##// END OF EJS Templates
py3: replace sys.platform with pycompat.sysplatform (part 2 of 2)
Pulkit Goyal -
r30642:e995f00a default
parent child Browse files
Show More
@@ -1,194 +1,194 b''
1 1 # win32mbcs.py -- MBCS filename support for Mercurial
2 2 #
3 3 # Copyright (c) 2008 Shun-ichi Goto <shunichi.goto@gmail.com>
4 4 #
5 5 # Version: 0.3
6 6 # Author: Shun-ichi Goto <shunichi.goto@gmail.com>
7 7 #
8 8 # This software may be used and distributed according to the terms of the
9 9 # GNU General Public License version 2 or any later version.
10 10 #
11 11
12 12 '''allow the use of MBCS paths with problematic encodings
13 13
14 14 Some MBCS encodings are not good for some path operations (i.e.
15 15 splitting path, case conversion, etc.) with its encoded bytes. We call
16 16 such a encoding (i.e. shift_jis and big5) as "problematic encoding".
17 17 This extension can be used to fix the issue with those encodings by
18 18 wrapping some functions to convert to Unicode string before path
19 19 operation.
20 20
21 21 This extension is useful for:
22 22
23 23 - Japanese Windows users using shift_jis encoding.
24 24 - Chinese Windows users using big5 encoding.
25 25 - All users who use a repository with one of problematic encodings on
26 26 case-insensitive file system.
27 27
28 28 This extension is not needed for:
29 29
30 30 - Any user who use only ASCII chars in path.
31 31 - Any user who do not use any of problematic encodings.
32 32
33 33 Note that there are some limitations on using this extension:
34 34
35 35 - You should use single encoding in one repository.
36 36 - If the repository path ends with 0x5c, .hg/hgrc cannot be read.
37 37 - win32mbcs is not compatible with fixutf8 extension.
38 38
39 39 By default, win32mbcs uses encoding.encoding decided by Mercurial.
40 40 You can specify the encoding by config option::
41 41
42 42 [win32mbcs]
43 43 encoding = sjis
44 44
45 45 It is useful for the users who want to commit with UTF-8 log message.
46 46 '''
47 47 from __future__ import absolute_import
48 48
49 49 import os
50 50 import sys
51 51
52 52 from mercurial.i18n import _
53 53 from mercurial import (
54 54 encoding,
55 55 error,
56 56 pycompat,
57 57 )
58 58
59 59 # Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for
60 60 # extensions which SHIP WITH MERCURIAL. Non-mainline extensions should
61 61 # be specifying the version(s) of Mercurial they are tested with, or
62 62 # leave the attribute unspecified.
63 63 testedwith = 'ships-with-hg-core'
64 64
65 65 _encoding = None # see extsetup
66 66
67 67 def decode(arg):
68 68 if isinstance(arg, str):
69 69 uarg = arg.decode(_encoding)
70 70 if arg == uarg.encode(_encoding):
71 71 return uarg
72 72 raise UnicodeError("Not local encoding")
73 73 elif isinstance(arg, tuple):
74 74 return tuple(map(decode, arg))
75 75 elif isinstance(arg, list):
76 76 return map(decode, arg)
77 77 elif isinstance(arg, dict):
78 78 for k, v in arg.items():
79 79 arg[k] = decode(v)
80 80 return arg
81 81
82 82 def encode(arg):
83 83 if isinstance(arg, unicode):
84 84 return arg.encode(_encoding)
85 85 elif isinstance(arg, tuple):
86 86 return tuple(map(encode, arg))
87 87 elif isinstance(arg, list):
88 88 return map(encode, arg)
89 89 elif isinstance(arg, dict):
90 90 for k, v in arg.items():
91 91 arg[k] = encode(v)
92 92 return arg
93 93
94 94 def appendsep(s):
95 95 # ensure the path ends with os.sep, appending it if necessary.
96 96 try:
97 97 us = decode(s)
98 98 except UnicodeError:
99 99 us = s
100 100 if us and us[-1] not in ':/\\':
101 101 s += pycompat.ossep
102 102 return s
103 103
104 104
105 105 def basewrapper(func, argtype, enc, dec, args, kwds):
106 106 # check check already converted, then call original
107 107 for arg in args:
108 108 if isinstance(arg, argtype):
109 109 return func(*args, **kwds)
110 110
111 111 try:
112 112 # convert string arguments, call func, then convert back the
113 113 # return value.
114 114 return enc(func(*dec(args), **dec(kwds)))
115 115 except UnicodeError:
116 116 raise error.Abort(_("[win32mbcs] filename conversion failed with"
117 117 " %s encoding\n") % (_encoding))
118 118
119 119 def wrapper(func, args, kwds):
120 120 return basewrapper(func, unicode, encode, decode, args, kwds)
121 121
122 122
123 123 def reversewrapper(func, args, kwds):
124 124 return basewrapper(func, str, decode, encode, args, kwds)
125 125
126 126 def wrapperforlistdir(func, args, kwds):
127 127 # Ensure 'path' argument ends with os.sep to avoids
128 128 # misinterpreting last 0x5c of MBCS 2nd byte as path separator.
129 129 if args:
130 130 args = list(args)
131 131 args[0] = appendsep(args[0])
132 132 if 'path' in kwds:
133 133 kwds['path'] = appendsep(kwds['path'])
134 134 return func(*args, **kwds)
135 135
136 136 def wrapname(name, wrapper):
137 137 module, name = name.rsplit('.', 1)
138 138 module = sys.modules[module]
139 139 func = getattr(module, name)
140 140 def f(*args, **kwds):
141 141 return wrapper(func, args, kwds)
142 142 f.__name__ = func.__name__
143 143 setattr(module, name, f)
144 144
145 145 # List of functions to be wrapped.
146 146 # NOTE: os.path.dirname() and os.path.basename() are safe because
147 147 # they use result of os.path.split()
148 148 funcs = '''os.path.join os.path.split os.path.splitext
149 149 os.path.normpath os.makedirs mercurial.util.endswithsep
150 150 mercurial.util.splitpath mercurial.util.fscasesensitive
151 151 mercurial.util.fspath mercurial.util.pconvert mercurial.util.normpath
152 152 mercurial.util.checkwinfilename mercurial.util.checkosfilename
153 153 mercurial.util.split'''
154 154
155 155 # These functions are required to be called with local encoded string
156 156 # because they expects argument is local encoded string and cause
157 157 # problem with unicode string.
158 158 rfuncs = '''mercurial.encoding.upper mercurial.encoding.lower'''
159 159
160 160 # List of Windows specific functions to be wrapped.
161 161 winfuncs = '''os.path.splitunc'''
162 162
163 163 # codec and alias names of sjis and big5 to be faked.
164 164 problematic_encodings = '''big5 big5-tw csbig5 big5hkscs big5-hkscs
165 165 hkscs cp932 932 ms932 mskanji ms-kanji shift_jis csshiftjis shiftjis
166 166 sjis s_jis shift_jis_2004 shiftjis2004 sjis_2004 sjis2004
167 167 shift_jisx0213 shiftjisx0213 sjisx0213 s_jisx0213 950 cp950 ms950 '''
168 168
169 169 def extsetup(ui):
170 170 # TODO: decide use of config section for this extension
171 171 if ((not os.path.supports_unicode_filenames) and
172 (sys.platform != 'cygwin')):
172 (pycompat.sysplatform != 'cygwin')):
173 173 ui.warn(_("[win32mbcs] cannot activate on this platform.\n"))
174 174 return
175 175 # determine encoding for filename
176 176 global _encoding
177 177 _encoding = ui.config('win32mbcs', 'encoding', encoding.encoding)
178 178 # fake is only for relevant environment.
179 179 if _encoding.lower() in problematic_encodings.split():
180 180 for f in funcs.split():
181 181 wrapname(f, wrapper)
182 182 if pycompat.osname == 'nt':
183 183 for f in winfuncs.split():
184 184 wrapname(f, wrapper)
185 185 wrapname("mercurial.osutil.listdir", wrapperforlistdir)
186 186 # wrap functions to be called with local byte string arguments
187 187 for f in rfuncs.split():
188 188 wrapname(f, reversewrapper)
189 189 # Check sys.args manually instead of using ui.debug() because
190 190 # command line options is not yet applied when
191 191 # extensions.loadall() is called.
192 192 if '--debug' in sys.argv:
193 193 ui.write(("[win32mbcs] activated with encoding: %s\n")
194 194 % _encoding)
@@ -1,652 +1,652 b''
1 1 # posix.py - Posix utility function implementations for Mercurial
2 2 #
3 3 # Copyright 2005-2009 Matt Mackall <mpm@selenic.com> and others
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 fcntl
12 12 import getpass
13 13 import grp
14 14 import os
15 15 import pwd
16 16 import re
17 17 import select
18 18 import stat
19 19 import sys
20 20 import tempfile
21 21 import unicodedata
22 22
23 23 from .i18n import _
24 24 from . import (
25 25 encoding,
26 26 pycompat,
27 27 )
28 28
29 29 posixfile = open
30 30 normpath = os.path.normpath
31 31 samestat = os.path.samestat
32 32 try:
33 33 oslink = os.link
34 34 except AttributeError:
35 35 # Some platforms build Python without os.link on systems that are
36 36 # vaguely unix-like but don't have hardlink support. For those
37 37 # poor souls, just say we tried and that it failed so we fall back
38 38 # to copies.
39 39 def oslink(src, dst):
40 40 raise OSError(errno.EINVAL,
41 41 'hardlinks not supported: %s to %s' % (src, dst))
42 42 unlink = os.unlink
43 43 rename = os.rename
44 44 removedirs = os.removedirs
45 45 expandglobs = False
46 46
47 47 umask = os.umask(0)
48 48 os.umask(umask)
49 49
50 50 def split(p):
51 51 '''Same as posixpath.split, but faster
52 52
53 53 >>> import posixpath
54 54 >>> for f in ['/absolute/path/to/file',
55 55 ... 'relative/path/to/file',
56 56 ... 'file_alone',
57 57 ... 'path/to/directory/',
58 58 ... '/multiple/path//separators',
59 59 ... '/file_at_root',
60 60 ... '///multiple_leading_separators_at_root',
61 61 ... '']:
62 62 ... assert split(f) == posixpath.split(f), f
63 63 '''
64 64 ht = p.rsplit('/', 1)
65 65 if len(ht) == 1:
66 66 return '', p
67 67 nh = ht[0].rstrip('/')
68 68 if nh:
69 69 return nh, ht[1]
70 70 return ht[0] + '/', ht[1]
71 71
72 72 def openhardlinks():
73 73 '''return true if it is safe to hold open file handles to hardlinks'''
74 74 return True
75 75
76 76 def nlinks(name):
77 77 '''return number of hardlinks for the given file'''
78 78 return os.lstat(name).st_nlink
79 79
80 80 def parsepatchoutput(output_line):
81 81 """parses the output produced by patch and returns the filename"""
82 82 pf = output_line[14:]
83 if os.sys.platform == 'OpenVMS':
83 if pycompat.sysplatform == 'OpenVMS':
84 84 if pf[0] == '`':
85 85 pf = pf[1:-1] # Remove the quotes
86 86 else:
87 87 if pf.startswith("'") and pf.endswith("'") and " " in pf:
88 88 pf = pf[1:-1] # Remove the quotes
89 89 return pf
90 90
91 91 def sshargs(sshcmd, host, user, port):
92 92 '''Build argument list for ssh'''
93 93 args = user and ("%s@%s" % (user, host)) or host
94 94 return port and ("%s -p %s" % (args, port)) or args
95 95
96 96 def isexec(f):
97 97 """check whether a file is executable"""
98 98 return (os.lstat(f).st_mode & 0o100 != 0)
99 99
100 100 def setflags(f, l, x):
101 101 s = os.lstat(f).st_mode
102 102 if l:
103 103 if not stat.S_ISLNK(s):
104 104 # switch file to link
105 105 fp = open(f)
106 106 data = fp.read()
107 107 fp.close()
108 108 os.unlink(f)
109 109 try:
110 110 os.symlink(data, f)
111 111 except OSError:
112 112 # failed to make a link, rewrite file
113 113 fp = open(f, "w")
114 114 fp.write(data)
115 115 fp.close()
116 116 # no chmod needed at this point
117 117 return
118 118 if stat.S_ISLNK(s):
119 119 # switch link to file
120 120 data = os.readlink(f)
121 121 os.unlink(f)
122 122 fp = open(f, "w")
123 123 fp.write(data)
124 124 fp.close()
125 125 s = 0o666 & ~umask # avoid restatting for chmod
126 126
127 127 sx = s & 0o100
128 128 if x and not sx:
129 129 # Turn on +x for every +r bit when making a file executable
130 130 # and obey umask.
131 131 os.chmod(f, s | (s & 0o444) >> 2 & ~umask)
132 132 elif not x and sx:
133 133 # Turn off all +x bits
134 134 os.chmod(f, s & 0o666)
135 135
136 136 def copymode(src, dst, mode=None):
137 137 '''Copy the file mode from the file at path src to dst.
138 138 If src doesn't exist, we're using mode instead. If mode is None, we're
139 139 using umask.'''
140 140 try:
141 141 st_mode = os.lstat(src).st_mode & 0o777
142 142 except OSError as inst:
143 143 if inst.errno != errno.ENOENT:
144 144 raise
145 145 st_mode = mode
146 146 if st_mode is None:
147 147 st_mode = ~umask
148 148 st_mode &= 0o666
149 149 os.chmod(dst, st_mode)
150 150
151 151 def checkexec(path):
152 152 """
153 153 Check whether the given path is on a filesystem with UNIX-like exec flags
154 154
155 155 Requires a directory (like /foo/.hg)
156 156 """
157 157
158 158 # VFAT on some Linux versions can flip mode but it doesn't persist
159 159 # a FS remount. Frequently we can detect it if files are created
160 160 # with exec bit on.
