Show More
@@ -1,629 +1,629 b'' | |||
|
1 | 1 | # store.py - repository store handling for Mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2008 Matt Mackall <mpm@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import errno |
|
11 | 11 | import hashlib |
|
12 | 12 | import os |
|
13 | 13 | import stat |
|
14 | 14 | |
|
15 | 15 | from .i18n import _ |
|
16 | 16 | from . import ( |
|
17 | 17 | error, |
|
18 | 18 | node, |
|
19 | 19 | policy, |
|
20 | 20 | pycompat, |
|
21 | 21 | util, |
|
22 | 22 | vfs as vfsmod, |
|
23 | 23 | ) |
|
24 | 24 | |
|
25 | 25 | parsers = policy.importmod(r'parsers') |
|
26 | 26 | |
|
27 | 27 | def _matchtrackedpath(path, matcher): |
|
28 | 28 | """parses a fncache entry and returns whether the entry is tracking a path |
|
29 | 29 | matched by matcher or not. |
|
30 | 30 | |
|
31 | 31 | If matcher is None, returns True""" |
|
32 | 32 | |
|
33 | 33 | if matcher is None: |
|
34 | 34 | return True |
|
35 | 35 | path = decodedir(path) |
|
36 | 36 | if path.startswith('data/'): |
|
37 | 37 | return matcher(path[len('data/'):-len('.i')]) |
|
38 | 38 | elif path.startswith('meta/'): |
|
39 | 39 | return matcher.visitdir(path[len('meta/'):-len('/00manifest.i')] or '.') |
|
40 | 40 | |
|
41 | 41 | raise error.ProgrammingError("cannot decode path %s" % path) |
|
42 | 42 | |
|
43 | 43 | # This avoids a collision between a file named foo and a dir named |
|
44 | 44 | # foo.i or foo.d |
|
45 | 45 | def _encodedir(path): |
|
46 | 46 | ''' |
|
47 | 47 | >>> _encodedir(b'data/foo.i') |
|
48 | 48 | 'data/foo.i' |
|
49 | 49 | >>> _encodedir(b'data/foo.i/bla.i') |
|
50 | 50 | 'data/foo.i.hg/bla.i' |
|
51 | 51 | >>> _encodedir(b'data/foo.i.hg/bla.i') |
|
52 | 52 | 'data/foo.i.hg.hg/bla.i' |
|
53 | 53 | >>> _encodedir(b'data/foo.i\\ndata/foo.i/bla.i\\ndata/foo.i.hg/bla.i\\n') |
|
54 | 54 | 'data/foo.i\\ndata/foo.i.hg/bla.i\\ndata/foo.i.hg.hg/bla.i\\n' |
|
55 | 55 | ''' |
|
56 | 56 | return (path |
|
57 | 57 | .replace(".hg/", ".hg.hg/") |
|
58 | 58 | .replace(".i/", ".i.hg/") |
|
59 | 59 | .replace(".d/", ".d.hg/")) |
|
60 | 60 | |
|
61 | 61 | encodedir = getattr(parsers, 'encodedir', _encodedir) |
|
62 | 62 | |
|
63 | 63 | def decodedir(path): |
|
64 | 64 | ''' |
|
65 | 65 | >>> decodedir(b'data/foo.i') |
|
66 | 66 | 'data/foo.i' |
|
67 | 67 | >>> decodedir(b'data/foo.i.hg/bla.i') |
|
68 | 68 | 'data/foo.i/bla.i' |
|
69 | 69 | >>> decodedir(b'data/foo.i.hg.hg/bla.i') |
|
70 | 70 | 'data/foo.i.hg/bla.i' |
|
71 | 71 | ''' |
|
72 | 72 | if ".hg/" not in path: |
|
73 | 73 | return path |
|
74 | 74 | return (path |
|
75 | 75 | .replace(".d.hg/", ".d/") |
|
76 | 76 | .replace(".i.hg/", ".i/") |
|
77 | 77 | .replace(".hg.hg/", ".hg/")) |
|
78 | 78 | |
|
79 | 79 | def _reserved(): |
|
80 | 80 | ''' characters that are problematic for filesystems |
|
81 | 81 | |
|
82 | 82 | * ascii escapes (0..31) |
|
83 | 83 | * ascii hi (126..255) |
|
84 | 84 | * windows specials |
|
85 | 85 | |
|
86 | 86 | these characters will be escaped by encodefunctions |
|
87 | 87 | ''' |
|
88 | 88 | winreserved = [ord(x) for x in u'\\:*?"<>|'] |
|
89 | 89 | for x in range(32): |
|
90 | 90 | yield x |
|
91 | 91 | for x in range(126, 256): |
|
92 | 92 | yield x |
|
93 | 93 | for x in winreserved: |
|
94 | 94 | yield x |
|
95 | 95 | |
|
96 | 96 | def _buildencodefun(): |
|
97 | 97 | ''' |
|
98 | 98 | >>> enc, dec = _buildencodefun() |
|
99 | 99 | |
|
100 | 100 | >>> enc(b'nothing/special.txt') |
|
101 | 101 | 'nothing/special.txt' |
|
102 | 102 | >>> dec(b'nothing/special.txt') |
|
103 | 103 | 'nothing/special.txt' |
|
104 | 104 | |
|
105 | 105 | >>> enc(b'HELLO') |
|
106 | 106 | '_h_e_l_l_o' |
|
107 | 107 | >>> dec(b'_h_e_l_l_o') |
|
108 | 108 | 'HELLO' |
|
109 | 109 | |
|
110 | 110 | >>> enc(b'hello:world?') |
|
111 | 111 | 'hello~3aworld~3f' |
|
112 | 112 | >>> dec(b'hello~3aworld~3f') |
|
113 | 113 | 'hello:world?' |
|
114 | 114 | |
|
115 | 115 | >>> enc(b'the\\x07quick\\xADshot') |
|
116 | 116 | 'the~07quick~adshot' |
|
117 | 117 | >>> dec(b'the~07quick~adshot') |
|
118 | 118 | 'the\\x07quick\\xadshot' |
|
119 | 119 | ''' |
|
120 | 120 | e = '_' |
|
121 | 121 | xchr = pycompat.bytechr |
|
122 | 122 | asciistr = list(map(xchr, range(127))) |
|
123 | 123 | capitals = list(range(ord("A"), ord("Z") + 1)) |
|
124 | 124 | |
|
125 | 125 | cmap = dict((x, x) for x in asciistr) |
|
126 | 126 | for x in _reserved(): |
|
127 | 127 | cmap[xchr(x)] = "~%02x" % x |
|
128 | 128 | for x in capitals + [ord(e)]: |
|
129 | 129 | cmap[xchr(x)] = e + xchr(x).lower() |
|
130 | 130 | |
|
131 | 131 | dmap = {} |
|
132 | 132 | for k, v in cmap.iteritems(): |
|
133 | 133 | dmap[v] = k |
|
134 | 134 | def decode(s): |
|
135 | 135 | i = 0 |
|
136 | 136 | while i < len(s): |
|
137 | 137 | for l in pycompat.xrange(1, 4): |
|
138 | 138 | try: |
|
139 | 139 | yield dmap[s[i:i + l]] |
|
140 | 140 | i += l |
|
141 | 141 | break |
|
142 | 142 | except KeyError: |
|
143 | 143 | pass |
|
144 | 144 | else: |
|
145 | 145 | raise KeyError |
|
146 | 146 | return (lambda s: ''.join([cmap[s[c:c + 1]] |
|
147 | 147 | for c in pycompat.xrange(len(s))]), |
|
148 | 148 | lambda s: ''.join(list(decode(s)))) |
|
149 | 149 | |
|
150 | 150 | _encodefname, _decodefname = _buildencodefun() |
|
151 | 151 | |
|
152 | 152 | def encodefilename(s): |
|
153 | 153 | ''' |
|
154 | 154 | >>> encodefilename(b'foo.i/bar.d/bla.hg/hi:world?/HELLO') |
|
155 | 155 | 'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o' |
|
156 | 156 | ''' |
|
157 | 157 | return _encodefname(encodedir(s)) |
|
158 | 158 | |
|
159 | 159 | def decodefilename(s): |
|
160 | 160 | ''' |
|
161 | 161 | >>> decodefilename(b'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o') |
|
162 | 162 | 'foo.i/bar.d/bla.hg/hi:world?/HELLO' |
|
163 | 163 | ''' |
|
164 | 164 | return decodedir(_decodefname(s)) |
|
165 | 165 | |
|
166 | 166 | def _buildlowerencodefun(): |
|
167 | 167 | ''' |
|
168 | 168 | >>> f = _buildlowerencodefun() |
|
169 | 169 | >>> f(b'nothing/special.txt') |
|
170 | 170 | 'nothing/special.txt' |
|
171 | 171 | >>> f(b'HELLO') |
|
172 | 172 | 'hello' |
|
173 | 173 | >>> f(b'hello:world?') |
|
174 | 174 | 'hello~3aworld~3f' |
|
175 | 175 | >>> f(b'the\\x07quick\\xADshot') |
|
176 | 176 | 'the~07quick~adshot' |
|
177 | 177 | ''' |
|
178 | 178 | xchr = pycompat.bytechr |
|
179 | 179 | cmap = dict([(xchr(x), xchr(x)) for x in pycompat.xrange(127)]) |
|
180 | 180 | for x in _reserved(): |
|
181 | 181 | cmap[xchr(x)] = "~%02x" % x |
|
182 | 182 | for x in range(ord("A"), ord("Z") + 1): |
|
183 | 183 | cmap[xchr(x)] = xchr(x).lower() |
|
184 | 184 | def lowerencode(s): |
|
185 | 185 | return "".join([cmap[c] for c in pycompat.iterbytestr(s)]) |
|
186 | 186 | return lowerencode |
|
187 | 187 | |
|
188 | 188 | lowerencode = getattr(parsers, 'lowerencode', None) or _buildlowerencodefun() |
|
189 | 189 | |
|
