Show More
@@ -1,637 +1,639 b'' | |||
|
1 | 1 | # Copyright 2009-2010 Gregory P. Ward |
|
2 | 2 | # Copyright 2009-2010 Intelerad Medical Systems Incorporated |
|
3 | 3 | # Copyright 2010-2011 Fog Creek Software |
|
4 | 4 | # Copyright 2010-2011 Unity Technologies |
|
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 | '''largefiles utility code: must not import other modules in this package.''' |
|
10 | 10 | |
|
11 | 11 | import os |
|
12 | 12 | import platform |
|
13 | 13 | import stat |
|
14 | 14 | import copy |
|
15 | 15 | |
|
16 | 16 | from mercurial import dirstate, httpconnection, match as match_, util, scmutil |
|
17 | 17 | from mercurial.i18n import _ |
|
18 | 18 | from mercurial import node, error |
|
19 | 19 | |
|
20 | 20 | shortname = '.hglf' |
|
21 | 21 | shortnameslash = shortname + '/' |
|
22 | 22 | longname = 'largefiles' |
|
23 | 23 | |
|
24 | 24 | |
|
25 | 25 | # -- Private worker functions ------------------------------------------ |
|
26 | 26 | |
|
27 | 27 | def getminsize(ui, assumelfiles, opt, default=10): |
|
28 | 28 | lfsize = opt |
|
29 | 29 | if not lfsize and assumelfiles: |
|
30 | 30 | lfsize = ui.config(longname, 'minsize', default=default) |
|
31 | 31 | if lfsize: |
|
32 | 32 | try: |
|
33 | 33 | lfsize = float(lfsize) |
|
34 | 34 | except ValueError: |
|
35 | 35 | raise error.Abort(_('largefiles: size must be number (not %s)\n') |
|
36 | 36 | % lfsize) |
|
37 | 37 | if lfsize is None: |
|
38 | 38 | raise error.Abort(_('minimum size for largefiles must be specified')) |
|
39 | 39 | return lfsize |
|
40 | 40 | |
|
41 | 41 | def link(src, dest): |
|
42 | 42 | util.makedirs(os.path.dirname(dest)) |
|
43 | 43 | try: |
|
44 | 44 | util.oslink(src, dest) |
|
45 | 45 | except OSError: |
|
46 | 46 | # if hardlinks fail, fallback on atomic copy |
|
47 | 47 | dst = util.atomictempfile(dest) |
|
48 | 48 | for chunk in util.filechunkiter(open(src, 'rb')): |
|
49 | 49 | dst.write(chunk) |
|
50 | 50 | dst.close() |
|
51 | 51 | os.chmod(dest, os.stat(src).st_mode) |
|
52 | 52 | |
|
53 | 53 | def usercachepath(ui, hash): |
|
54 | 54 | path = ui.configpath(longname, 'usercache', None) |
|
55 | 55 | if path: |
|
56 | 56 | path = os.path.join(path, hash) |
|
57 | 57 | else: |
|
58 | 58 | if os.name == 'nt': |
|
59 | 59 | appdata = os.getenv('LOCALAPPDATA', os.getenv('APPDATA')) |
|
60 | 60 | if appdata: |
|
61 | 61 | path = os.path.join(appdata, longname, hash) |
|
62 | 62 | elif platform.system() == 'Darwin': |
|
63 | 63 | home = os.getenv('HOME') |
|
64 | 64 | if home: |
|
65 | 65 | path = os.path.join(home, 'Library', 'Caches', |
|
66 | 66 | longname, hash) |
|
67 | 67 | elif os.name == 'posix': |
|
68 | 68 | path = os.getenv('XDG_CACHE_HOME') |
|
69 | 69 | if path: |
|
70 | 70 | path = os.path.join(path, longname, hash) |
|
71 | 71 | else: |
|
72 | 72 | home = os.getenv('HOME') |
|
73 | 73 | if home: |
|
74 | 74 | path = os.path.join(home, '.cache', longname, hash) |
|
75 | 75 | else: |
|
76 | 76 | raise error.Abort(_('unknown operating system: %s\n') % os.name) |
|
77 | 77 | return path |
|
78 | 78 | |
|
79 | 79 | def inusercache(ui, hash): |
|
80 | 80 | path = usercachepath(ui, hash) |
|
81 | 81 | return path and os.path.exists(path) |
|
82 | 82 | |
|
83 | 83 | def findfile(repo, hash): |
|
84 | 84 | path, exists = findstorepath(repo, hash) |
|
85 | 85 | if exists: |
|
86 | 86 | repo.ui.note(_('found %s in store\n') % hash) |
|
87 | 87 | return path |
|
88 | 88 | elif inusercache(repo.ui, hash): |
|
89 | 89 | repo.ui.note(_('found %s in system cache\n') % hash) |
|
90 | 90 | path = storepath(repo, hash) |
|
91 | 91 | link(usercachepath(repo.ui, hash), path) |
|
92 | 92 | return path |
|
93 | 93 | return None |
|
94 | 94 | |
|
95 | 95 | class largefilesdirstate(dirstate.dirstate): |
|
96 | 96 | def __getitem__(self, key): |
|
97 | 97 | return super(largefilesdirstate, self).__getitem__(unixpath(key)) |
|
98 | 98 | def normal(self, f): |
|
99 | 99 | return super(largefilesdirstate, self).normal(unixpath(f)) |
|
100 | 100 | def remove(self, f): |
|
101 | 101 | return super(largefilesdirstate, self).remove(unixpath(f)) |
|
102 | 102 | def add(self, f): |
|
103 | 103 | return super(largefilesdirstate, self).add(unixpath(f)) |
|
104 | 104 | def drop(self, f): |
|
105 | 105 | return super(largefilesdirstate, self).drop(unixpath(f)) |
|
106 | 106 | def forget(self, f): |
|
107 | 107 | return super(largefilesdirstate, self).forget(unixpath(f)) |
|
108 | 108 | def normallookup(self, f): |
|
109 | 109 | return super(largefilesdirstate, self).normallookup(unixpath(f)) |
|
110 | 110 | def _ignore(self, f): |
|
111 | 111 | return False |
|
112 | 112 | def write(self, tr=False): |
|
113 | 113 | # (1) disable PENDING mode always |
|
114 | 114 | # (lfdirstate isn't yet managed as a part of the transaction) |
|
115 | 115 | # (2) avoid develwarn 'use dirstate.write with ....' |
|