161 161
162 162 try:
163 163 EXECFLAGS = stat.S_IXUSR | stat.S_IXGRP | stat.S_IXOTH
164 164 cachedir = os.path.join(path, '.hg', 'cache')
165 165 if os.path.isdir(cachedir):
166 166 checkisexec = os.path.join(cachedir, 'checkisexec')
167 167 checknoexec = os.path.join(cachedir, 'checknoexec')
168 168
169 169 try:
170 170 m = os.stat(checkisexec).st_mode
171 171 except OSError as e:
172 172 if e.errno != errno.ENOENT:
173 173 raise
174 174 # checkisexec does not exist - fall through ...
175 175 else:
176 176 # checkisexec exists, check if it actually is exec
177 177 if m & EXECFLAGS != 0:
178 178 # ensure checkisexec exists, check it isn't exec
179 179 try:
180 180 m = os.stat(checknoexec).st_mode
181 181 except OSError as e:
182 182 if e.errno != errno.ENOENT:
183 183 raise
184 184 file(checknoexec, 'w').close() # might fail
185 185 m = os.stat(checknoexec).st_mode
186 186 if m & EXECFLAGS == 0:
187 187 # check-exec is exec and check-no-exec is not exec
188 188 return True
189 189 # checknoexec exists but is exec - delete it
190 190 os.unlink(checknoexec)
191 191 # checkisexec exists but is not exec - delete it
192 192 os.unlink(checkisexec)
193 193
194 194 # check using one file, leave it as checkisexec
195 195 checkdir = cachedir
196 196 else:
197 197 # check directly in path and don't leave checkisexec behind
198 198 checkdir = path
199 199 checkisexec = None
200 200 fh, fn = tempfile.mkstemp(dir=checkdir, prefix='hg-checkexec-')
201 201 try:
202 202 os.close(fh)
203 203 m = os.stat(fn).st_mode
204 204 if m & EXECFLAGS == 0:
205 205 os.chmod(fn, m & 0o777 | EXECFLAGS)
206 206 if os.stat(fn).st_mode & EXECFLAGS != 0:
207 207 if checkisexec is not None:
208 208 os.rename(fn, checkisexec)
209 209 fn = None
210 210 return True
211 211 finally:
212 212 if fn is not None:
213 213 os.unlink(fn)
214 214 except (IOError, OSError):
215 215 # we don't care, the user probably won't be able to commit anyway
216 216 return False
217 217
218 218 def checklink(path):
219 219 """check whether the given path is on a symlink-capable filesystem"""
220 220 # mktemp is not racy because symlink creation will fail if the
221 221 # file already exists
222 222 while True:
223 223 cachedir = os.path.join(path, '.hg', 'cache')
224 224 checklink = os.path.join(cachedir, 'checklink')
225 225 # try fast path, read only
226 226 if os.path.islink(checklink):
227 227 return True
228 228 if os.path.isdir(cachedir):
229 229 checkdir = cachedir
230 230 else:
231 231 checkdir = path
232 232 cachedir = None
233 233 name = tempfile.mktemp(dir=checkdir, prefix='checklink-')
234 234 try:
235 235 fd = None
236 236 if cachedir is None:
237 237 fd = tempfile.NamedTemporaryFile(dir=checkdir,
238 238 prefix='hg-checklink-')
239 239 target = os.path.basename(fd.name)
240 240 else:
241 241 # create a fixed file to link to; doesn't matter if it
242 242 # already exists.
243 243 target = 'checklink-target'
244 244 open(os.path.join(cachedir, target), 'w').close()
245 245 try:
246 246 os.symlink(target, name)
247 247 if cachedir is None:
248 248 os.unlink(name)
249 249 else:
250 250 try:
251 251 os.rename(name, checklink)
252 252 except OSError:
253 253 os.unlink(name)
254 254 return True
255 255 except OSError as inst:
256 256 # link creation might race, try again
257 257 if inst[0] == errno.EEXIST:
258 258 continue
259 259 raise
260 260 finally:
261 261 if fd is not None:
262 262 fd.close()
263 263 except AttributeError:
264 264 return False
265 265 except OSError as inst:
266 266 # sshfs might report failure while successfully creating the link
267 267 if inst[0] == errno.EIO and os.path.exists(name):
268 268 os.unlink(name)
269 269 return False
270 270
271 271 def checkosfilename(path):
272 272 '''Check that the base-relative path is a valid filename on this platform.
273 273 Returns None if the path is ok, or a UI string describing the problem.'''
274 274 pass # on posix platforms, every path is ok
275 275
276 276 def setbinary(fd):
277 277 pass
278 278
279 279 def pconvert(path):
280 280 return path
281 281
282 282 def localpath(path):
283 283 return path
284 284
285 285 def samefile(fpath1, fpath2):
286 286 """Returns whether path1 and path2 refer to the same file. This is only
287 287 guaranteed to work for files, not directories."""
288 288 return os.path.samefile(fpath1, fpath2)
289 289
290 290 def samedevice(fpath1, fpath2):
291 291 """Returns whether fpath1 and fpath2 are on the same device. This is only
292 292 guaranteed to work for files, not directories."""
293 293 st1 = os.lstat(fpath1)
294 294 st2 = os.lstat(fpath2)
295 295 return st1.st_dev == st2.st_dev
296 296
297 297 # os.path.normcase is a no-op, which doesn't help us on non-native filesystems
298 298 def normcase(path):
299 299 return path.lower()
300 300
301 301 # what normcase does to ASCII strings
302 302 normcasespec = encoding.normcasespecs.lower
303 303 # fallback normcase function for non-ASCII strings
304 304 normcasefallback = normcase
305 305
306 306 if pycompat.sysplatform == 'darwin':
307 307
308 308 def normcase(path):
309 309 '''
310 310 Normalize a filename for OS X-compatible comparison:
311 311 - escape-encode invalid characters
312 312 - decompose to NFD
313 313 - lowercase
314 314 - omit ignored characters [200c-200f, 202a-202e, 206a-206f,feff]
315 315
316 316 >>> normcase('UPPER')
317 317 'upper'
318 318 >>> normcase('Caf\xc3\xa9')
319 319 'cafe\\xcc\\x81'
320 320 >>> normcase('\xc3\x89')
321 321 'e\\xcc\\x81'
322 322 >>> normcase('\xb8\xca\xc3\xca\xbe\xc8.JPG') # issue3918
323 323 '%b8%ca%c3\\xca\\xbe%c8.jpg'
324 324 '''
325 325
326 326 try:
327 327 return encoding.asciilower(path) # exception for non-ASCII
328 328 except UnicodeDecodeError:
329 329 return normcasefallback(path)
330 330
331 331 normcasespec = encoding.normcasespecs.lower
332 332
333 333 def normcasefallback(path):
334 334 try:
335 335 u = path.decode('utf-8')
336 336 except UnicodeDecodeError:
337 337 # OS X percent-encodes any bytes that aren't valid utf-8
338 338 s = ''
339 339 pos = 0
340 340 l = len(path)
341 341 while pos < l:
342 342 try:
343 343 c = encoding.getutf8char(path, pos)
344 344 pos += len(c)
345 345 except ValueError:
346 346 c = '%%%02X' % ord(path[pos])
347 347 pos += 1
348 348 s += c
349 349
350 350 u = s.decode('utf-8')
351 351
352 352 # Decompose then lowercase (HFS+ technote specifies lower)
353 353 enc = unicodedata.normalize('NFD', u).lower().encode('utf-8')
354 354 # drop HFS+ ignored characters
355 355 return encoding.hfsignoreclean(enc)
356 356
357 357 if pycompat.sysplatform == 'cygwin':
358 358 # workaround for cygwin, in which mount point part of path is
359 359 # treated as case sensitive, even though underlying NTFS is case
360 360 # insensitive.
361 361
362 362 # default mount points
363 363 cygwinmountpoints = sorted([
364 364 "/usr/bin",
365 365 "/usr/lib",
366 366 "/cygdrive",
367 367 ], reverse=True)
368 368
369 369 # use upper-ing as normcase as same as NTFS workaround
370 370 def normcase(path):
371 371 pathlen = len(path)
372 372 if (pathlen == 0) or (path[0] != pycompat.ossep):
373 373 # treat as relative
374 374 return encoding.upper(path)
375 375
376 376 # to preserve case of mountpoint part
377 377 for mp in cygwinmountpoints:
378 378 if not path.startswith(mp):
379 379 continue
380 380
381 381 mplen = len(mp)
382 382 if mplen == pathlen: # mount point itself
383 383 return mp
384 384 if path[mplen] == pycompat.ossep:
385 385 return mp + encoding.upper(path[mplen:])
386 386
387 387 return encoding.upper(path)
388 388
389 389 normcasespec = encoding.normcasespecs.other
390 390 normcasefallback = normcase
391 391
392 392 # Cygwin translates native ACLs to POSIX permissions,
393 393 # but these translations are not supported by native
394 394 # tools, so the exec bit tends to be set erroneously.
395 395 # Therefore, disable executable bit access on Cygwin.
396 396 def checkexec(path):
397 397 return False
398 398
399 399 # Similarly, Cygwin's symlink emulation is likely to create
400 400 # problems when Mercurial is used from both Cygwin and native
401 401 # Windows, with other native tools, or on shared volumes
402 402 def checklink(path):
403 403 return False
404 404
405 405 _needsshellquote = None
406 406 def shellquote(s):
407 if os.sys.platform == 'OpenVMS':
407 if pycompat.sysplatform == 'OpenVMS':
408 408 return '"%s"' % s
409 409 global _needsshellquote
410 410 if _needsshellquote is None:
411 411 _needsshellquote = re.compile(r'[^a-zA-Z0-9._/+-]').search
412 412 if s and not _needsshellquote(s):
413 413 # "s" shouldn't have to be quoted
414 414 return s
415 415 else:
416 416 return "'%s'" % s.replace("'", "'\\''")
417 417
418 418 def quotecommand(cmd):
419 419 return cmd
420 420
421 421 def popen(command, mode='r'):
422 422 return os.popen(command, mode)
423 423
424 424 def testpid(pid):
425 425 '''return False if pid dead, True if running or not sure'''
426 if os.sys.platform == 'OpenVMS':
426 if pycompat.sysplatform == 'OpenVMS':
427 427 return True
428 428 try:
429 429 os.kill(pid, 0)
430 430 return True
431 431 except OSError as inst:
432 432 return inst.errno != errno.ESRCH
433 433
434 434 def explainexit(code):
435 435 """return a 2-tuple (desc, code) describing a subprocess status
436 436 (codes from kill are negative - not os.system/wait encoding)"""
437 437 if code >= 0:
438 438 return _("exited with status %d") % code, code
439 439 return _("killed by signal %d") % -code, -code
440 440
441 441 def isowner(st):
442 442 """Return True if the stat object st is from the current user."""
443 443 return st.st_uid == os.getuid()
444 444
445 445 def findexe(command):
446 446 '''Find executable for command searching like which does.
447 447 If command is a basename then PATH is searched for command.
448 448 PATH isn't searched if command is an absolute or relative path.
449 449 If command isn't found None is returned.'''
450 450 if pycompat.sysplatform == 'OpenVMS':
451 451 return command
452 452
453 453 def findexisting(executable):
454 454 'Will return executable if existing file'
455 455 if os.path.isfile(executable) and os.access(executable, os.X_OK):
456 456 return executable
457 457 return None
458 458
459 459 if pycompat.ossep in command:
460 460 return findexisting(command)
461 461
462 462 if pycompat.sysplatform == 'plan9':
463 463 return findexisting(os.path.join('/bin', command))
464 464
465 465 for path in encoding.environ.get('PATH', '').split(pycompat.ospathsep):
466 466 executable = findexisting(os.path.join(path, command))
467 467 if executable is not None:
468 468 return executable
469 469 return None
470 470
471 471 def setsignalhandler():
472 472 pass
473 473
474 474 _wantedkinds = set([stat.S_IFREG, stat.S_IFLNK])
475 475
476 476 def statfiles(files):
477 477 '''Stat each file in files. Yield each stat, or None if a file does not
478 478 exist or has a type we don't care about.'''
479 479 lstat = os.lstat
480 480 getkind = stat.S_IFMT
481 481 for nf in files:
482 482 try:
483 483 st = lstat(nf)
484 484 if getkind(st.st_mode) not in _wantedkinds:
485 485 st = None
486 486 except OSError as err:
487 487 if err.errno not in (errno.ENOENT, errno.ENOTDIR):
488 488 raise
489 489 st = None
490 490 yield st
491 491
492 492 def getuser():
493 493 '''return name of current user'''
494 494 return getpass.getuser()
495 495
496 496 def username(uid=None):
497 497 """Return the name of the user with the given uid.
498 498
499 499 If uid is None, return the name of the current user."""
500 500
501 501 if uid is None:
502 502 uid = os.getuid()
503 503 try:
504 504 return pwd.getpwuid(uid)[0]
505 505 except KeyError:
506 506 return str(uid)
507 507
508 508 def groupname(gid=None):
509 509 """Return the name of the group with the given gid.
510 510
511 511 If gid is None, return the name of the current group."""
512 512
513 513 if gid is None:
514 514 gid = os.getgid()
515 515 try:
516 516 return grp.getgrgid(gid)[0]
517 517 except KeyError:
518 518 return str(gid)
519 519
520 520 def groupmembers(name):
521 521 """Return the list of members of the group with the given
522 522 name, KeyError if the group does not exist.
523 523 """
524 524 return list(grp.getgrnam(name).gr_mem)
525 525
526 526 def spawndetached(args):
527 527 return os.spawnvp(os.P_NOWAIT | getattr(os, 'P_DETACH', 0),
528 528 args[0], args)
529 529
530 530 def gethgcmd():
531 531 return sys.argv[:1]
532 532
533 533 def makedir(path, notindexed):
534 534 os.mkdir(path)
535 535
536 536 def unlinkpath(f, ignoremissing=False):
537 537 """unlink and remove the directory if it is empty"""
538 538 try:
539 539 os.unlink(f)
540 540 except OSError as e:
541 541 if not (ignoremissing and e.errno == errno.ENOENT):
542 542 raise
543 543 # try removing directories that might now be empty
544 544 try:
545 545 os.removedirs(os.path.dirname(f))
546 546 except OSError:
547 547 pass
548 548
549 549 def lookupreg(key, name=None, scope=None):
550 550 return None
551 551
552 552 def hidewindow():
553 553 """Hide current shell window.
