|
210 | 210 | yield chunkheader(len(chunk)) |
|
211 | 211 | pos = 0 |
|
212 | 212 | while pos < len(chunk): |
|
213 | 213 | next = pos + 2**20 |
|
214 | 214 | yield chunk[pos:next] |
|
215 | 215 | pos = next |
|
216 | 216 | yield closechunk() |
|
217 | 217 | |
|
218 | 218 | class headerlessfixup(object): |
|
219 | 219 | def __init__(self, fh, h): |
|
220 | 220 | self._h = h |
|
221 | 221 | self._fh = fh |
|
222 | 222 | def read(self, n): |
|
223 | 223 | if self._h: |
|
224 | 224 | d, self._h = self._h[:n], self._h[n:] |
|
225 | 225 | if len(d) < n: |
|
226 | 226 | d += readexactly(self._fh, n - len(d)) |
|
227 | 227 | return d |
|
228 | 228 | return readexactly(self._fh, n) |
|
229 | 229 | |
|
230 | 230 | class cg1packer(object): |
|
231 | 231 | deltaheader = _CHANGEGROUPV1_DELTA_HEADER |
|
232 | 232 | def __init__(self, repo, bundlecaps=None): |
|
233 | 233 | """Given a source repo, construct a bundler. |
|
234 | 234 | |
|
235 | 235 | bundlecaps is optional and can be used to specify the set of |
|
236 | 236 | capabilities which can be used to build the bundle. |
|
237 | 237 | """ |
|
238 | 238 | # Set of capabilities we can use to build the bundle. |
|
239 | 239 | if bundlecaps is None: |
|
240 | 240 | bundlecaps = set() |
|
241 | 241 | self._bundlecaps = bundlecaps |
|
242 | 242 | self._changelog = repo.changelog |
|
243 | 243 | self._manifest = repo.manifest |
|
244 | 244 | reorder = repo.ui.config('bundle', 'reorder', 'auto') |
|
245 | 245 | if reorder == 'auto': |
|
246 | 246 | reorder = None |
|
247 | 247 | else: |
|
248 | 248 | reorder = util.parsebool(reorder) |
|
249 | 249 | self._repo = repo |
|
250 | 250 | self._reorder = reorder |
|
251 | 251 | self._progress = repo.ui.progress |
|
252 | 252 | def close(self): |
|
253 | 253 | return closechunk() |
|
254 | 254 | |
|
255 | 255 | def fileheader(self, fname): |
|
256 | 256 | return chunkheader(len(fname)) + fname |
|
257 | 257 | |
|
258 | 258 | def group(self, nodelist, revlog, lookup, units=None, reorder=None): |
|
259 | 259 | """Calculate a delta group, yielding a sequence of changegroup chunks |
|
260 | 260 | (strings). |
|
261 | 261 | |
|
262 | 262 | Given a list of changeset revs, return a set of deltas and |
|
263 | 263 | metadata corresponding to nodes. The first delta is |
|
264 | 264 | first parent(nodelist[0]) -> nodelist[0], the receiver is |
|
265 | 265 | guaranteed to have this parent as it has all history before |
|
266 | 266 | these changesets. In the case firstparent is nullrev the |
|
267 | 267 | changegroup starts with a full revision. |
|
268 | 268 | |
|
269 | 269 | If units is not None, progress detail will be generated, units specifies |
|
270 | 270 | the type of revlog that is touched (changelog, manifest, etc.). |
|
271 | 271 | """ |
|
272 | 272 | # if we don't have any revisions touched by these changesets, bail |
|
273 | 273 | if len(nodelist) == 0: |
|
274 | 274 | yield self.close() |
|
275 | 275 | return |
|
276 | 276 | |
|
277 | 277 | # for generaldelta revlogs, we linearize the revs; this will both be |
|
278 | 278 | # much quicker and generate a much smaller bundle |
|