116 | 116 | super(largefilesdirstate, self).write(None) |
|
117 | 117 | |
|
118 | 118 | def openlfdirstate(ui, repo, create=True): |
|
119 | 119 | ''' |
|
120 | 120 | Return a dirstate object that tracks largefiles: i.e. its root is |
|
121 | 121 | the repo root, but it is saved in .hg/largefiles/dirstate. |
|
122 | 122 | ''' |
|
123 | 123 | lfstoredir = repo.join(longname) |
|
124 | 124 | opener = scmutil.opener(lfstoredir) |
|
125 | 125 | lfdirstate = largefilesdirstate(opener, ui, repo.root, |
|
126 | 126 | repo.dirstate._validate) |
|
127 | 127 | |
|
128 | 128 | # If the largefiles dirstate does not exist, populate and create |
|
129 | 129 | # it. This ensures that we create it on the first meaningful |
|
130 | 130 | # largefiles operation in a new clone. |
|
131 | 131 | if create and not os.path.exists(os.path.join(lfstoredir, 'dirstate')): |
|
132 | 132 | matcher = getstandinmatcher(repo) |
|
133 | 133 | standins = repo.dirstate.walk(matcher, [], False, False) |
|
134 | 134 | |
|
135 | 135 | if len(standins) > 0: |
|
136 | 136 | util.makedirs(lfstoredir) |
|
137 | 137 | |
|
138 | 138 | for standin in standins: |
|
139 | 139 | lfile = splitstandin(standin) |
|
140 | 140 | lfdirstate.normallookup(lfile) |
|
141 | 141 | return lfdirstate |
|
142 | 142 | |
|
143 | 143 | def lfdirstatestatus(lfdirstate, repo): |
|
144 | 144 | wctx = repo['.'] |
|
145 | 145 | match = match_.always(repo.root, repo.getcwd()) |
|
146 | 146 | unsure, s = lfdirstate.status(match, [], False, False, False) |
|
147 | 147 | modified, clean = s.modified, s.clean |
|
148 | 148 | for lfile in unsure: |
|
149 | 149 | try: |
|
150 | 150 | fctx = wctx[standin(lfile)] |
|
151 | 151 | except LookupError: |
|
152 | 152 | fctx = None |
|
153 | 153 | if not fctx or fctx.data().strip() != hashfile(repo.wjoin(lfile)): |
|
154 | 154 | modified.append(lfile) |
|
155 | 155 | else: |
|
156 | 156 | clean.append(lfile) |
|
157 | 157 | lfdirstate.normal(lfile) |
|
158 | 158 | return s |
|
159 | 159 | |
|
160 | 160 | def listlfiles(repo, rev=None, matcher=None): |
|
161 | 161 | '''return a list of largefiles in the working copy or the |
|
162 | 162 | specified changeset''' |
|
163 | 163 | |
|
164 | 164 | if matcher is None: |
|
165 | 165 | matcher = getstandinmatcher(repo) |
|
166 | 166 | |
|
167 | 167 | # ignore unknown files in working directory |
|
168 | 168 | return [splitstandin(f) |
|
169 | 169 | for f in repo[rev].walk(matcher) |
|
170 | 170 | if rev is not None or repo.dirstate[f] != '?'] |
|
171 | 171 | |
|
172 | 172 | def instore(repo, hash, forcelocal=False): |
|
173 | 173 | return os.path.exists(storepath(repo, hash, forcelocal)) |
|
174 | 174 | |
|
175 | 175 | def storepath(repo, hash, forcelocal=False): |
|
176 | 176 | if not forcelocal and repo.shared(): |
|
177 | 177 | return repo.vfs.reljoin(repo.sharedpath, longname, hash) |
|
178 | 178 | return repo.join(longname, hash) |
|
179 | 179 | |
|
180 | 180 | def findstorepath(repo, hash): |
|
181 | 181 | '''Search through the local store path(s) to find the file for the given |
|
182 | 182 | hash. If the file is not found, its path in the primary store is returned. |
|
183 | 183 | The return value is a tuple of (path, exists(path)). |
|
184 | 184 | ''' |
|
185 | 185 | # For shared repos, the primary store is in the share source. But for |
|
186 | 186 | # backward compatibility, force a lookup in the local store if it wasn't |
|
187 | 187 | # found in the share source. |
|
188 | 188 | path = storepath(repo, hash, False) |
|
189 | 189 | |
|
190 | 190 | if instore(repo, hash): |
|
191 | 191 | return (path, True) |
|
192 | 192 | elif repo.shared() and instore(repo, hash, True): |
|
193 | 193 | return storepath(repo, hash, True) |
|
194 | 194 | |
|
195 | 195 | return (path, False) |
|
196 | 196 | |
|
197 | 197 | def copyfromcache(repo, hash, filename): |
|
198 | 198 | '''Copy the specified largefile from the repo or system cache to |
|
199 | 199 | filename in the repository. Return true on success or false if the |
|
200 | 200 | file was not found in either cache (which should not happened: |
|
201 | 201 | this is meant to be called only after ensuring that the needed |
|
202 | 202 | largefile exists in the cache).''' |
|
203 | 203 | path = findfile(repo, hash) |
|
204 | 204 | if path is None: |
|
205 | 205 | return False |
|
206 | 206 | util.makedirs(os.path.dirname(repo.wjoin(filename))) |
|
207 | 207 | # The write may fail before the file is fully written, but we |
|
208 | 208 | # don't use atomic writes in the working copy. |
|
209 | 209 | dest = repo.wjoin(filename) |
|
210 | 210 | with open(path, 'rb') as srcfd: |
|
211 | 211 | with open(dest, 'wb') as destfd: |
|
212 | 212 | gothash = copyandhash(srcfd, destfd) |
|
213 | 213 | if gothash != hash: |
|
214 | 214 | repo.ui.warn(_('%s: data corruption in %s with hash %s\n') |
|
215 | 215 | % (filename, path, gothash)) |
|
216 | 216 | util.unlink(dest) |
|
217 | 217 | return False |
