Show More
@@ -1,664 +1,665 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 | from __future__ import absolute_import |
|
11 | 11 | |
|
12 | 12 | import copy |
|
13 | 13 | import hashlib |
|
14 | 14 | import os |
|
15 | 15 | import platform |
|
16 | 16 | import stat |
|
17 | 17 | |
|
18 | 18 | from mercurial.i18n import _ |
|
19 | 19 | |
|
20 | 20 | from mercurial import ( |
|
21 | 21 | dirstate, |
|
22 | 22 | error, |
|
23 | 23 | httpconnection, |
|
24 | 24 | match as matchmod, |
|
25 | 25 | node, |
|
26 | 26 | pycompat, |
|
27 | 27 | scmutil, |
|
28 | 28 | util, |
|
29 | 29 | ) |
|
30 | 30 | |
|
31 | 31 | shortname = '.hglf' |
|
32 | 32 | shortnameslash = shortname + '/' |
|
33 | 33 | longname = 'largefiles' |
|
34 | 34 | |
|
35 | 35 | # -- Private worker functions ------------------------------------------ |
|
36 | 36 | |
|
37 | 37 | def getminsize(ui, assumelfiles, opt, default=10): |
|
38 | 38 | lfsize = opt |
|
39 | 39 | if not lfsize and assumelfiles: |
|
40 | 40 | lfsize = ui.config(longname, 'minsize', default=default) |
|
41 | 41 | if lfsize: |
|
42 | 42 | try: |
|
43 | 43 | lfsize = float(lfsize) |
|
44 | 44 | except ValueError: |
|
45 | 45 | raise error.Abort(_('largefiles: size must be number (not %s)\n') |
|
46 | 46 | % lfsize) |
|
47 | 47 | if lfsize is None: |
|
48 | 48 | raise error.Abort(_('minimum size for largefiles must be specified')) |
|
49 | 49 | return lfsize |
|
50 | 50 | |
|
51 | 51 | def link(src, dest): |
|
52 | 52 | """Try to create hardlink - if that fails, efficiently make a copy.""" |
|
53 | 53 | util.makedirs(os.path.dirname(dest)) |
|
54 | 54 | try: |
|
55 | 55 | util.oslink(src, dest) |
|
56 | 56 | except OSError: |
|
57 | 57 | # if hardlinks fail, fallback on atomic copy |
|
58 | 58 | with open(src, 'rb') as srcf: |
|
59 | 59 | with util.atomictempfile(dest) as dstf: |
|
60 | 60 | for chunk in util.filechunkiter(srcf): |
|
61 | 61 | dstf.write(chunk) |
|
62 | 62 | os.chmod(dest, os.stat(src).st_mode) |
|
63 | 63 | |
|
64 | 64 | def usercachepath(ui, hash): |
|
65 | 65 | '''Return the correct location in the "global" largefiles cache for a file |
|
66 | 66 | with the given hash. |
|
67 | 67 | This cache is used for sharing of largefiles across repositories - both |
|
68 | 68 | to preserve download bandwidth and storage space.''' |
|
69 | 69 | return os.path.join(_usercachedir(ui), hash) |
|
70 | 70 | |
|
71 | 71 | def _usercachedir(ui): |
|
72 | 72 | '''Return the location of the "global" largefiles cache.''' |
|
73 | 73 | path = ui.configpath(longname, 'usercache', None) |
|
74 | 74 | if path: |
|
75 | 75 | return path |
|
76 | 76 | if pycompat.osname == 'nt': |
|
77 |
appdata = |
|
|
77 | appdata = pycompat.osgetenv('LOCALAPPDATA',\ | |
|
78 | pycompat.osgetenv('APPDATA')) | |
|
78 | 79 | if appdata: |
|
79 | 80 | return os.path.join(appdata, longname) |
|
80 | 81 | elif platform.system() == 'Darwin': |
|
81 |
home = |
|
|
82 | home = pycompat.osgetenv('HOME') | |
|
82 | 83 | if home: |
|
83 | 84 | return os.path.join(home, 'Library', 'Caches', longname) |
|
84 | 85 | elif pycompat.osname == 'posix': |
|
85 |
path = |
|
|
86 | path = pycompat.osgetenv('XDG_CACHE_HOME') | |
|
86 | 87 | if path: |
|
87 | 88 | return os.path.join(path, longname) |
|
88 |
home = |
|
|
89 | home = pycompat.osgetenv('HOME') | |
|
89 | 90 | if home: |
|
90 | 91 | return os.path.join(home, '.cache', longname) |
|
91 | 92 | else: |
|
92 | 93 | raise error.Abort(_('unknown operating system: %s\n') |
|
93 | 94 | % pycompat.osname) |
|
94 | 95 | raise error.Abort(_('unknown %s usercache location') % longname) |
|
95 | 96 | |
|
96 | 97 | def inusercache(ui, hash): |
|
97 | 98 | path = usercachepath(ui, hash) |
|
98 | 99 | return os.path.exists(path) |
|
99 | 100 | |
|
100 | 101 | def findfile(repo, hash): |
|
101 | 102 | '''Return store path of the largefile with the specified hash. |
|
102 | 103 | As a side effect, the file might be linked from user cache. |
|
103 | 104 | Return None if the file can't be found locally.''' |
|
104 | 105 | path, exists = findstorepath(repo, hash) |
|
105 | 106 | if exists: |
|
106 | 107 | repo.ui.note(_('found %s in store\n') % hash) |
|
107 | 108 | return path |
|
108 | 109 | elif inusercache(repo.ui, hash): |
|
109 | 110 | repo.ui.note(_('found %s in system cache\n') % hash) |
|
110 | 111 | path = storepath(repo, hash) |
|
111 | 112 | link(usercachepath(repo.ui, hash), path) |
|
112 | 113 | return path |
|
113 | 114 | return None |
|
114 | 115 | |
|
115 | 116 | class largefilesdirstate(dirstate.dirstate): |
|
116 | 117 | def __getitem__(self, key): |
|
117 | 118 | return super(largefilesdirstate, self).__getitem__(unixpath(key)) |
|
118 | 119 | def normal(self, f): |
|
119 | 120 | return super(largefilesdirstate, self).normal(unixpath(f)) |
|
120 | 121 | def remove(self, f): |
|
121 | 122 | return super(largefilesdirstate, self).remove(unixpath(f)) |
|
122 | 123 | def add(self, f): |
|
123 | 124 | return super(largefilesdirstate, self).add(unixpath(f)) |
|
124 | 125 | def drop(self, f): |
|
125 | 126 | return super(largefilesdirstate, self).drop(unixpath(f)) |
|
126 | 127 | def forget(self, f): |
|
127 | 128 | return super(largefilesdirstate, self).forget(unixpath(f)) |
|
128 | 129 | def normallookup(self, f): |
|
129 | 130 | return super(largefilesdirstate, self).normallookup(unixpath(f)) |
|
130 | 131 | def _ignore(self, f): |
|
131 | 132 | return False |
|
132 | 133 | def write(self, tr=False): |
|
133 | 134 | # (1) disable PENDING mode always |
|
134 | 135 | # (lfdirstate isn't yet managed as a part of the transaction) |
|
135 | 136 | # (2) avoid develwarn 'use dirstate.write with ....' |
|
136 | 137 | super(largefilesdirstate, self).write(None) |
|
137 | 138 | |
|
138 | 139 | def openlfdirstate(ui, repo, create=True): |
|
139 | 140 | ''' |
|
140 | 141 | Return a dirstate object that tracks largefiles: i.e. its root is |
|
141 | 142 | the repo root, but it is saved in .hg/largefiles/dirstate. |
|
142 | 143 | ''' |
|
143 | 144 | vfs = repo.vfs |
|
144 | 145 | lfstoredir = longname |
|
145 | 146 | opener = scmutil.opener(vfs.join(lfstoredir)) |
|
146 | 147 | lfdirstate = largefilesdirstate(opener, ui, repo.root, |
|
147 | 148 | repo.dirstate._validate) |
|
148 | 149 | |
|
149 | 150 | # If the largefiles dirstate does not exist, populate and create |
|
150 | 151 | # it. This ensures that we create it on the first meaningful |
|
151 | 152 | # largefiles operation in a new clone. |
|
152 | 153 | if create and not vfs.exists(vfs.join(lfstoredir, 'dirstate')): |
|
153 | 154 | matcher = getstandinmatcher(repo) |
|
154 | 155 | standins = repo.dirstate.walk(matcher, [], False, False) |
|
155 | 156 | |
|
156 | 157 | if len(standins) > 0: |
|
157 | 158 | vfs.makedirs(lfstoredir) |
|
158 | 159 | |
|
159 | 160 | for standin in standins: |
|
160 | 161 | lfile = splitstandin(standin) |
|
161 | 162 | lfdirstate.normallookup(lfile) |
|
162 | 163 | return lfdirstate |
|
163 | 164 | |
|
164 | 165 | def lfdirstatestatus(lfdirstate, repo): |
|
165 | 166 | wctx = repo['.'] |
|
166 | 167 | match = matchmod.always(repo.root, repo.getcwd()) |
|
167 | 168 | unsure, s = lfdirstate.status(match, [], False, False, False) |
|
168 | 169 | modified, clean = s.modified, s.clean |
|
169 | 170 | for lfile in unsure: |
|
170 | 171 | try: |
|