190 | 190 | # Windows reserved names: con, prn, aux, nul, com1..com9, lpt1..lpt9 |
|
191 | 191 | _winres3 = ('aux', 'con', 'prn', 'nul') # length 3 |
|
192 | 192 | _winres4 = ('com', 'lpt') # length 4 (with trailing 1..9) |
|
193 | 193 | def _auxencode(path, dotencode): |
|
194 | 194 | ''' |
|
195 | 195 | Encodes filenames containing names reserved by Windows or which end in |
|
196 | 196 | period or space. Does not touch other single reserved characters c. |
|
197 | 197 | Specifically, c in '\\:*?"<>|' or ord(c) <= 31 are *not* encoded here. |
|
198 | 198 | Additionally encodes space or period at the beginning, if dotencode is |
|
199 | 199 | True. Parameter path is assumed to be all lowercase. |
|
200 | 200 | A segment only needs encoding if a reserved name appears as a |
|
201 | 201 | basename (e.g. "aux", "aux.foo"). A directory or file named "foo.aux" |
|
202 | 202 | doesn't need encoding. |
|
203 | 203 | |
|
204 | 204 | >>> s = b'.foo/aux.txt/txt.aux/con/prn/nul/foo.' |
|
205 | 205 | >>> _auxencode(s.split(b'/'), True) |
|
206 | 206 | ['~2efoo', 'au~78.txt', 'txt.aux', 'co~6e', 'pr~6e', 'nu~6c', 'foo~2e'] |
|
207 | 207 | >>> s = b'.com1com2/lpt9.lpt4.lpt1/conprn/com0/lpt0/foo.' |
|
208 | 208 | >>> _auxencode(s.split(b'/'), False) |
|
209 | 209 | ['.com1com2', 'lp~749.lpt4.lpt1', 'conprn', 'com0', 'lpt0', 'foo~2e'] |
|
210 | 210 | >>> _auxencode([b'foo. '], True) |
|
211 | 211 | ['foo.~20'] |
|
212 | 212 | >>> _auxencode([b' .foo'], True) |
|
213 | 213 | ['~20.foo'] |
|
214 | 214 | ''' |
|
215 | 215 | for i, n in enumerate(path): |
|
216 | 216 | if not n: |
|
217 | 217 | continue |
|
218 | 218 | if dotencode and n[0] in '. ': |
|
219 | 219 | n = "~%02x" % ord(n[0:1]) + n[1:] |
|
220 | 220 | path[i] = n |
|
221 | 221 | else: |
|
222 | 222 | l = n.find('.') |
|
223 | 223 | if l == -1: |
|
224 | 224 | l = len(n) |
|
225 | 225 | if ((l == 3 and n[:3] in _winres3) or |
|
226 | 226 | (l == 4 and n[3:4] <= '9' and n[3:4] >= '1' |
|
227 | 227 | and n[:3] in _winres4)): |
|
228 | 228 | # encode third letter ('aux' -> 'au~78') |
|
229 | 229 | ec = "~%02x" % ord(n[2:3]) |
|
230 | 230 | n = n[0:2] + ec + n[3:] |
|
231 | 231 | path[i] = n |
|
232 | 232 | if n[-1] in '. ': |
|
233 | 233 | # encode last period or space ('foo...' -> 'foo..~2e') |
|
234 | 234 | path[i] = n[:-1] + "~%02x" % ord(n[-1:]) |
|
235 | 235 | return path |
|
236 | 236 | |
|
237 | 237 | _maxstorepathlen = 120 |
|
238 | 238 | _dirprefixlen = 8 |
|
239 | 239 | _maxshortdirslen = 8 * (_dirprefixlen + 1) - 4 |
|
240 | 240 | |
|
241 | 241 | def _hashencode(path, dotencode): |
|
242 | 242 | digest = node.hex(hashlib.sha1(path).digest()) |
|
243 | 243 | le = lowerencode(path[5:]).split('/') # skips prefix 'data/' or 'meta/' |
|
244 | 244 | parts = _auxencode(le, dotencode) |
|
245 | 245 | basename = parts[-1] |
|
246 | 246 | _root, ext = os.path.splitext(basename) |
|
247 | 247 | sdirs = [] |
|
248 | 248 | sdirslen = 0 |
|
249 | 249 | for p in parts[:-1]: |
|
250 | 250 | d = p[:_dirprefixlen] |
|
251 | 251 | if d[-1] in '. ': |
|
252 | 252 | # Windows can't access dirs ending in period or space |
|
253 | 253 | d = d[:-1] + '_' |
|
254 | 254 | if sdirslen == 0: |
|
255 | 255 | t = len(d) |
|
256 | 256 | else: |
|
257 | 257 | t = sdirslen + 1 + len(d) |
|
258 | 258 | if t > _maxshortdirslen: |
|
259 | 259 | break |
|
260 | 260 | sdirs.append(d) |
|
261 | 261 | sdirslen = t |
|
262 | 262 | dirs = '/'.join(sdirs) |
|
263 | 263 | if len(dirs) > 0: |
|
264 | 264 | dirs += '/' |
|
265 | 265 | res = 'dh/' + dirs + digest + ext |
|
266 | 266 | spaceleft = _maxstorepathlen - len(res) |
|
267 | 267 | if spaceleft > 0: |
|
268 | 268 | filler = basename[:spaceleft] |
|
269 | 269 | res = 'dh/' + dirs + filler + digest + ext |
|
270 | 270 | return res |
|
271 | 271 | |
|
272 | 272 | def _hybridencode(path, dotencode): |
|
273 | 273 | '''encodes path with a length limit |
|
274 | 274 | |
|
275 | 275 | Encodes all paths that begin with 'data/', according to the following. |
|
276 | 276 | |
|
277 | 277 | Default encoding (reversible): |
|
278 | 278 | |
|
279 | 279 | Encodes all uppercase letters 'X' as '_x'. All reserved or illegal |
|
280 | 280 | characters are encoded as '~xx', where xx is the two digit hex code |
|
281 | 281 | of the character (see encodefilename). |
|
282 | 282 | Relevant path components consisting of Windows reserved filenames are |
|
283 | 283 | masked by encoding the third character ('aux' -> 'au~78', see _auxencode). |
|
284 | 284 | |
|
285 | 285 | Hashed encoding (not reversible): |
|
286 | 286 | |
|
287 | 287 | If the default-encoded path is longer than _maxstorepathlen, a |
|
288 | 288 | non-reversible hybrid hashing of the path is done instead. |
|
289 | 289 | This encoding uses up to _dirprefixlen characters of all directory |
|
290 | 290 | levels of the lowerencoded path, but not more levels than can fit into |
|
291 | 291 | _maxshortdirslen. |
|
292 | 292 | Then follows the filler followed by the sha digest of the full path. |
|
293 | 293 | The filler is the beginning of the basename of the lowerencoded path |
|
294 | 294 | (the basename is everything after the last path separator). The filler |
|
295 | 295 | is as long as possible, filling in characters from the basename until |
|
296 | 296 | the encoded path has _maxstorepathlen characters (or all chars of the |
|
297 | 297 | basename have been taken). |
|
298 | 298 | The extension (e.g. '.i' or '.d') is preserved. |
|
299 | 299 | |
|
300 | 300 | The string 'data/' at the beginning is replaced with 'dh/', if the hashed |
|
301 | 301 | encoding was used. |
|
302 | 302 | ''' |
|
303 | 303 | path = encodedir(path) |
|
304 | 304 | ef = _encodefname(path).split('/') |
|
305 | 305 | res = '/'.join(_auxencode(ef, dotencode)) |
|
306 | 306 | if len(res) > _maxstorepathlen: |
|
307 | 307 | res = _hashencode(path, dotencode) |
|
308 | 308 | return res |
|
309 | 309 | |
|
310 | 310 | def _pathencode(path): |
|
311 | 311 | de = encodedir(path) |
|
312 | 312 | if len(path) > _maxstorepathlen: |
|
313 | 313 | return _hashencode(de, True) |
|
314 | 314 | ef = _encodefname(de).split('/') |
|
315 | 315 | res = '/'.join(_auxencode(ef, True)) |
|
316 | 316 | if len(res) > _maxstorepathlen: |
|
317 | 317 | return _hashencode(de, True) |
|
318 | 318 | return res |
|
319 | 319 | |
|
320 | 320 | _pathencode = getattr(parsers, 'pathencode', _pathencode) |
|
321 | 321 | |
|
322 | 322 | def _plainhybridencode(f): |
|
323 | 323 | return _hybridencode(f, False) |
|
324 | 324 | |
|
325 | 325 | def _calcmode(vfs): |
|
326 | 326 | try: |
|
327 | 327 | # files in .hg/ will be created using this mode |
|
328 | 328 | mode = vfs.stat().st_mode |
|
329 | 329 | # avoid some useless chmods |
|
330 | 330 | if (0o777 & ~util.umask) == (0o777 & mode): |
|
331 | 331 | mode = None |
|
332 | 332 | except OSError: |
|
333 | 333 | mode = None |