|
218 | 218 | return True |
|
219 | 219 | |
|
220 | 220 | def copytostore(repo, rev, file, uploaded=False): |
|
221 | 221 | hash = readstandin(repo, file, rev) |
|
222 | 222 | if instore(repo, hash): |
|
223 | 223 | return |
|
224 | 224 | absfile = repo.wjoin(file) |
|
225 | 225 | if os.path.exists(absfile): |
|
226 | 226 | copytostoreabsolute(repo, absfile, hash) |
|
227 | 227 | else: |
|
228 | 228 | repo.ui.warn(_("%s: largefile %s not available from local store\n") % |
|
229 | 229 | (file, hash)) |
|
230 | 230 | |
|
231 | 231 | def copyalltostore(repo, node): |
|
232 | 232 | '''Copy all largefiles in a given revision to the store''' |
|
233 | 233 | |
|
234 | 234 | ctx = repo[node] |
|
235 | 235 | for filename in ctx.files(): |
|
236 | 236 | if isstandin(filename) and filename in ctx.manifest(): |
|
237 | 237 | realfile = splitstandin(filename) |
|
238 | 238 | copytostore(repo, ctx.node(), realfile) |
|
239 | 239 | |
|
240 | 240 | |
|
241 | 241 | def copytostoreabsolute(repo, file, hash): |
|
242 | 242 | if inusercache(repo.ui, hash): |
|
243 | 243 | link(usercachepath(repo.ui, hash), storepath(repo, hash)) |
|
244 | 244 | else: |
|
245 | 245 | util.makedirs(os.path.dirname(storepath(repo, hash))) |
|
246 | 246 | dst = util.atomictempfile(storepath(repo, hash), |
|
247 | 247 | createmode=repo.store.createmode) |
|
248 | 248 | for chunk in util.filechunkiter(open(file, 'rb')): |
|
249 | 249 | dst.write(chunk) |
|
250 | 250 | dst.close() |
|
251 | 251 | linktousercache(repo, hash) |
|
252 | 252 | |
|
253 | 253 | def linktousercache(repo, hash): |
|
254 | 254 | path = usercachepath(repo.ui, hash) |
|
255 | 255 | if path: |
|
256 | 256 | link(storepath(repo, hash), path) |
|
257 | 257 | |
|
258 | 258 | def getstandinmatcher(repo, rmatcher=None): |
|
259 | 259 | '''Return a match object that applies rmatcher to the standin directory''' |
|
260 | 260 | standindir = repo.wjoin(shortname) |
|
261 | 261 | |
|
262 | 262 | # no warnings about missing files or directories |
|
263 | 263 | badfn = lambda f, msg: None |
|
264 | 264 | |
|
265 | 265 | if rmatcher and not rmatcher.always(): |
|
266 | 266 | pats = [os.path.join(standindir, pat) for pat in rmatcher.files()] |
|
267 | 267 | if not pats: |
|
268 | 268 | pats = [standindir] |
|
269 | 269 | match = scmutil.match(repo[None], pats, badfn=badfn) |
|
270 | 270 | # if pats is empty, it would incorrectly always match, so clear _always |
|
271 | 271 | match._always = False |
|
272 | 272 | else: |
|
273 | 273 | # no patterns: relative to repo root |
|
274 | 274 | match = scmutil.match(repo[None], [standindir], badfn=badfn) |
|
275 | 275 | return match |
|
276 | 276 | |
|
277 | 277 | def composestandinmatcher(repo, rmatcher): |
|
278 | 278 | '''Return a matcher that accepts standins corresponding to the |
|
279 | 279 | files accepted by rmatcher. Pass the list of files in the matcher |
|
280 | 280 | as the paths specified by the user.''' |
|
281 | 281 | smatcher = getstandinmatcher(repo, rmatcher) |
|
282 | 282 | isstandin = smatcher.matchfn |
|
283 | 283 | def composedmatchfn(f): |
|
284 | 284 | return isstandin(f) and rmatcher.matchfn(splitstandin(f)) |
|
285 | 285 | smatcher.matchfn = composedmatchfn |
|
286 | 286 | |
|
287 | 287 | return smatcher |
|
288 | 288 | |
|
289 | 289 | def standin(filename): |
|
290 | 290 | '''Return the repo-relative path to the standin for the specified big |
|
291 | 291 | file.''' |
|
292 | 292 | # Notes: |
|
293 | 293 | # 1) Some callers want an absolute path, but for instance addlargefiles |
|
294 | 294 | # needs it repo-relative so it can be passed to repo[None].add(). So |
|
295 | 295 | # leave it up to the caller to use repo.wjoin() to get an absolute path. |
|
296 | 296 | # 2) Join with '/' because that's what dirstate always uses, even on |
|
297 | 297 | # Windows. Change existing separator to '/' first in case we are |
|
298 | 298 | # passed filenames from an external source (like the command line). |
|
299 | 299 | return shortnameslash + util.pconvert(filename) |
|
300 | 300 | |
|
301 | 301 | def isstandin(filename): |
|
302 | 302 | '''Return true if filename is a big file standin. filename must be |
|
303 | 303 | in Mercurial's internal form (slash-separated).''' |
|
304 | 304 | return filename.startswith(shortnameslash) |
|
305 | 305 | |
|
306 | 306 | def splitstandin(filename): |
|
307 | 307 | # Split on / because that's what dirstate always uses, even on Windows. |
|
308 | 308 | # Change local separator to / first just in case we are passed filenames |
|
309 | 309 | # from an external source (like the command line). |
|
310 | 310 | bits = util.pconvert(filename).split('/', 1) |
|
311 | 311 | if len(bits) == 2 and bits[0] == shortname: |
|
312 | 312 | return bits[1] |
|
313 | 313 | else: |
|
314 | 314 | return None |
|
315 | 315 | |
|
316 | 316 | def updatestandin(repo, standin): |
|
317 | 317 | file = repo.wjoin(splitstandin(standin)) |