171 | 172 | fctx = wctx[standin(lfile)] |
|
172 | 173 | except LookupError: |
|
173 | 174 | fctx = None |
|
174 | 175 | if not fctx or fctx.data().strip() != hashfile(repo.wjoin(lfile)): |
|
175 | 176 | modified.append(lfile) |
|
176 | 177 | else: |
|
177 | 178 | clean.append(lfile) |
|
178 | 179 | lfdirstate.normal(lfile) |
|
179 | 180 | return s |
|
180 | 181 | |
|
181 | 182 | def listlfiles(repo, rev=None, matcher=None): |
|
182 | 183 | '''return a list of largefiles in the working copy or the |
|
183 | 184 | specified changeset''' |
|
184 | 185 | |
|
185 | 186 | if matcher is None: |
|
186 | 187 | matcher = getstandinmatcher(repo) |
|
187 | 188 | |
|
188 | 189 | # ignore unknown files in working directory |
|
189 | 190 | return [splitstandin(f) |
|
190 | 191 | for f in repo[rev].walk(matcher) |
|
191 | 192 | if rev is not None or repo.dirstate[f] != '?'] |
|
192 | 193 | |
|
193 | 194 | def instore(repo, hash, forcelocal=False): |
|
194 | 195 | '''Return true if a largefile with the given hash exists in the store''' |
|
195 | 196 | return os.path.exists(storepath(repo, hash, forcelocal)) |
|
196 | 197 | |
|
197 | 198 | def storepath(repo, hash, forcelocal=False): |
|
198 | 199 | '''Return the correct location in the repository largefiles store for a |
|
199 | 200 | file with the given hash.''' |
|
200 | 201 | if not forcelocal and repo.shared(): |
|
201 | 202 | return repo.vfs.reljoin(repo.sharedpath, longname, hash) |
|
202 | 203 | return repo.join(longname, hash) |
|
203 | 204 | |
|
204 | 205 | def findstorepath(repo, hash): |
|
205 | 206 | '''Search through the local store path(s) to find the file for the given |
|
206 | 207 | hash. If the file is not found, its path in the primary store is returned. |
|
207 | 208 | The return value is a tuple of (path, exists(path)). |
|
208 | 209 | ''' |
|
209 | 210 | # For shared repos, the primary store is in the share source. But for |
|
210 | 211 | # backward compatibility, force a lookup in the local store if it wasn't |
|
211 | 212 | # found in the share source. |
|
212 | 213 | path = storepath(repo, hash, False) |
|
213 | 214 | |
|
214 | 215 | if instore(repo, hash): |
|
215 | 216 | return (path, True) |
|
216 | 217 | elif repo.shared() and instore(repo, hash, True): |
|
217 | 218 | return storepath(repo, hash, True), True |
|
218 | 219 | |
|
219 | 220 | return (path, False) |
|
220 | 221 | |
|
221 | 222 | def copyfromcache(repo, hash, filename): |
|
222 | 223 | '''Copy the specified largefile from the repo or system cache to |
|
223 | 224 | filename in the repository. Return true on success or false if the |
|
224 | 225 | file was not found in either cache (which should not happened: |
|
225 | 226 | this is meant to be called only after ensuring that the needed |
|
226 | 227 | largefile exists in the cache).''' |
|
227 | 228 | wvfs = repo.wvfs |
|
228 | 229 | path = findfile(repo, hash) |
|
229 | 230 | if path is None: |
|
230 | 231 | return False |
|
231 | 232 | wvfs.makedirs(wvfs.dirname(wvfs.join(filename))) |
|
232 | 233 | # The write may fail before the file is fully written, but we |
|
233 | 234 | # don't use atomic writes in the working copy. |
|
234 | 235 | with open(path, 'rb') as srcfd: |
|
235 | 236 | with wvfs(filename, 'wb') as destfd: |
|
236 | 237 | gothash = copyandhash( |
|
237 | 238 | util.filechunkiter(srcfd), destfd) |
|
238 | 239 | if gothash != hash: |
|
239 | 240 | repo.ui.warn(_('%s: data corruption in %s with hash %s\n') |
|
240 | 241 | % (filename, path, gothash)) |
|
241 | 242 | wvfs.unlink(filename) |
|
242 | 243 | return False |
|
243 | 244 | return True |
|
244 | 245 | |
|
245 | 246 | def copytostore(repo, rev, file, uploaded=False): |
|
246 | 247 | wvfs = repo.wvfs |
|
247 | 248 | hash = readstandin(repo, file, rev) |
|
248 | 249 | if instore(repo, hash): |
|
249 | 250 | return |
|
250 | 251 | if wvfs.exists(file): |
|
251 | 252 | copytostoreabsolute(repo, wvfs.join(file), hash) |
|
252 | 253 | else: |
|
253 | 254 | repo.ui.warn(_("%s: largefile %s not available from local store\n") % |
|
254 | 255 | (file, hash)) |
|
255 | 256 | |
|
256 | 257 | def copyalltostore(repo, node): |
|
257 | 258 | '''Copy all largefiles in a given revision to the store''' |
|
258 | 259 | |
|
259 | 260 | ctx = repo[node] |
|
260 | 261 | for filename in ctx.files(): |
|
261 | 262 | if isstandin(filename) and filename in ctx.manifest(): |
|
262 | 263 | realfile = splitstandin(filename) |
|
263 | 264 | copytostore(repo, ctx.node(), realfile) |
|
264 | 265 | |
|
265 | 266 | def copytostoreabsolute(repo, file, hash): |
|
266 | 267 | if inusercache(repo.ui, hash): |
|
267 | 268 | link(usercachepath(repo.ui, hash), storepath(repo, hash)) |
|
268 | 269 | else: |
|
269 | 270 | util.makedirs(os.path.dirname(storepath(repo, hash))) |
|
270 | 271 | with open(file, 'rb') as srcf: |
|
271 | 272 | with util.atomictempfile(storepath(repo, hash), |
|
272 | 273 | createmode=repo.store.createmode) as dstf: |
|
273 | 274 | for chunk in util.filechunkiter(srcf): |
|
274 | 275 | dstf.write(chunk) |
|
275 | 276 | linktousercache(repo, hash) |
|
276 | 277 | |
|
277 | 278 | def linktousercache(repo, hash): |
|
278 | 279 | '''Link / copy the largefile with the specified hash from the store |
|
279 | 280 | to the cache.''' |
|
280 | 281 | path = usercachepath(repo.ui, hash) |
|
281 | 282 | link(storepath(repo, hash), path) |
|
282 | 283 | |
|
283 | 284 | def getstandinmatcher(repo, rmatcher=None): |
|
284 | 285 | '''Return a match object that applies rmatcher to the standin directory''' |
|
285 | 286 | wvfs = repo.wvfs |
|
286 | 287 | standindir = shortname |
|
287 | 288 | |
|
288 | 289 | # no warnings about missing files or directories |
|
289 | 290 | badfn = lambda f, msg: None |
|
290 | 291 | |
|
291 | 292 | if rmatcher and not rmatcher.always(): |
|
292 | 293 | pats = [wvfs.join(standindir, pat) for pat in rmatcher.files()] |
|
293 | 294 | if not pats: |
|
294 | 295 | pats = [wvfs.join(standindir)] |
|
295 | 296 | match = scmutil.match(repo[None], pats, badfn=badfn) |
|
296 | 297 | # if pats is empty, it would incorrectly always match, so clear _always |
|
297 | 298 | match._always = False |
|
298 | 299 | else: |
|
299 | 300 | # no patterns: relative to repo root |
|
300 | 301 | match = scmutil.match(repo[None], [wvfs.join(standindir)], badfn=badfn) |
|
301 | 302 | return match |
|
302 | 303 | |
|
303 | 304 | def composestandinmatcher(repo, rmatcher): |
|
304 | 305 | '''Return a matcher that accepts standins corresponding to the |
|
305 | 306 | files accepted by rmatcher. Pass the list of files in the matcher |
|
306 | 307 | as the paths specified by the user.''' |
|
307 | 308 | smatcher = getstandinmatcher(repo, rmatcher) |
|
308 | 309 | isstandin = smatcher.matchfn |
|
309 | 310 | def composedmatchfn(f): |
|
310 | 311 | return isstandin(f) and rmatcher.matchfn(splitstandin(f)) |
|
311 | 312 | smatcher.matchfn = composedmatchfn |
|
312 | 313 | |
|
313 | 314 | return smatcher |
|
314 | 315 | |
|
315 | 316 | def standin(filename): |
|
316 | 317 | '''Return the repo-relative path to the standin for the specified big |
|
317 | 318 | file.''' |
|
318 | 319 | # Notes: |
|
319 | 320 | # 1) Some callers want an absolute path, but for instance addlargefiles |
|
320 | 321 | # needs it repo-relative so it can be passed to repo[None].add(). So |
|