279 | 279 | if (revlog._generaldelta and reorder is not False) or reorder: |
|
280 | 280 | dag = dagutil.revlogdag(revlog) |
|
281 | 281 | revs = set(revlog.rev(n) for n in nodelist) |
|
282 | 282 | revs = dag.linearize(revs) |
|
283 | 283 | else: |
|
284 | 284 | revs = sorted([revlog.rev(n) for n in nodelist]) |
|
285 | 285 | |
|
286 | 286 | # add the parent of the first rev |
|
287 | 287 | p = revlog.parentrevs(revs[0])[0] |
|
288 | 288 | revs.insert(0, p) |
|
289 | 289 | |
|
290 | 290 | # build deltas |
|
291 | 291 | total = len(revs) - 1 |
|
292 | 292 | msgbundling = _('bundling') |
|
293 | 293 | for r in xrange(len(revs) - 1): |
|
294 | 294 | if units is not None: |
|
295 | 295 | self._progress(msgbundling, r + 1, unit=units, total=total) |
|
296 | 296 | prev, curr = revs[r], revs[r + 1] |
|
297 | 297 | linknode = lookup(revlog.node(curr)) |
|
298 | 298 | for c in self.revchunk(revlog, curr, prev, linknode): |
|
299 | 299 | yield c |
|
300 | 300 | |
|
301 | 301 | yield self.close() |
|
302 | 302 | |
|
303 | 303 | # filter any nodes that claim to be part of the known set |
|
304 | 304 | def prune(self, revlog, missing, commonrevs, source): |
|
305 | 305 | rr, rl = revlog.rev, revlog.linkrev |
|
306 | 306 | return [n for n in missing if rl(rr(n)) not in commonrevs] |
|
307 | 307 | |
|
308 | 308 | def generate(self, commonrevs, clnodes, fastpathlinkrev, source): |
|
309 | 309 | '''yield a sequence of changegroup chunks (strings)''' |
|
310 | 310 | repo = self._repo |
|
311 | 311 | cl = self._changelog |
|
312 | 312 | mf = self._manifest |
|
313 | 313 | reorder = self._reorder |
|
314 | 314 | progress = self._progress |
|
315 | 315 | |
|
316 | 316 | # for progress output |
|
317 | 317 | msgbundling = _('bundling') |
|
318 | 318 | |
|
319 | 319 | mfs = {} # needed manifests |
|
320 | 320 | fnodes = {} # needed file nodes |
|
321 | 321 | changedfiles = set() |
|
322 | 322 | |
|
323 | 323 | # Callback for the changelog, used to collect changed files and manifest |
|
324 | 324 | # nodes. |
|
325 | 325 | # Returns the linkrev node (identity in the changelog case). |
|
326 | 326 | def lookupcl(x): |
|
327 | 327 | c = cl.read(x) |
|
328 | 328 | changedfiles.update(c[3]) |
|
329 | 329 | # record the first changeset introducing this manifest version |
|
330 | 330 | mfs.setdefault(c[0], x) |
|
331 | 331 | return x |
|
332 | 332 | |
|
333 | 333 | # Callback for the manifest, used to collect linkrevs for filelog |
|
334 | 334 | # revisions. |
|
335 | 335 | # Returns the linkrev node (collected in lookupcl). |
|
336 | 336 | def lookupmf(x): |
|
337 | 337 | clnode = mfs[x] |
|
338 | 338 | if not fastpathlinkrev: |
|
339 | 339 | mdata = mf.readfast(x) |
|
340 | 340 | for f, n in mdata.iteritems(): |
|
341 | 341 | if f in changedfiles: |
|
342 | 342 | # record the first changeset introducing this filelog |
|
343 | 343 | # version |
|
344 | 344 | fnodes[f].setdefault(n, clnode) |
|
345 | 345 | return clnode |
|
346 | 346 | |
|
347 | 347 | for chunk in self.group(clnodes, cl, lookupcl, units=_('changesets'), |
|
348 | 348 | reorder=reorder): |
|
349 | 349 | yield chunk |
|
350 | 350 | progress(msgbundling, None) |
|