|
318 | 318 | if os.path.exists(file): |
|
319 | 319 | hash = hashfile(file) |
|
320 | 320 | executable = getexecutable(file) |
|
321 | 321 | writestandin(repo, standin, hash, executable) |
|
322 | else: | |
|
323 | raise error.Abort(_('%s: file not found!') % splitstandin(standin)) | |
|
322 | 324 | |
|
323 | 325 | def readstandin(repo, filename, node=None): |
|
324 | 326 | '''read hex hash from standin for filename at given node, or working |
|
325 | 327 | directory if no node is given''' |
|
326 | 328 | return repo[node][standin(filename)].data().strip() |
|
327 | 329 | |
|
328 | 330 | def writestandin(repo, standin, hash, executable): |
|
329 | 331 | '''write hash to <repo.root>/<standin>''' |
|
330 | 332 | repo.wwrite(standin, hash + '\n', executable and 'x' or '') |
|
331 | 333 | |
|
332 | 334 | def copyandhash(instream, outfile): |
|
333 | 335 | '''Read bytes from instream (iterable) and write them to outfile, |
|
334 | 336 | computing the SHA-1 hash of the data along the way. Return the hash.''' |
|
335 | 337 | hasher = util.sha1('') |
|
336 | 338 | for data in instream: |
|
337 | 339 | hasher.update(data) |
|
338 | 340 | outfile.write(data) |
|
339 | 341 | return hasher.hexdigest() |
|
340 | 342 | |
|
341 | 343 | def hashrepofile(repo, file): |
|
342 | 344 | return hashfile(repo.wjoin(file)) |
|
343 | 345 | |
|
344 | 346 | def hashfile(file): |
|
345 | 347 | if not os.path.exists(file): |
|
346 | 348 | return '' |
|
347 | 349 | hasher = util.sha1('') |
|
348 | 350 | fd = open(file, 'rb') |
|
349 | 351 | for data in util.filechunkiter(fd, 128 * 1024): |
|
350 | 352 | hasher.update(data) |
|
351 | 353 | fd.close() |
|
352 | 354 | return hasher.hexdigest() |
|
353 | 355 | |
|
354 | 356 | def getexecutable(filename): |
|
355 | 357 | mode = os.stat(filename).st_mode |
|
356 | 358 | return ((mode & stat.S_IXUSR) and |
|
357 | 359 | (mode & stat.S_IXGRP) and |
|
358 | 360 | (mode & stat.S_IXOTH)) |
|
359 | 361 | |
|
360 | 362 | def urljoin(first, second, *arg): |
|
361 | 363 | def join(left, right): |
|
362 | 364 | if not left.endswith('/'): |
|
363 | 365 | left += '/' |
|
364 | 366 | if right.startswith('/'): |
|
365 | 367 | right = right[1:] |
|
366 | 368 | return left + right |
|
367 | 369 | |
|
368 | 370 | url = join(first, second) |
|
369 | 371 | for a in arg: |
|
370 | 372 | url = join(url, a) |
|
371 | 373 | return url |
|
372 | 374 | |
|
373 | 375 | def hexsha1(data): |
|
374 | 376 | """hexsha1 returns the hex-encoded sha1 sum of the data in the file-like |
|
375 | 377 | object data""" |
|
376 | 378 | h = util.sha1() |
|
377 | 379 | for chunk in util.filechunkiter(data): |
|
378 | 380 | h.update(chunk) |
|
379 | 381 | return h.hexdigest() |
|
380 | 382 | |
|
381 | 383 | def httpsendfile(ui, filename): |
|
382 | 384 | return httpconnection.httpsendfile(ui, filename, 'rb') |
|
383 | 385 | |
|
384 | 386 | def unixpath(path): |
|
385 | 387 | '''Return a version of path normalized for use with the lfdirstate.''' |
|
386 | 388 | return util.pconvert(os.path.normpath(path)) |
|
387 | 389 | |
|
388 | 390 | def islfilesrepo(repo): |
|
389 | 391 | if ('largefiles' in repo.requirements and |
|
390 | 392 | any(shortnameslash in f[0] for f in repo.store.datafiles())): |
|
391 | 393 | return True |
|
392 | 394 | |
|
393 | 395 | return any(openlfdirstate(repo.ui, repo, False)) |
|
394 | 396 | |
|
395 | 397 | class storeprotonotcapable(Exception): |
|
396 | 398 | def __init__(self, storetypes): |
|
397 | 399 | self.storetypes = storetypes |
|
398 | 400 | |
|
399 | 401 | def getstandinsstate(repo): |
|
400 | 402 | standins = [] |
|
401 | 403 | matcher = getstandinmatcher(repo) |
|
402 | 404 | for standin in repo.dirstate.walk(matcher, [], False, False): |
|
403 | 405 | lfile = splitstandin(standin) |
|
404 | 406 | try: |
|
405 | 407 | hash = readstandin(repo, lfile) |
|
406 | 408 | except IOError: |
|
407 | 409 | hash = None |
|
408 | 410 | standins.append((lfile, hash)) |
|
409 | 411 | return standins |
|
410 | 412 | |
|
411 | 413 | def synclfdirstate(repo, lfdirstate, lfile, normallookup): |
|
412 | 414 | lfstandin = standin(lfile) |
|
413 | 415 | if lfstandin in repo.dirstate: |
|
414 | 416 | stat = repo.dirstate._map[lfstandin] |
|
415 | 417 | state, mtime = stat[0], stat[3] |
|
416 | 418 | else: |
|
417 | 419 | state, mtime = '?', -1 |
|
418 | 420 | if state == 'n': |
|
419 | 421 | if (normallookup or mtime < 0 or |
|
420 | 422 | not os.path.exists(repo.wjoin(lfile))): |
|
421 | 423 | # state 'n' doesn't ensure 'clean' in this case |
|
422 | 424 | lfdirstate.normallookup(lfile) |
|
423 | 425 | else: |
|
424 | 426 | lfdirstate.normal(lfile) |
|
425 | 427 | elif state == 'm': |
|
426 | 428 | lfdirstate.normallookup(lfile) |
|
427 | 429 | elif state == 'r': |
|
428 | 430 | lfdirstate.remove(lfile) |
|
429 | 431 | elif state == 'a': |
|
430 | 432 | lfdirstate.add(lfile) |
|
431 | 433 | elif state == '?': |