|
334 | 334 | return mode |
|
335 | 335 | |
|
336 | 336 | _data = ('narrowspec data meta 00manifest.d 00manifest.i' |
|
337 | 337 | ' 00changelog.d 00changelog.i phaseroots obsstore') |
|
338 | 338 | |
|
339 | 339 | def isrevlog(f, kind, st): |
|
340 | 340 | return kind == stat.S_IFREG and f[-2:] in ('.i', '.d') |
|
341 | 341 | |
|
342 | 342 | class basicstore(object): |
|
343 | 343 | '''base class for local repository stores''' |
|
344 | 344 | def __init__(self, path, vfstype): |
|
345 | 345 | vfs = vfstype(path) |
|
346 | 346 | self.path = vfs.base |
|
347 | 347 | self.createmode = _calcmode(vfs) |
|
348 | 348 | vfs.createmode = self.createmode |
|
349 | 349 | self.rawvfs = vfs |
|
350 | 350 | self.vfs = vfsmod.filtervfs(vfs, encodedir) |
|
351 | 351 | self.opener = self.vfs |
|
352 | 352 | |
|
353 | 353 | def join(self, f): |
|
354 | 354 | return self.path + '/' + encodedir(f) |
|
355 | 355 | |
|
356 | 356 | def _walk(self, relpath, recurse, filefilter=isrevlog): |
|
357 | 357 | '''yields (unencoded, encoded, size)''' |
|
358 | 358 | path = self.path |
|
359 | 359 | if relpath: |
|
360 | 360 | path += '/' + relpath |
|
361 | 361 | striplen = len(self.path) + 1 |
|
362 | 362 | l = [] |
|
363 | 363 | if self.rawvfs.isdir(path): |
|
364 | 364 | visit = [path] |
|
365 | 365 | readdir = self.rawvfs.readdir |
|
366 | 366 | while visit: |
|
367 | 367 | p = visit.pop() |
|
368 | 368 | for f, kind, st in readdir(p, stat=True): |
|
369 | 369 | fp = p + '/' + f |
|
370 | 370 | if filefilter(f, kind, st): |
|
371 | 371 | n = util.pconvert(fp[striplen:]) |
|
372 | 372 | l.append((decodedir(n), n, st.st_size)) |
|
373 | 373 | elif kind == stat.S_IFDIR and recurse: |
|
374 | 374 | visit.append(fp) |
|
375 | 375 | l.sort() |
|
376 | 376 | return l |
|
377 | 377 | |
|
378 | 378 | def datafiles(self, matcher=None): |
|
379 | 379 | return self._walk('data', True) + self._walk('meta', True) |
|
380 | 380 | |
|
381 | 381 | def topfiles(self): |
|
382 | 382 | # yield manifest before changelog |
|
383 | 383 | return reversed(self._walk('', False)) |
|
384 | 384 | |
|
385 | 385 | def walk(self, matcher=None): |
|
386 | 386 | '''yields (unencoded, encoded, size) |
|
387 | 387 | |
|
388 | 388 | if a matcher is passed, storage files of only those tracked paths |
|
389 | 389 | are passed with matches the matcher |
|
390 | 390 | ''' |
|
391 | 391 | # yield data files first |
|
392 | 392 | for x in self.datafiles(matcher): |
|
393 | 393 | yield x |
|
394 | 394 | for x in self.topfiles(): |
|
395 | 395 | yield x |
|
396 | 396 | |
|
397 | 397 | def copylist(self): |
|
398 | 398 | return ['requires'] + _data.split() |
|
399 | 399 | |
|
400 | 400 | def write(self, tr): |
|
401 | 401 | pass |
|
402 | 402 | |
|
403 | 403 | def invalidatecaches(self): |
|
404 | 404 | pass |
|
405 | 405 | |
|
406 | 406 | def markremoved(self, fn): |
|
407 | 407 | pass |
|
408 | 408 | |
|
409 | 409 | def __contains__(self, path): |
|
410 | 410 | '''Checks if the store contains path''' |
|
411 | 411 | path = "/".join(("data", path)) |
|
412 | 412 | # file? |
|
413 | 413 | if self.vfs.exists(path + ".i"): |
|
414 | 414 | return True |
|
415 | 415 | # dir? |
|
416 | 416 | if not path.endswith("/"): |
|
417 | 417 | path = path + "/" |
|
418 | 418 | return self.vfs.exists(path) |
|
419 | 419 | |
|
420 | 420 | class encodedstore(basicstore): |
|
421 | 421 | def __init__(self, path, vfstype): |
|
422 | 422 | vfs = vfstype(path + '/store') |
|
423 | 423 | self.path = vfs.base |
|
424 | 424 | self.createmode = _calcmode(vfs) |
|
425 | 425 | vfs.createmode = self.createmode |
|
426 | 426 | self.rawvfs = vfs |
|
427 | 427 | self.vfs = vfsmod.filtervfs(vfs, encodefilename) |
|
428 | 428 | self.opener = self.vfs |
|
429 | 429 | |
|
430 | 430 | def datafiles(self, matcher=None): |
|
431 | 431 | for a, b, size in super(encodedstore, self).datafiles(): |
|
432 | 432 | try: |
|
433 | 433 | a = decodefilename(a) |
|
434 | 434 | except KeyError: |
|
435 | 435 | a = None |
|
436 | 436 | if a is not None and not _matchtrackedpath(a, matcher): |
|
437 | 437 | continue |
|
438 | 438 | yield a, b, size |
|
439 | 439 | |
|
440 | 440 | def join(self, f): |
|
441 | 441 | return self.path + '/' + encodefilename(f) |
|
442 | 442 | |
|
443 | 443 | def copylist(self): |
|
444 | 444 | return (['requires', '00changelog.i'] + |
|
445 | 445 | ['store/' + f for f in _data.split()]) |
|
446 | 446 | |
|
447 | 447 | class fncache(object): |
|
448 | 448 | # the filename used to be partially encoded |
|
449 | 449 | # hence the encodedir/decodedir dance |
|
450 | 450 | def __init__(self, vfs): |
|
451 | 451 | self.vfs = vfs |
|
452 | 452 | self.entries = None |
|
453 | 453 | self._dirty = False |
|
454 | 454 | # set of new additions to fncache |
|
455 | 455 | self.addls = set() |
|
456 | 456 | |
|
457 | 457 | def _load(self): |
|
458 | 458 | '''fill the entries from the fncache file''' |
|
459 | 459 | self._dirty = False |
|
460 | 460 | try: |
|
461 | 461 | fp = self.vfs('fncache', mode='rb') |
|
462 | 462 | except IOError: |
|
463 | 463 | # skip nonexistent file |
|
464 | 464 | self.entries = set() |
|
465 | 465 | return |
|
466 | 466 | self.entries = set(decodedir(fp.read()).splitlines()) |
|
467 | 467 | if '' in self.entries: |
|
468 | 468 | fp.seek(0) |
|
469 | 469 | for n, line in enumerate(util.iterfile(fp)): |
|
470 | 470 | if not line.rstrip('\n'): |
|
471 | 471 | t = _('invalid entry in fncache, line %d') % (n + 1) |
|
472 | 472 | raise error.Abort(t) |
|
473 | 473 | fp.close() |
|
474 | 474 | |
|
475 | 475 | def write(self, tr): |
|
476 | 476 | if self._dirty: |
|
477 | 477 | assert self.entries is not None |
|
478 | 478 | self.entries = self.entries | self.addls |
|
479 | 479 | self.addls = set() |
|
480 | 480 | tr.addbackup('fncache') |
|
481 | 481 | fp = self.vfs('fncache', mode='wb', atomictemp=True) |
|
482 | 482 | if self.entries: |
|
483 | 483 | fp.write(encodedir('\n'.join(self.entries) + '\n')) |
|
484 | 484 | fp.close() |
|
485 | 485 | self._dirty = False |
|
486 | 486 | if self.addls: |
|
487 | 487 | # if we have just new entries, let's append them to the fncache |
|
488 | 488 | tr.addbackup('fncache') |
|
489 | 489 | fp = self.vfs('fncache', mode='ab', atomictemp=True) |
|
490 | 490 | if self.addls: |
|
491 | 491 | fp.write(encodedir('\n'.join(self.addls) + '\n')) |
|
492 | 492 | fp.close() |
|
493 | 493 | self.entries = None |
|
494 | 494 | self.addls = set() |
|
495 | 495 | |
|
496 | 496 | def add(self, fn): |
|
497 | 497 | if self.entries is None: |
|
498 | 498 | self._load() |
|
499 | 499 | if fn not in self.entries: |
|
500 | 500 | self.addls.add(fn) |
|
501 | 501 | |
|
502 | 502 | def remove(self, fn): |
|
503 | 503 | if self.entries is None: |
|
504 | 504 | self._load() |
|
505 | 505 | if fn in self.addls: |
|
506 | 506 | self.addls.remove(fn) |
|
507 | 507 | return |