351 | 351 | |
|
352 | 352 | for f in changedfiles: |
|
353 | 353 | fnodes[f] = {} |
|
354 | 354 | mfnodes = self.prune(mf, mfs, commonrevs, source) |
|
355 | 355 | for chunk in self.group(mfnodes, mf, lookupmf, units=_('manifests'), |
|
356 | 356 | reorder=reorder): |
|
357 | 357 | yield chunk |
|
358 | 358 | progress(msgbundling, None) |
|
359 | 359 | |
|
360 | 360 | mfs.clear() |
|
361 | 361 | needed = set(cl.rev(x) for x in clnodes) |
|
362 | 362 | |
|
363 | 363 | def linknodes(filerevlog, fname): |
|
364 | 364 | if fastpathlinkrev: |
|
365 | 365 | llr = filerevlog.linkrev |
|
366 | 366 | def genfilenodes(): |
|
367 | 367 | for r in filerevlog: |
|
368 | 368 | linkrev = llr(r) |
|
369 | 369 | if linkrev in needed: |
|
370 | 370 | yield filerevlog.node(r), cl.node(linkrev) |
|
371 | 371 | fnodes[fname] = dict(genfilenodes()) |
|
372 | 372 | return fnodes.get(fname, {}) |
|
373 | 373 | |
|
374 | 374 | for chunk in self.generatefiles(changedfiles, linknodes, commonrevs, |
|
375 | 375 | source): |
|
376 | 376 | yield chunk |
|
377 | 377 | |
|
378 | 378 | yield self.close() |
|
379 | 379 | progress(msgbundling, None) |
|
380 | 380 | |
|
381 | 381 | if clnodes: |
|
382 | 382 | repo.hook('outgoing', node=hex(clnodes[0]), source=source) |
|
383 | 383 | |
|
384 | 384 | def generatefiles(self, changedfiles, linknodes, commonrevs, source): |
|
385 | 385 | repo = self._repo |
|
386 | 386 | progress = self._progress |
|
387 | 387 | reorder = self._reorder |
|
388 | 388 | msgbundling = _('bundling') |
|
389 | 389 | |
|
390 | 390 | total = len(changedfiles) |
|
391 | 391 | # for progress output |
|
392 | 392 | msgfiles = _('files') |
|
393 | 393 | for i, fname in enumerate(sorted(changedfiles)): |
|
394 | 394 | filerevlog = repo.file(fname) |
|
395 | 395 | if not filerevlog: |
|
396 | 396 | raise util.Abort(_("empty or missing revlog for %s") % fname) |
|
397 | 397 | |
|
398 | 398 | linkrevnodes = linknodes(filerevlog, fname) |
|
399 | 399 | # Lookup for filenodes, we collected the linkrev nodes above in the |
|
400 | 400 | # fastpath case and with lookupmf in the slowpath case. |
|
401 | 401 | def lookupfilelog(x): |
|
402 | 402 | return linkrevnodes[x] |
|
403 | 403 | |
|
404 | 404 | filenodes = self.prune(filerevlog, linkrevnodes, commonrevs, source) |
|
405 | 405 | if filenodes: |
|
406 | 406 | progress(msgbundling, i + 1, item=fname, unit=msgfiles, |
|
407 | 407 | total=total) |
|
408 | 408 | yield self.fileheader(fname) |
|
409 | 409 | for chunk in self.group(filenodes, filerevlog, lookupfilelog, |
|
410 | 410 | reorder=reorder): |
|
411 | 411 | yield chunk |
|
412 | 412 | |
|
413 | 413 | def revchunk(self, revlog, rev, prev, linknode): |
|
414 | 414 | node = revlog.node(rev) |
|
415 | 415 | p1, p2 = revlog.parentrevs(rev) |
|
416 | 416 | base = prev |
|
417 | 417 | |
|
418 | 418 | prefix = '' |
|
419 | 419 | if base == nullrev: |
|
420 | 420 | delta = revlog.revision(node) |
|
421 | 421 | prefix = mdiff.trivialdiffheader(len(delta)) |
|
422 | 422 | else: |
|
423 | 423 | delta = revlog.revdiff(base, rev) |
|
424 | 424 | p1n, p2n = revlog.parents(node) |
|
425 | 425 | basenode = revlog.node(base) |