554 554
555 555 Used to hide the window opened when starting asynchronous
556 556 child process under Windows, unneeded on other systems.
557 557 """
558 558 pass
559 559
560 560 class cachestat(object):
561 561 def __init__(self, path):
562 562 self.stat = os.stat(path)
563 563
564 564 def cacheable(self):
565 565 return bool(self.stat.st_ino)
566 566
567 567 __hash__ = object.__hash__
568 568
569 569 def __eq__(self, other):
570 570 try:
571 571 # Only dev, ino, size, mtime and atime are likely to change. Out
572 572 # of these, we shouldn't compare atime but should compare the
573 573 # rest. However, one of the other fields changing indicates
574 574 # something fishy going on, so return False if anything but atime
575 575 # changes.
576 576 return (self.stat.st_mode == other.stat.st_mode and
577 577 self.stat.st_ino == other.stat.st_ino and
578 578 self.stat.st_dev == other.stat.st_dev and
579 579 self.stat.st_nlink == other.stat.st_nlink and
580 580 self.stat.st_uid == other.stat.st_uid and
581 581 self.stat.st_gid == other.stat.st_gid and
582 582 self.stat.st_size == other.stat.st_size and
583 583 self.stat.st_mtime == other.stat.st_mtime and
584 584 self.stat.st_ctime == other.stat.st_ctime)
585 585 except AttributeError:
586 586 return False
587 587
588 588 def __ne__(self, other):
589 589 return not self == other
590 590
591 591 def executablepath():
592 592 return None # available on Windows only
593 593
594 594 def statislink(st):
595 595 '''check whether a stat result is a symlink'''
596 596 return st and stat.S_ISLNK(st.st_mode)
597 597
598 598 def statisexec(st):
599 599 '''check whether a stat result is an executable file'''
600 600 return st and (st.st_mode & 0o100 != 0)
601 601
602 602 def poll(fds):
603 603 """block until something happens on any file descriptor
604 604
605 605 This is a generic helper that will check for any activity
606 606 (read, write. exception) and return the list of touched files.
607 607
608 608 In unsupported cases, it will raise a NotImplementedError"""
609 609 try:
610 610 res = select.select(fds, fds, fds)
611 611 except ValueError: # out of range file descriptor
612 612 raise NotImplementedError()
613 613 return sorted(list(set(sum(res, []))))
614 614
615 615 def readpipe(pipe):
616 616 """Read all available data from a pipe."""
617 617 # We can't fstat() a pipe because Linux will always report 0.
618 618 # So, we set the pipe to non-blocking mode and read everything
619 619 # that's available.
620 620 flags = fcntl.fcntl(pipe, fcntl.F_GETFL)
621 621 flags |= os.O_NONBLOCK
622 622 oldflags = fcntl.fcntl(pipe, fcntl.F_SETFL, flags)
623 623
624 624 try:
625 625 chunks = []
626 626 while True:
627 627 try:
628 628 s = pipe.read()
629 629 if not s:
630 630 break
631 631 chunks.append(s)
632 632 except IOError:
633 633 break
634 634
635 635 return ''.join(chunks)
636 636 finally:
637 637 fcntl.fcntl(pipe, fcntl.F_SETFL, oldflags)
638 638
639 639 def bindunixsocket(sock, path):
640 640 """Bind the UNIX domain socket to the specified path"""
641 641 # use relative path instead of full path at bind() if possible, since
642 642 # AF_UNIX path has very small length limit (107 chars) on common
643 643 # platforms (see sys/un.h)
644 644 dirname, basename = os.path.split(path)
645 645 bakwdfd = None
646 646 if dirname:
647 647 bakwdfd = os.open('.', os.O_DIRECTORY)
648 648 os.chdir(dirname)
649 649 sock.bind(basename)
650 650 if bakwdfd:
651 651 os.fchdir(bakwdfd)
652 652 os.close(bakwdfd)
@@ -1,366 +1,365 b''
1 1 # osutil.py - pure Python version of osutil.c
2 2 #
3 3 # Copyright 2009 Matt Mackall <mpm@selenic.com> and others
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 ctypes
11 11 import ctypes.util
12 12 import os
13 13 import socket
14 14 import stat as statmod
15 import sys
16 15
17 16 from . import (
18 17 policy,
19 18 pycompat,
20 19 )
21 20
22 21 modulepolicy = policy.policy
23 22 policynocffi = policy.policynocffi
24 23
25 24 def _mode_to_kind(mode):
26 25 if statmod.S_ISREG(mode):
27 26 return statmod.S_IFREG
28 27 if statmod.S_ISDIR(mode):
29 28 return statmod.S_IFDIR
30 29 if statmod.S_ISLNK(mode):
31 30 return statmod.S_IFLNK
32 31 if statmod.S_ISBLK(mode):
33 32 return statmod.S_IFBLK
34 33 if statmod.S_ISCHR(mode):
35 34 return statmod.S_IFCHR
36 35 if statmod.S_ISFIFO(mode):
37 36 return statmod.S_IFIFO
38 37 if statmod.S_ISSOCK(mode):
39 38 return statmod.S_IFSOCK
40 39 return mode
41 40
42 41 def listdirpure(path, stat=False, skip=None):
43 42 '''listdir(path, stat=False) -> list_of_tuples
44 43
45 44 Return a sorted list containing information about the entries
46 45 in the directory.
47 46
48 47 If stat is True, each element is a 3-tuple:
49 48
50 49 (name, type, stat object)
51 50
52 51 Otherwise, each element is a 2-tuple:
53 52
54 53 (name, type)
55 54 '''
56 55 result = []
57 56 prefix = path
58 57 if not prefix.endswith(pycompat.ossep):
59 58 prefix += pycompat.ossep
60 59 names = os.listdir(path)
61 60 names.sort()
62 61 for fn in names:
63 62 st = os.lstat(prefix + fn)
64 63 if fn == skip and statmod.S_ISDIR(st.st_mode):
65 64 return []
66 65 if stat:
67 66 result.append((fn, _mode_to_kind(st.st_mode), st))
68 67 else:
69 68 result.append((fn, _mode_to_kind(st.st_mode)))
70 69 return result
71 70
72 71 ffi = None
73 if modulepolicy not in policynocffi and sys.platform == 'darwin':
72 if modulepolicy not in policynocffi and pycompat.sysplatform == 'darwin':
74 73 try:
75 74 from _osutil_cffi import ffi, lib
76 75 except ImportError:
77 76 if modulepolicy == 'cffi': # strict cffi import
78 77 raise
79 78
80 if sys.platform == 'darwin' and ffi is not None:
79 if pycompat.sysplatform == 'darwin' and ffi is not None:
81 80 listdir_batch_size = 4096
82 81 # tweakable number, only affects performance, which chunks
83 82 # of bytes do we get back from getattrlistbulk
84 83
85 84 attrkinds = [None] * 20 # we need the max no for enum VXXX, 20 is plenty
86 85
87 86 attrkinds[lib.VREG] = statmod.S_IFREG
88 87 attrkinds[lib.VDIR] = statmod.S_IFDIR
89 88 attrkinds[lib.VLNK] = statmod.S_IFLNK
90 89 attrkinds[lib.VBLK] = statmod.S_IFBLK
91 90 attrkinds[lib.VCHR] = statmod.S_IFCHR
92 91 attrkinds[lib.VFIFO] = statmod.S_IFIFO
93 92 attrkinds[lib.VSOCK] = statmod.S_IFSOCK
94 93
95 94 class stat_res(object):
96 95 def __init__(self, st_mode, st_mtime, st_size):
97 96 self.st_mode = st_mode
98 97 self.st_mtime = st_mtime
99 98 self.st_size = st_size
100 99
101 100 tv_sec_ofs = ffi.offsetof("struct timespec", "tv_sec")
102 101 buf = ffi.new("char[]", listdir_batch_size)
103 102
104 103 def listdirinternal(dfd, req, stat, skip):
105 104 ret = []
106 105 while True:
107 106 r = lib.getattrlistbulk(dfd, req, buf, listdir_batch_size, 0)
108 107 if r == 0:
109 108 break
110 109 if r == -1:
111 110 raise OSError(ffi.errno, os.strerror(ffi.errno))
112 111 cur = ffi.cast("val_attrs_t*", buf)
113 112 for i in range(r):
114 113 lgt = cur.length
115 114 assert lgt == ffi.cast('uint32_t*', cur)[0]
116 115 ofs = cur.name_info.attr_dataoffset
117 116 str_lgt = cur.name_info.attr_length
118 117 base_ofs = ffi.offsetof('val_attrs_t', 'name_info')
119 118 name = str(ffi.buffer(ffi.cast("char*", cur) + base_ofs + ofs,
120 119 str_lgt - 1))
121 120 tp = attrkinds[cur.obj_type]
122 121 if name == "." or name == "..":
123 122 continue
124 123 if skip == name and tp == statmod.S_ISDIR:
125 124 return []
126 125 if stat:
127 126 mtime = cur.mtime.tv_sec
128 127 mode = (cur.accessmask & ~lib.S_IFMT)| tp
129 128 ret.append((name, tp, stat_res(st_mode=mode, st_mtime=mtime,
130 129 st_size=cur.datalength)))
131 130 else:
132 131 ret.append((name, tp))
133 132 cur = ffi.cast("val_attrs_t*", int(ffi.cast("intptr_t", cur))
134 133 + lgt)
135 134 return ret
136 135
137 136 def listdir(path, stat=False, skip=None):
138 137 req = ffi.new("struct attrlist*")
139 138 req.bitmapcount = lib.ATTR_BIT_MAP_COUNT
140 139 req.commonattr = (lib.ATTR_CMN_RETURNED_ATTRS |
141 140 lib.ATTR_CMN_NAME |
142 141 lib.ATTR_CMN_OBJTYPE |
143 142 lib.ATTR_CMN_ACCESSMASK |
144 143 lib.ATTR_CMN_MODTIME)
145 144 req.fileattr = lib.ATTR_FILE_DATALENGTH
146 145 dfd = lib.open(path, lib.O_RDONLY, 0)
147 146 if dfd == -1:
148 147 raise OSError(ffi.errno, os.strerror(ffi.errno))
149 148
150 149 try:
151 150 ret = listdirinternal(dfd, req, stat, skip)
152 151 finally:
153 152 try:
154 153 lib.close(dfd)
155 154 except BaseException:
156 155 pass # we ignore all the errors from closing, not
157 156 # much we can do about that
158 157 return ret
159 158 else:
160 159 listdir = listdirpure
161 160
162 161 if pycompat.osname != 'nt':
163 162 posixfile = open
164 163
165 164 _SCM_RIGHTS = 0x01
166 165 _socklen_t = ctypes.c_uint
167 166
168 if sys.platform == 'linux2':
167 if pycompat.sysplatform.startswith('linux'):
169 168 # socket.h says "the type should be socklen_t but the definition of
170 169 # the kernel is incompatible with this."
171 170 _cmsg_len_t = ctypes.c_size_t
172 171 _msg_controllen_t = ctypes.c_size_t
173 172 _msg_iovlen_t = ctypes.c_size_t
174 173 else:
175 174 _cmsg_len_t = _socklen_t
176 175 _msg_controllen_t = _socklen_t
177 176 _msg_iovlen_t = ctypes.c_int
178 177
179 178 class _iovec(ctypes.Structure):
180 179 _fields_ = [
181 180 (u'iov_base', ctypes.c_void_p),
182 181 (u'iov_len', ctypes.c_size_t),
183 182 ]
184 183
185 184 class _msghdr(ctypes.Structure):
186 185 _fields_ = [
187 186 (u'msg_name', ctypes.c_void_p),
188 187 (u'msg_namelen', _socklen_t),
189 188 (u'msg_iov', ctypes.POINTER(_iovec)),
190 189 (u'msg_iovlen', _msg_iovlen_t),
191 190 (u'msg_control', ctypes.c_void_p),
192 191 (u'msg_controllen', _msg_controllen_t),
193 192 (u'msg_flags', ctypes.c_int),
194 193 ]
195 194
196 195 class _cmsghdr(ctypes.Structure):
197 196 _fields_ = [
198 197 (u'cmsg_len', _cmsg_len_t),
199 198 (u'cmsg_level', ctypes.c_int),
200 199 (u'cmsg_type', ctypes.c_int),
201 200 (u'cmsg_data', ctypes.c_ubyte * 0),
202 201 ]
203 202
204 203 _libc = ctypes.CDLL(ctypes.util.find_library(u'c'), use_errno=True)
205 204 _recvmsg = getattr(_libc, 'recvmsg', None)
206 205 if _recvmsg:
207 206 _recvmsg.restype = getattr(ctypes, 'c_ssize_t', ctypes.c_long)
208 207 _recvmsg.argtypes = (ctypes.c_int, ctypes.POINTER(_msghdr),
209 208 ctypes.c_int)
210 209 else:
211 210 # recvmsg isn't always provided by libc; such systems are unsupported
212 211 def _recvmsg(sockfd, msg, flags):
213 212 raise NotImplementedError('unsupported platform')
214 213
215 214 def _CMSG_FIRSTHDR(msgh):
216 215 if msgh.msg_controllen < ctypes.sizeof(_cmsghdr):
217 216 return
218 217 cmsgptr = ctypes.cast(msgh.msg_control, ctypes.POINTER(_cmsghdr))
219 218 return cmsgptr.contents
220 219
221 220 # The pure version is less portable than the native version because the
222 221 # handling of socket ancillary data heavily depends on C preprocessor.
223 222 # Also, some length fields are wrongly typed in Linux kernel.