|
508 | 508 | try: |
|
509 | 509 | self.entries.remove(fn) |
|
510 | 510 | self._dirty = True |
|
511 | 511 | except KeyError: |
|
512 | 512 | pass |
|
513 | 513 | |
|
514 | 514 | def __contains__(self, fn): |
|
515 | 515 | if fn in self.addls: |
|
516 | 516 | return True |
|
517 | 517 | if self.entries is None: |
|
518 | 518 | self._load() |
|
519 | 519 | return fn in self.entries |
|
520 | 520 | |
|
521 | 521 | def __iter__(self): |
|
522 | 522 | if self.entries is None: |
|
523 | 523 | self._load() |
|
524 | 524 | return iter(self.entries | self.addls) |
|
525 | 525 | |
|
526 |
class _fncachevfs(vfsmod. |
|
|
526 | class _fncachevfs(vfsmod.proxyvfs): | |
|
527 | 527 | def __init__(self, vfs, fnc, encode): |
|
528 | 528 | vfsmod.proxyvfs.__init__(self, vfs) |
|
529 | 529 | self.fncache = fnc |
|
530 | 530 | self.encode = encode |
|
531 | 531 | |
|
532 | 532 | def __call__(self, path, mode='r', *args, **kw): |
|
533 | 533 | encoded = self.encode(path) |
|
534 | 534 | if mode not in ('r', 'rb') and (path.startswith('data/') or |
|
535 | 535 | path.startswith('meta/')): |
|
536 | 536 | # do not trigger a fncache load when adding a file that already is |
|
537 | 537 | # known to exist. |
|
538 | 538 | notload = self.fncache.entries is None and self.vfs.exists(encoded) |
|
539 | 539 | if notload and 'a' in mode and not self.vfs.stat(encoded).st_size: |
|
540 | 540 | # when appending to an existing file, if the file has size zero, |
|
541 | 541 | # it should be considered as missing. Such zero-size files are |
|
542 | 542 | # the result of truncation when a transaction is aborted. |
|
543 | 543 | notload = False |
|
544 | 544 | if not notload: |
|
545 | 545 | self.fncache.add(path) |
|
546 | 546 | return self.vfs(encoded, mode, *args, **kw) |
|
547 | 547 | |
|
548 | 548 | def join(self, path): |
|
549 | 549 | if path: |
|
550 | 550 | return self.vfs.join(self.encode(path)) |
|
551 | 551 | else: |
|
552 | 552 | return self.vfs.join(path) |
|
553 | 553 | |
|
554 | 554 | class fncachestore(basicstore): |
|
555 | 555 | def __init__(self, path, vfstype, dotencode): |
|
556 | 556 | if dotencode: |
|
557 | 557 | encode = _pathencode |
|
558 | 558 | else: |
|
559 | 559 | encode = _plainhybridencode |
|
560 | 560 | self.encode = encode |
|
561 | 561 | vfs = vfstype(path + '/store') |
|
562 | 562 | self.path = vfs.base |
|
563 | 563 | self.pathsep = self.path + '/' |
|
564 | 564 | self.createmode = _calcmode(vfs) |
|
565 | 565 | vfs.createmode = self.createmode |
|
566 | 566 | self.rawvfs = vfs |
|
567 | 567 | fnc = fncache(vfs) |
|
568 | 568 | self.fncache = fnc |
|
569 | 569 | self.vfs = _fncachevfs(vfs, fnc, encode) |
|
570 | 570 | self.opener = self.vfs |
|
571 | 571 | |
|
572 | 572 | def join(self, f): |
|
573 | 573 | return self.pathsep + self.encode(f) |
|
574 | 574 | |
|
575 | 575 | def getsize(self, path): |
|
576 | 576 | return self.rawvfs.stat(path).st_size |
|
577 | 577 | |
|
578 | 578 | def datafiles(self, matcher=None): |
|
579 | 579 | for f in sorted(self.fncache): |
|
580 | 580 | if not _matchtrackedpath(f, matcher): |
|
581 | 581 | continue |
|
582 | 582 | ef = self.encode(f) |
|
583 | 583 | try: |
|
584 | 584 | yield f, ef, self.getsize(ef) |
|
585 | 585 | except OSError as err: |
|
586 | 586 | if err.errno != errno.ENOENT: |
|
587 | 587 | raise |
|
588 | 588 | |
|
589 | 589 | def copylist(self): |
|
590 | 590 | d = ('narrowspec data meta dh fncache phaseroots obsstore' |
|
591 | 591 | ' 00manifest.d 00manifest.i 00changelog.d 00changelog.i') |
|
592 | 592 | return (['requires', '00changelog.i'] + |
|
593 | 593 | ['store/' + f for f in d.split()]) |
|
594 | 594 | |
|
595 | 595 | def write(self, tr): |
|
596 | 596 | self.fncache.write(tr) |
|
597 | 597 | |
|
598 | 598 | def invalidatecaches(self): |
|
599 | 599 | self.fncache.entries = None |
|
600 | 600 | self.fncache.addls = set() |
|
601 | 601 | |
|
602 | 602 | def markremoved(self, fn): |
|
603 | 603 | self.fncache.remove(fn) |
|
604 | 604 | |
|
605 | 605 | def _exists(self, f): |
|
606 | 606 | ef = self.encode(f) |
|
607 | 607 | try: |
|
608 | 608 | self.getsize(ef) |
|
609 | 609 | return True |
|
610 | 610 | except OSError as err: |
|
611 | 611 | if err.errno != errno.ENOENT: |
|
612 | 612 | raise |
|
613 | 613 | # nonexistent entry |
|
614 | 614 | return False |
|
615 | 615 | |
|
616 | 616 | def __contains__(self, path): |
|
617 | 617 | '''Checks if the store contains path''' |
|
618 | 618 | path = "/".join(("data", path)) |
|
619 | 619 | # check for files (exact match) |
|
620 | 620 | e = path + '.i' |
|
621 | 621 | if e in self.fncache and self._exists(e): |
|
622 | 622 | return True |
|
623 | 623 | # now check for directories (prefix match) |
|
624 | 624 | if not path.endswith('/'): |
|
625 | 625 | path += '/' |
|
626 | 626 | for e in self.fncache: |
|
627 | 627 | if e.startswith(path) and self._exists(e): |
|
628 | 628 | return True |
|
629 | 629 | return False |
@@ -1,671 +1,671 b'' | |||
|
1 | 1 | # vfs.py - Mercurial 'vfs' classes |
|
2 | 2 | # |
|
3 | 3 | # Copyright Matt Mackall <mpm@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | from __future__ import absolute_import |
|
8 | 8 | |
|
9 | 9 | import contextlib |
|
10 | 10 | import errno |
|
11 | 11 | import os |
|
12 | 12 | import shutil |
|
13 | 13 | import stat |
|
14 | 14 | import threading |
|
15 | 15 | |
|
16 | 16 | from .i18n import _ |
|
17 | 17 | from . import ( |
|
18 | 18 | encoding, |
|
19 | 19 | error, |
|
20 | 20 | pathutil, |
|
21 | 21 | pycompat, |
|
22 | 22 | util, |
|
23 | 23 | ) |
|
24 | 24 | |
|
25 | 25 | def _avoidambig(path, oldstat): |
|
26 | 26 | """Avoid file stat ambiguity forcibly |
|
27 | 27 | |
|
28 | 28 | This function causes copying ``path`` file, if it is owned by |
|
29 | 29 | another (see issue5418 and issue5584 for detail). |
|
30 | 30 | """ |
|
31 | 31 | def checkandavoid(): |
|
32 | 32 | newstat = util.filestat.frompath(path) |
|
33 | 33 | # return whether file stat ambiguity is (already) avoided |
|
34 | 34 | return (not newstat.isambig(oldstat) or |
|
35 | 35 | newstat.avoidambig(path, oldstat)) |
|
36 | 36 | if not checkandavoid(): |
|
37 | 37 | # simply copy to change owner of path to get privilege to |
|
38 | 38 | # advance mtime (see issue5418) |
|
39 | 39 | util.rename(util.mktempcopy(path), path) |
|
40 | 40 | checkandavoid() |
|
41 | 41 | |
|
42 | 42 | class abstractvfs(object): |
|
43 | 43 | """Abstract base class; cannot be instantiated""" |
|
44 | 44 | |
|
45 | 45 | def __init__(self, *args, **kwargs): |
|
46 | 46 | '''Prevent instantiation; don't call this from subclasses.''' |
|
47 | 47 | raise NotImplementedError('attempted instantiating ' + str(type(self))) |
|
48 | 48 | |
|
49 | 49 | def _auditpath(self, path, mode): |
|
50 | 50 | pass |
|
51 | 51 | |
|
52 | 52 | def tryread(self, path): |
|
53 | 53 | '''gracefully return an empty string for missing files''' |