|
426 | 426 | meta = self.builddeltaheader(node, p1n, p2n, basenode, linknode) |
|
427 | 427 | meta += prefix |
|
428 | 428 | l = len(meta) + len(delta) |
|
429 | 429 | yield chunkheader(l) |
|
430 | 430 | yield meta |
|
431 | 431 | yield delta |
|
432 | 432 | def builddeltaheader(self, node, p1n, p2n, basenode, linknode): |
|
433 | 433 | # do nothing with basenode, it is implicitly the previous one in HG10 |
|
434 | 434 | return struct.pack(self.deltaheader, node, p1n, p2n, linknode) |
|
435 | 435 | |
|
436 | packermap = {'01': (cg1packer, cg1unpacker)} | |
|
437 | ||
|
436 | 438 | def _changegroupinfo(repo, nodes, source): |
|
437 | 439 | if repo.ui.verbose or source == 'bundle': |
|
438 | 440 | repo.ui.status(_("%d changesets found\n") % len(nodes)) |
|
439 | 441 | if repo.ui.debugflag: |
|
440 | 442 | repo.ui.debug("list of changesets:\n") |
|
441 | 443 | for node in nodes: |
|
442 | 444 | repo.ui.debug("%s\n" % hex(node)) |
|
443 | 445 | |
|
444 | 446 | def getsubset(repo, outgoing, bundler, source, fastpath=False): |
|
445 | 447 | repo = repo.unfiltered() |
|
446 | 448 | commonrevs = outgoing.common |
|
447 | 449 | csets = outgoing.missing |
|
448 | 450 | heads = outgoing.missingheads |
|
449 | 451 | # We go through the fast path if we get told to, or if all (unfiltered |
|
450 | 452 | # heads have been requested (since we then know there all linkrevs will |
|
451 | 453 | # be pulled by the client). |
|
452 | 454 | heads.sort() |
|
453 | 455 | fastpathlinkrev = fastpath or ( |
|
454 | 456 | repo.filtername is None and heads == sorted(repo.heads())) |
|
455 | 457 | |
|
456 | 458 | repo.hook('preoutgoing', throw=True, source=source) |
|
457 | 459 | _changegroupinfo(repo, csets, source) |
|
458 | 460 | gengroup = bundler.generate(commonrevs, csets, fastpathlinkrev, source) |
|
459 | 461 | return cg1unpacker(util.chunkbuffer(gengroup), 'UN') |
|
460 | 462 | |
|
461 | 463 | def changegroupsubset(repo, roots, heads, source): |
|
462 | 464 | """Compute a changegroup consisting of all the nodes that are |
|
463 | 465 | descendants of any of the roots and ancestors of any of the heads. |
|
464 | 466 | Return a chunkbuffer object whose read() method will return |
|
465 | 467 | successive changegroup chunks. |
|
466 | 468 | |
|
467 | 469 | It is fairly complex as determining which filenodes and which |
|
468 | 470 | manifest nodes need to be included for the changeset to be complete |
|
469 | 471 | is non-trivial. |
|
470 | 472 | |
|
471 | 473 | Another wrinkle is doing the reverse, figuring out which changeset in |
|
472 | 474 | the changegroup a particular filenode or manifestnode belongs to. |
|
473 | 475 | """ |
|
474 | 476 | cl = repo.changelog |
|
475 | 477 | if not roots: |
|
476 | 478 | roots = [nullid] |
|
477 | 479 | # TODO: remove call to nodesbetween. |
|
478 | 480 | csets, roots, heads = cl.nodesbetween(roots, heads) |
|
479 | 481 | discbases = [] |
|
480 | 482 | for n in roots: |
|
481 | 483 | discbases.extend([p for p in cl.parents(n) if p != nullid]) |
|
482 | 484 | outgoing = discovery.outgoing(cl, discbases, heads) |
|
483 | 485 | bundler = cg1packer(repo) |
|