321 | 322 | # leave it up to the caller to use repo.wjoin() to get an absolute path. |
|
322 | 323 | # 2) Join with '/' because that's what dirstate always uses, even on |
|
323 | 324 | # Windows. Change existing separator to '/' first in case we are |
|
324 | 325 | # passed filenames from an external source (like the command line). |
|
325 | 326 | return shortnameslash + util.pconvert(filename) |
|
326 | 327 | |
|
327 | 328 | def isstandin(filename): |
|
328 | 329 | '''Return true if filename is a big file standin. filename must be |
|
329 | 330 | in Mercurial's internal form (slash-separated).''' |
|
330 | 331 | return filename.startswith(shortnameslash) |
|
331 | 332 | |
|
332 | 333 | def splitstandin(filename): |
|
333 | 334 | # Split on / because that's what dirstate always uses, even on Windows. |
|
334 | 335 | # Change local separator to / first just in case we are passed filenames |
|
335 | 336 | # from an external source (like the command line). |
|
336 | 337 | bits = util.pconvert(filename).split('/', 1) |
|
337 | 338 | if len(bits) == 2 and bits[0] == shortname: |
|
338 | 339 | return bits[1] |
|
339 | 340 | else: |
|
340 | 341 | return None |
|
341 | 342 | |
|
342 | 343 | def updatestandin(repo, standin): |
|
343 | 344 | file = repo.wjoin(splitstandin(standin)) |
|
344 | 345 | if repo.wvfs.exists(splitstandin(standin)): |
|
345 | 346 | hash = hashfile(file) |
|
346 | 347 | executable = getexecutable(file) |
|
347 | 348 | writestandin(repo, standin, hash, executable) |
|
348 | 349 | else: |
|
349 | 350 | raise error.Abort(_('%s: file not found!') % splitstandin(standin)) |
|
350 | 351 | |
|
351 | 352 | def readstandin(repo, filename, node=None): |
|
352 | 353 | '''read hex hash from standin for filename at given node, or working |
|
353 | 354 | directory if no node is given''' |
|
354 | 355 | return repo[node][standin(filename)].data().strip() |
|
355 | 356 | |
|
356 | 357 | def writestandin(repo, standin, hash, executable): |
|
357 | 358 | '''write hash to <repo.root>/<standin>''' |
|
358 | 359 | repo.wwrite(standin, hash + '\n', executable and 'x' or '') |
|
359 | 360 | |
|
360 | 361 | def copyandhash(instream, outfile): |
|
361 | 362 | '''Read bytes from instream (iterable) and write them to outfile, |
|
362 | 363 | computing the SHA-1 hash of the data along the way. Return the hash.''' |
|
363 | 364 | hasher = hashlib.sha1('') |
|
364 | 365 | for data in instream: |
|
365 | 366 | hasher.update(data) |
|
366 | 367 | outfile.write(data) |
|
367 | 368 | return hasher.hexdigest() |
|
368 | 369 | |
|
369 | 370 | def hashrepofile(repo, file): |
|
370 | 371 | return hashfile(repo.wjoin(file)) |
|
371 | 372 | |
|
372 | 373 | def hashfile(file): |
|
373 | 374 | if not os.path.exists(file): |
|
374 | 375 | return '' |
|
375 | 376 | hasher = hashlib.sha1('') |
|
376 | 377 | with open(file, 'rb') as fd: |
|
377 | 378 | for data in util.filechunkiter(fd): |
|
378 | 379 | hasher.update(data) |
|
379 | 380 | return hasher.hexdigest() |
|
380 | 381 | |
|
381 | 382 | def getexecutable(filename): |
|
382 | 383 | mode = os.stat(filename).st_mode |
|
383 | 384 | return ((mode & stat.S_IXUSR) and |
|
384 | 385 | (mode & stat.S_IXGRP) and |
|
385 | 386 | (mode & stat.S_IXOTH)) |
|
386 | 387 | |
|
387 | 388 | def urljoin(first, second, *arg): |
|
388 | 389 | def join(left, right): |
|
389 | 390 | if not left.endswith('/'): |
|
390 | 391 | left += '/' |
|
391 | 392 | if right.startswith('/'): |
|
392 | 393 | right = right[1:] |
|
393 | 394 | return left + right |
|
394 | 395 | |
|
395 | 396 | url = join(first, second) |
|
396 | 397 | for a in arg: |
|
397 | 398 | url = join(url, a) |
|
398 | 399 | return url |
|
399 | 400 | |
|
400 | 401 | def hexsha1(data): |
|
401 | 402 | """hexsha1 returns the hex-encoded sha1 sum of the data in the file-like |
|
402 | 403 | object data""" |
|
403 | 404 | h = hashlib.sha1() |
|
404 | 405 | for chunk in util.filechunkiter(data): |
|
405 | 406 | h.update(chunk) |
|
406 | 407 | return h.hexdigest() |
|
407 | 408 | |
|
408 | 409 | def httpsendfile(ui, filename): |
|
409 | 410 | return httpconnection.httpsendfile(ui, filename, 'rb') |
|
410 | 411 | |
|
411 | 412 | def unixpath(path): |
|
412 | 413 | '''Return a version of path normalized for use with the lfdirstate.''' |
|
413 | 414 | return util.pconvert(os.path.normpath(path)) |
|
414 | 415 | |
|
415 | 416 | def islfilesrepo(repo): |
|
416 | 417 | '''Return true if the repo is a largefile repo.''' |
|
417 | 418 | if ('largefiles' in repo.requirements and |
|
418 | 419 | any(shortnameslash in f[0] for f in repo.store.datafiles())): |
|
419 | 420 | return True |
|
420 | 421 | |
|
421 | 422 | return any(openlfdirstate(repo.ui, repo, False)) |
|
422 | 423 | |
|
423 | 424 | class storeprotonotcapable(Exception): |
|
424 | 425 | def __init__(self, storetypes): |
|
425 | 426 | self.storetypes = storetypes |
|
426 | 427 | |
|
427 | 428 | def getstandinsstate(repo): |
|
428 | 429 | standins = [] |
|
429 | 430 | matcher = getstandinmatcher(repo) |
|
430 | 431 | for standin in repo.dirstate.walk(matcher, [], False, False): |
|
431 | 432 | lfile = splitstandin(standin) |
|
432 | 433 | try: |
|
433 | 434 | hash = readstandin(repo, lfile) |
|
434 | 435 | except IOError: |
|
435 | 436 | hash = None |
|
436 | 437 | standins.append((lfile, hash)) |
|
437 | 438 | return standins |
|
438 | 439 | |
|
439 | 440 | def synclfdirstate(repo, lfdirstate, lfile, normallookup): |
|
440 | 441 | lfstandin = standin(lfile) |
|
441 | 442 | if lfstandin in repo.dirstate: |
|
442 | 443 | stat = repo.dirstate._map[lfstandin] |
|
443 | 444 | state, mtime = stat[0], stat[3] |
|
444 | 445 | else: |
|
445 | 446 | state, mtime = '?', -1 |
|
446 | 447 | if state == 'n': |
|
447 | 448 | if (normallookup or mtime < 0 or |
|
448 | 449 | not repo.wvfs.exists(lfile)): |
|
449 | 450 | # state 'n' doesn't ensure 'clean' in this case |
|
450 | 451 | lfdirstate.normallookup(lfile) |
|
451 | 452 | else: |
|
452 | 453 | lfdirstate.normal(lfile) |
|
453 | 454 | elif state == 'm': |
|
454 | 455 | lfdirstate.normallookup(lfile) |
|
455 | 456 | elif state == 'r': |
|
456 | 457 | lfdirstate.remove(lfile) |
|
457 | 458 | elif state == 'a': |
|
458 | 459 | lfdirstate.add(lfile) |
|
459 | 460 | elif state == '?': |
|
460 | 461 | lfdirstate.drop(lfile) |
|
461 | 462 | |
|
462 | 463 | def markcommitted(orig, ctx, node): |
|
463 | 464 | repo = ctx.repo() |
|
464 | 465 | |
|
465 | 466 | orig(node) |
|
466 | 467 | |
|
467 | 468 | # ATTENTION: "ctx.files()" may differ from "repo[node].files()" |
|
468 | 469 | # because files coming from the 2nd parent are omitted in the latter. |
|
469 | 470 | # |
|
470 | 471 | # The former should be used to get targets of "synclfdirstate", |
|
471 | 472 | # because such files: |
|
472 | 473 | # - are marked as "a" by "patch.patch()" (e.g. via transplant), and |
|
473 | 474 | # - have to be marked as "n" after commit, but |
|
474 | 475 | # - aren't listed in "repo[node].files()" |
|
475 | 476 | |
|
476 | 477 | lfdirstate = openlfdirstate(repo.ui, repo) |
|
477 | 478 | for f in ctx.files(): |
|
478 | 479 | if isstandin(f): |
|
479 | 480 | lfile = splitstandin(f) |
|
480 | 481 | synclfdirstate(repo, lfdirstate, lfile, False) |
|
481 | 482 | lfdirstate.write() |
|
482 | 483 | |
|
483 | 484 | # As part of committing, copy all of the largefiles into the cache. |