|
54 | 54 | try: |
|
55 | 55 | return self.read(path) |
|
56 | 56 | except IOError as inst: |
|
57 | 57 | if inst.errno != errno.ENOENT: |
|
58 | 58 | raise |
|
59 | 59 | return "" |
|
60 | 60 | |
|
61 | 61 | def tryreadlines(self, path, mode='rb'): |
|
62 | 62 | '''gracefully return an empty array for missing files''' |
|
63 | 63 | try: |
|
64 | 64 | return self.readlines(path, mode=mode) |
|
65 | 65 | except IOError as inst: |
|
66 | 66 | if inst.errno != errno.ENOENT: |
|
67 | 67 | raise |
|
68 | 68 | return [] |
|
69 | 69 | |
|
70 | 70 | @util.propertycache |
|
71 | 71 | def open(self): |
|
72 | 72 | '''Open ``path`` file, which is relative to vfs root. |
|
73 | 73 | |
|
74 | 74 | Newly created directories are marked as "not to be indexed by |
|
75 | 75 | the content indexing service", if ``notindexed`` is specified |
|
76 | 76 | for "write" mode access. |
|
77 | 77 | ''' |
|
78 | 78 | return self.__call__ |
|
79 | 79 | |
|
80 | 80 | def read(self, path): |
|
81 | 81 | with self(path, 'rb') as fp: |
|
82 | 82 | return fp.read() |
|
83 | 83 | |
|
84 | 84 | def readlines(self, path, mode='rb'): |
|
85 | 85 | with self(path, mode=mode) as fp: |
|
86 | 86 | return fp.readlines() |
|
87 | 87 | |
|
88 | 88 | def write(self, path, data, backgroundclose=False, **kwargs): |
|
89 | 89 | with self(path, 'wb', backgroundclose=backgroundclose, **kwargs) as fp: |
|
90 | 90 | return fp.write(data) |
|
91 | 91 | |
|
92 | 92 | def writelines(self, path, data, mode='wb', notindexed=False): |
|
93 | 93 | with self(path, mode=mode, notindexed=notindexed) as fp: |
|
94 | 94 | return fp.writelines(data) |
|
95 | 95 | |
|
96 | 96 | def append(self, path, data): |
|
97 | 97 | with self(path, 'ab') as fp: |
|
98 | 98 | return fp.write(data) |
|
99 | 99 | |
|
100 | 100 | def basename(self, path): |
|
101 | 101 | """return base element of a path (as os.path.basename would do) |
|
102 | 102 | |
|
103 | 103 | This exists to allow handling of strange encoding if needed.""" |
|
104 | 104 | return os.path.basename(path) |
|
105 | 105 | |
|
106 | 106 | def chmod(self, path, mode): |
|
107 | 107 | return os.chmod(self.join(path), mode) |
|
108 | 108 | |
|
109 | 109 | def dirname(self, path): |
|
110 | 110 | """return dirname element of a path (as os.path.dirname would do) |
|
111 | 111 | |
|
112 | 112 | This exists to allow handling of strange encoding if needed.""" |
|
113 | 113 | return os.path.dirname(path) |
|
114 | 114 | |
|
115 | 115 | def exists(self, path=None): |
|
116 | 116 | return os.path.exists(self.join(path)) |
|
117 | 117 | |
|
118 | 118 | def fstat(self, fp): |
|
119 | 119 | return util.fstat(fp) |
|
120 | 120 | |
|
121 | 121 | def isdir(self, path=None): |
|
122 | 122 | return os.path.isdir(self.join(path)) |
|
123 | 123 | |
|
124 | 124 | def isfile(self, path=None): |
|
125 | 125 | return os.path.isfile(self.join(path)) |
|
126 | 126 | |
|
127 | 127 | def islink(self, path=None): |
|
128 | 128 | return os.path.islink(self.join(path)) |
|
129 | 129 | |
|
130 | 130 | def isfileorlink(self, path=None): |
|
131 | 131 | '''return whether path is a regular file or a symlink |
|
132 | 132 | |
|
133 | 133 | Unlike isfile, this doesn't follow symlinks.''' |
|
134 | 134 | try: |
|
135 | 135 | st = self.lstat(path) |
|
136 | 136 | except OSError: |
|
137 | 137 | return False |
|
138 | 138 | mode = st.st_mode |
|
139 | 139 | return stat.S_ISREG(mode) or stat.S_ISLNK(mode) |
|
140 | 140 | |
|
141 | 141 | def reljoin(self, *paths): |
|
142 | 142 | """join various elements of a path together (as os.path.join would do) |
|
143 | 143 | |
|
144 | 144 | The vfs base is not injected so that path stay relative. This exists |
|
145 | 145 | to allow handling of strange encoding if needed.""" |
|
146 | 146 | return os.path.join(*paths) |
|
147 | 147 | |
|
148 | 148 | def split(self, path): |
|
149 | 149 | """split top-most element of a path (as os.path.split would do) |
|
150 | 150 | |
|
151 | 151 | This exists to allow handling of strange encoding if needed.""" |
|
152 | 152 | return os.path.split(path) |
|
153 | 153 | |
|
154 | 154 | def lexists(self, path=None): |
|
155 | 155 | return os.path.lexists(self.join(path)) |
|
156 | 156 | |
|
157 | 157 | def lstat(self, path=None): |
|
158 | 158 | return os.lstat(self.join(path)) |
|
159 | 159 | |
|
160 | 160 | def listdir(self, path=None): |
|
161 | 161 | return os.listdir(self.join(path)) |
|
162 | 162 | |
|
163 | 163 | def makedir(self, path=None, notindexed=True): |
|
164 | 164 | return util.makedir(self.join(path), notindexed) |
|
165 | 165 | |
|
166 | 166 | def makedirs(self, path=None, mode=None): |
|
167 | 167 | return util.makedirs(self.join(path), mode) |
|
168 | 168 | |
|
169 | 169 | def makelock(self, info, path): |
|
170 | 170 | return util.makelock(info, self.join(path)) |
|
171 | 171 | |
|
172 | 172 | def mkdir(self, path=None): |
|
173 | 173 | return os.mkdir(self.join(path)) |
|
174 | 174 | |
|
175 | 175 | def mkstemp(self, suffix='', prefix='tmp', dir=None): |
|
176 | 176 | fd, name = pycompat.mkstemp(suffix=suffix, prefix=prefix, |
|
177 | 177 | dir=self.join(dir)) |
|
178 | 178 | dname, fname = util.split(name) |
|
179 | 179 | if dir: |
|
180 | 180 | return fd, os.path.join(dir, fname) |
|
181 | 181 | else: |
|
182 | 182 | return fd, fname |
|
183 | 183 | |
|
184 | 184 | def readdir(self, path=None, stat=None, skip=None): |
|
185 | 185 | return util.listdir(self.join(path), stat, skip) |
|
186 | 186 | |
|
187 | 187 | def readlock(self, path): |
|
188 | 188 | return util.readlock(self.join(path)) |
|
189 | 189 | |
|
190 | 190 | def rename(self, src, dst, checkambig=False): |
|
191 | 191 | """Rename from src to dst |
|
192 | 192 | |
|
193 | 193 | checkambig argument is used with util.filestat, and is useful |
|
194 | 194 | only if destination file is guarded by any lock |
|
195 | 195 | (e.g. repo.lock or repo.wlock). |
|
196 | 196 | |
|
197 | 197 | To avoid file stat ambiguity forcibly, checkambig=True involves |
|
198 | 198 | copying ``src`` file, if it is owned by another. Therefore, use |
|
199 | 199 | checkambig=True only in limited cases (see also issue5418 and |
|
200 | 200 | issue5584 for detail). |
|
201 | 201 | """ |
|
202 | 202 | self._auditpath(dst, 'w') |
|
203 | 203 | srcpath = self.join(src) |
|
204 | 204 | dstpath = self.join(dst) |
|
205 | 205 | oldstat = checkambig and util.filestat.frompath(dstpath) |
|
206 | 206 | if oldstat and oldstat.stat: |
|
207 | 207 | ret = util.rename(srcpath, dstpath) |
|
208 | 208 | _avoidambig(dstpath, oldstat) |
|
209 | 209 | return ret |
|
210 | 210 | return util.rename(srcpath, dstpath) |
|
211 | 211 | |
|
212 | 212 | def readlink(self, path): |
|
213 | 213 | return util.readlink(self.join(path)) |
|
214 | 214 | |
|
215 | 215 | def removedirs(self, path=None): |
|