224 223 def recvfds(sockfd):
225 224 """receive list of file descriptors via socket"""
226 225 dummy = (ctypes.c_ubyte * 1)()
227 226 iov = _iovec(ctypes.cast(dummy, ctypes.c_void_p), ctypes.sizeof(dummy))
228 227 cbuf = ctypes.create_string_buffer(256)
229 228 msgh = _msghdr(None, 0,
230 229 ctypes.pointer(iov), 1,
231 230 ctypes.cast(cbuf, ctypes.c_void_p), ctypes.sizeof(cbuf),
232 231 0)
233 232 r = _recvmsg(sockfd, ctypes.byref(msgh), 0)
234 233 if r < 0:
235 234 e = ctypes.get_errno()
236 235 raise OSError(e, os.strerror(e))
237 236 # assumes that the first cmsg has fds because it isn't easy to write
238 237 # portable CMSG_NXTHDR() with ctypes.
239 238 cmsg = _CMSG_FIRSTHDR(msgh)
240 239 if not cmsg:
241 240 return []
242 241 if (cmsg.cmsg_level != socket.SOL_SOCKET or
243 242 cmsg.cmsg_type != _SCM_RIGHTS):
244 243 return []
245 244 rfds = ctypes.cast(cmsg.cmsg_data, ctypes.POINTER(ctypes.c_int))
246 245 rfdscount = ((cmsg.cmsg_len - _cmsghdr.cmsg_data.offset) /
247 246 ctypes.sizeof(ctypes.c_int))
248 247 return [rfds[i] for i in xrange(rfdscount)]
249 248
250 249 else:
251 250 import msvcrt
252 251
253 252 _kernel32 = ctypes.windll.kernel32
254 253
255 254 _DWORD = ctypes.c_ulong
256 255 _LPCSTR = _LPSTR = ctypes.c_char_p
257 256 _HANDLE = ctypes.c_void_p
258 257
259 258 _INVALID_HANDLE_VALUE = _HANDLE(-1).value
260 259
261 260 # CreateFile
262 261 _FILE_SHARE_READ = 0x00000001
263 262 _FILE_SHARE_WRITE = 0x00000002
264 263 _FILE_SHARE_DELETE = 0x00000004
265 264
266 265 _CREATE_ALWAYS = 2
267 266 _OPEN_EXISTING = 3
268 267 _OPEN_ALWAYS = 4
269 268
270 269 _GENERIC_READ = 0x80000000
271 270 _GENERIC_WRITE = 0x40000000
272 271
273 272 _FILE_ATTRIBUTE_NORMAL = 0x80
274 273
275 274 # open_osfhandle flags
276 275 _O_RDONLY = 0x0000
277 276 _O_RDWR = 0x0002
278 277 _O_APPEND = 0x0008
279 278
280 279 _O_TEXT = 0x4000
281 280 _O_BINARY = 0x8000
282 281
283 282 # types of parameters of C functions used (required by pypy)
284 283
285 284 _kernel32.CreateFileA.argtypes = [_LPCSTR, _DWORD, _DWORD, ctypes.c_void_p,
286 285 _DWORD, _DWORD, _HANDLE]
287 286 _kernel32.CreateFileA.restype = _HANDLE
288 287
289 288 def _raiseioerror(name):
290 289 err = ctypes.WinError()
291 290 raise IOError(err.errno, '%s: %s' % (name, err.strerror))
292 291
293 292 class posixfile(object):
294 293 '''a file object aiming for POSIX-like semantics
295 294
296 295 CPython's open() returns a file that was opened *without* setting the
297 296 _FILE_SHARE_DELETE flag, which causes rename and unlink to abort.
298 297 This even happens if any hardlinked copy of the file is in open state.
299 298 We set _FILE_SHARE_DELETE here, so files opened with posixfile can be
300 299 renamed and deleted while they are held open.
301 300 Note that if a file opened with posixfile is unlinked, the file
302 301 remains but cannot be opened again or be recreated under the same name,
303 302 until all reading processes have closed the file.'''
304 303
305 304 def __init__(self, name, mode='r', bufsize=-1):
306 305 if 'b' in mode:
307 306 flags = _O_BINARY
308 307 else:
309 308 flags = _O_TEXT
310 309
311 310 m0 = mode[0]
312 311 if m0 == 'r' and '+' not in mode:
313 312 flags |= _O_RDONLY
314 313 access = _GENERIC_READ
315 314 else:
316 315 # work around http://support.microsoft.com/kb/899149 and
317 316 # set _O_RDWR for 'w' and 'a', even if mode has no '+'
318 317 flags |= _O_RDWR
319 318 access = _GENERIC_READ | _GENERIC_WRITE
320 319
321 320 if m0 == 'r':
322 321 creation = _OPEN_EXISTING
323 322 elif m0 == 'w':
324 323 creation = _CREATE_ALWAYS
325 324 elif m0 == 'a':
326 325 creation = _OPEN_ALWAYS
327 326 flags |= _O_APPEND
328 327 else:
329 328 raise ValueError("invalid mode: %s" % mode)
330 329
331 330 fh = _kernel32.CreateFileA(name, access,
332 331 _FILE_SHARE_READ | _FILE_SHARE_WRITE | _FILE_SHARE_DELETE,
333 332 None, creation, _FILE_ATTRIBUTE_NORMAL, None)
334 333 if fh == _INVALID_HANDLE_VALUE:
335 334 _raiseioerror(name)
336 335
337 336 fd = msvcrt.open_osfhandle(fh, flags)
338 337 if fd == -1:
339 338 _kernel32.CloseHandle(fh)
340 339 _raiseioerror(name)
341 340
342 341 f = os.fdopen(fd, mode, bufsize)
343 342 # unfortunately, f.name is '<fdopen>' at this point -- so we store
344 343 # the name on this wrapper. We cannot just assign to f.name,
345 344 # because that attribute is read-only.
346 345 object.__setattr__(self, 'name', name)
347 346 object.__setattr__(self, '_file', f)
348 347
349 348 def __iter__(self):
350 349 return self._file
351 350
352 351 def __getattr__(self, name):
353 352 return getattr(self._file, name)
354 353
355 354 def __setattr__(self, name, value):
356 355 '''mimics the read-only attributes of Python file objects
357 356 by raising 'TypeError: readonly attribute' if someone tries:
358 357 f = posixfile('foo.txt')
359 358 f.name = 'bla' '''
360 359 return self._file.__setattr__(name, value)
361 360
362 361 def __enter__(self):
363 362 return self._file.__enter__()
364 363
365 364 def __exit__(self, exc_type, exc_value, exc_tb):
366 365 return self._file.__exit__(exc_type, exc_value, exc_tb)
@@ -1,3246 +1,3246 b''
1 1 # util.py - Mercurial utility functions and platform specific implementations
2 2 #
3 3 # Copyright 2005 K. Thananchayan <thananck@yahoo.com>
4 4 # Copyright 2005-2007 Matt Mackall <mpm@selenic.com>
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 """Mercurial utility functions and platform specific implementations.
11 11
12 12 This contains helper routines that are independent of the SCM core and
13 13 hide platform-specific details from the core.
14 14 """
15 15
16 16 from __future__ import absolute_import
17 17
18 18 import bz2
19 19 import calendar
20 20 import collections
21 21 import datetime
22 22 import errno
23 23 import gc
24 24 import hashlib
25 25 import imp
26 26 import os
27 27 import platform as pyplatform
28 28 import re as remod
29 29 import shutil
30 30 import signal
31 31 import socket
32 32 import stat
33 33 import string
34 34 import subprocess
35 35 import sys
36 36 import tempfile
37 37 import textwrap
38 38 import time
39 39 import traceback
40 40 import zlib
41 41
42 42 from . import (
43 43 encoding,
44 44 error,
45 45 i18n,
46 46 osutil,
47 47 parsers,
48 48 pycompat,
49 49 )
50 50
51 51 empty = pycompat.empty
52 52 httplib = pycompat.httplib
53 53 httpserver = pycompat.httpserver
54 54 pickle = pycompat.pickle
55 55 queue = pycompat.queue
56 56 socketserver = pycompat.socketserver
57 57 stderr = pycompat.stderr
58 58 stdin = pycompat.stdin
59 59 stdout = pycompat.stdout
60 60 stringio = pycompat.stringio
61 61 urlerr = pycompat.urlerr
62 62 urlparse = pycompat.urlparse
63 63 urlreq = pycompat.urlreq
64 64 xmlrpclib = pycompat.xmlrpclib
65 65
66 66 if pycompat.osname == 'nt':
67 67 from . import windows as platform
68 68 stdout = platform.winstdout(pycompat.stdout)
69 69 else:
70 70 from . import posix as platform
71 71
72 72 _ = i18n._
73 73
74 74 bindunixsocket = platform.bindunixsocket
75 75 cachestat = platform.cachestat
76 76 checkexec = platform.checkexec
77 77 checklink = platform.checklink
78 78 copymode = platform.copymode
79 79 executablepath = platform.executablepath
80 80 expandglobs = platform.expandglobs
81 81 explainexit = platform.explainexit
82 82 findexe = platform.findexe
83 83 gethgcmd = platform.gethgcmd
84 84 getuser = platform.getuser
85 85 getpid = os.getpid
86 86 groupmembers = platform.groupmembers
87 87 groupname = platform.groupname
88 88 hidewindow = platform.hidewindow
89 89 isexec = platform.isexec
90 90 isowner = platform.isowner
91 91 localpath = platform.localpath
92 92 lookupreg = platform.lookupreg
93 93 makedir = platform.makedir
94 94 nlinks = platform.nlinks
95 95 normpath = platform.normpath
96 96 normcase = platform.normcase
97 97 normcasespec = platform.normcasespec
98 98 normcasefallback = platform.normcasefallback
99 99 openhardlinks = platform.openhardlinks
100 100 oslink = platform.oslink
101 101 parsepatchoutput = platform.parsepatchoutput
102 102 pconvert = platform.pconvert
103 103 poll = platform.poll
104 104 popen = platform.popen
105 105 posixfile = platform.posixfile
106 106 quotecommand = platform.quotecommand
107 107 readpipe = platform.readpipe
108 108 rename = platform.rename
109 109 removedirs = platform.removedirs
110 110 samedevice = platform.samedevice
111 111 samefile = platform.samefile
112 112 samestat = platform.samestat
113 113 setbinary = platform.setbinary
114 114 setflags = platform.setflags
115 115 setsignalhandler = platform.setsignalhandler
116 116 shellquote = platform.shellquote
117 117 spawndetached = platform.spawndetached
118 118 split = platform.split
119 119 sshargs = platform.sshargs
120 120 statfiles = getattr(osutil, 'statfiles', platform.statfiles)
121 121 statisexec = platform.statisexec
122 122 statislink = platform.statislink
123 123 testpid = platform.testpid
124 124 umask = platform.umask
125 125 unlink = platform.unlink
126 126 unlinkpath = platform.unlinkpath
127 127 username = platform.username
128 128
129 129 # Python compatibility
130 130
131 131 _notset = object()
132 132
133 133 # disable Python's problematic floating point timestamps (issue4836)
134 134 # (Python hypocritically says you shouldn't change this behavior in
135 135 # libraries, and sure enough Mercurial is not a library.)
136 136 os.stat_float_times(False)
137 137
138 138 def safehasattr(thing, attr):
139 139 return getattr(thing, attr, _notset) is not _notset
140 140
141 141 DIGESTS = {
142 142 'md5': hashlib.md5,
143 143 'sha1': hashlib.sha1,
144 144 'sha512': hashlib.sha512,
145 145 }
146 146 # List of digest types from strongest to weakest
147 147 DIGESTS_BY_STRENGTH = ['sha512', 'sha1', 'md5']
148 148
149 149 for k in DIGESTS_BY_STRENGTH:
150 150 assert k in DIGESTS
151 151
152 152 class digester(object):
153 153 """helper to compute digests.
154 154
155 155 This helper can be used to compute one or more digests given their name.
156 156
157 157 >>> d = digester(['md5', 'sha1'])
158 158 >>> d.update('foo')
159 159 >>> [k for k in sorted(d)]
160 160 ['md5', 'sha1']
161 161 >>> d['md5']
162 162 'acbd18db4cc2f85cedef654fccc4a4d8'
163 163 >>> d['sha1']
164 164 '0beec7b5ea3f0fdbc95d0dd47f3c5bc275da8a33'
165 165 >>> digester.preferred(['md5', 'sha1'])
166 166 'sha1'
167 167 """
168 168
169 169 def __init__(self, digests, s=''):
170 170 self._hashes = {}
171 171 for k in digests:
172 172 if k not in DIGESTS:
173 173 raise Abort(_('unknown digest type: %s') % k)
174 174 self._hashes[k] = DIGESTS[k]()
175 175 if s:
176 176 self.update(s)
177 177
178 178 def update(self, data):
179 179 for h in self._hashes.values():
180 180 h.update(data)
181 181
182 182 def __getitem__(self, key):
183 183 if key not in DIGESTS:
184 184 raise Abort(_('unknown digest type: %s') % k)
185 185 return self._hashes[key].hexdigest()
186 186
187 187 def __iter__(self):
188 188 return iter(self._hashes)
189 189
190 190 @staticmethod
191 191 def preferred(supported):
192 192 """returns the strongest digest type in both supported and DIGESTS."""
193 193
194 194 for k in DIGESTS_BY_STRENGTH:
195 195 if k in supported:
196 196 return k
197 197 return None
198 198
199 199 class digestchecker(object):
200 200 """file handle wrapper that additionally checks content against a given
201 201 size and digests.
202 202
203 203 d = digestchecker(fh, size, {'md5': '...'})
204 204
205 205 When multiple digests are given, all of them are validated.