|
484 | 485 | copyalltostore(repo, node) |
|
485 | 486 | |
|
486 | 487 | def getlfilestoupdate(oldstandins, newstandins): |
|
487 | 488 | changedstandins = set(oldstandins).symmetric_difference(set(newstandins)) |
|
488 | 489 | filelist = [] |
|
489 | 490 | for f in changedstandins: |
|
490 | 491 | if f[0] not in filelist: |
|
491 | 492 | filelist.append(f[0]) |
|
492 | 493 | return filelist |
|
493 | 494 | |
|
494 | 495 | def getlfilestoupload(repo, missing, addfunc): |
|
495 | 496 | for i, n in enumerate(missing): |
|
496 | 497 | repo.ui.progress(_('finding outgoing largefiles'), i, |
|
497 | 498 | unit=_('revisions'), total=len(missing)) |
|
498 | 499 | parents = [p for p in repo[n].parents() if p != node.nullid] |
|
499 | 500 | |
|
500 | 501 | oldlfstatus = repo.lfstatus |
|
501 | 502 | repo.lfstatus = False |
|
502 | 503 | try: |
|
503 | 504 | ctx = repo[n] |
|
504 | 505 | finally: |
|
505 | 506 | repo.lfstatus = oldlfstatus |
|
506 | 507 | |
|
507 | 508 | files = set(ctx.files()) |
|
508 | 509 | if len(parents) == 2: |
|
509 | 510 | mc = ctx.manifest() |
|
510 | 511 | mp1 = ctx.parents()[0].manifest() |
|
511 | 512 | mp2 = ctx.parents()[1].manifest() |
|
512 | 513 | for f in mp1: |
|
513 | 514 | if f not in mc: |
|
514 | 515 | files.add(f) |
|
515 | 516 | for f in mp2: |
|
516 | 517 | if f not in mc: |
|
517 | 518 | files.add(f) |
|
518 | 519 | for f in mc: |
|
519 | 520 | if mc[f] != mp1.get(f, None) or mc[f] != mp2.get(f, None): |
|
520 | 521 | files.add(f) |
|
521 | 522 | for fn in files: |
|
522 | 523 | if isstandin(fn) and fn in ctx: |
|
523 | 524 | addfunc(fn, ctx[fn].data().strip()) |
|
524 | 525 | repo.ui.progress(_('finding outgoing largefiles'), None) |
|
525 | 526 | |
|
526 | 527 | def updatestandinsbymatch(repo, match): |
|
527 | 528 | '''Update standins in the working directory according to specified match |
|
528 | 529 | |
|
529 | 530 | This returns (possibly modified) ``match`` object to be used for |
|
530 | 531 | subsequent commit process. |
|
531 | 532 | ''' |
|
532 | 533 | |
|
533 | 534 | ui = repo.ui |
|
534 | 535 | |
|
535 | 536 | # Case 1: user calls commit with no specific files or |
|
536 | 537 | # include/exclude patterns: refresh and commit all files that |
|
537 | 538 | # are "dirty". |
|
538 | 539 | if match is None or match.always(): |
|
539 | 540 | # Spend a bit of time here to get a list of files we know |
|
540 | 541 | # are modified so we can compare only against those. |
|
541 | 542 | # It can cost a lot of time (several seconds) |
|
542 | 543 | # otherwise to update all standins if the largefiles are |
|
543 | 544 | # large. |
|
544 | 545 | lfdirstate = openlfdirstate(ui, repo) |
|
545 | 546 | dirtymatch = matchmod.always(repo.root, repo.getcwd()) |
|
546 | 547 | unsure, s = lfdirstate.status(dirtymatch, [], False, False, |
|
547 | 548 | False) |
|
548 | 549 | modifiedfiles = unsure + s.modified + s.added + s.removed |
|
549 | 550 | lfiles = listlfiles(repo) |
|
550 | 551 | # this only loops through largefiles that exist (not |
|
551 | 552 | # removed/renamed) |
|
552 | 553 | for lfile in lfiles: |
|
553 | 554 | if lfile in modifiedfiles: |
|
554 | 555 | if repo.wvfs.exists(standin(lfile)): |
|
555 | 556 | # this handles the case where a rebase is being |
|
556 | 557 | # performed and the working copy is not updated |
|
557 | 558 | # yet. |
|
558 | 559 | if repo.wvfs.exists(lfile): |
|
559 | 560 | updatestandin(repo, |
|
560 | 561 | standin(lfile)) |
|
561 | 562 | |
|
562 | 563 | return match |
|
563 | 564 | |
|
564 | 565 | lfiles = listlfiles(repo) |
|
565 | 566 | match._files = repo._subdirlfs(match.files(), lfiles) |
|
566 | 567 | |
|
567 | 568 | # Case 2: user calls commit with specified patterns: refresh |
|
568 | 569 | # any matching big files. |
|
569 | 570 | smatcher = composestandinmatcher(repo, match) |
|
570 | 571 | standins = repo.dirstate.walk(smatcher, [], False, False) |
|
571 | 572 | |
|
572 | 573 | # No matching big files: get out of the way and pass control to |
|
573 | 574 | # the usual commit() method. |
|
574 | 575 | if not standins: |
|
575 | 576 | return match |
|
576 | 577 | |
|
577 | 578 | # Refresh all matching big files. It's possible that the |
|
578 | 579 | # commit will end up failing, in which case the big files will |
|
579 | 580 | # stay refreshed. No harm done: the user modified them and |
|
580 | 581 | # asked to commit them, so sooner or later we're going to |
|
581 | 582 | # refresh the standins. Might as well leave them refreshed. |
|
582 | 583 | lfdirstate = openlfdirstate(ui, repo) |
|
583 | 584 | for fstandin in standins: |
|
584 | 585 | lfile = splitstandin(fstandin) |
|
585 | 586 | if lfdirstate[lfile] != 'r': |
|
586 | 587 | updatestandin(repo, fstandin) |
|
587 | 588 | |
|
588 | 589 | # Cook up a new matcher that only matches regular files or |
|
589 | 590 | # standins corresponding to the big files requested by the |
|
590 | 591 | # user. Have to modify _files to prevent commit() from |
|
591 | 592 | # complaining "not tracked" for big files. |
|
592 | 593 | match = copy.copy(match) |
|
593 | 594 | origmatchfn = match.matchfn |
|
594 | 595 | |
|
595 | 596 | # Check both the list of largefiles and the list of |
|
596 | 597 | # standins because if a largefile was removed, it |
|
597 | 598 | # won't be in the list of largefiles at this point |
|
598 | 599 | match._files += sorted(standins) |
|
599 | 600 | |
|
600 | 601 | actualfiles = [] |
|
601 | 602 | for f in match._files: |
|
602 | 603 | fstandin = standin(f) |
|
603 | 604 | |
|
604 | 605 | # For largefiles, only one of the normal and standin should be |
|
605 | 606 | # committed (except if one of them is a remove). In the case of a |
|
606 | 607 | # standin removal, drop the normal file if it is unknown to dirstate. |
|
607 | 608 | # Thus, skip plain largefile names but keep the standin. |
|
608 | 609 | if f in lfiles or fstandin in standins: |
|
609 | 610 | if repo.dirstate[fstandin] != 'r': |
|
610 | 611 | if repo.dirstate[f] != 'r': |
|
611 | 612 | continue |
|
612 | 613 | elif repo.dirstate[f] == '?': |
|
613 | 614 | continue |
|
614 | 615 | |
|
615 | 616 | actualfiles.append(f) |
|
616 | 617 | match._files = actualfiles |
|
617 | 618 | |
|
618 | 619 | def matchfn(f): |
|
619 | 620 | if origmatchfn(f): |
|
620 | 621 | return f not in lfiles |
|
621 | 622 | else: |
|
622 | 623 | return f in standins |
|
623 | 624 | |
|
624 | 625 | match.matchfn = matchfn |
|
625 | 626 | |
|
626 | 627 | return match |
|
627 | 628 | |
|
628 | 629 | class automatedcommithook(object): |
|
629 | 630 | '''Stateful hook to update standins at the 1st commit of resuming |
|
630 | 631 | |
|
631 | 632 | For efficiency, updating standins in the working directory should |
|
632 | 633 | be avoided while automated committing (like rebase, transplant and |
|
633 | 634 | so on), because they should be updated before committing. |
|
634 | 635 | |
|
635 | 636 | But the 1st commit of resuming automated committing (e.g. ``rebase |
|
636 | 637 | --continue``) should update them, because largefiles may be |
|
637 | 638 | modified manually. |
|
638 | 639 | ''' |
|
639 | 640 | def __init__(self, resuming): |
|
640 | 641 | self.resuming = resuming |
|
641 | 642 | |
|
642 | 643 | def __call__(self, repo, match): |
|