484 | 486 | return getsubset(repo, outgoing, bundler, source) |
|
485 | 487 | |
|
486 | 488 | def getlocalchangegroup(repo, source, outgoing, bundlecaps=None): |
|
487 | 489 | """Like getbundle, but taking a discovery.outgoing as an argument. |
|
488 | 490 | |
|
489 | 491 | This is only implemented for local repos and reuses potentially |
|
490 | 492 | precomputed sets in outgoing.""" |
|
491 | 493 | if not outgoing.missing: |
|
492 | 494 | return None |
|
493 | 495 | bundler = cg1packer(repo, bundlecaps) |
|
494 | 496 | return getsubset(repo, outgoing, bundler, source) |
|
495 | 497 | |
|
496 | 498 | def _computeoutgoing(repo, heads, common): |
|
497 | 499 | """Computes which revs are outgoing given a set of common |
|
498 | 500 | and a set of heads. |
|
499 | 501 | |
|
500 | 502 | This is a separate function so extensions can have access to |
|
501 | 503 | the logic. |
|
502 | 504 | |
|
503 | 505 | Returns a discovery.outgoing object. |
|
504 | 506 | """ |
|
505 | 507 | cl = repo.changelog |
|
506 | 508 | if common: |
|
507 | 509 | hasnode = cl.hasnode |
|
508 | 510 | common = [n for n in common if hasnode(n)] |
|
509 | 511 | else: |
|
510 | 512 | common = [nullid] |
|
511 | 513 | if not heads: |
|
512 | 514 | heads = cl.heads() |
|
513 | 515 | return discovery.outgoing(cl, common, heads) |
|
514 | 516 | |
|
515 | 517 | def getchangegroup(repo, source, heads=None, common=None, bundlecaps=None): |
|
516 | 518 | """Like changegroupsubset, but returns the set difference between the |
|
517 | 519 | ancestors of heads and the ancestors common. |
|
518 | 520 | |
|
519 | 521 | If heads is None, use the local heads. If common is None, use [nullid]. |
|
520 | 522 | |
|
521 | 523 | The nodes in common might not all be known locally due to the way the |
|
522 | 524 | current discovery protocol works. |
|
523 | 525 | """ |
|
524 | 526 | outgoing = _computeoutgoing(repo, heads, common) |
|
525 | 527 | return getlocalchangegroup(repo, source, outgoing, bundlecaps=bundlecaps) |
|
526 | 528 | |
|
527 | 529 | def changegroup(repo, basenodes, source): |
|
528 | 530 | # to avoid a race we use changegroupsubset() (issue1320) |
|
529 | 531 | return changegroupsubset(repo, basenodes, repo.heads(), source) |
|
530 | 532 | |
|
531 | 533 | def addchangegroupfiles(repo, source, revmap, trp, pr, needfiles): |
|
532 | 534 | revisions = 0 |
|
533 | 535 | files = 0 |
|
534 | 536 | while True: |
|
535 | 537 | chunkdata = source.filelogheader() |
|
536 | 538 | if not chunkdata: |
|
537 | 539 | break |
|
538 | 540 | f = chunkdata["filename"] |
|
539 | 541 | repo.ui.debug("adding %s revisions\n" % f) |
|
540 | 542 | pr() |
|
541 | 543 | fl = repo.file(f) |
|
542 | 544 | o = len(fl) |
|
543 | 545 | if not fl.addgroup(source, revmap, trp): |
|
544 | 546 | raise util.Abort(_("received file revlog group is empty")) |
|
545 | 547 | revisions += len(fl) - o |
|
546 | 548 | files += 1 |
|
547 | 549 | if f in needfiles: |
|
548 | 550 | needs = needfiles[f] |
|
549 | 551 | for new in xrange(o, len(fl)): |
|
550 | 552 | n = fl.node(new) |
|
551 | 553 | if n in needs: |
|
552 | 554 | needs.remove(n) |
|
553 | 555 | else: |