206 206 """
207 207
208 208 def __init__(self, fh, size, digests):
209 209 self._fh = fh
210 210 self._size = size
211 211 self._got = 0
212 212 self._digests = dict(digests)
213 213 self._digester = digester(self._digests.keys())
214 214
215 215 def read(self, length=-1):
216 216 content = self._fh.read(length)
217 217 self._digester.update(content)
218 218 self._got += len(content)
219 219 return content
220 220
221 221 def validate(self):
222 222 if self._size != self._got:
223 223 raise Abort(_('size mismatch: expected %d, got %d') %
224 224 (self._size, self._got))
225 225 for k, v in self._digests.items():
226 226 if v != self._digester[k]:
227 227 # i18n: first parameter is a digest name
228 228 raise Abort(_('%s mismatch: expected %s, got %s') %
229 229 (k, v, self._digester[k]))
230 230
231 231 try:
232 232 buffer = buffer
233 233 except NameError:
234 234 if not pycompat.ispy3:
235 235 def buffer(sliceable, offset=0):
236 236 return sliceable[offset:]
237 237 else:
238 238 def buffer(sliceable, offset=0):
239 239 return memoryview(sliceable)[offset:]
240 240
241 241 closefds = pycompat.osname == 'posix'
242 242
243 243 _chunksize = 4096
244 244
245 245 class bufferedinputpipe(object):
246 246 """a manually buffered input pipe
247 247
248 248 Python will not let us use buffered IO and lazy reading with 'polling' at
249 249 the same time. We cannot probe the buffer state and select will not detect
250 250 that data are ready to read if they are already buffered.
251 251
252 252 This class let us work around that by implementing its own buffering
253 253 (allowing efficient readline) while offering a way to know if the buffer is
254 254 empty from the output (allowing collaboration of the buffer with polling).
255 255
256 256 This class lives in the 'util' module because it makes use of the 'os'
257 257 module from the python stdlib.
258 258 """
259 259
260 260 def __init__(self, input):
261 261 self._input = input
262 262 self._buffer = []
263 263 self._eof = False
264 264 self._lenbuf = 0
265 265
266 266 @property
267 267 def hasbuffer(self):
268 268 """True is any data is currently buffered
269 269
270 270 This will be used externally a pre-step for polling IO. If there is
271 271 already data then no polling should be set in place."""
272 272 return bool(self._buffer)
273 273
274 274 @property
275 275 def closed(self):
276 276 return self._input.closed
277 277
278 278 def fileno(self):
279 279 return self._input.fileno()
280 280
281 281 def close(self):
282 282 return self._input.close()
283 283
284 284 def read(self, size):
285 285 while (not self._eof) and (self._lenbuf < size):
286 286 self._fillbuffer()
287 287 return self._frombuffer(size)
288 288
289 289 def readline(self, *args, **kwargs):
290 290 if 1 < len(self._buffer):
291 291 # this should not happen because both read and readline end with a
292 292 # _frombuffer call that collapse it.
293 293 self._buffer = [''.join(self._buffer)]
294 294 self._lenbuf = len(self._buffer[0])
295 295 lfi = -1
296 296 if self._buffer:
297 297 lfi = self._buffer[-1].find('\n')
298 298 while (not self._eof) and lfi < 0:
299 299 self._fillbuffer()
300 300 if self._buffer:
301 301 lfi = self._buffer[-1].find('\n')
302 302 size = lfi + 1
303 303 if lfi < 0: # end of file
304 304 size = self._lenbuf
305 305 elif 1 < len(self._buffer):
306 306 # we need to take previous chunks into account
307 307 size += self._lenbuf - len(self._buffer[-1])
308 308 return self._frombuffer(size)
309 309
310 310 def _frombuffer(self, size):
311 311 """return at most 'size' data from the buffer
312 312
313 313 The data are removed from the buffer."""
314 314 if size == 0 or not self._buffer:
315 315 return ''
316 316 buf = self._buffer[0]
317 317 if 1 < len(self._buffer):
318 318 buf = ''.join(self._buffer)
319 319
320 320 data = buf[:size]
321 321 buf = buf[len(data):]
322 322 if buf:
323 323 self._buffer = [buf]
324 324 self._lenbuf = len(buf)
325 325 else:
326 326 self._buffer = []
327 327 self._lenbuf = 0
328 328 return data
329 329
330 330 def _fillbuffer(self):
331 331 """read data to the buffer"""
332 332 data = os.read(self._input.fileno(), _chunksize)
333 333 if not data:
334 334 self._eof = True
335 335 else:
336 336 self._lenbuf += len(data)
337 337 self._buffer.append(data)
338 338
339 339 def popen2(cmd, env=None, newlines=False):
340 340 # Setting bufsize to -1 lets the system decide the buffer size.
341 341 # The default for bufsize is 0, meaning unbuffered. This leads to
342 342 # poor performance on Mac OS X: http://bugs.python.org/issue4194
343 343 p = subprocess.Popen(cmd, shell=True, bufsize=-1,
344 344 close_fds=closefds,
345 345 stdin=subprocess.PIPE, stdout=subprocess.PIPE,
346 346 universal_newlines=newlines,
347 347 env=env)
348 348 return p.stdin, p.stdout
349 349
350 350 def popen3(cmd, env=None, newlines=False):
351 351 stdin, stdout, stderr, p = popen4(cmd, env, newlines)
352 352 return stdin, stdout, stderr
353 353
354 354 def popen4(cmd, env=None, newlines=False, bufsize=-1):
355 355 p = subprocess.Popen(cmd, shell=True, bufsize=bufsize,
356 356 close_fds=closefds,
357 357 stdin=subprocess.PIPE, stdout=subprocess.PIPE,
358 358 stderr=subprocess.PIPE,
359 359 universal_newlines=newlines,
360 360 env=env)
361 361 return p.stdin, p.stdout, p.stderr, p
362 362
363 363 def version():
364 364 """Return version information if available."""
365 365 try:
366 366 from . import __version__
367 367 return __version__.version
368 368 except ImportError:
369 369 return 'unknown'
370 370
371 371 def versiontuple(v=None, n=4):
372 372 """Parses a Mercurial version string into an N-tuple.
373 373
374 374 The version string to be parsed is specified with the ``v`` argument.
375 375 If it isn't defined, the current Mercurial version string will be parsed.
376 376
377 377 ``n`` can be 2, 3, or 4. Here is how some version strings map to
378 378 returned values:
379 379
380 380 >>> v = '3.6.1+190-df9b73d2d444'
381 381 >>> versiontuple(v, 2)
382 382 (3, 6)
383 383 >>> versiontuple(v, 3)
384 384 (3, 6, 1)
385 385 >>> versiontuple(v, 4)
386 386 (3, 6, 1, '190-df9b73d2d444')
387 387
388 388 >>> versiontuple('3.6.1+190-df9b73d2d444+20151118')
389 389 (3, 6, 1, '190-df9b73d2d444+20151118')
390 390
391 391 >>> v = '3.6'
392 392 >>> versiontuple(v, 2)
393 393 (3, 6)
394 394 >>> versiontuple(v, 3)
395 395 (3, 6, None)
396 396 >>> versiontuple(v, 4)
397 397 (3, 6, None, None)
398 398
399 399 >>> v = '3.9-rc'
400 400 >>> versiontuple(v, 2)
401 401 (3, 9)
402 402 >>> versiontuple(v, 3)
403 403 (3, 9, None)
404 404 >>> versiontuple(v, 4)
405 405 (3, 9, None, 'rc')
406 406
407 407 >>> v = '3.9-rc+2-02a8fea4289b'
408 408 >>> versiontuple(v, 2)
409 409 (3, 9)
410 410 >>> versiontuple(v, 3)
411 411 (3, 9, None)
412 412 >>> versiontuple(v, 4)
413 413 (3, 9, None, 'rc+2-02a8fea4289b')
414 414 """
415 415 if not v:
416 416 v = version()
417 417 parts = remod.split('[\+-]', v, 1)
418 418 if len(parts) == 1:
419 419 vparts, extra = parts[0], None
420 420 else:
421 421 vparts, extra = parts
422 422
423 423 vints = []
424 424 for i in vparts.split('.'):
425 425 try:
426 426 vints.append(int(i))
427 427 except ValueError:
428 428 break
429 429 # (3, 6) -> (3, 6, None)
430 430 while len(vints) < 3:
431 431 vints.append(None)
432 432
433 433 if n == 2:
434 434 return (vints[0], vints[1])
435 435 if n == 3:
436 436 return (vints[0], vints[1], vints[2])
437 437 if n == 4:
438 438 return (vints[0], vints[1], vints[2], extra)
439 439
440 440 # used by parsedate
441 441 defaultdateformats = (
442 442 '%Y-%m-%dT%H:%M:%S', # the 'real' ISO8601
443 443 '%Y-%m-%dT%H:%M', # without seconds
444 444 '%Y-%m-%dT%H%M%S', # another awful but legal variant without :
445 445 '%Y-%m-%dT%H%M', # without seconds
446 446 '%Y-%m-%d %H:%M:%S', # our common legal variant
447 447 '%Y-%m-%d %H:%M', # without seconds
448 448 '%Y-%m-%d %H%M%S', # without :
449 449 '%Y-%m-%d %H%M', # without seconds
450 450 '%Y-%m-%d %I:%M:%S%p',
451 451 '%Y-%m-%d %H:%M',
452 452 '%Y-%m-%d %I:%M%p',
453 453 '%Y-%m-%d',
454 454 '%m-%d',
455 455 '%m/%d',
456 456 '%m/%d/%y',
457 457 '%m/%d/%Y',
458 458 '%a %b %d %H:%M:%S %Y',
459 459 '%a %b %d %I:%M:%S%p %Y',
460 460 '%a, %d %b %Y %H:%M:%S', # GNU coreutils "/bin/date --rfc-2822"
461 461 '%b %d %H:%M:%S %Y',
462 462 '%b %d %I:%M:%S%p %Y',
463 463 '%b %d %H:%M:%S',
464 464 '%b %d %I:%M:%S%p',
465 465 '%b %d %H:%M',
466 466 '%b %d %I:%M%p',
467 467 '%b %d %Y',
468 468 '%b %d',
469 469 '%H:%M:%S',
470 470 '%I:%M:%S%p',
471 471 '%H:%M',
472 472 '%I:%M%p',
473 473 )
474 474
475 475 extendeddateformats = defaultdateformats + (
476 476 "%Y",
477 477 "%Y-%m",
478 478 "%b",
479 479 "%b %Y",
480 480 )
481 481
482 482 def cachefunc(func):
483 483 '''cache the result of function calls'''
484 484 # XXX doesn't handle keywords args
485 485 if func.__code__.co_argcount == 0:
486 486 cache = []
487 487 def f():
488 488 if len(cache) == 0:
489 489 cache.append(func())
490 490 return cache[0]
491 491 return f
492 492 cache = {}
493 493 if func.__code__.co_argcount == 1:
494 494 # we gain a small amount of time because
495 495 # we don't need to pack/unpack the list
496 496 def f(arg):
497 497 if arg not in cache:
498 498 cache[arg] = func(arg)
499 499 return cache[arg]
500 500 else:
501 501 def f(*args):
502 502 if args not in cache:
503 503 cache[args] = func(*args)
504 504 return cache[args]
505 505
506 506 return f
507 507
508 508 class sortdict(dict):
509 509 '''a simple sorted dictionary'''
510 510 def __init__(self, data=None):
511 511 self._list = []
512 512 if data:
513 513 self.update(data)
514 514 def copy(self):
515 515 return sortdict(self)
516 516 def __setitem__(self, key, val):
517 517 if key in self:
518 518 self._list.remove(key)
519 519 self._list.append(key)
520 520 dict.__setitem__(self, key, val)
521 521 def __iter__(self):
522 522 return self._list.__iter__()
523 523 def update(self, src):
524 524 if isinstance(src, dict):
525 525 src = src.iteritems()
526 526 for k, v in src:
527 527 self[k] = v
528 528 def clear(self):
529 529 dict.clear(self)
530 530 self._list = []
531 531 def items(self):
532 532 return [(k, self[k]) for k in self._list]
533 533 def __delitem__(self, key):
534 534 dict.__delitem__(self, key)
535 535 self._list.remove(key)
536 536 def pop(self, key, *args, **kwargs):
537 537 dict.pop(self, key, *args, **kwargs)
538 538 try:
539 539 self._list.remove(key)
540 540 except ValueError:
541 541 pass
542 542 def keys(self):
543 543 return self._list
544 544 def iterkeys(self):
545 545 return self._list.__iter__()
546 546 def iteritems(self):
547 547 for k in self._list:
548 548 yield k, self[k]
549 549 def insert(self, index, key, val):
550 550 self._list.insert(index, key)
551 551 dict.__setitem__(self, key, val)
552 552 def __repr__(self):
553 553 if not self:
554 554 return '%s()' % self.__class__.__name__
555 555 return '%s(%r)' % (self.__class__.__name__, self.items())
556 556
557 557 class _lrucachenode(object):
558 558 """A node in a doubly linked list.
559 559
560 560 Holds a reference to nodes on either side as well as a key-value
561 561 pair for the dictionary entry.
562 562 """
563 563 __slots__ = (u'next', u'prev', u'key', u'value')
564 564
565 565 def __init__(self):
566 566 self.next = None
567 567 self.prev = None
568 568
569 569 self.key = _notset
570 570 self.value = None
571 571
572 572 def markempty(self):
573 573 """Mark the node as emptied."""
574 574 self.key = _notset
575 575
576 576 class lrucachedict(object):
577 577 """Dict that caches most recent accesses and sets.
578 578
579 579 The dict consists of an actual backing dict - indexed by original
580 580 key - and a doubly linked circular list defining the order of entries in
581 581 the cache.
582 582
583 583 The head node is the newest entry in the cache. If the cache is full,
584 584 we recycle head.prev and make it the new head. Cache accesses result in
585 585 the node being moved to before the existing head and being marked as the
586 586 new head node.
587 587 """
588 588 def __init__(self, max):
589 589 self._cache = {}
590 590
591 591 self._head = head = _lrucachenode()
592 592 head.prev = head
593 593 head.next = head
594 594 self._size = 1
595 595 self._capacity = max
596 596
597 597 def __len__(self):
598 598 return len(self._cache)
599 599
600 600 def __contains__(self, k):
601 601 return k in self._cache
602 602
603 603 def __iter__(self):
604 604 # We don't have to iterate in cache order, but why not.