643 | 644 | if self.resuming: |
|
644 | 645 | self.resuming = False # avoids updating at subsequent commits |
|
645 | 646 | return updatestandinsbymatch(repo, match) |
|
646 | 647 | else: |
|
647 | 648 | return match |
|
648 | 649 | |
|
649 | 650 | def getstatuswriter(ui, repo, forcibly=None): |
|
650 | 651 | '''Return the function to write largefiles specific status out |
|
651 | 652 | |
|
652 | 653 | If ``forcibly`` is ``None``, this returns the last element of |
|
653 | 654 | ``repo._lfstatuswriters`` as "default" writer function. |
|
654 | 655 | |
|
655 | 656 | Otherwise, this returns the function to always write out (or |
|
656 | 657 | ignore if ``not forcibly``) status. |
|
657 | 658 | ''' |
|
658 | 659 | if forcibly is None and util.safehasattr(repo, '_largefilesenabled'): |
|
659 | 660 | return repo._lfstatuswriters[-1] |
|
660 | 661 | else: |
|
661 | 662 | if forcibly: |
|
662 | 663 | return ui.status # forcibly WRITE OUT |
|
663 | 664 | else: |
|
664 | 665 | return lambda *msg, **opts: None # forcibly IGNORE |
@@ -1,176 +1,176 b'' | |||
|
1 | 1 | # profiling.py - profiling functions |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2016 Gregory Szorc <gregory.szorc@gmail.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from __future__ import absolute_import, print_function |
|
9 | 9 | |
|
10 | 10 | import contextlib |
|
11 | import os | |
|
12 | 11 | import time |
|
13 | 12 | |
|
14 | 13 | from .i18n import _ |
|
15 | 14 | from . import ( |
|
16 | 15 | error, |
|
16 | pycompat, | |
|
17 | 17 | util, |
|
18 | 18 | ) |
|
19 | 19 | |
|
20 | 20 | @contextlib.contextmanager |
|
21 | 21 | def lsprofile(ui, fp): |
|
22 | 22 | format = ui.config('profiling', 'format', default='text') |
|
23 | 23 | field = ui.config('profiling', 'sort', default='inlinetime') |
|
24 | 24 | limit = ui.configint('profiling', 'limit', default=30) |
|
25 | 25 | climit = ui.configint('profiling', 'nested', default=0) |
|
26 | 26 | |
|
27 | 27 | if format not in ['text', 'kcachegrind']: |
|
28 | 28 | ui.warn(_("unrecognized profiling format '%s'" |
|
29 | 29 | " - Ignored\n") % format) |
|
30 | 30 | format = 'text' |
|
31 | 31 | |
|
32 | 32 | try: |
|
33 | 33 | from . import lsprof |
|
34 | 34 | except ImportError: |
|
35 | 35 | raise error.Abort(_( |
|
36 | 36 | 'lsprof not available - install from ' |
|
37 | 37 | 'http://codespeak.net/svn/user/arigo/hack/misc/lsprof/')) |
|
38 | 38 | p = lsprof.Profiler() |
|
39 | 39 | p.enable(subcalls=True) |
|
40 | 40 | try: |
|
41 | 41 | yield |
|
42 | 42 | finally: |
|
43 | 43 | p.disable() |
|
44 | 44 | |
|
45 | 45 | if format == 'kcachegrind': |
|
46 | 46 | from . import lsprofcalltree |
|
47 | 47 | calltree = lsprofcalltree.KCacheGrind(p) |
|
48 | 48 | calltree.output(fp) |
|
49 | 49 | else: |
|
50 | 50 | # format == 'text' |
|
51 | 51 | stats = lsprof.Stats(p.getstats()) |
|
52 | 52 | stats.sort(field) |
|
53 | 53 | stats.pprint(limit=limit, file=fp, climit=climit) |
|
54 | 54 | |
|
55 | 55 | @contextlib.contextmanager |
|
56 | 56 | def flameprofile(ui, fp): |
|
57 | 57 | try: |
|
58 | 58 | from flamegraph import flamegraph |
|
59 | 59 | except ImportError: |
|
60 | 60 | raise error.Abort(_( |
|
61 | 61 | 'flamegraph not available - install from ' |
|
62 | 62 | 'https://github.com/evanhempel/python-flamegraph')) |
|
63 | 63 | # developer config: profiling.freq |
|
64 | 64 | freq = ui.configint('profiling', 'freq', default=1000) |
|
65 | 65 | filter_ = None |
|
66 | 66 | collapse_recursion = True |
|
67 | 67 | thread = flamegraph.ProfileThread(fp, 1.0 / freq, |
|
68 | 68 | filter_, collapse_recursion) |
|
69 | 69 | start_time = time.clock() |
|
70 | 70 | try: |
|
71 | 71 | thread.start() |
|
72 | 72 | yield |
|
73 | 73 | finally: |
|
74 | 74 | thread.stop() |
|
75 | 75 | thread.join() |
|
76 | 76 | print('Collected %d stack frames (%d unique) in %2.2f seconds.' % ( |
|
77 | 77 | time.clock() - start_time, thread.num_frames(), |
|
78 | 78 | thread.num_frames(unique=True))) |
|
79 | 79 | |
|
80 | 80 | @contextlib.contextmanager |
|
81 | 81 | def statprofile(ui, fp): |
|
82 | 82 | from . import statprof |
|
83 | 83 | |
|
84 | 84 | freq = ui.configint('profiling', 'freq', default=1000) |
|
85 | 85 | if freq > 0: |
|
86 | 86 | # Cannot reset when profiler is already active. So silently no-op. |
|
87 | 87 | if statprof.state.profile_level == 0: |
|
88 | 88 | statprof.reset(freq) |
|
89 | 89 | else: |
|
90 | 90 | ui.warn(_("invalid sampling frequency '%s' - ignoring\n") % freq) |
|
91 | 91 | |
|
92 | 92 | statprof.start(mechanism='thread') |
|
93 | 93 | |
|
94 | 94 | try: |
|
95 | 95 | yield |
|
96 | 96 | finally: |
|
97 | 97 | data = statprof.stop() |
|
98 | 98 | |
|
99 | 99 | profformat = ui.config('profiling', 'statformat', 'hotpath') |
|
100 | 100 | |
|
101 | 101 | formats = { |
|
102 | 102 | 'byline': statprof.DisplayFormats.ByLine, |
|
103 | 103 | 'bymethod': statprof.DisplayFormats.ByMethod, |
|
104 | 104 | 'hotpath': statprof.DisplayFormats.Hotpath, |
|
105 | 105 | 'json': statprof.DisplayFormats.Json, |
|
106 | 106 | } |
|
107 | 107 | |
|
108 | 108 | if profformat in formats: |
|
109 | 109 | displayformat = formats[profformat] |
|
110 | 110 | else: |
|
111 | 111 | ui.warn(_('unknown profiler output format: %s\n') % profformat) |
|
112 | 112 | displayformat = statprof.DisplayFormats.Hotpath |
|
113 | 113 | |
|
114 | 114 | statprof.display(fp, data=data, format=displayformat) |
|
115 | 115 | |
|
116 | 116 | @contextlib.contextmanager |
|
117 | 117 | def profile(ui): |
|
118 | 118 | """Start profiling. |
|
119 | 119 | |
|
120 | 120 | Profiling is active when the context manager is active. When the context |
|
121 | 121 | manager exits, profiling results will be written to the configured output. |
|
122 | 122 | """ |
|
123 |
profiler = |
|
|
123 | profiler = pycompat.osgetenv('HGPROF') | |
|
124 | 124 | if profiler is None: |
|
125 | 125 | profiler = ui.config('profiling', 'type', default='stat') |
|
126 | 126 | if profiler not in ('ls', 'stat', 'flame'): |
|
127 | 127 | ui.warn(_("unrecognized profiler '%s' - ignored\n") % profiler) |
|
128 | 128 | profiler = 'stat' |
|
129 | 129 | |
|
130 | 130 | output = ui.config('profiling', 'output') |
|
131 | 131 | |
|
132 | 132 | if output == 'blackbox': |
|
133 | 133 | fp = util.stringio() |
|
134 | 134 | elif output: |
|
135 | 135 | path = ui.expandpath(output) |
|
136 | 136 | fp = open(path, 'wb') |
|
137 | 137 | else: |
|
138 | 138 | fp = ui.ferr |
|
139 | 139 | |
|
140 | 140 | try: |
|
141 | 141 | if profiler == 'ls': |
|
142 | 142 | proffn = lsprofile |
|
143 | 143 | elif profiler == 'flame': |
|
144 | 144 | proffn = flameprofile |
|
145 | 145 | else: |
|
146 | 146 | proffn = statprofile |
|
147 | 147 | |
|
148 | 148 | with proffn(ui, fp): |
|
149 | 149 | yield |
|
150 | 150 | |
|
151 | 151 | finally: |
|
152 | 152 | if output: |
|
153 | 153 | if output == 'blackbox': |
|
154 | 154 | val = 'Profile:\n%s' % fp.getvalue() |
|
155 | 155 | # ui.log treats the input as a format string, |
|
156 | 156 | # so we need to escape any % signs. |
|
157 | 157 | val = val.replace('%', '%%') |
|
158 | 158 | ui.log('profile', val) |
|
159 | 159 | fp.close() |
|
160 | 160 | |
|
161 | 161 | @contextlib.contextmanager |
|
162 | 162 | def maybeprofile(ui): |
|
163 | 163 | """Profile if enabled, else do nothing. |
|
164 | 164 | |