|
554 | 556 | raise util.Abort( |
|
555 | 557 | _("received spurious file revlog entry")) |
|
556 | 558 | if not needs: |
|
557 | 559 | del needfiles[f] |
|
558 | 560 | repo.ui.progress(_('files'), None) |
|
559 | 561 | |
|
560 | 562 | for f, needs in needfiles.iteritems(): |
|
561 | 563 | fl = repo.file(f) |
|
562 | 564 | for n in needs: |
|
563 | 565 | try: |
|
564 | 566 | fl.rev(n) |
|
565 | 567 | except error.LookupError: |
|
566 | 568 | raise util.Abort( |
|
567 | 569 | _('missing file data for %s:%s - run hg verify') % |
|
568 | 570 | (f, hex(n))) |
|
569 | 571 | |
|
570 | 572 | return revisions, files |
|
571 | 573 | |
|
572 | 574 | def addchangegroup(repo, source, srctype, url, emptyok=False, |
|
573 | 575 | targetphase=phases.draft): |
|
574 | 576 | """Add the changegroup returned by source.read() to this repo. |
|
575 | 577 | srctype is a string like 'push', 'pull', or 'unbundle'. url is |
|
576 | 578 | the URL of the repo where this changegroup is coming from. |
|
577 | 579 | |
|
578 | 580 | Return an integer summarizing the change to this repo: |
|
579 | 581 | - nothing changed or no source: 0 |
|
580 | 582 | - more heads than before: 1+added heads (2..n) |
|
581 | 583 | - fewer heads than before: -1-removed heads (-2..-n) |
|
582 | 584 | - number of heads stays the same: 1 |
|
583 | 585 | """ |
|
584 | 586 | repo = repo.unfiltered() |
|
585 | 587 | def csmap(x): |
|
586 | 588 | repo.ui.debug("add changeset %s\n" % short(x)) |
|
587 | 589 | return len(cl) |
|
588 | 590 | |
|
589 | 591 | def revmap(x): |
|
590 | 592 | return cl.rev(x) |
|
591 | 593 | |
|
592 | 594 | if not source: |
|
593 | 595 | return 0 |
|
594 | 596 | |
|
595 | 597 | changesets = files = revisions = 0 |
|
596 | 598 | efiles = set() |
|
597 | 599 | |
|
598 | 600 | # write changelog data to temp files so concurrent readers will not see |
|
599 | 601 | # inconsistent view |
|
600 | 602 | cl = repo.changelog |
|
601 | 603 | cl.delayupdate() |
|
602 | 604 | oldheads = cl.heads() |
|
603 | 605 | |
|
604 | 606 | tr = repo.transaction("\n".join([srctype, util.hidepassword(url)])) |
|
605 | 607 | # The transaction could have been created before and already carries source |
|
606 | 608 | # information. In this case we use the top level data. We overwrite the |
|
607 | 609 | # argument because we need to use the top level value (if they exist) in |
|
608 | 610 | # this function. |
|
609 | 611 | srctype = tr.hookargs.setdefault('source', srctype) |
|
610 | 612 | url = tr.hookargs.setdefault('url', url) |
|
611 | 613 | try: |
|
612 | 614 | repo.hook('prechangegroup', throw=True, **tr.hookargs) |
|
613 | 615 | |
|
614 | 616 | trp = weakref.proxy(tr) |
|
615 | 617 | # pull off the changeset group |
|
616 | 618 | repo.ui.status(_("adding changesets\n")) |
|
617 | 619 | clstart = len(cl) |
|
618 | 620 | class prog(object): |
|
619 | 621 | step = _('changesets') |
|
620 | 622 | count = 1 |
|
621 | 623 | ui = repo.ui |
|
622 | 624 | total = None |
|
623 | 625 | def __call__(repo): |
|
624 | 626 | repo.ui.progress(repo.step, repo.count, unit=_('chunks'), |
|
625 | 627 | total=repo.total) |
|
626 | 628 | repo.count += 1 |