216 | 216 | """Remove a leaf directory and all empty intermediate ones |
|
217 | 217 | """ |
|
218 | 218 | return util.removedirs(self.join(path)) |
|
219 | 219 | |
|
220 | 220 | def rmdir(self, path=None): |
|
221 | 221 | """Remove an empty directory.""" |
|
222 | 222 | return os.rmdir(self.join(path)) |
|
223 | 223 | |
|
224 | 224 | def rmtree(self, path=None, ignore_errors=False, forcibly=False): |
|
225 | 225 | """Remove a directory tree recursively |
|
226 | 226 | |
|
227 | 227 | If ``forcibly``, this tries to remove READ-ONLY files, too. |
|
228 | 228 | """ |
|
229 | 229 | if forcibly: |
|
230 | 230 | def onerror(function, path, excinfo): |
|
231 | 231 | if function is not os.remove: |
|
232 | 232 | raise |
|
233 | 233 | # read-only files cannot be unlinked under Windows |
|
234 | 234 | s = os.stat(path) |
|
235 | 235 | if (s.st_mode & stat.S_IWRITE) != 0: |
|
236 | 236 | raise |
|
237 | 237 | os.chmod(path, stat.S_IMODE(s.st_mode) | stat.S_IWRITE) |
|
238 | 238 | os.remove(path) |
|
239 | 239 | else: |
|
240 | 240 | onerror = None |
|
241 | 241 | return shutil.rmtree(self.join(path), |
|
242 | 242 | ignore_errors=ignore_errors, onerror=onerror) |
|
243 | 243 | |
|
244 | 244 | def setflags(self, path, l, x): |
|
245 | 245 | return util.setflags(self.join(path), l, x) |
|
246 | 246 | |
|
247 | 247 | def stat(self, path=None): |
|
248 | 248 | return os.stat(self.join(path)) |
|
249 | 249 | |
|
250 | 250 | def unlink(self, path=None): |
|
251 | 251 | return util.unlink(self.join(path)) |
|
252 | 252 | |
|
253 | 253 | def tryunlink(self, path=None): |
|
254 | 254 | """Attempt to remove a file, ignoring missing file errors.""" |
|
255 | 255 | util.tryunlink(self.join(path)) |
|
256 | 256 | |
|
257 | 257 | def unlinkpath(self, path=None, ignoremissing=False, rmdir=True): |
|
258 | 258 | return util.unlinkpath(self.join(path), ignoremissing=ignoremissing, |
|
259 | 259 | rmdir=rmdir) |
|
260 | 260 | |
|
261 | 261 | def utime(self, path=None, t=None): |
|
262 | 262 | return os.utime(self.join(path), t) |
|
263 | 263 | |
|
264 | 264 | def walk(self, path=None, onerror=None): |
|
265 | 265 | """Yield (dirpath, dirs, files) tuple for each directories under path |
|
266 | 266 | |
|
267 | 267 | ``dirpath`` is relative one from the root of this vfs. This |
|
268 | 268 | uses ``os.sep`` as path separator, even you specify POSIX |
|
269 | 269 | style ``path``. |
|
270 | 270 | |
|
271 | 271 | "The root of this vfs" is represented as empty ``dirpath``. |
|
272 | 272 | """ |
|
273 | 273 | root = os.path.normpath(self.join(None)) |
|
274 | 274 | # when dirpath == root, dirpath[prefixlen:] becomes empty |
|
275 | 275 | # because len(dirpath) < prefixlen. |
|
276 | 276 | prefixlen = len(pathutil.normasprefix(root)) |
|
277 | 277 | for dirpath, dirs, files in os.walk(self.join(path), onerror=onerror): |
|
278 | 278 | yield (dirpath[prefixlen:], dirs, files) |
|
279 | 279 | |
|
280 | 280 | @contextlib.contextmanager |
|
281 | 281 | def backgroundclosing(self, ui, expectedcount=-1): |
|
282 | 282 | """Allow files to be closed asynchronously. |
|
283 | 283 | |
|
284 | 284 | When this context manager is active, ``backgroundclose`` can be passed |
|
285 | 285 | to ``__call__``/``open`` to result in the file possibly being closed |
|
286 | 286 | asynchronously, on a background thread. |
|
287 | 287 | """ |
|
288 | 288 | # Sharing backgroundfilecloser between threads is complex and using |
|
289 | 289 | # multiple instances puts us at risk of running out of file descriptors |
|
290 | 290 | # only allow to use backgroundfilecloser when in main thread. |
|
291 | 291 | if not isinstance(threading.currentThread(), threading._MainThread): |
|
292 | 292 | yield |
|
293 | 293 | return |
|
294 | 294 | vfs = getattr(self, 'vfs', self) |
|
295 | 295 | if getattr(vfs, '_backgroundfilecloser', None): |
|
296 | 296 | raise error.Abort( |
|
297 | 297 | _('can only have 1 active background file closer')) |
|
298 | 298 | |
|
299 | 299 | with backgroundfilecloser(ui, expectedcount=expectedcount) as bfc: |
|
300 | 300 | try: |
|
301 | 301 | vfs._backgroundfilecloser = bfc |
|
302 | 302 | yield bfc |
|
303 | 303 | finally: |
|
304 | 304 | vfs._backgroundfilecloser = None |
|
305 | 305 | |
|
306 | 306 | class vfs(abstractvfs): |
|
307 | 307 | '''Operate files relative to a base directory |
|
308 | 308 | |
|
309 | 309 | This class is used to hide the details of COW semantics and |
|
310 | 310 | remote file access from higher level code. |
|
311 | 311 | |
|
312 | 312 | 'cacheaudited' should be enabled only if (a) vfs object is short-lived, or |
|
313 | 313 | (b) the base directory is managed by hg and considered sort-of append-only. |
|
314 | 314 | See pathutil.pathauditor() for details. |
|
315 | 315 | ''' |
|
316 | 316 | def __init__(self, base, audit=True, cacheaudited=False, expandpath=False, |
|
317 | 317 | realpath=False): |
|
318 | 318 | if expandpath: |
|
319 | 319 | base = util.expandpath(base) |
|
320 | 320 | if realpath: |
|
321 | 321 | base = os.path.realpath(base) |
|
322 | 322 | self.base = base |
|
323 | 323 | self._audit = audit |
|
324 | 324 | if audit: |
|
325 | 325 | self.audit = pathutil.pathauditor(self.base, cached=cacheaudited) |
|
326 | 326 | else: |
|
327 | 327 | self.audit = (lambda path, mode=None: True) |
|
328 | 328 | self.createmode = None |
|
329 | 329 | self._trustnlink = None |
|
330 | 330 | |
|
331 | 331 | @util.propertycache |
|
332 | 332 | def _cansymlink(self): |
|
333 | 333 | return util.checklink(self.base) |
|
334 | 334 | |
|
335 | 335 | @util.propertycache |
|
336 | 336 | def _chmod(self): |
|
337 | 337 | return util.checkexec(self.base) |
|
338 | 338 | |
|
339 | 339 | def _fixfilemode(self, name): |
|
340 | 340 | if self.createmode is None or not self._chmod: |
|
341 | 341 | return |
|
342 | 342 | os.chmod(name, self.createmode & 0o666) |
|
343 | 343 | |
|
344 | 344 | def _auditpath(self, path, mode): |
|
345 | 345 | if self._audit: |
|
346 | 346 | if os.path.isabs(path) and path.startswith(self.base): |
|
347 | 347 | path = os.path.relpath(path, self.base) |
|
348 | 348 | r = util.checkosfilename(path) |
|
349 | 349 | if r: |
|
350 | 350 | raise error.Abort("%s: %r" % (r, path)) |
|
351 | 351 | self.audit(path, mode=mode) |
|
352 | 352 | |
|
353 | 353 | def __call__(self, path, mode="r", atomictemp=False, notindexed=False, |
|
354 | 354 | backgroundclose=False, checkambig=False, auditpath=True, |
|
355 | 355 | makeparentdirs=True): |
|
356 | 356 | '''Open ``path`` file, which is relative to vfs root. |
|
357 | 357 | |
|
358 | 358 | By default, parent directories are created as needed. Newly created |
|
359 | 359 | directories are marked as "not to be indexed by the content indexing |
|
360 | 360 | service", if ``notindexed`` is specified for "write" mode access. |
|
361 | 361 | Set ``makeparentdirs=False`` to not create directories implicitly. |