|
165 | 165 | This context manager can be used to optionally profile if profiling |
|
166 | 166 | is enabled. Otherwise, it does nothing. |
|
167 | 167 | |
|
168 | 168 | The purpose of this context manager is to make calling code simpler: |
|
169 | 169 | just use a single code path for calling into code you may want to profile |
|
170 | 170 | and this function determines whether to start profiling. |
|
171 | 171 | """ |
|
172 | 172 | if ui.configbool('profiling', 'enabled'): |
|
173 | 173 | with profile(ui): |
|
174 | 174 | yield |
|
175 | 175 | else: |
|
176 | 176 | yield |
@@ -1,479 +1,481 b'' | |||
|
1 | 1 | # url.py - HTTP handling for mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2005, 2006, 2007, 2008 Matt Mackall <mpm@selenic.com> |
|
4 | 4 | # Copyright 2006, 2007 Alexis S. L. Carvalho <alexis@cecm.usp.br> |
|
5 | 5 | # Copyright 2006 Vadim Gelfer <vadim.gelfer@gmail.com> |
|
6 | 6 | # |
|
7 | 7 | # This software may be used and distributed according to the terms of the |
|
8 | 8 | # GNU General Public License version 2 or any later version. |
|
9 | 9 | |
|
10 | 10 | from __future__ import absolute_import |
|
11 | 11 | |
|
12 | 12 | import base64 |
|
13 | 13 | import os |
|
14 | 14 | import socket |
|
15 | 15 | |
|
16 | 16 | from .i18n import _ |
|
17 | 17 | from . import ( |
|
18 | 18 | error, |
|
19 | 19 | httpconnection as httpconnectionmod, |
|
20 | 20 | keepalive, |
|
21 | pycompat, | |
|
21 | 22 | sslutil, |
|
22 | 23 | util, |
|
23 | 24 | ) |
|
24 | 25 | |
|
25 | 26 | httplib = util.httplib |
|
26 | 27 | stringio = util.stringio |
|
27 | 28 | urlerr = util.urlerr |
|
28 | 29 | urlreq = util.urlreq |
|
29 | 30 | |
|
30 | 31 | class passwordmgr(object): |
|
31 | 32 | def __init__(self, ui, passwddb): |
|
32 | 33 | self.ui = ui |
|
33 | 34 | self.passwddb = passwddb |
|
34 | 35 | |
|
35 | 36 | def add_password(self, realm, uri, user, passwd): |
|
36 | 37 | return self.passwddb.add_password(realm, uri, user, passwd) |
|
37 | 38 | |
|
38 | 39 | def find_user_password(self, realm, authuri): |
|
39 | 40 | authinfo = self.passwddb.find_user_password(realm, authuri) |
|
40 | 41 | user, passwd = authinfo |
|
41 | 42 | if user and passwd: |
|
42 | 43 | self._writedebug(user, passwd) |
|
43 | 44 | return (user, passwd) |
|
44 | 45 | |
|
45 | 46 | if not user or not passwd: |
|
46 | 47 | res = httpconnectionmod.readauthforuri(self.ui, authuri, user) |
|
47 | 48 | if res: |
|
48 | 49 | group, auth = res |
|
49 | 50 | user, passwd = auth.get('username'), auth.get('password') |
|
50 | 51 | self.ui.debug("using auth.%s.* for authentication\n" % group) |
|
51 | 52 | if not user or not passwd: |
|
52 | 53 | u = util.url(authuri) |
|
53 | 54 | u.query = None |
|
54 | 55 | if not self.ui.interactive(): |
|
55 | 56 | raise error.Abort(_('http authorization required for %s') % |
|
56 | 57 | util.hidepassword(str(u))) |
|
57 | 58 | |
|
58 | 59 | self.ui.write(_("http authorization required for %s\n") % |
|
59 | 60 | util.hidepassword(str(u))) |
|
60 | 61 | self.ui.write(_("realm: %s\n") % realm) |
|
61 | 62 | if user: |
|
62 | 63 | self.ui.write(_("user: %s\n") % user) |
|
63 | 64 | else: |
|
64 | 65 | user = self.ui.prompt(_("user:"), default=None) |
|
65 | 66 | |
|
66 | 67 | if not passwd: |
|
67 | 68 | passwd = self.ui.getpass() |
|
68 | 69 | |
|
69 | 70 | self.passwddb.add_password(realm, authuri, user, passwd) |
|
70 | 71 | self._writedebug(user, passwd) |
|
71 | 72 | return (user, passwd) |
|
72 | 73 | |
|
73 | 74 | def _writedebug(self, user, passwd): |
|
74 | 75 | msg = _('http auth: user %s, password %s\n') |
|
75 | 76 | self.ui.debug(msg % (user, passwd and '*' * len(passwd) or 'not set')) |
|
76 | 77 | |
|
77 | 78 | def find_stored_password(self, authuri): |
|
78 | 79 | return self.passwddb.find_user_password(None, authuri) |
|
79 | 80 | |
|
80 | 81 | class proxyhandler(urlreq.proxyhandler): |
|
81 | 82 | def __init__(self, ui): |
|
82 |
proxyurl = ui.config("http_proxy", "host") or |
|
|
83 | proxyurl = (ui.config("http_proxy", "host") or | |
|
84 | pycompat.osgetenv('http_proxy')) | |
|
83 | 85 | # XXX proxyauthinfo = None |
|
84 | 86 | |
|
85 | 87 | if proxyurl: |
|
86 | 88 | # proxy can be proper url or host[:port] |
|
87 | 89 | if not (proxyurl.startswith('http:') or |
|
88 | 90 | proxyurl.startswith('https:')): |
|
89 | 91 | proxyurl = 'http://' + proxyurl + '/' |
|
90 | 92 | proxy = util.url(proxyurl) |
|
91 | 93 | if not proxy.user: |
|
92 | 94 | proxy.user = ui.config("http_proxy", "user") |
|
93 | 95 | proxy.passwd = ui.config("http_proxy", "passwd") |
|
94 | 96 | |
|
95 | 97 | # see if we should use a proxy for this url |
|
96 | 98 | no_list = ["localhost", "127.0.0.1"] |
|
97 | 99 | no_list.extend([p.lower() for |
|
98 | 100 | p in ui.configlist("http_proxy", "no")]) |
|
99 | 101 | no_list.extend([p.strip().lower() for |
|
100 |
p in |
|
|
102 | p in pycompat.osgetenv("no_proxy", '').split(',') | |
|
101 | 103 | if p.strip()]) |
|
102 | 104 | # "http_proxy.always" config is for running tests on localhost |
|
103 | 105 | if ui.configbool("http_proxy", "always"): |
|
104 | 106 | self.no_list = [] |
|
105 | 107 | else: |
|
106 | 108 | self.no_list = no_list |
|
107 | 109 | |
|
108 | 110 | proxyurl = str(proxy) |
|
109 | 111 | proxies = {'http': proxyurl, 'https': proxyurl} |
|
110 | 112 | ui.debug('proxying through http://%s:%s\n' % |
|
111 | 113 | (proxy.host, proxy.port)) |
|
112 | 114 | else: |
|
113 | 115 | proxies = {} |
|
114 | 116 | |
|
115 | 117 | urlreq.proxyhandler.__init__(self, proxies) |
|
116 | 118 | self.ui = ui |
|
117 | 119 | |
|
118 | 120 | def proxy_open(self, req, proxy, type_): |
|
119 | 121 | host = req.get_host().split(':')[0] |
|
120 | 122 | for e in self.no_list: |
|
121 | 123 | if host == e: |
|
122 | 124 | return None |
|
123 | 125 | if e.startswith('*.') and host.endswith(e[2:]): |
|
124 | 126 | return None |
|
125 | 127 | if e.startswith('.') and host.endswith(e[1:]): |
|
126 | 128 | return None |
|
127 | 129 | |
|
128 | 130 | return urlreq.proxyhandler.proxy_open(self, req, proxy, type_) |
|
129 | 131 | |
|
130 | 132 | def _gen_sendfile(orgsend): |
|
131 | 133 | def _sendfile(self, data): |
|
132 | 134 | # send a file |
|
133 | 135 | if isinstance(data, httpconnectionmod.httpsendfile): |
|
134 | 136 | # if auth required, some data sent twice, so rewind here |
|
135 | 137 | data.seek(0) |
|
136 | 138 | for chunk in util.filechunkiter(data): |
|
137 | 139 | orgsend(self, chunk) |
|
138 | 140 | else: |
|
139 | 141 | orgsend(self, data) |
|
140 | 142 | return _sendfile |
|
141 | 143 | |
|
142 | 144 | has_https = util.safehasattr(urlreq, 'httpshandler') |
|
143 | 145 | |
|
144 | 146 | class httpconnection(keepalive.HTTPConnection): |
|
145 | 147 | # must be able to send big bundle as stream. |
|
146 | 148 | send = _gen_sendfile(keepalive.HTTPConnection.send) |
|
147 | 149 | |
|
148 | 150 | def getresponse(self): |
|
149 | 151 | proxyres = getattr(self, 'proxyres', None) |
|
150 | 152 | if proxyres: |
|
151 | 153 | if proxyres.will_close: |
|
152 | 154 | self.close() |
|
153 | 155 | self.proxyres = None |
|
154 | 156 | return proxyres |
|
155 | 157 | return keepalive.HTTPConnection.getresponse(self) |
|
156 | 158 | |
|
157 | 159 | # general transaction handler to support different ways to handle |
|