|
627 | 629 | pr = prog() |
|
628 | 630 | source.callback = pr |
|
629 | 631 | |
|
630 | 632 | source.changelogheader() |
|
631 | 633 | srccontent = cl.addgroup(source, csmap, trp) |
|
632 | 634 | if not (srccontent or emptyok): |
|
633 | 635 | raise util.Abort(_("received changelog group is empty")) |
|
634 | 636 | clend = len(cl) |
|
635 | 637 | changesets = clend - clstart |
|
636 | 638 | for c in xrange(clstart, clend): |
|
637 | 639 | efiles.update(repo[c].files()) |
|
638 | 640 | efiles = len(efiles) |
|
639 | 641 | repo.ui.progress(_('changesets'), None) |
|
640 | 642 | |
|
641 | 643 | # pull off the manifest group |
|
642 | 644 | repo.ui.status(_("adding manifests\n")) |
|
643 | 645 | pr.step = _('manifests') |
|
644 | 646 | pr.count = 1 |
|
645 | 647 | pr.total = changesets # manifests <= changesets |
|
646 | 648 | # no need to check for empty manifest group here: |
|
647 | 649 | # if the result of the merge of 1 and 2 is the same in 3 and 4, |
|
648 | 650 | # no new manifest will be created and the manifest group will |
|
649 | 651 | # be empty during the pull |
|
650 | 652 | source.manifestheader() |
|
651 | 653 | repo.manifest.addgroup(source, revmap, trp) |
|
652 | 654 | repo.ui.progress(_('manifests'), None) |
|
653 | 655 | |
|
654 | 656 | needfiles = {} |
|
655 | 657 | if repo.ui.configbool('server', 'validate', default=False): |
|
656 | 658 | # validate incoming csets have their manifests |
|
657 | 659 | for cset in xrange(clstart, clend): |
|
658 | 660 | mfest = repo.changelog.read(repo.changelog.node(cset))[0] |
|
659 | 661 | mfest = repo.manifest.readdelta(mfest) |
|
660 | 662 | # store file nodes we must see |
|
661 | 663 | for f, n in mfest.iteritems(): |
|
662 | 664 | needfiles.setdefault(f, set()).add(n) |
|
663 | 665 | |
|
664 | 666 | # process the files |
|
665 | 667 | repo.ui.status(_("adding file changes\n")) |
|
666 | 668 | pr.step = _('files') |
|
667 | 669 | pr.count = 1 |
|
668 | 670 | pr.total = efiles |
|
669 | 671 | source.callback = None |
|
670 | 672 | |
|
671 | 673 | newrevs, newfiles = addchangegroupfiles(repo, source, revmap, trp, pr, |
|
672 | 674 | needfiles) |
|
673 | 675 | revisions += newrevs |
|
674 | 676 | files += newfiles |
|
675 | 677 | |
|
676 | 678 | dh = 0 |
|
677 | 679 | if oldheads: |
|
678 | 680 | heads = cl.heads() |
|
679 | 681 | dh = len(heads) - len(oldheads) |
|
680 | 682 | for h in heads: |
|
681 | 683 | if h not in oldheads and repo[h].closesbranch(): |
|
682 | 684 | dh -= 1 |
|
683 | 685 | htext = "" |
|
684 | 686 | if dh: |
|
685 | 687 | htext = _(" (%+d heads)") % dh |
|
686 | 688 | |
|
687 | 689 | repo.ui.status(_("added %d changesets" |
|
688 | 690 | " with %d changes to %d files%s\n") |
|
689 | 691 | % (changesets, revisions, files, htext)) |
|
690 | 692 | repo.invalidatevolatilesets() |
|
691 | 693 | |
|
692 | 694 | if changesets > 0: |
|
693 | 695 | p = lambda: cl.writepending() and repo.root or "" |
|
694 | 696 | if 'node' not in tr.hookargs: |
|
695 | 697 | tr.hookargs['node'] = hex(cl.node(clstart)) |
|
696 | 698 | hookargs = dict(tr.hookargs) |
|
697 | 699 | else: |
|
698 | 700 | hookargs = dict(tr.hookargs) |