|
432 | 434 | lfdirstate.drop(lfile) |
|
433 | 435 | |
|
434 | 436 | def markcommitted(orig, ctx, node): |
|
435 | 437 | repo = ctx.repo() |
|
436 | 438 | |
|
437 | 439 | orig(node) |
|
438 | 440 | |
|
439 | 441 | # ATTENTION: "ctx.files()" may differ from "repo[node].files()" |
|
440 | 442 | # because files coming from the 2nd parent are omitted in the latter. |
|
441 | 443 | # |
|
442 | 444 | # The former should be used to get targets of "synclfdirstate", |
|
443 | 445 | # because such files: |
|
444 | 446 | # - are marked as "a" by "patch.patch()" (e.g. via transplant), and |
|
445 | 447 | # - have to be marked as "n" after commit, but |
|
446 | 448 | # - aren't listed in "repo[node].files()" |
|
447 | 449 | |
|
448 | 450 | lfdirstate = openlfdirstate(repo.ui, repo) |
|
449 | 451 | for f in ctx.files(): |
|
450 | 452 | if isstandin(f): |
|
451 | 453 | lfile = splitstandin(f) |
|
452 | 454 | synclfdirstate(repo, lfdirstate, lfile, False) |
|
453 | 455 | lfdirstate.write() |
|
454 | 456 | |
|
455 | 457 | # As part of committing, copy all of the largefiles into the cache. |
|
456 | 458 | copyalltostore(repo, node) |
|
457 | 459 | |
|
458 | 460 | def getlfilestoupdate(oldstandins, newstandins): |
|
459 | 461 | changedstandins = set(oldstandins).symmetric_difference(set(newstandins)) |
|
460 | 462 | filelist = [] |
|
461 | 463 | for f in changedstandins: |
|
462 | 464 | if f[0] not in filelist: |
|
463 | 465 | filelist.append(f[0]) |
|
464 | 466 | return filelist |
|
465 | 467 | |
|
466 | 468 | def getlfilestoupload(repo, missing, addfunc): |
|
467 | 469 | for i, n in enumerate(missing): |
|
468 | 470 | repo.ui.progress(_('finding outgoing largefiles'), i, |
|
469 | 471 | unit=_('revision'), total=len(missing)) |
|
470 | 472 | parents = [p for p in repo.changelog.parents(n) if p != node.nullid] |
|
471 | 473 | |
|
472 | 474 | oldlfstatus = repo.lfstatus |
|
473 | 475 | repo.lfstatus = False |
|
474 | 476 | try: |
|
475 | 477 | ctx = repo[n] |
|
476 | 478 | finally: |
|
477 | 479 | repo.lfstatus = oldlfstatus |
|
478 | 480 | |
|
479 | 481 | files = set(ctx.files()) |
|
480 | 482 | if len(parents) == 2: |
|
481 | 483 | mc = ctx.manifest() |
|
482 | 484 | mp1 = ctx.parents()[0].manifest() |
|
483 | 485 | mp2 = ctx.parents()[1].manifest() |
|
484 | 486 | for f in mp1: |
|
485 | 487 | if f not in mc: |
|
486 | 488 | files.add(f) |
|
487 | 489 | for f in mp2: |
|
488 | 490 | if f not in mc: |
|
489 | 491 | files.add(f) |
|
490 | 492 | for f in mc: |
|
491 | 493 | if mc[f] != mp1.get(f, None) or mc[f] != mp2.get(f, None): |
|
492 | 494 | files.add(f) |
|
493 | 495 | for fn in files: |
|
494 | 496 | if isstandin(fn) and fn in ctx: |
|
495 | 497 | addfunc(fn, ctx[fn].data().strip()) |
|
496 | 498 | repo.ui.progress(_('finding outgoing largefiles'), None) |
|
497 | 499 | |
|
498 | 500 | def updatestandinsbymatch(repo, match): |
|
499 | 501 | '''Update standins in the working directory according to specified match |
|
500 | 502 | |
|
501 | 503 | This returns (possibly modified) ``match`` object to be used for |
|
502 | 504 | subsequent commit process. |
|
503 | 505 | ''' |
|
504 | 506 | |
|
505 | 507 | ui = repo.ui |
|
506 | 508 | |
|
507 | 509 | # Case 1: user calls commit with no specific files or |
|
508 | 510 | # include/exclude patterns: refresh and commit all files that |
|
509 | 511 | # are "dirty". |
|
510 | 512 | if match is None or match.always(): |
|
511 | 513 | # Spend a bit of time here to get a list of files we know |
|
512 | 514 | # are modified so we can compare only against those. |
|
513 | 515 | # It can cost a lot of time (several seconds) |
|
514 | 516 | # otherwise to update all standins if the largefiles are |
|
515 | 517 | # large. |
|
516 | 518 | lfdirstate = openlfdirstate(ui, repo) |
|
517 | 519 | dirtymatch = match_.always(repo.root, repo.getcwd()) |
|
518 | 520 | unsure, s = lfdirstate.status(dirtymatch, [], False, False, |
|
519 | 521 | False) |
|
520 | 522 | modifiedfiles = unsure + s.modified + s.added + s.removed |
|
521 | 523 | lfiles = listlfiles(repo) |
|
522 | 524 | # this only loops through largefiles that exist (not |
|
523 | 525 | # removed/renamed) |
|
524 | 526 | for lfile in lfiles: |
|
525 | 527 | if lfile in modifiedfiles: |
|
526 | 528 | if os.path.exists( |
|
527 | 529 | repo.wjoin(standin(lfile))): |
|
528 | 530 | # this handles the case where a rebase is being |
|
529 | 531 | # performed and the working copy is not updated |
|
530 | 532 | # yet. |
|
531 | 533 | if os.path.exists(repo.wjoin(lfile)): |
|
532 | 534 | updatestandin(repo, |
|
533 | 535 | standin(lfile)) |
|
534 | 536 | |
|
535 | 537 | return match |
|
536 | 538 | |
|
537 | 539 | lfiles = listlfiles(repo) |
|
538 | 540 | match._files = repo._subdirlfs(match.files(), lfiles) |
|
539 | 541 | |
|
540 | 542 | # Case 2: user calls commit with specified patterns: refresh |
|
541 | 543 | # any matching big files. |