158 | 160 | # HTTPS proxying before and after Python 2.6.3. |
|
159 | 161 | def _generic_start_transaction(handler, h, req): |
|
160 | 162 | tunnel_host = getattr(req, '_tunnel_host', None) |
|
161 | 163 | if tunnel_host: |
|
162 | 164 | if tunnel_host[:7] not in ['http://', 'https:/']: |
|
163 | 165 | tunnel_host = 'https://' + tunnel_host |
|
164 | 166 | new_tunnel = True |
|
165 | 167 | else: |
|
166 | 168 | tunnel_host = req.get_selector() |
|
167 | 169 | new_tunnel = False |
|
168 | 170 | |
|
169 | 171 | if new_tunnel or tunnel_host == req.get_full_url(): # has proxy |
|
170 | 172 | u = util.url(tunnel_host) |
|
171 | 173 | if new_tunnel or u.scheme == 'https': # only use CONNECT for HTTPS |
|
172 | 174 | h.realhostport = ':'.join([u.host, (u.port or '443')]) |
|
173 | 175 | h.headers = req.headers.copy() |
|
174 | 176 | h.headers.update(handler.parent.addheaders) |
|
175 | 177 | return |
|
176 | 178 | |
|
177 | 179 | h.realhostport = None |
|
178 | 180 | h.headers = None |
|
179 | 181 | |
|
180 | 182 | def _generic_proxytunnel(self): |
|
181 | 183 | proxyheaders = dict( |
|
182 | 184 | [(x, self.headers[x]) for x in self.headers |
|
183 | 185 | if x.lower().startswith('proxy-')]) |
|
184 | 186 | self.send('CONNECT %s HTTP/1.0\r\n' % self.realhostport) |
|
185 | 187 | for header in proxyheaders.iteritems(): |
|
186 | 188 | self.send('%s: %s\r\n' % header) |
|
187 | 189 | self.send('\r\n') |
|
188 | 190 | |
|
189 | 191 | # majority of the following code is duplicated from |
|
190 | 192 | # httplib.HTTPConnection as there are no adequate places to |
|
191 | 193 | # override functions to provide the needed functionality |
|
192 | 194 | res = self.response_class(self.sock, |
|
193 | 195 | strict=self.strict, |
|
194 | 196 | method=self._method) |
|
195 | 197 | |
|
196 | 198 | while True: |
|
197 | 199 | version, status, reason = res._read_status() |
|
198 | 200 | if status != httplib.CONTINUE: |
|
199 | 201 | break |
|
200 | 202 | # skip lines that are all whitespace |
|
201 | 203 | list(iter(lambda: res.fp.readline().strip(), '')) |
|
202 | 204 | res.status = status |
|
203 | 205 | res.reason = reason.strip() |
|
204 | 206 | |
|
205 | 207 | if res.status == 200: |
|
206 | 208 | # skip lines until we find a blank line |
|
207 | 209 | list(iter(res.fp.readline, '\r\n')) |
|
208 | 210 | return True |
|
209 | 211 | |
|
210 | 212 | if version == 'HTTP/1.0': |
|
211 | 213 | res.version = 10 |
|
212 | 214 | elif version.startswith('HTTP/1.'): |
|
213 | 215 | res.version = 11 |
|
214 | 216 | elif version == 'HTTP/0.9': |
|
215 | 217 | res.version = 9 |
|
216 | 218 | else: |
|
217 | 219 | raise httplib.UnknownProtocol(version) |
|
218 | 220 | |
|
219 | 221 | if res.version == 9: |
|
220 | 222 | res.length = None |
|
221 | 223 | res.chunked = 0 |
|
222 | 224 | res.will_close = 1 |
|
223 | 225 | res.msg = httplib.HTTPMessage(stringio()) |
|
224 | 226 | return False |
|
225 | 227 | |
|
226 | 228 | res.msg = httplib.HTTPMessage(res.fp) |
|
227 | 229 | res.msg.fp = None |
|
228 | 230 | |
|
229 | 231 | # are we using the chunked-style of transfer encoding? |
|
230 | 232 | trenc = res.msg.getheader('transfer-encoding') |
|
231 | 233 | if trenc and trenc.lower() == "chunked": |
|
232 | 234 | res.chunked = 1 |
|
233 | 235 | res.chunk_left = None |
|
234 | 236 | else: |
|
235 | 237 | res.chunked = 0 |
|
236 | 238 | |
|
237 | 239 | # will the connection close at the end of the response? |
|
238 | 240 | res.will_close = res._check_close() |
|
239 | 241 | |
|
240 | 242 | # do we have a Content-Length? |
|
241 | 243 | # NOTE: RFC 2616, section 4.4, #3 says we ignore this if |
|
242 | 244 | # transfer-encoding is "chunked" |
|
243 | 245 | length = res.msg.getheader('content-length') |
|
244 | 246 | if length and not res.chunked: |
|
245 | 247 | try: |
|
246 | 248 | res.length = int(length) |
|
247 | 249 | except ValueError: |
|
248 | 250 | res.length = None |
|
249 | 251 | else: |
|
250 | 252 | if res.length < 0: # ignore nonsensical negative lengths |
|
251 | 253 | res.length = None |
|
252 | 254 | else: |
|
253 | 255 | res.length = None |
|
254 | 256 | |
|
255 | 257 | # does the body have a fixed length? (of zero) |
|
256 | 258 | if (status == httplib.NO_CONTENT or status == httplib.NOT_MODIFIED or |
|
257 | 259 | 100 <= status < 200 or # 1xx codes |
|
258 | 260 | res._method == 'HEAD'): |
|
259 | 261 | res.length = 0 |
|
260 | 262 | |
|
261 | 263 | # if the connection remains open, and we aren't using chunked, and |
|
262 | 264 | # a content-length was not provided, then assume that the connection |
|
263 | 265 | # WILL close. |
|
264 | 266 | if (not res.will_close and |
|
265 | 267 | not res.chunked and |
|
266 | 268 | res.length is None): |
|
267 | 269 | res.will_close = 1 |
|
268 | 270 | |
|
269 | 271 | self.proxyres = res |
|
270 | 272 | |
|
271 | 273 | return False |
|
272 | 274 | |
|
273 | 275 | class httphandler(keepalive.HTTPHandler): |
|
274 | 276 | def http_open(self, req): |
|
275 | 277 | return self.do_open(httpconnection, req) |
|
276 | 278 | |
|
277 | 279 | def _start_transaction(self, h, req): |
|
278 | 280 | _generic_start_transaction(self, h, req) |
|
279 | 281 | return keepalive.HTTPHandler._start_transaction(self, h, req) |
|
280 | 282 | |
|
281 | 283 | if has_https: |
|
282 | 284 | class httpsconnection(httplib.HTTPConnection): |
|
283 | 285 | response_class = keepalive.HTTPResponse |
|
284 | 286 | default_port = httplib.HTTPS_PORT |
|
285 | 287 | # must be able to send big bundle as stream. |
|
286 | 288 | send = _gen_sendfile(keepalive.safesend) |
|
287 | 289 | getresponse = keepalive.wrapgetresponse(httplib.HTTPConnection) |
|
288 | 290 | |
|
289 | 291 | def __init__(self, host, port=None, key_file=None, cert_file=None, |
|
290 | 292 | *args, **kwargs): |
|
291 | 293 | httplib.HTTPConnection.__init__(self, host, port, *args, **kwargs) |
|
292 | 294 | self.key_file = key_file |
|
293 | 295 | self.cert_file = cert_file |
|
294 | 296 | |
|
295 | 297 | def connect(self): |
|
296 | 298 | self.sock = socket.create_connection((self.host, self.port)) |
|
297 | 299 | |
|
298 | 300 | host = self.host |
|
299 | 301 | if self.realhostport: # use CONNECT proxy |
|
300 | 302 | _generic_proxytunnel(self) |
|
301 | 303 | host = self.realhostport.rsplit(':', 1)[0] |
|
302 | 304 | self.sock = sslutil.wrapsocket( |
|
303 | 305 | self.sock, self.key_file, self.cert_file, ui=self.ui, |
|
304 | 306 | serverhostname=host) |
|
305 | 307 | sslutil.validatesocket(self.sock) |
|
306 | 308 | |
|
307 | 309 | class httpshandler(keepalive.KeepAliveHandler, urlreq.httpshandler): |
|
308 | 310 | def __init__(self, ui): |
|
309 | 311 | keepalive.KeepAliveHandler.__init__(self) |
|
310 | 312 | urlreq.httpshandler.__init__(self) |
|
311 | 313 | self.ui = ui |
|
312 | 314 | self.pwmgr = passwordmgr(self.ui, |
|
313 | 315 | self.ui.httppasswordmgrdb) |
|
314 | 316 | |
|
315 | 317 | def _start_transaction(self, h, req): |
|
316 | 318 | _generic_start_transaction(self, h, req) |
|
317 | 319 | return keepalive.KeepAliveHandler._start_transaction(self, h, req) |
|
318 | 320 | |
|
319 | 321 | def https_open(self, req): |
|
320 | 322 | # req.get_full_url() does not contain credentials and we may |
|
321 | 323 | # need them to match the certificates. |
|
322 | 324 | url = req.get_full_url() |
|