605 605 n = self._head
606 606 for i in range(len(self._cache)):
607 607 yield n.key
608 608 n = n.next
609 609
610 610 def __getitem__(self, k):
611 611 node = self._cache[k]
612 612 self._movetohead(node)
613 613 return node.value
614 614
615 615 def __setitem__(self, k, v):
616 616 node = self._cache.get(k)
617 617 # Replace existing value and mark as newest.
618 618 if node is not None:
619 619 node.value = v
620 620 self._movetohead(node)
621 621 return
622 622
623 623 if self._size < self._capacity:
624 624 node = self._addcapacity()
625 625 else:
626 626 # Grab the last/oldest item.
627 627 node = self._head.prev
628 628
629 629 # At capacity. Kill the old entry.
630 630 if node.key is not _notset:
631 631 del self._cache[node.key]
632 632
633 633 node.key = k
634 634 node.value = v
635 635 self._cache[k] = node
636 636 # And mark it as newest entry. No need to adjust order since it
637 637 # is already self._head.prev.
638 638 self._head = node
639 639
640 640 def __delitem__(self, k):
641 641 node = self._cache.pop(k)
642 642 node.markempty()
643 643
644 644 # Temporarily mark as newest item before re-adjusting head to make
645 645 # this node the oldest item.
646 646 self._movetohead(node)
647 647 self._head = node.next
648 648
649 649 # Additional dict methods.
650 650
651 651 def get(self, k, default=None):
652 652 try:
653 653 return self._cache[k].value
654 654 except KeyError:
655 655 return default
656 656
657 657 def clear(self):
658 658 n = self._head
659 659 while n.key is not _notset:
660 660 n.markempty()
661 661 n = n.next
662 662
663 663 self._cache.clear()
664 664
665 665 def copy(self):
666 666 result = lrucachedict(self._capacity)
667 667 n = self._head.prev
668 668 # Iterate in oldest-to-newest order, so the copy has the right ordering
669 669 for i in range(len(self._cache)):
670 670 result[n.key] = n.value
671 671 n = n.prev
672 672 return result
673 673
674 674 def _movetohead(self, node):
675 675 """Mark a node as the newest, making it the new head.
676 676
677 677 When a node is accessed, it becomes the freshest entry in the LRU
678 678 list, which is denoted by self._head.
679 679
680 680 Visually, let's make ``N`` the new head node (* denotes head):
681 681
682 682 previous/oldest <-> head <-> next/next newest
683 683
684 684 ----<->--- A* ---<->-----
685 685 | |
686 686 E <-> D <-> N <-> C <-> B
687 687
688 688 To:
689 689
690 690 ----<->--- N* ---<->-----
691 691 | |
692 692 E <-> D <-> C <-> B <-> A
693 693
694 694 This requires the following moves:
695 695
696 696 C.next = D (node.prev.next = node.next)
697 697 D.prev = C (node.next.prev = node.prev)
698 698 E.next = N (head.prev.next = node)
699 699 N.prev = E (node.prev = head.prev)
700 700 N.next = A (node.next = head)
701 701 A.prev = N (head.prev = node)
702 702 """
703 703 head = self._head
704 704 # C.next = D
705 705 node.prev.next = node.next
706 706 # D.prev = C
707 707 node.next.prev = node.prev
708 708 # N.prev = E
709 709 node.prev = head.prev
710 710 # N.next = A
711 711 # It is tempting to do just "head" here, however if node is
712 712 # adjacent to head, this will do bad things.
713 713 node.next = head.prev.next
714 714 # E.next = N
715 715 node.next.prev = node
716 716 # A.prev = N
717 717 node.prev.next = node
718 718
719 719 self._head = node
720 720
721 721 def _addcapacity(self):
722 722 """Add a node to the circular linked list.
723 723
724 724 The new node is inserted before the head node.
725 725 """
726 726 head = self._head
727 727 node = _lrucachenode()
728 728 head.prev.next = node
729 729 node.prev = head.prev
730 730 node.next = head
731 731 head.prev = node
732 732 self._size += 1
733 733 return node
734 734
735 735 def lrucachefunc(func):
736 736 '''cache most recent results of function calls'''
737 737 cache = {}
738 738 order = collections.deque()
739 739 if func.__code__.co_argcount == 1:
740 740 def f(arg):
741 741 if arg not in cache:
742 742 if len(cache) > 20:
743 743 del cache[order.popleft()]
744 744 cache[arg] = func(arg)
745 745 else:
746 746 order.remove(arg)
747 747 order.append(arg)
748 748 return cache[arg]
749 749 else:
750 750 def f(*args):
751 751 if args not in cache:
752 752 if len(cache) > 20:
753 753 del cache[order.popleft()]
754 754 cache[args] = func(*args)
755 755 else:
756 756 order.remove(args)
757 757 order.append(args)
758 758 return cache[args]
759 759
760 760 return f
761 761
762 762 class propertycache(object):
763 763 def __init__(self, func):
764 764 self.func = func
765 765 self.name = func.__name__
766 766 def __get__(self, obj, type=None):
767 767 result = self.func(obj)
768 768 self.cachevalue(obj, result)
769 769 return result
770 770
771 771 def cachevalue(self, obj, value):
772 772 # __dict__ assignment required to bypass __setattr__ (eg: repoview)
773 773 obj.__dict__[self.name] = value
774 774
775 775 def pipefilter(s, cmd):
776 776 '''filter string S through command CMD, returning its output'''
777 777 p = subprocess.Popen(cmd, shell=True, close_fds=closefds,
778 778 stdin=subprocess.PIPE, stdout=subprocess.PIPE)
779 779 pout, perr = p.communicate(s)
780 780 return pout
781 781
782 782 def tempfilter(s, cmd):
783 783 '''filter string S through a pair of temporary files with CMD.
784 784 CMD is used as a template to create the real command to be run,
785 785 with the strings INFILE and OUTFILE replaced by the real names of
786 786 the temporary files generated.'''
787 787 inname, outname = None, None
788 788 try:
789 789 infd, inname = tempfile.mkstemp(prefix='hg-filter-in-')
790 790 fp = os.fdopen(infd, 'wb')
791 791 fp.write(s)
792 792 fp.close()
793 793 outfd, outname = tempfile.mkstemp(prefix='hg-filter-out-')
794 794 os.close(outfd)
795 795 cmd = cmd.replace('INFILE', inname)
796 796 cmd = cmd.replace('OUTFILE', outname)
797 797 code = os.system(cmd)
798 if sys.platform == 'OpenVMS' and code & 1:
798 if pycompat.sysplatform == 'OpenVMS' and code & 1:
799 799 code = 0
800 800 if code:
801 801 raise Abort(_("command '%s' failed: %s") %
802 802 (cmd, explainexit(code)))
803 803 return readfile(outname)
804 804 finally:
805 805 try:
806 806 if inname:
807 807 os.unlink(inname)
808 808 except OSError:
809 809 pass
810 810 try:
811 811 if outname:
812 812 os.unlink(outname)
813 813 except OSError:
814 814 pass
815 815
816 816 filtertable = {
817 817 'tempfile:': tempfilter,
818 818 'pipe:': pipefilter,
819 819 }
820 820
821 821 def filter(s, cmd):
822 822 "filter a string through a command that transforms its input to its output"
823 823 for name, fn in filtertable.iteritems():
824 824 if cmd.startswith(name):
825 825 return fn(s, cmd[len(name):].lstrip())
826 826 return pipefilter(s, cmd)
827 827
828 828 def binary(s):
829 829 """return true if a string is binary data"""
830 830 return bool(s and '\0' in s)
831 831
832 832 def increasingchunks(source, min=1024, max=65536):
833 833 '''return no less than min bytes per chunk while data remains,
834 834 doubling min after each chunk until it reaches max'''
835 835 def log2(x):
836 836 if not x:
837 837 return 0
838 838 i = 0
839 839 while x:
840 840 x >>= 1
841 841 i += 1
842 842 return i - 1
843 843
844 844 buf = []
845 845 blen = 0
846 846 for chunk in source:
847 847 buf.append(chunk)
848 848 blen += len(chunk)
849 849 if blen >= min:
850 850 if min < max:
851 851 min = min << 1
852 852 nmin = 1 << log2(blen)
853 853 if nmin > min:
854 854 min = nmin
855 855 if min > max:
856 856 min = max
857 857 yield ''.join(buf)
858 858 blen = 0
859 859 buf = []
860 860 if buf:
861 861 yield ''.join(buf)
862 862
863 863 Abort = error.Abort
864 864
865 865 def always(fn):
866 866 return True
867 867
868 868 def never(fn):
869 869 return False
870 870
871 871 def nogc(func):
872 872 """disable garbage collector
873 873
874 874 Python's garbage collector triggers a GC each time a certain number of
875 875 container objects (the number being defined by gc.get_threshold()) are
876 876 allocated even when marked not to be tracked by the collector. Tracking has
877 877 no effect on when GCs are triggered, only on what objects the GC looks
878 878 into. As a workaround, disable GC while building complex (huge)
879 879 containers.
880 880
881 881 This garbage collector issue have been fixed in 2.7.
882 882 """
883 883 if sys.version_info >= (2, 7):
884 884 return func
885 885 def wrapper(*args, **kwargs):
886 886 gcenabled = gc.isenabled()
887 887 gc.disable()
888 888 try:
889 889 return func(*args, **kwargs)
890 890 finally:
891 891 if gcenabled:
892 892 gc.enable()
893 893 return wrapper
894 894
895 895 def pathto(root, n1, n2):
896 896 '''return the relative path from one place to another.
897 897 root should use os.sep to separate directories
898 898 n1 should use os.sep to separate directories
899 899 n2 should use "/" to separate directories
900 900 returns an os.sep-separated path.
901 901
902 902 If n1 is a relative path, it's assumed it's
903 903 relative to root.
904 904 n2 should always be relative to root.
905 905 '''
906 906 if not n1:
907 907 return localpath(n2)
908 908 if os.path.isabs(n1):
909 909 if os.path.splitdrive(root)[0] != os.path.splitdrive(n1)[0]:
910 910 return os.path.join(root, localpath(n2))
911 911 n2 = '/'.join((pconvert(root), n2))
912 912 a, b = splitpath(n1), n2.split('/')
913 913 a.reverse()
914 914 b.reverse()
915 915 while a and b and a[-1] == b[-1]:
916 916 a.pop()
917 917 b.pop()
918 918 b.reverse()
919 919 return pycompat.ossep.join((['..'] * len(a)) + b) or '.'
920 920
921 921 def mainfrozen():
922 922 """return True if we are a frozen executable.
923 923
924 924 The code supports py2exe (most common, Windows only) and tools/freeze
925 925 (portable, not much used).
926 926 """
927 927 return (safehasattr(sys, "frozen") or # new py2exe
928 928 safehasattr(sys, "importers") or # old py2exe
929 929 imp.is_frozen(u"__main__")) # tools/freeze
930 930
931 931 # the location of data files matching the source code
932 932 if mainfrozen() and getattr(sys, 'frozen', None) != 'macosx_app':
933 933 # executable version (py2exe) doesn't support __file__
934 934 datapath = os.path.dirname(sys.executable)
935 935 else:
936 936 datapath = os.path.dirname(__file__)
937 937
938 938 if not isinstance(datapath, bytes):
939 939 datapath = pycompat.fsencode(datapath)
940 940
941 941 i18n.setdatapath(datapath)
942 942
943 943 _hgexecutable = None
944 944
945 945 def hgexecutable():
946 946 """return location of the 'hg' executable.
947 947
948 948 Defaults to $HG or 'hg' in the search path.
949 949 """
950 950 if _hgexecutable is None:
951 951 hg = encoding.environ.get('HG')
952 952 mainmod = sys.modules['__main__']
953 953 if hg:
954 954 _sethgexecutable(hg)
955 955 elif mainfrozen():
956 956 if getattr(sys, 'frozen', None) == 'macosx_app':
957 957 # Env variable set by py2app
958 958 _sethgexecutable(encoding.environ['EXECUTABLEPATH'])
959 959 else:
960 960 _sethgexecutable(sys.executable)
961 961 elif os.path.basename(getattr(mainmod, '__file__', '')) == 'hg':
962 962 _sethgexecutable(mainmod.__file__)
963 963 else:
964 964 exe = findexe('hg') or os.path.basename(sys.argv[0])
965 965 _sethgexecutable(exe)
966 966 return _hgexecutable
967 967
968 968 def _sethgexecutable(path):
969 969 """set location of the 'hg' executable"""
970 970 global _hgexecutable
971 971 _hgexecutable = path
972 972
973 973 def _isstdout(f):
974 974 fileno = getattr(f, 'fileno', None)
975 975 return fileno and fileno() == sys.__stdout__.fileno()
976 976
977 977 def system(cmd, environ=None, cwd=None, onerr=None, errprefix=None, out=None):
978 978 '''enhanced shell command execution.
979 979 run with environment maybe modified, maybe in different dir.
980 980
981 981 if command fails and onerr is None, return status, else raise onerr
982 982 object as exception.
983 983
984 984 if out is specified, it is assumed to be a file-like object that has a
985 985 write() method. stdout and stderr will be redirected to out.'''