|
699 | 701 | hookargs['node'] = hex(cl.node(clstart)) |
|
700 | 702 | repo.hook('pretxnchangegroup', throw=True, pending=p, **hookargs) |
|
701 | 703 | |
|
702 | 704 | added = [cl.node(r) for r in xrange(clstart, clend)] |
|
703 | 705 | publishing = repo.ui.configbool('phases', 'publish', True) |
|
704 | 706 | if srctype in ('push', 'serve'): |
|
705 | 707 | # Old servers can not push the boundary themselves. |
|
706 | 708 | # New servers won't push the boundary if changeset already |
|
707 | 709 | # exists locally as secret |
|
708 | 710 | # |
|
709 | 711 | # We should not use added here but the list of all change in |
|
710 | 712 | # the bundle |
|
711 | 713 | if publishing: |
|
712 | 714 | phases.advanceboundary(repo, tr, phases.public, srccontent) |
|
713 | 715 | else: |
|
714 | 716 | # Those changesets have been pushed from the outside, their |
|
715 | 717 | # phases are going to be pushed alongside. Therefor |
|
716 | 718 | # `targetphase` is ignored. |
|
717 | 719 | phases.advanceboundary(repo, tr, phases.draft, srccontent) |
|
718 | 720 | phases.retractboundary(repo, tr, phases.draft, added) |
|
719 | 721 | elif srctype != 'strip': |
|
720 | 722 | # publishing only alter behavior during push |
|
721 | 723 | # |
|
722 | 724 | # strip should not touch boundary at all |
|
723 | 725 | phases.retractboundary(repo, tr, targetphase, added) |
|
724 | 726 | |
|
725 | 727 | # make changelog see real files again |
|
726 | 728 | cl.finalize(trp) |
|
727 | 729 | |
|
728 | 730 | tr.close() |
|
729 | 731 | |
|
730 | 732 | if changesets > 0: |
|
731 | 733 | if srctype != 'strip': |
|
732 | 734 | # During strip, branchcache is invalid but coming call to |
|
733 | 735 | # `destroyed` will repair it. |
|
734 | 736 | # In other case we can safely update cache on disk. |
|
735 | 737 | branchmap.updatecache(repo.filtered('served')) |
|
736 | 738 | |
|
737 | 739 | def runhooks(): |
|
738 | 740 | # These hooks run when the lock releases, not when the |
|
739 | 741 | # transaction closes. So it's possible for the changelog |
|
740 | 742 | # to have changed since we last saw it. |
|
741 | 743 | if clstart >= len(repo): |
|
742 | 744 | return |
|
743 | 745 | |
|
744 | 746 | # forcefully update the on-disk branch cache |
|
745 | 747 | repo.ui.debug("updating the branch cache\n") |
|
746 | 748 | repo.hook("changegroup", **hookargs) |
|
747 | 749 | |
|
748 | 750 | for n in added: |
|
749 | 751 | args = hookargs.copy() |
|
750 | 752 | args['node'] = hex(n) |
|
751 | 753 | repo.hook("incoming", **args) |
|
752 | 754 | |
|
753 | 755 | newheads = [h for h in repo.heads() if h not in oldheads] |
|
754 | 756 | repo.ui.log("incoming", |
|
755 | 757 | "%s incoming changes - new heads: %s\n", |
|
756 | 758 | len(added), |
|
757 | 759 | ', '.join([hex(c[:6]) for c in newheads])) |
|
758 | 760 | repo._afterlock(runhooks) |
|
759 | 761 | |
|
760 | 762 | finally: |
|
761 | 763 | tr.release() |
|
762 | 764 | # never return 0 here: |
|
763 | 765 | if dh < 0: |
|
764 | 766 | return dh - 1 |
|
765 | 767 | else: |
|
766 | 768 | return dh + 1 |
General Comments 0
You need to be logged in to leave comments.
Login now