|
362 | 362 | |
|
363 | 363 | If ``backgroundclose`` is passed, the file may be closed asynchronously. |
|
364 | 364 | It can only be used if the ``self.backgroundclosing()`` context manager |
|
365 | 365 | is active. This should only be specified if the following criteria hold: |
|
366 | 366 | |
|
367 | 367 | 1. There is a potential for writing thousands of files. Unless you |
|
368 | 368 | are writing thousands of files, the performance benefits of |
|
369 | 369 | asynchronously closing files is not realized. |
|
370 | 370 | 2. Files are opened exactly once for the ``backgroundclosing`` |
|
371 | 371 | active duration and are therefore free of race conditions between |
|
372 | 372 | closing a file on a background thread and reopening it. (If the |
|
373 | 373 | file were opened multiple times, there could be unflushed data |
|
374 | 374 | because the original file handle hasn't been flushed/closed yet.) |
|
375 | 375 | |
|
376 | 376 | ``checkambig`` argument is passed to atomictemplfile (valid |
|
377 | 377 | only for writing), and is useful only if target file is |
|
378 | 378 | guarded by any lock (e.g. repo.lock or repo.wlock). |
|
379 | 379 | |
|
380 | 380 | To avoid file stat ambiguity forcibly, checkambig=True involves |
|
381 | 381 | copying ``path`` file opened in "append" mode (e.g. for |
|
382 | 382 | truncation), if it is owned by another. Therefore, use |
|
383 | 383 | combination of append mode and checkambig=True only in limited |
|
384 | 384 | cases (see also issue5418 and issue5584 for detail). |
|
385 | 385 | ''' |
|
386 | 386 | if auditpath: |
|
387 | 387 | self._auditpath(path, mode) |
|
388 | 388 | f = self.join(path) |
|
389 | 389 | |
|
390 | 390 | if "b" not in mode: |
|
391 | 391 | mode += "b" # for that other OS |
|
392 | 392 | |
|
393 | 393 | nlink = -1 |
|
394 | 394 | if mode not in ('r', 'rb'): |
|
395 | 395 | dirname, basename = util.split(f) |
|
396 | 396 | # If basename is empty, then the path is malformed because it points |
|
397 | 397 | # to a directory. Let the posixfile() call below raise IOError. |
|
398 | 398 | if basename: |
|
399 | 399 | if atomictemp: |
|
400 | 400 | if makeparentdirs: |
|
401 | 401 | util.makedirs(dirname, self.createmode, notindexed) |
|
402 | 402 | return util.atomictempfile(f, mode, self.createmode, |
|
403 | 403 | checkambig=checkambig) |
|
404 | 404 | try: |
|
405 | 405 | if 'w' in mode: |
|
406 | 406 | util.unlink(f) |
|
407 | 407 | nlink = 0 |
|
408 | 408 | else: |
|
409 | 409 | # nlinks() may behave differently for files on Windows |
|
410 | 410 | # shares if the file is open. |
|
411 | 411 | with util.posixfile(f): |
|
412 | 412 | nlink = util.nlinks(f) |
|
413 | 413 | if nlink < 1: |
|
414 | 414 | nlink = 2 # force mktempcopy (issue1922) |
|
415 | 415 | except (OSError, IOError) as e: |
|
416 | 416 | if e.errno != errno.ENOENT: |
|
417 | 417 | raise |
|
418 | 418 | nlink = 0 |
|
419 | 419 | if makeparentdirs: |
|
420 | 420 | util.makedirs(dirname, self.createmode, notindexed) |
|
421 | 421 | if nlink > 0: |
|
422 | 422 | if self._trustnlink is None: |
|
423 | 423 | self._trustnlink = nlink > 1 or util.checknlink(f) |
|
424 | 424 | if nlink > 1 or not self._trustnlink: |
|
425 | 425 | util.rename(util.mktempcopy(f), f) |
|
426 | 426 | fp = util.posixfile(f, mode) |
|
427 | 427 | if nlink == 0: |
|
428 | 428 | self._fixfilemode(f) |
|
429 | 429 | |
|
430 | 430 | if checkambig: |
|
431 | 431 | if mode in ('r', 'rb'): |
|
432 | 432 | raise error.Abort(_('implementation error: mode %s is not' |
|
433 | 433 | ' valid for checkambig=True') % mode) |
|
434 | 434 | fp = checkambigatclosing(fp) |
|
435 | 435 | |
|
436 | 436 | if (backgroundclose and |
|
437 | 437 | isinstance(threading.currentThread(), threading._MainThread)): |
|
438 | 438 | if not self._backgroundfilecloser: |
|
439 | 439 | raise error.Abort(_('backgroundclose can only be used when a ' |
|
440 | 440 | 'backgroundclosing context manager is active') |
|
441 | 441 | ) |
|
442 | 442 | |
|
443 | 443 | fp = delayclosedfile(fp, self._backgroundfilecloser) |
|
444 | 444 | |
|
445 | 445 | return fp |
|
446 | 446 | |
|
447 | 447 | def symlink(self, src, dst): |
|
448 | 448 | self.audit(dst) |
|
449 | 449 | linkname = self.join(dst) |
|
450 | 450 | util.tryunlink(linkname) |
|
451 | 451 | |
|
452 | 452 | util.makedirs(os.path.dirname(linkname), self.createmode) |
|
453 | 453 | |
|
454 | 454 | if self._cansymlink: |
|
455 | 455 | try: |
|
456 | 456 | os.symlink(src, linkname) |
|
457 | 457 | except OSError as err: |
|
458 | 458 | raise OSError(err.errno, _('could not symlink to %r: %s') % |
|
459 | 459 | (src, encoding.strtolocal(err.strerror)), |
|
460 | 460 | linkname) |
|
461 | 461 | else: |
|
462 | 462 | self.write(dst, src) |
|
463 | 463 | |
|
464 | 464 | def join(self, path, *insidef): |
|
465 | 465 | if path: |
|
466 | 466 | return os.path.join(self.base, path, *insidef) |
|
467 | 467 | else: |
|
468 | 468 | return self.base |
|
469 | 469 | |
|
470 | 470 | opener = vfs |
|
471 | 471 | |
|
472 |
class proxyvfs( |
|
|
472 | class proxyvfs(abstractvfs): | |
|
473 | 473 | def __init__(self, vfs): |
|
474 | 474 | self.vfs = vfs |
|
475 | 475 | |
|
476 | 476 | @property |
|
477 | 477 | def options(self): |
|
478 | 478 | return self.vfs.options |
|
479 | 479 | |
|
480 | 480 | @options.setter |
|
481 | 481 | def options(self, value): |
|
482 | 482 | self.vfs.options = value |
|
483 | 483 | |
|
484 |
class filtervfs( |
|
|
484 | class filtervfs(proxyvfs, abstractvfs): | |
|
485 | 485 | '''Wrapper vfs for filtering filenames with a function.''' |
|
486 | 486 | |
|
487 | 487 | def __init__(self, vfs, filter): |
|
488 | 488 | proxyvfs.__init__(self, vfs) |
|
489 | 489 | self._filter = filter |
|
490 | 490 | |
|
491 | 491 | def __call__(self, path, *args, **kwargs): |
|
492 | 492 | return self.vfs(self._filter(path), *args, **kwargs) |
|
493 | 493 | |
|
494 | 494 | def join(self, path, *insidef): |
|
495 | 495 | if path: |
|
496 | 496 | return self.vfs.join(self._filter(self.vfs.reljoin(path, *insidef))) |
|
497 | 497 | else: |
|
498 | 498 | return self.vfs.join(path) |
|
499 | 499 | |
|
500 | 500 | filteropener = filtervfs |
|
501 | 501 | |
|
502 |
class readonlyvfs( |
|
|
502 | class readonlyvfs(proxyvfs): | |
|
503 | 503 | '''Wrapper vfs preventing any writing.''' |
|
504 | 504 | |
|
505 | 505 | def __init__(self, vfs): |
|
506 | 506 | proxyvfs.__init__(self, vfs) |
|
507 | 507 | |
|
508 | 508 | def __call__(self, path, mode='r', *args, **kw): |
|
509 | 509 | if mode not in ('r', 'rb'): |
|
510 | 510 | raise error.Abort(_('this vfs is read only')) |
|
511 | 511 | return self.vfs(path, mode, *args, **kw) |
|
512 | 512 | |
|
513 | 513 | def join(self, path, *insidef): |
|
514 | 514 | return self.vfs.join(path, *insidef) |
|
515 | 515 | |
|
516 | 516 | class closewrapbase(object): |
|