986 986 if environ is None:
987 987 environ = {}
988 988 try:
989 989 stdout.flush()
990 990 except Exception:
991 991 pass
992 992 def py2shell(val):
993 993 'convert python object into string that is useful to shell'
994 994 if val is None or val is False:
995 995 return '0'
996 996 if val is True:
997 997 return '1'
998 998 return str(val)
999 999 origcmd = cmd
1000 1000 cmd = quotecommand(cmd)
1001 if sys.platform == 'plan9' and (sys.version_info[0] == 2
1001 if pycompat.sysplatform == 'plan9' and (sys.version_info[0] == 2
1002 1002 and sys.version_info[1] < 7):
1003 1003 # subprocess kludge to work around issues in half-baked Python
1004 1004 # ports, notably bichued/python:
1005 1005 if not cwd is None:
1006 1006 os.chdir(cwd)
1007 1007 rc = os.system(cmd)
1008 1008 else:
1009 1009 env = dict(encoding.environ)
1010 1010 env.update((k, py2shell(v)) for k, v in environ.iteritems())
1011 1011 env['HG'] = hgexecutable()
1012 1012 if out is None or _isstdout(out):
1013 1013 rc = subprocess.call(cmd, shell=True, close_fds=closefds,
1014 1014 env=env, cwd=cwd)
1015 1015 else:
1016 1016 proc = subprocess.Popen(cmd, shell=True, close_fds=closefds,
1017 1017 env=env, cwd=cwd, stdout=subprocess.PIPE,
1018 1018 stderr=subprocess.STDOUT)
1019 1019 for line in iter(proc.stdout.readline, ''):
1020 1020 out.write(line)
1021 1021 proc.wait()
1022 1022 rc = proc.returncode
1023 if sys.platform == 'OpenVMS' and rc & 1:
1023 if pycompat.sysplatform == 'OpenVMS' and rc & 1:
1024 1024 rc = 0
1025 1025 if rc and onerr:
1026 1026 errmsg = '%s %s' % (os.path.basename(origcmd.split(None, 1)[0]),
1027 1027 explainexit(rc)[0])
1028 1028 if errprefix:
1029 1029 errmsg = '%s: %s' % (errprefix, errmsg)
1030 1030 raise onerr(errmsg)
1031 1031 return rc
1032 1032
1033 1033 def checksignature(func):
1034 1034 '''wrap a function with code to check for calling errors'''
1035 1035 def check(*args, **kwargs):
1036 1036 try:
1037 1037 return func(*args, **kwargs)
1038 1038 except TypeError:
1039 1039 if len(traceback.extract_tb(sys.exc_info()[2])) == 1:
1040 1040 raise error.SignatureError
1041 1041 raise
1042 1042
1043 1043 return check
1044 1044
1045 1045 def copyfile(src, dest, hardlink=False, copystat=False, checkambig=False):
1046 1046 '''copy a file, preserving mode and optionally other stat info like
1047 1047 atime/mtime
1048 1048
1049 1049 checkambig argument is used with filestat, and is useful only if
1050 1050 destination file is guarded by any lock (e.g. repo.lock or
1051 1051 repo.wlock).
1052 1052
1053 1053 copystat and checkambig should be exclusive.
1054 1054 '''
1055 1055 assert not (copystat and checkambig)
1056 1056 oldstat = None
1057 1057 if os.path.lexists(dest):
1058 1058 if checkambig:
1059 1059 oldstat = checkambig and filestat(dest)
1060 1060 unlink(dest)
1061 1061 # hardlinks are problematic on CIFS, quietly ignore this flag
1062 1062 # until we find a way to work around it cleanly (issue4546)
1063 1063 if False and hardlink:
1064 1064 try:
1065 1065 oslink(src, dest)
1066 1066 return
1067 1067 except (IOError, OSError):
1068 1068 pass # fall back to normal copy
1069 1069 if os.path.islink(src):
1070 1070 os.symlink(os.readlink(src), dest)
1071 1071 # copytime is ignored for symlinks, but in general copytime isn't needed
1072 1072 # for them anyway
1073 1073 else:
1074 1074 try:
1075 1075 shutil.copyfile(src, dest)
1076 1076 if copystat:
1077 1077 # copystat also copies mode
1078 1078 shutil.copystat(src, dest)
1079 1079 else:
1080 1080 shutil.copymode(src, dest)
1081 1081 if oldstat and oldstat.stat:
1082 1082 newstat = filestat(dest)
1083 1083 if newstat.isambig(oldstat):
1084 1084 # stat of copied file is ambiguous to original one
1085 1085 advanced = (oldstat.stat.st_mtime + 1) & 0x7fffffff
1086 1086 os.utime(dest, (advanced, advanced))
1087 1087 except shutil.Error as inst:
1088 1088 raise Abort(str(inst))
1089 1089
1090 1090 def copyfiles(src, dst, hardlink=None, progress=lambda t, pos: None):
1091 1091 """Copy a directory tree using hardlinks if possible."""
1092 1092 num = 0
1093 1093
1094 1094 if hardlink is None:
1095 1095 hardlink = (os.stat(src).st_dev ==
1096 1096 os.stat(os.path.dirname(dst)).st_dev)
1097 1097 if hardlink:
1098 1098 topic = _('linking')
1099 1099 else:
1100 1100 topic = _('copying')
1101 1101
1102 1102 if os.path.isdir(src):
1103 1103 os.mkdir(dst)
1104 1104 for name, kind in osutil.listdir(src):
1105 1105 srcname = os.path.join(src, name)
1106 1106 dstname = os.path.join(dst, name)
1107 1107 def nprog(t, pos):
1108 1108 if pos is not None:
1109 1109 return progress(t, pos + num)
1110 1110 hardlink, n = copyfiles(srcname, dstname, hardlink, progress=nprog)
1111 1111 num += n
1112 1112 else:
1113 1113 if hardlink:
1114 1114 try:
1115 1115 oslink(src, dst)
1116 1116 except (IOError, OSError):
1117 1117 hardlink = False
1118 1118 shutil.copy(src, dst)
1119 1119 else:
1120 1120 shutil.copy(src, dst)
1121 1121 num += 1
1122 1122 progress(topic, num)
1123 1123 progress(topic, None)
1124 1124
1125 1125 return hardlink, num
1126 1126
1127 1127 _winreservednames = '''con prn aux nul
1128 1128 com1 com2 com3 com4 com5 com6 com7 com8 com9
1129 1129 lpt1 lpt2 lpt3 lpt4 lpt5 lpt6 lpt7 lpt8 lpt9'''.split()
1130 1130 _winreservedchars = ':*?"<>|'
1131 1131 def checkwinfilename(path):
1132 1132 r'''Check that the base-relative path is a valid filename on Windows.
1133 1133 Returns None if the path is ok, or a UI string describing the problem.
1134 1134
1135 1135 >>> checkwinfilename("just/a/normal/path")
1136 1136 >>> checkwinfilename("foo/bar/con.xml")
1137 1137 "filename contains 'con', which is reserved on Windows"
1138 1138 >>> checkwinfilename("foo/con.xml/bar")
1139 1139 "filename contains 'con', which is reserved on Windows"
1140 1140 >>> checkwinfilename("foo/bar/xml.con")
1141 1141 >>> checkwinfilename("foo/bar/AUX/bla.txt")
1142 1142 "filename contains 'AUX', which is reserved on Windows"
1143 1143 >>> checkwinfilename("foo/bar/bla:.txt")
1144 1144 "filename contains ':', which is reserved on Windows"
1145 1145 >>> checkwinfilename("foo/bar/b\07la.txt")
1146 1146 "filename contains '\\x07', which is invalid on Windows"
1147 1147 >>> checkwinfilename("foo/bar/bla ")
1148 1148 "filename ends with ' ', which is not allowed on Windows"
1149 1149 >>> checkwinfilename("../bar")
1150 1150 >>> checkwinfilename("foo\\")
1151 1151 "filename ends with '\\', which is invalid on Windows"
1152 1152 >>> checkwinfilename("foo\\/bar")
1153 1153 "directory name ends with '\\', which is invalid on Windows"
1154 1154 '''
1155 1155 if path.endswith('\\'):
1156 1156 return _("filename ends with '\\', which is invalid on Windows")
1157 1157 if '\\/' in path:
1158 1158 return _("directory name ends with '\\', which is invalid on Windows")
1159 1159 for n in path.replace('\\', '/').split('/'):
1160 1160 if not n:
1161 1161 continue
1162 1162 for c in n:
1163 1163 if c in _winreservedchars:
1164 1164 return _("filename contains '%s', which is reserved "
1165 1165 "on Windows") % c
1166 1166 if ord(c) <= 31:
1167 1167 return _("filename contains %r, which is invalid "
1168 1168 "on Windows") % c
1169 1169 base = n.split('.')[0]
1170 1170 if base and base.lower() in _winreservednames:
1171 1171 return _("filename contains '%s', which is reserved "
1172 1172 "on Windows") % base
1173 1173 t = n[-1]
1174 1174 if t in '. ' and n not in '..':
1175 1175 return _("filename ends with '%s', which is not allowed "
1176 1176 "on Windows") % t
1177 1177
1178 1178 if pycompat.osname == 'nt':
1179 1179 checkosfilename = checkwinfilename
1180 1180 else:
1181 1181 checkosfilename = platform.checkosfilename
1182 1182
1183 1183 def makelock(info, pathname):
1184 1184 try:
1185 1185 return os.symlink(info, pathname)
1186 1186 except OSError as why:
1187 1187 if why.errno == errno.EEXIST:
1188 1188 raise
1189 1189 except AttributeError: # no symlink in os
1190 1190 pass
1191 1191
1192 1192 ld = os.open(pathname, os.O_CREAT | os.O_WRONLY | os.O_EXCL)
1193 1193 os.write(ld, info)
1194 1194 os.close(ld)
1195 1195
1196 1196 def readlock(pathname):
1197 1197 try:
1198 1198 return os.readlink(pathname)
1199 1199 except OSError as why:
1200 1200 if why.errno not in (errno.EINVAL, errno.ENOSYS):
1201 1201 raise
1202 1202 except AttributeError: # no symlink in os
1203 1203 pass
1204 1204 fp = posixfile(pathname)
1205 1205 r = fp.read()
1206 1206 fp.close()
1207 1207 return r
1208 1208
1209 1209 def fstat(fp):
1210 1210 '''stat file object that may not have fileno method.'''
1211 1211 try:
1212 1212 return os.fstat(fp.fileno())
1213 1213 except AttributeError:
1214 1214 return os.stat(fp.name)
1215 1215
1216 1216 # File system features
1217 1217
1218 1218 def fscasesensitive(path):
1219 1219 """
1220 1220 Return true if the given path is on a case-sensitive filesystem
1221 1221
1222 1222 Requires a path (like /foo/.hg) ending with a foldable final
1223 1223 directory component.
1224 1224 """
1225 1225 s1 = os.lstat(path)
1226 1226 d, b = os.path.split(path)
1227 1227 b2 = b.upper()
1228 1228 if b == b2:
1229 1229 b2 = b.lower()
1230 1230 if b == b2:
1231 1231 return True # no evidence against case sensitivity
1232 1232 p2 = os.path.join(d, b2)
1233 1233 try:
1234 1234 s2 = os.lstat(p2)
1235 1235 if s2 == s1:
1236 1236 return False
1237 1237 return True
1238 1238 except OSError:
1239 1239 return True
1240 1240
1241 1241 try:
1242 1242 import re2
1243 1243 _re2 = None
1244 1244 except ImportError:
1245 1245 _re2 = False
1246 1246
1247 1247 class _re(object):
1248 1248 def _checkre2(self):
1249 1249 global _re2
1250 1250 try:
1251 1251 # check if match works, see issue3964
1252 1252 _re2 = bool(re2.match(r'\[([^\[]+)\]', '[ui]'))
1253 1253 except ImportError:
1254 1254 _re2 = False
1255 1255
1256 1256 def compile(self, pat, flags=0):
1257 1257 '''Compile a regular expression, using re2 if possible
1258 1258
1259 1259 For best performance, use only re2-compatible regexp features. The
1260 1260 only flags from the re module that are re2-compatible are
1261 1261 IGNORECASE and MULTILINE.'''
1262 1262 if _re2 is None:
1263 1263 self._checkre2()
1264 1264 if _re2 and (flags & ~(remod.IGNORECASE | remod.MULTILINE)) == 0:
1265 1265 if flags & remod.IGNORECASE:
1266 1266 pat = '(?i)' + pat
1267 1267 if flags & remod.MULTILINE:
1268 1268 pat = '(?m)' + pat
1269 1269 try:
1270 1270 return re2.compile(pat)
1271 1271 except re2.error:
1272 1272 pass
1273 1273 return remod.compile(pat, flags)
1274 1274
1275 1275 @propertycache
1276 1276 def escape(self):
1277 1277 '''Return the version of escape corresponding to self.compile.
1278 1278
1279 1279 This is imperfect because whether re2 or re is used for a particular
1280 1280 function depends on the flags, etc, but it's the best we can do.
1281 1281 '''
1282 1282 global _re2
1283 1283 if _re2 is None:
1284 1284 self._checkre2()
1285 1285 if _re2:
1286 1286 return re2.escape
1287 1287 else:
1288 1288 return remod.escape
1289 1289
1290 1290 re = _re()
1291 1291
1292 1292 _fspathcache = {}
1293 1293 def fspath(name, root):
1294 1294 '''Get name in the case stored in the filesystem
1295 1295
1296 1296 The name should be relative to root, and be normcase-ed for efficiency.
1297 1297
1298 1298 Note that this function is unnecessary, and should not be
1299 1299 called, for case-sensitive filesystems (simply because it's expensive).
1300 1300
1301 1301 The root should be normcase-ed, too.
1302 1302 '''
1303 1303 def _makefspathcacheentry(dir):
1304 1304 return dict((normcase(n), n) for n in os.listdir(dir))
1305 1305
1306 1306 seps = pycompat.ossep
1307 1307 if pycompat.osaltsep:
1308 1308 seps = seps + pycompat.osaltsep
1309 1309 # Protect backslashes. This gets silly very quickly.