323 | 325 | user, password = self.pwmgr.find_stored_password(url) |
|
324 | 326 | res = httpconnectionmod.readauthforuri(self.ui, url, user) |
|
325 | 327 | if res: |
|
326 | 328 | group, auth = res |
|
327 | 329 | self.auth = auth |
|
328 | 330 | self.ui.debug("using auth.%s.* for authentication\n" % group) |
|
329 | 331 | else: |
|
330 | 332 | self.auth = None |
|
331 | 333 | return self.do_open(self._makeconnection, req) |
|
332 | 334 | |
|
333 | 335 | def _makeconnection(self, host, port=None, *args, **kwargs): |
|
334 | 336 | keyfile = None |
|
335 | 337 | certfile = None |
|
336 | 338 | |
|
337 | 339 | if len(args) >= 1: # key_file |
|
338 | 340 | keyfile = args[0] |
|
339 | 341 | if len(args) >= 2: # cert_file |
|
340 | 342 | certfile = args[1] |
|
341 | 343 | args = args[2:] |
|
342 | 344 | |
|
343 | 345 | # if the user has specified different key/cert files in |
|
344 | 346 | # hgrc, we prefer these |
|
345 | 347 | if self.auth and 'key' in self.auth and 'cert' in self.auth: |
|
346 | 348 | keyfile = self.auth['key'] |
|
347 | 349 | certfile = self.auth['cert'] |
|
348 | 350 | |
|
349 | 351 | conn = httpsconnection(host, port, keyfile, certfile, *args, |
|
350 | 352 | **kwargs) |
|
351 | 353 | conn.ui = self.ui |
|
352 | 354 | return conn |
|
353 | 355 | |
|
354 | 356 | class httpdigestauthhandler(urlreq.httpdigestauthhandler): |
|
355 | 357 | def __init__(self, *args, **kwargs): |
|
356 | 358 | urlreq.httpdigestauthhandler.__init__(self, *args, **kwargs) |
|
357 | 359 | self.retried_req = None |
|
358 | 360 | |
|
359 | 361 | def reset_retry_count(self): |
|
360 | 362 | # Python 2.6.5 will call this on 401 or 407 errors and thus loop |
|
361 | 363 | # forever. We disable reset_retry_count completely and reset in |
|
362 | 364 | # http_error_auth_reqed instead. |
|
363 | 365 | pass |
|
364 | 366 | |
|
365 | 367 | def http_error_auth_reqed(self, auth_header, host, req, headers): |
|
366 | 368 | # Reset the retry counter once for each request. |
|
367 | 369 | if req is not self.retried_req: |
|
368 | 370 | self.retried_req = req |
|
369 | 371 | self.retried = 0 |
|
370 | 372 | return urlreq.httpdigestauthhandler.http_error_auth_reqed( |
|
371 | 373 | self, auth_header, host, req, headers) |
|
372 | 374 | |
|
373 | 375 | class httpbasicauthhandler(urlreq.httpbasicauthhandler): |
|
374 | 376 | def __init__(self, *args, **kwargs): |
|
375 | 377 | self.auth = None |
|
376 | 378 | urlreq.httpbasicauthhandler.__init__(self, *args, **kwargs) |
|
377 | 379 | self.retried_req = None |
|
378 | 380 | |
|
379 | 381 | def http_request(self, request): |
|
380 | 382 | if self.auth: |
|
381 | 383 | request.add_unredirected_header(self.auth_header, self.auth) |
|
382 | 384 | |
|
383 | 385 | return request |
|
384 | 386 | |
|
385 | 387 | def https_request(self, request): |
|
386 | 388 | if self.auth: |
|
387 | 389 | request.add_unredirected_header(self.auth_header, self.auth) |
|
388 | 390 | |
|
389 | 391 | return request |
|
390 | 392 | |
|
391 | 393 | def reset_retry_count(self): |
|
392 | 394 | # Python 2.6.5 will call this on 401 or 407 errors and thus loop |
|
393 | 395 | # forever. We disable reset_retry_count completely and reset in |
|
394 | 396 | # http_error_auth_reqed instead. |
|
395 | 397 | pass |
|
396 | 398 | |
|
397 | 399 | def http_error_auth_reqed(self, auth_header, host, req, headers): |
|
398 | 400 | # Reset the retry counter once for each request. |
|
399 | 401 | if req is not self.retried_req: |
|
400 | 402 | self.retried_req = req |
|
401 | 403 | self.retried = 0 |
|
402 | 404 | return urlreq.httpbasicauthhandler.http_error_auth_reqed( |
|
403 | 405 | self, auth_header, host, req, headers) |
|
404 | 406 | |
|
405 | 407 | def retry_http_basic_auth(self, host, req, realm): |
|
406 | 408 | user, pw = self.passwd.find_user_password(realm, req.get_full_url()) |
|
407 | 409 | if pw is not None: |
|
408 | 410 | raw = "%s:%s" % (user, pw) |
|
409 | 411 | auth = 'Basic %s' % base64.b64encode(raw).strip() |
|
410 | 412 | if req.get_header(self.auth_header, None) == auth: |
|
411 | 413 | return None |
|
412 | 414 | self.auth = auth |
|
413 | 415 | req.add_unredirected_header(self.auth_header, auth) |
|
414 | 416 | return self.parent.open(req) |
|
415 | 417 | else: |
|
416 | 418 | return None |
|
417 | 419 | |
|
418 | 420 | handlerfuncs = [] |
|
419 | 421 | |
|
420 | 422 | def opener(ui, authinfo=None): |
|
421 | 423 | ''' |
|
422 | 424 | construct an opener suitable for urllib2 |
|
423 | 425 | authinfo will be added to the password manager |
|
424 | 426 | ''' |
|
425 | 427 | # experimental config: ui.usehttp2 |
|
426 | 428 | if ui.configbool('ui', 'usehttp2', False): |
|
427 | 429 | handlers = [ |
|
428 | 430 | httpconnectionmod.http2handler( |
|
429 | 431 | ui, |
|
430 | 432 | passwordmgr(ui, ui.httppasswordmgrdb)) |
|
431 | 433 | ] |
|
432 | 434 | else: |
|
433 | 435 | handlers = [httphandler()] |
|
434 | 436 | if has_https: |
|
435 | 437 | handlers.append(httpshandler(ui)) |
|
436 | 438 | |
|
437 | 439 | handlers.append(proxyhandler(ui)) |
|
438 | 440 | |
|
439 | 441 | passmgr = passwordmgr(ui, ui.httppasswordmgrdb) |
|
440 | 442 | if authinfo is not None: |
|
441 | 443 | realm, uris, user, passwd = authinfo |
|
442 | 444 | saveduser, savedpass = passmgr.find_stored_password(uris[0]) |
|
443 | 445 | if user != saveduser or passwd: |
|
444 | 446 | passmgr.add_password(realm, uris, user, passwd) |
|
445 | 447 | ui.debug('http auth: user %s, password %s\n' % |
|
446 | 448 | (user, passwd and '*' * len(passwd) or 'not set')) |
|
447 | 449 | |
|
448 | 450 | handlers.extend((httpbasicauthhandler(passmgr), |
|
449 | 451 | httpdigestauthhandler(passmgr))) |
|
450 | 452 | handlers.extend([h(ui, passmgr) for h in handlerfuncs]) |
|
451 | 453 | opener = urlreq.buildopener(*handlers) |
|
452 | 454 | |
|
453 | 455 | # The user agent should should *NOT* be used by servers for e.g. |
|
454 | 456 | # protocol detection or feature negotiation: there are other |
|
455 | 457 | # facilities for that. |
|
456 | 458 | # |
|
457 | 459 | # "mercurial/proto-1.0" was the original user agent string and |
|
458 | 460 | # exists for backwards compatibility reasons. |
|
459 | 461 | # |
|
460 | 462 | # The "(Mercurial %s)" string contains the distribution |
|
461 | 463 | # name and version. Other client implementations should choose their |
|
462 | 464 | # own distribution name. Since servers should not be using the user |
|
463 | 465 | # agent string for anything, clients should be able to define whatever |
|
464 | 466 | # user agent they deem appropriate. |
|
465 | 467 | agent = 'mercurial/proto-1.0 (Mercurial %s)' % util.version() |
|
466 | 468 | opener.addheaders = [('User-agent', agent)] |
|
467 | 469 | opener.addheaders.append(('Accept', 'application/mercurial-0.1')) |
|
468 | 470 | return opener |
|
469 | 471 | |
|
470 | 472 | def open(ui, url_, data=None): |
|
471 | 473 | u = util.url(url_) |
|
472 | 474 | if u.scheme: |
|
473 | 475 | u.scheme = u.scheme.lower() |
|
474 | 476 | url_, authinfo = u.authinfo() |
|
475 | 477 | else: |
|
476 | 478 | path = util.normpath(os.path.abspath(url_)) |
|
477 | 479 | url_ = 'file://' + urlreq.pathname2url(path) |
|
478 | 480 | authinfo = None |
|
479 | 481 | return opener(ui, authinfo).open(url_, data) |
General Comments 0
You need to be logged in to leave comments.
Login now