|
542 | 544 | smatcher = composestandinmatcher(repo, match) |
|
543 | 545 | standins = repo.dirstate.walk(smatcher, [], False, False) |
|
544 | 546 | |
|
545 | 547 | # No matching big files: get out of the way and pass control to |
|
546 | 548 | # the usual commit() method. |
|
547 | 549 | if not standins: |
|
548 | 550 | return match |
|
549 | 551 | |
|
550 | 552 | # Refresh all matching big files. It's possible that the |
|
551 | 553 | # commit will end up failing, in which case the big files will |
|
552 | 554 | # stay refreshed. No harm done: the user modified them and |
|
553 | 555 | # asked to commit them, so sooner or later we're going to |
|
554 | 556 | # refresh the standins. Might as well leave them refreshed. |
|
555 | 557 | lfdirstate = openlfdirstate(ui, repo) |
|
556 | 558 | for fstandin in standins: |
|
557 | 559 | lfile = splitstandin(fstandin) |
|
558 | 560 | if lfdirstate[lfile] != 'r': |
|
559 | 561 | updatestandin(repo, fstandin) |
|
560 | 562 | |
|
561 | 563 | # Cook up a new matcher that only matches regular files or |
|
562 | 564 | # standins corresponding to the big files requested by the |
|
563 | 565 | # user. Have to modify _files to prevent commit() from |
|
564 | 566 | # complaining "not tracked" for big files. |
|
565 | 567 | match = copy.copy(match) |
|
566 | 568 | origmatchfn = match.matchfn |
|
567 | 569 | |
|
568 | 570 | # Check both the list of largefiles and the list of |
|
569 | 571 | # standins because if a largefile was removed, it |
|
570 | 572 | # won't be in the list of largefiles at this point |
|
571 | 573 | match._files += sorted(standins) |
|
572 | 574 | |
|
573 | 575 | actualfiles = [] |
|
574 | 576 | for f in match._files: |
|
575 | 577 | fstandin = standin(f) |
|
576 | 578 | |
|
577 | 579 | # For largefiles, only one of the normal and standin should be |
|
578 | 580 | # committed (except if one of them is a remove). In the case of a |
|
579 | 581 | # standin removal, drop the normal file if it is unknown to dirstate. |
|
580 | 582 | # Thus, skip plain largefile names but keep the standin. |
|
581 | 583 | if f in lfiles or fstandin in standins: |
|
582 | 584 | if repo.dirstate[fstandin] != 'r': |
|
583 | 585 | if repo.dirstate[f] != 'r': |
|
584 | 586 | continue |
|
585 | 587 | elif repo.dirstate[f] == '?': |
|
586 | 588 | continue |
|
587 | 589 | |
|
588 | 590 | actualfiles.append(f) |
|
589 | 591 | match._files = actualfiles |
|
590 | 592 | |
|
591 | 593 | def matchfn(f): |
|
592 | 594 | if origmatchfn(f): |
|
593 | 595 | return f not in lfiles |
|
594 | 596 | else: |
|
595 | 597 | return f in standins |
|
596 | 598 | |
|
597 | 599 | match.matchfn = matchfn |
|
598 | 600 | |
|
599 | 601 | return match |
|
600 | 602 | |
|
601 | 603 | class automatedcommithook(object): |
|
602 | 604 | '''Stateful hook to update standins at the 1st commit of resuming |
|
603 | 605 | |
|
604 | 606 | For efficiency, updating standins in the working directory should |
|
605 | 607 | be avoided while automated committing (like rebase, transplant and |
|
606 | 608 | so on), because they should be updated before committing. |
|
607 | 609 | |
|
608 | 610 | But the 1st commit of resuming automated committing (e.g. ``rebase |
|
609 | 611 | --continue``) should update them, because largefiles may be |
|
610 | 612 | modified manually. |
|
611 | 613 | ''' |
|
612 | 614 | def __init__(self, resuming): |
|
613 | 615 | self.resuming = resuming |
|
614 | 616 | |
|
615 | 617 | def __call__(self, repo, match): |
|
616 | 618 | if self.resuming: |
|
617 | 619 | self.resuming = False # avoids updating at subsequent commits |
|
618 | 620 | return updatestandinsbymatch(repo, match) |
|
619 | 621 | else: |
|
620 | 622 | return match |
|
621 | 623 | |
|
622 | 624 | def getstatuswriter(ui, repo, forcibly=None): |
|
623 | 625 | '''Return the function to write largefiles specific status out |
|
624 | 626 | |
|
625 | 627 | If ``forcibly`` is ``None``, this returns the last element of |
|
626 | 628 | ``repo._lfstatuswriters`` as "default" writer function. |
|
627 | 629 | |
|
628 | 630 | Otherwise, this returns the function to always write out (or |
|
629 | 631 | ignore if ``not forcibly``) status. |
|
630 | 632 | ''' |
|
631 | 633 | if forcibly is None and util.safehasattr(repo, '_largefilesenabled'): |
|
632 | 634 | return repo._lfstatuswriters[-1] |
|
633 | 635 | else: |
|
634 | 636 | if forcibly: |
|
635 | 637 | return ui.status # forcibly WRITE OUT |
|
636 | 638 | else: |
|
637 | 639 | return lambda *msg, **opts: None # forcibly IGNORE |
@@ -1,231 +1,237 b'' | |||
|
1 | 1 | Create user cache directory |
|
2 | 2 | |
|
3 | 3 | $ USERCACHE=`pwd`/cache; export USERCACHE |
|
4 | 4 | $ cat <<EOF >> ${HGRCPATH} |
|
5 | 5 | > [extensions] |
|
6 | 6 | > hgext.largefiles= |
|
7 | 7 | > [largefiles] |
|
8 | 8 | > usercache=${USERCACHE} |
|
9 | 9 | > EOF |
|
10 | 10 | $ mkdir -p ${USERCACHE} |
|
11 | 11 | |
|