517 | 517 | """Base class of wrapper, which hooks closing |
|
518 | 518 | |
|
519 | 519 | Do not instantiate outside of the vfs layer. |
|
520 | 520 | """ |
|
521 | 521 | def __init__(self, fh): |
|
522 | 522 | object.__setattr__(self, r'_origfh', fh) |
|
523 | 523 | |
|
524 | 524 | def __getattr__(self, attr): |
|
525 | 525 | return getattr(self._origfh, attr) |
|
526 | 526 | |
|
527 | 527 | def __setattr__(self, attr, value): |
|
528 | 528 | return setattr(self._origfh, attr, value) |
|
529 | 529 | |
|
530 | 530 | def __delattr__(self, attr): |
|
531 | 531 | return delattr(self._origfh, attr) |
|
532 | 532 | |
|
533 | 533 | def __enter__(self): |
|
534 | 534 | self._origfh.__enter__() |
|
535 | 535 | return self |
|
536 | 536 | |
|
537 | 537 | def __exit__(self, exc_type, exc_value, exc_tb): |
|
538 | 538 | raise NotImplementedError('attempted instantiating ' + str(type(self))) |
|
539 | 539 | |
|
540 | 540 | def close(self): |
|
541 | 541 | raise NotImplementedError('attempted instantiating ' + str(type(self))) |
|
542 | 542 | |
|
543 | 543 | class delayclosedfile(closewrapbase): |
|
544 | 544 | """Proxy for a file object whose close is delayed. |
|
545 | 545 | |
|
546 | 546 | Do not instantiate outside of the vfs layer. |
|
547 | 547 | """ |
|
548 | 548 | def __init__(self, fh, closer): |
|
549 | 549 | super(delayclosedfile, self).__init__(fh) |
|
550 | 550 | object.__setattr__(self, r'_closer', closer) |
|
551 | 551 | |
|
552 | 552 | def __exit__(self, exc_type, exc_value, exc_tb): |
|
553 | 553 | self._closer.close(self._origfh) |
|
554 | 554 | |
|
555 | 555 | def close(self): |
|
556 | 556 | self._closer.close(self._origfh) |
|
557 | 557 | |
|
558 | 558 | class backgroundfilecloser(object): |
|
559 | 559 | """Coordinates background closing of file handles on multiple threads.""" |
|
560 | 560 | def __init__(self, ui, expectedcount=-1): |
|
561 | 561 | self._running = False |
|
562 | 562 | self._entered = False |
|
563 | 563 | self._threads = [] |
|
564 | 564 | self._threadexception = None |
|
565 | 565 | |
|
566 | 566 | # Only Windows/NTFS has slow file closing. So only enable by default |
|
567 | 567 | # on that platform. But allow to be enabled elsewhere for testing. |
|
568 | 568 | defaultenabled = pycompat.iswindows |
|
569 | 569 | enabled = ui.configbool('worker', 'backgroundclose', defaultenabled) |
|
570 | 570 | |
|
571 | 571 | if not enabled: |
|
572 | 572 | return |
|
573 | 573 | |
|
574 | 574 | # There is overhead to starting and stopping the background threads. |
|
575 | 575 | # Don't do background processing unless the file count is large enough |
|
576 | 576 | # to justify it. |
|
577 | 577 | minfilecount = ui.configint('worker', 'backgroundcloseminfilecount') |
|
578 | 578 | # FUTURE dynamically start background threads after minfilecount closes. |
|
579 | 579 | # (We don't currently have any callers that don't know their file count) |
|
580 | 580 | if expectedcount > 0 and expectedcount < minfilecount: |
|
581 | 581 | return |
|
582 | 582 | |
|
583 | 583 | maxqueue = ui.configint('worker', 'backgroundclosemaxqueue') |
|
584 | 584 | threadcount = ui.configint('worker', 'backgroundclosethreadcount') |
|
585 | 585 | |
|
586 | 586 | ui.debug('starting %d threads for background file closing\n' % |
|
587 | 587 | threadcount) |
|
588 | 588 | |
|
589 | 589 | self._queue = pycompat.queue.Queue(maxsize=maxqueue) |
|
590 | 590 | self._running = True |
|
591 | 591 | |
|
592 | 592 | for i in range(threadcount): |
|
593 | 593 | t = threading.Thread(target=self._worker, name='backgroundcloser') |
|
594 | 594 | self._threads.append(t) |
|
595 | 595 | t.start() |
|
596 | 596 | |
|
597 | 597 | def __enter__(self): |
|
598 | 598 | self._entered = True |
|
599 | 599 | return self |
|
600 | 600 | |
|
601 | 601 | def __exit__(self, exc_type, exc_value, exc_tb): |
|
602 | 602 | self._running = False |
|
603 | 603 | |
|
604 | 604 | # Wait for threads to finish closing so open files don't linger for |
|
605 | 605 | # longer than lifetime of context manager. |
|
606 | 606 | for t in self._threads: |
|
607 | 607 | t.join() |
|
608 | 608 | |
|
609 | 609 | def _worker(self): |
|
610 | 610 | """Main routine for worker thread.""" |
|
611 | 611 | while True: |
|
612 | 612 | try: |
|
613 | 613 | fh = self._queue.get(block=True, timeout=0.100) |
|
614 | 614 | # Need to catch or the thread will terminate and |
|
615 | 615 | # we could orphan file descriptors. |
|
616 | 616 | try: |
|
617 | 617 | fh.close() |
|
618 | 618 | except Exception as e: |
|
619 | 619 | # Stash so can re-raise from main thread later. |
|
620 | 620 | self._threadexception = e |
|
621 | 621 | except pycompat.queue.Empty: |
|
622 | 622 | if not self._running: |
|
623 | 623 | break |
|
624 | 624 | |
|
625 | 625 | def close(self, fh): |
|
626 | 626 | """Schedule a file for closing.""" |
|
627 | 627 | if not self._entered: |
|
628 | 628 | raise error.Abort(_('can only call close() when context manager ' |
|
629 | 629 | 'active')) |
|
630 | 630 | |
|
631 | 631 | # If a background thread encountered an exception, raise now so we fail |
|
632 | 632 | # fast. Otherwise we may potentially go on for minutes until the error |
|
633 | 633 | # is acted on. |
|
634 | 634 | if self._threadexception: |
|
635 | 635 | e = self._threadexception |
|
636 | 636 | self._threadexception = None |
|
637 | 637 | raise e |
|
638 | 638 | |
|
639 | 639 | # If we're not actively running, close synchronously. |
|
640 | 640 | if not self._running: |
|
641 | 641 | fh.close() |
|
642 | 642 | return |
|
643 | 643 | |
|
644 | 644 | self._queue.put(fh, block=True, timeout=None) |
|
645 | 645 | |
|
646 | 646 | class checkambigatclosing(closewrapbase): |
|
647 | 647 | """Proxy for a file object, to avoid ambiguity of file stat |
|
648 | 648 | |
|
649 | 649 | See also util.filestat for detail about "ambiguity of file stat". |
|
650 | 650 | |
|
651 | 651 | This proxy is useful only if the target file is guarded by any |
|
652 | 652 | lock (e.g. repo.lock or repo.wlock) |
|
653 | 653 | |
|
654 | 654 | Do not instantiate outside of the vfs layer. |
|
655 | 655 | """ |
|
656 | 656 | def __init__(self, fh): |
|
657 | 657 | super(checkambigatclosing, self).__init__(fh) |
|
658 | 658 | object.__setattr__(self, r'_oldstat', util.filestat.frompath(fh.name)) |
|
659 | 659 | |
|
660 | 660 | def _checkambig(self): |
|
661 | 661 | oldstat = self._oldstat |
|
662 | 662 | if oldstat.stat: |
|
663 | 663 | _avoidambig(self._origfh.name, oldstat) |
|
664 | 664 | |
|
665 | 665 | def __exit__(self, exc_type, exc_value, exc_tb): |
|
666 | 666 | self._origfh.__exit__(exc_type, exc_value, exc_tb) |
|
667 | 667 | self._checkambig() |
|
668 | 668 | |
|
669 | 669 | def close(self): |
|
670 | 670 | self._origfh.close() |
|
671 | 671 | self._checkambig() |
General Comments 0
You need to be logged in to leave comments.
Login now