12 | 12 | Create source repo, and commit adding largefile. |
|
13 | 13 | |
|
14 | 14 | $ hg init src |
|
15 | 15 | $ cd src |
|
16 | 16 | $ echo large > large |
|
17 | 17 | $ hg add --large large |
|
18 | 18 | $ hg commit -m 'add largefile' |
|
19 | 19 | $ hg rm large |
|
20 | 20 | $ hg commit -m 'branchhead without largefile' large |
|
21 | 21 | $ hg up -qr 0 |
|
22 | $ rm large | |
|
23 | $ echo "0000000000000000000000000000000000000000" > .hglf/large | |
|
24 | $ hg commit -m 'commit missing file with corrupt standin' large | |
|
25 | abort: large: file not found! | |
|
26 | [255] | |
|
27 | $ hg up -Cqr 0 | |
|
22 | 28 | $ cd .. |
|
23 | 29 | |
|
24 | 30 | Discard all cached largefiles in USERCACHE |
|
25 | 31 | |
|
26 | 32 | $ rm -rf ${USERCACHE} |
|
27 | 33 | |
|
28 | 34 | Create mirror repo, and pull from source without largefile: |
|
29 | 35 | "pull" is used instead of "clone" for suppression of (1) updating to |
|
30 | 36 | tip (= caching largefile from source repo), and (2) recording source |
|
31 | 37 | repo as "default" path in .hg/hgrc. |
|
32 | 38 | |
|
33 | 39 | $ hg init mirror |
|
34 | 40 | $ cd mirror |
|
35 | 41 | $ hg pull ../src |
|
36 | 42 | pulling from ../src |
|
37 | 43 | requesting all changes |
|
38 | 44 | adding changesets |
|
39 | 45 | adding manifests |
|
40 | 46 | adding file changes |
|
41 | 47 | added 2 changesets with 1 changes to 1 files |
|
42 | 48 | (run 'hg update' to get a working copy) |
|
43 | 49 | |
|
44 | 50 | Update working directory to "tip", which requires largefile("large"), |
|
45 | 51 | but there is no cache file for it. So, hg must treat it as |
|
46 | 52 | "missing"(!) file. |
|
47 | 53 | |
|
48 | 54 | $ hg update -r0 |
|
49 | 55 | getting changed largefiles |
|
50 | 56 | large: largefile 7f7097b041ccf68cc5561e9600da4655d21c6d18 not available from file:/*/$TESTTMP/mirror (glob) |
|
51 | 57 | 0 largefiles updated, 0 removed |
|
52 | 58 | 1 files updated, 0 files merged, 0 files removed, 0 files unresolved |
|
53 | 59 | $ hg status |
|
54 | 60 | ! large |
|
55 | 61 | |
|
56 | 62 | Update working directory to null: this cleanup .hg/largefiles/dirstate |
|
57 | 63 | |
|
58 | 64 | $ hg update null |
|
59 | 65 | getting changed largefiles |
|
60 | 66 | 0 largefiles updated, 0 removed |
|
61 | 67 | 0 files updated, 0 files merged, 1 files removed, 0 files unresolved |
|
62 | 68 | |
|
63 | 69 | Update working directory to tip, again. |
|
64 | 70 | |
|
65 | 71 | $ hg update -r0 |
|
66 | 72 | getting changed largefiles |
|
67 | 73 | large: largefile 7f7097b041ccf68cc5561e9600da4655d21c6d18 not available from file:/*/$TESTTMP/mirror (glob) |
|
68 | 74 | 0 largefiles updated, 0 removed |
|
69 | 75 | 1 files updated, 0 files merged, 0 files removed, 0 files unresolved |
|
70 | 76 | $ hg status |
|
71 | 77 | ! large |
|
72 | 78 | $ cd .. |
|
73 | 79 | |
|
74 | 80 | Verify that largefiles from pulled branchheads are fetched, also to an empty repo |
|
75 | 81 | |
|
76 | 82 | $ hg init mirror2 |
|
77 | 83 | $ hg -R mirror2 pull src -r0 |
|
78 | 84 | pulling from src |
|
79 | 85 | adding changesets |
|
80 | 86 | adding manifests |
|
81 | 87 | adding file changes |
|
82 | 88 | added 1 changesets with 1 changes to 1 files |
|
83 | 89 | (run 'hg update' to get a working copy) |
|
84 | 90 | |
|
85 | 91 | #if unix-permissions |
|
86 | 92 | |
|
87 | 93 | Portable way to print file permissions: |
|
88 | 94 | |
|
89 | 95 | $ cat > ls-l.py <<EOF |
|
90 | 96 | > #!/usr/bin/env python |
|
91 | 97 | > import sys, os |
|
92 | 98 | > path = sys.argv[1] |
|
93 | 99 | > print '%03o' % (os.lstat(path).st_mode & 0777) |
|
94 | 100 | > EOF |
|
95 | 101 | $ chmod +x ls-l.py |
|
96 | 102 | |
|
97 | 103 | Test that files in .hg/largefiles inherit mode from .hg/store, not |
|
98 | 104 | from file in working copy: |
|
99 | 105 | |
|
100 | 106 | $ cd src |
|
101 | 107 | $ chmod 750 .hg/store |
|
102 | 108 | $ chmod 660 large |
|
103 | 109 | $ echo change >> large |
|
104 | 110 | $ hg commit -m change |
|
105 | 111 | created new head |
|
106 | 112 | $ ../ls-l.py .hg/largefiles/e151b474069de4ca6898f67ce2f2a7263adf8fea |
|
107 | 113 | 640 |
|
108 | 114 | |
|
109 | 115 | Test permission of with files in .hg/largefiles created by update: |
|
110 | 116 | |
|
111 | 117 | $ cd ../mirror |
|
112 | 118 | $ rm -r "$USERCACHE" .hg/largefiles # avoid links |
|
113 | 119 | $ chmod 750 .hg/store |
|
114 | 120 | $ hg pull ../src --update -q |
|
115 | 121 | $ ../ls-l.py .hg/largefiles/e151b474069de4ca6898f67ce2f2a7263adf8fea |
|
116 | 122 | 640 |
|
117 | 123 | |
|
118 | 124 | Test permission of files created by push: |
|
119 | 125 | |
|
120 | 126 | $ hg serve -R ../src -d -p $HGPORT --pid-file hg.pid \ |
|
121 | 127 | > --config "web.allow_push=*" --config web.push_ssl=no |
|
122 | 128 | $ cat hg.pid >> $DAEMON_PIDS |
|
123 | 129 | |
|
124 | 130 | $ echo change >> large |
|
125 | 131 | $ hg commit -m change |
|
126 | 132 | |
|
127 | 133 | $ rm -r "$USERCACHE" |
|
128 | 134 | |
|
129 | 135 | $ hg push -q http://localhost:$HGPORT/ |
|
130 | 136 | |
|
131 | 137 | $ ../ls-l.py ../src/.hg/largefiles/b734e14a0971e370408ab9bce8d56d8485e368a9 |
|
132 | 138 | 640 |
|
133 | 139 | |
|
134 | 140 | $ cd .. |
|
135 | 141 | |
|
136 | 142 | #endif |
|
137 | 143 | |
|
138 | 144 | Test issue 4053 (remove --after on a deleted, uncommitted file shouldn't say |
|
139 | 145 | it is missing, but a remove on a nonexistent unknown file still should. Same |
|
140 | 146 | for a forget.) |
|
141 | 147 | |
|
142 | 148 | $ cd src |
|
143 | 149 | $ touch x |
|
144 | 150 | $ hg add x |
|
145 | 151 | $ mv x y |
|
146 | 152 | $ hg remove -A x y ENOENT |
|
147 | 153 | ENOENT: * (glob) |
|
148 | 154 | not removing y: file is untracked |
|
149 | 155 | [1] |
|
150 | 156 | $ hg add y |
|
151 | 157 | $ mv y z |
|
152 | 158 | $ hg forget y z ENOENT |
|
153 | 159 | ENOENT: * (glob) |
|
154 | 160 | not removing z: file is already untracked |
|
155 | 161 | [1] |
|
156 | 162 | |
|
157 | 163 | Largefiles are accessible from the share's store |
|
158 | 164 | $ cd .. |
|
159 | 165 | $ hg share -q src share_dst --config extensions.share= |
|
160 | 166 | $ hg -R share_dst update -r0 |
|
161 | 167 | getting changed largefiles |
|
162 | 168 | 1 largefiles updated, 0 removed |
|
163 | 169 | 1 files updated, 0 files merged, 0 files removed, 0 files unresolved |
|
164 | 170 | |
|
165 | 171 | $ echo modified > share_dst/large |
|
166 | 172 | $ hg -R share_dst ci -m modified |
|
167 | 173 | created new head |
|
168 | 174 | |
|
169 | 175 | Only dirstate is in the local store for the share, and the largefile is in the |
|
170 | 176 | share source's local store. Avoid the extra largefiles added in the unix |
|
171 | 177 | conditional above. |
|
172 | 178 | $ hash=`hg -R share_dst cat share_dst/.hglf/large` |
|
173 | 179 | $ echo $hash |
|
174 | 180 | e2fb5f2139d086ded2cb600d5a91a196e76bf020 |
|
175 | 181 | |
|
176 | 182 | $ find share_dst/.hg/largefiles/* | sort |
|
177 | 183 | share_dst/.hg/largefiles/dirstate |
|
178 | 184 | |
|
179 | 185 | $ find src/.hg/largefiles/* | egrep "(dirstate|$hash)" | sort |
|
180 | 186 | src/.hg/largefiles/dirstate |
|
181 | 187 | src/.hg/largefiles/e2fb5f2139d086ded2cb600d5a91a196e76bf020 |
|
182 | 188 | |
|
183 | 189 | Inject corruption into the largefiles store and see how update handles that: |
|
184 | 190 | |
|
185 | 191 | $ cd src |
|
186 | 192 | $ hg up -qC |
|
187 | 193 | $ cat large |
|
188 | 194 | modified |
|
189 | 195 | $ rm large |
|
190 | 196 | $ cat .hglf/large |
|
191 | 197 | e2fb5f2139d086ded2cb600d5a91a196e76bf020 |
|
192 | 198 | $ mv .hg/largefiles/e2fb5f2139d086ded2cb600d5a91a196e76bf020 .. |
|
193 | 199 | $ echo corruption > .hg/largefiles/e2fb5f2139d086ded2cb600d5a91a196e76bf020 |
|
194 | 200 | $ hg up -C |
|
195 | 201 | getting changed largefiles |
|
196 | 202 | large: data corruption in $TESTTMP/src/.hg/largefiles/e2fb5f2139d086ded2cb600d5a91a196e76bf020 with hash 6a7bb2556144babe3899b25e5428123735bb1e27 (glob) |
|
197 | 203 | 0 largefiles updated, 0 removed |
|
198 | 204 | 0 files updated, 0 files merged, 0 files removed, 0 files unresolved |
|
199 | 205 | $ hg st |
|
200 | 206 | ! large |
|
201 | 207 | ? z |
|
202 | 208 | $ rm .hg/largefiles/e2fb5f2139d086ded2cb600d5a91a196e76bf020 |
|
203 | 209 | |
|
204 | 210 | #if serve |
|
205 | 211 | |
|
206 | 212 | Test coverage of error handling from putlfile: |
|
207 | 213 | |
|
208 | 214 | $ mkdir $TESTTMP/mirrorcache |
|
209 | 215 | $ hg serve -R ../mirror -d -p $HGPORT1 --pid-file hg.pid --config largefiles.usercache=$TESTTMP/mirrorcache |
|
210 | 216 | $ cat hg.pid >> $DAEMON_PIDS |
|
211 | 217 | |
|
212 | 218 | $ hg push http://localhost:$HGPORT1 -f --config files.usercache=nocache |
|
213 | 219 | pushing to http://localhost:$HGPORT1/ |
|
214 | 220 | searching for changes |
|
215 | 221 | abort: remotestore: could not open file $TESTTMP/src/.hg/largefiles/e2fb5f2139d086ded2cb600d5a91a196e76bf020: HTTP Error 403: ssl required |
|
216 | 222 | [255] |
|
217 | 223 | |
|
218 | 224 | $ rm .hg/largefiles/e2fb5f2139d086ded2cb600d5a91a196e76bf020 |
|
219 | 225 | |
|
220 | 226 | Test coverage of 'missing from store': |
|
221 | 227 | |
|
222 | 228 | $ hg serve -R ../mirror -d -p $HGPORT2 --pid-file hg.pid --config largefiles.usercache=$TESTTMP/mirrorcache --config "web.allow_push=*" --config web.push_ssl=no |
|
223 | 229 | $ cat hg.pid >> $DAEMON_PIDS |
|
224 | 230 | |
|
225 | 231 | $ hg push http://localhost:$HGPORT2 -f --config largefiles.usercache=nocache |
|
226 | 232 | pushing to http://localhost:$HGPORT2/ |
|
227 | 233 | searching for changes |
|
228 | 234 | abort: largefile e2fb5f2139d086ded2cb600d5a91a196e76bf020 missing from store (needs to be uploaded) |
|
229 | 235 | [255] |
|
230 | 236 | |
|
231 | 237 | #endif |
General Comments 0
You need to be logged in to leave comments.
Login now