1310 1310 seps.replace('\\','\\\\')
1311 1311 pattern = remod.compile(r'([^%s]+)|([%s]+)' % (seps, seps))
1312 1312 dir = os.path.normpath(root)
1313 1313 result = []
1314 1314 for part, sep in pattern.findall(name):
1315 1315 if sep:
1316 1316 result.append(sep)
1317 1317 continue
1318 1318
1319 1319 if dir not in _fspathcache:
1320 1320 _fspathcache[dir] = _makefspathcacheentry(dir)
1321 1321 contents = _fspathcache[dir]
1322 1322
1323 1323 found = contents.get(part)
1324 1324 if not found:
1325 1325 # retry "once per directory" per "dirstate.walk" which
1326 1326 # may take place for each patches of "hg qpush", for example
1327 1327 _fspathcache[dir] = contents = _makefspathcacheentry(dir)
1328 1328 found = contents.get(part)
1329 1329
1330 1330 result.append(found or part)
1331 1331 dir = os.path.join(dir, part)
1332 1332
1333 1333 return ''.join(result)
1334 1334
1335 1335 def checknlink(testfile):
1336 1336 '''check whether hardlink count reporting works properly'''
1337 1337
1338 1338 # testfile may be open, so we need a separate file for checking to
1339 1339 # work around issue2543 (or testfile may get lost on Samba shares)
1340 1340 f1 = testfile + ".hgtmp1"
1341 1341 if os.path.lexists(f1):
1342 1342 return False
1343 1343 try:
1344 1344 posixfile(f1, 'w').close()
1345 1345 except IOError:
1346 1346 try:
1347 1347 os.unlink(f1)
1348 1348 except OSError:
1349 1349 pass
1350 1350 return False
1351 1351
1352 1352 f2 = testfile + ".hgtmp2"
1353 1353 fd = None
1354 1354 try:
1355 1355 oslink(f1, f2)
1356 1356 # nlinks() may behave differently for files on Windows shares if
1357 1357 # the file is open.
1358 1358 fd = posixfile(f2)
1359 1359 return nlinks(f2) > 1
1360 1360 except OSError:
1361 1361 return False
1362 1362 finally:
1363 1363 if fd is not None:
1364 1364 fd.close()
1365 1365 for f in (f1, f2):
1366 1366 try:
1367 1367 os.unlink(f)
1368 1368 except OSError:
1369 1369 pass
1370 1370
1371 1371 def endswithsep(path):
1372 1372 '''Check path ends with os.sep or os.altsep.'''
1373 1373 return (path.endswith(pycompat.ossep)
1374 1374 or pycompat.osaltsep and path.endswith(pycompat.osaltsep))
1375 1375
1376 1376 def splitpath(path):
1377 1377 '''Split path by os.sep.
1378 1378 Note that this function does not use os.altsep because this is
1379 1379 an alternative of simple "xxx.split(os.sep)".
1380 1380 It is recommended to use os.path.normpath() before using this
1381 1381 function if need.'''
1382 1382 return path.split(pycompat.ossep)
1383 1383
1384 1384 def gui():
1385 1385 '''Are we running in a GUI?'''
1386 if sys.platform == 'darwin':
1386 if pycompat.sysplatform == 'darwin':
1387 1387 if 'SSH_CONNECTION' in encoding.environ:
1388 1388 # handle SSH access to a box where the user is logged in
1389 1389 return False
1390 1390 elif getattr(osutil, 'isgui', None):
1391 1391 # check if a CoreGraphics session is available
1392 1392 return osutil.isgui()
1393 1393 else:
1394 1394 # pure build; use a safe default
1395 1395 return True
1396 1396 else:
1397 1397 return pycompat.osname == "nt" or encoding.environ.get("DISPLAY")
1398 1398
1399 1399 def mktempcopy(name, emptyok=False, createmode=None):
1400 1400 """Create a temporary file with the same contents from name
1401 1401
1402 1402 The permission bits are copied from the original file.
1403 1403
1404 1404 If the temporary file is going to be truncated immediately, you
1405 1405 can use emptyok=True as an optimization.
1406 1406
1407 1407 Returns the name of the temporary file.
1408 1408 """
1409 1409 d, fn = os.path.split(name)
1410 1410 fd, temp = tempfile.mkstemp(prefix='.%s-' % fn, dir=d)
1411 1411 os.close(fd)
1412 1412 # Temporary files are created with mode 0600, which is usually not
1413 1413 # what we want. If the original file already exists, just copy
1414 1414 # its mode. Otherwise, manually obey umask.
1415 1415 copymode(name, temp, createmode)
1416 1416 if emptyok:
1417 1417 return temp
1418 1418 try:
1419 1419 try:
1420 1420 ifp = posixfile(name, "rb")
1421 1421 except IOError as inst:
1422 1422 if inst.errno == errno.ENOENT:
1423 1423 return temp
1424 1424 if not getattr(inst, 'filename', None):
1425 1425 inst.filename = name
1426 1426 raise
1427 1427 ofp = posixfile(temp, "wb")
1428 1428 for chunk in filechunkiter(ifp):
1429 1429 ofp.write(chunk)
1430 1430 ifp.close()
1431 1431 ofp.close()
1432 1432 except: # re-raises
1433 1433 try: os.unlink(temp)
1434 1434 except OSError: pass
1435 1435 raise
1436 1436 return temp
1437 1437
1438 1438 class filestat(object):
1439 1439 """help to exactly detect change of a file
1440 1440
1441 1441 'stat' attribute is result of 'os.stat()' if specified 'path'
1442 1442 exists. Otherwise, it is None. This can avoid preparative
1443 1443 'exists()' examination on client side of this class.
1444 1444 """
1445 1445 def __init__(self, path):
1446 1446 try:
1447 1447 self.stat = os.stat(path)
1448 1448 except OSError as err:
1449 1449 if err.errno != errno.ENOENT:
1450 1450 raise
1451 1451 self.stat = None
1452 1452
1453 1453 __hash__ = object.__hash__
1454 1454
1455 1455 def __eq__(self, old):
1456 1456 try:
1457 1457 # if ambiguity between stat of new and old file is
1458 1458 # avoided, comparison of size, ctime and mtime is enough
1459 1459 # to exactly detect change of a file regardless of platform
1460 1460 return (self.stat.st_size == old.stat.st_size and
1461 1461 self.stat.st_ctime == old.stat.st_ctime and
1462 1462 self.stat.st_mtime == old.stat.st_mtime)
1463 1463 except AttributeError:
1464 1464 return False
1465 1465
1466 1466 def isambig(self, old):
1467 1467 """Examine whether new (= self) stat is ambiguous against old one
1468 1468
1469 1469 "S[N]" below means stat of a file at N-th change:
1470 1470
1471 1471 - S[n-1].ctime < S[n].ctime: can detect change of a file
1472 1472 - S[n-1].ctime == S[n].ctime
1473 1473 - S[n-1].ctime < S[n].mtime: means natural advancing (*1)
1474 1474 - S[n-1].ctime == S[n].mtime: is ambiguous (*2)
1475 1475 - S[n-1].ctime > S[n].mtime: never occurs naturally (don't care)
1476 1476 - S[n-1].ctime > S[n].ctime: never occurs naturally (don't care)
1477 1477
1478 1478 Case (*2) above means that a file was changed twice or more at
1479 1479 same time in sec (= S[n-1].ctime), and comparison of timestamp
1480 1480 is ambiguous.
1481 1481
1482 1482 Base idea to avoid such ambiguity is "advance mtime 1 sec, if
1483 1483 timestamp is ambiguous".
1484 1484
1485 1485 But advancing mtime only in case (*2) doesn't work as
1486 1486 expected, because naturally advanced S[n].mtime in case (*1)
1487 1487 might be equal to manually advanced S[n-1 or earlier].mtime.
1488 1488
1489 1489 Therefore, all "S[n-1].ctime == S[n].ctime" cases should be
1490 1490 treated as ambiguous regardless of mtime, to avoid overlooking
1491 1491 by confliction between such mtime.
1492 1492
1493 1493 Advancing mtime "if isambig(oldstat)" ensures "S[n-1].mtime !=
1494 1494 S[n].mtime", even if size of a file isn't changed.
1495 1495 """
1496 1496 try:
1497 1497 return (self.stat.st_ctime == old.stat.st_ctime)
1498 1498 except AttributeError:
1499 1499 return False
1500 1500
1501 1501 def avoidambig(self, path, old):
1502 1502 """Change file stat of specified path to avoid ambiguity
1503 1503
1504 1504 'old' should be previous filestat of 'path'.
1505 1505
1506 1506 This skips avoiding ambiguity, if a process doesn't have
1507 1507 appropriate privileges for 'path'.
1508 1508 """
1509 1509 advanced = (old.stat.st_mtime + 1) & 0x7fffffff
1510 1510 try:
1511 1511 os.utime(path, (advanced, advanced))
1512 1512 except OSError as inst:
1513 1513 if inst.errno == errno.EPERM:
1514 1514 # utime() on the file created by another user causes EPERM,
1515 1515 # if a process doesn't have appropriate privileges
1516 1516 return
1517 1517 raise
1518 1518
1519 1519 def __ne__(self, other):
1520 1520 return not self == other
1521 1521
1522 1522 class atomictempfile(object):
1523 1523 '''writable file object that atomically updates a file
1524 1524
1525 1525 All writes will go to a temporary copy of the original file. Call
1526 1526 close() when you are done writing, and atomictempfile will rename
1527 1527 the temporary copy to the original name, making the changes
1528 1528 visible. If the object is destroyed without being closed, all your
1529 1529 writes are discarded.
1530 1530
1531 1531 checkambig argument of constructor is used with filestat, and is
1532 1532 useful only if target file is guarded by any lock (e.g. repo.lock
1533 1533 or repo.wlock).
1534 1534 '''
1535 1535 def __init__(self, name, mode='w+b', createmode=None, checkambig=False):
1536 1536 self.__name = name # permanent name
1537 1537 self._tempname = mktempcopy(name, emptyok=('w' in mode),
1538 1538 createmode=createmode)
1539 1539 self._fp = posixfile(self._tempname, mode)
1540 1540 self._checkambig = checkambig
1541 1541
1542 1542 # delegated methods
1543 1543 self.read = self._fp.read
1544 1544 self.write = self._fp.write
1545 1545 self.seek = self._fp.seek
1546 1546 self.tell = self._fp.tell
1547 1547 self.fileno = self._fp.fileno
1548 1548
1549 1549 def close(self):
1550 1550 if not self._fp.closed:
1551 1551 self._fp.close()
1552 1552 filename = localpath(self.__name)
1553 1553 oldstat = self._checkambig and filestat(filename)
1554 1554 if oldstat and oldstat.stat:
1555 1555 rename(self._tempname, filename)
1556 1556 newstat = filestat(filename)
1557 1557 if newstat.isambig(oldstat):
1558 1558 # stat of changed file is ambiguous to original one
1559 1559 advanced = (oldstat.stat.st_mtime + 1) & 0x7fffffff
1560 1560 os.utime(filename, (advanced, advanced))
1561 1561 else:
1562 1562 rename(self._tempname, filename)
1563 1563
1564 1564 def discard(self):
1565 1565 if not self._fp.closed:
1566 1566 try:
1567 1567 os.unlink(self._tempname)
1568 1568 except OSError:
1569 1569 pass
1570 1570 self._fp.close()
1571 1571
1572 1572 def __del__(self):
1573 1573 if safehasattr(self, '_fp'): # constructor actually did something
1574 1574 self.discard()
1575 1575
1576 1576 def __enter__(self):
1577 1577 return self
1578 1578
1579 1579 def __exit__(self, exctype, excvalue, traceback):
1580 1580 if exctype is not None:
1581 1581 self.discard()
1582 1582 else:
1583 1583 self.close()
1584 1584
1585 1585 def makedirs(name, mode=None, notindexed=False):
1586 1586 """recursive directory creation with parent mode inheritance
1587 1587
1588 1588 Newly created directories are marked as "not to be indexed by
1589 1589 the content indexing service", if ``notindexed`` is specified
1590 1590 for "write" mode access.
1591 1591 """
1592 1592 try:
1593 1593 makedir(name, notindexed)
1594 1594 except OSError as err:
1595 1595 if err.errno == errno.EEXIST:
1596 1596 return
1597 1597 if err.errno != errno.ENOENT or not name:
1598 1598 raise
1599 1599 parent = os.path.dirname(os.path.abspath(name))
1600 1600 if parent == name:
1601 1601 raise
1602 1602 makedirs(parent, mode, notindexed)
1603 1603 try:
1604 1604 makedir(name, notindexed)
1605 1605 except OSError as err:
1606 1606 # Catch EEXIST to handle races
1607 1607 if err.errno == errno.EEXIST:
1608 1608 return
1609 1609 raise
1610 1610 if mode is not None:
1611 1611 os.chmod(name, mode)
1612 1612
1613 1613 def readfile(path):
1614 1614 with open(path, 'rb') as fp:
1615 1615 return fp.read()
1616 1616
1617 1617 def writefile(path, text):
1618 1618 with open(path, 'wb') as fp:
1619 1619 fp.write(text)
1620 1620
1621 1621 def appendfile(path, text):
1622 1622 with open(path, 'ab') as fp:
1623 1623 fp.write(text)
1624 1624
1625 1625 class chunkbuffer(object):
1626 1626 """Allow arbitrary sized chunks of data to be efficiently read from an
1627 1627 iterator over chunks of arbitrary size."""
1628 1628
1629 1629 def __init__(self, in_iter):
1630 1630 """in_iter is the iterator that's iterating over the input chunks.
1631 1631 targetsize is how big a buffer to try to maintain."""
1632 1632 def splitbig(chunks):
1633 1633 for chunk in chunks:
1634 1634 if len(chunk) > 2**20:
1635 1635 pos = 0
1636 1636 while pos < len(chunk):
1637 1637 end = pos + 2 ** 18
1638 1638 yield chunk[pos:end]
1639 1639 pos = end
1640 1640 else:
1641 1641 yield chunk
1642 1642 self.iter = splitbig(in_iter)
1643 1643 self._queue = collections.deque()
1644 1644 self._chunkoffset = 0
1645 1645
1646 1646 def read(self, l=None):
1647 1647 """Read L bytes of data from the iterator of chunks of data.
1648 1648 Returns less than L bytes if the iterator runs dry.
1649 1649
1650 1650 If size parameter is omitted, read everything"""
1651 1651 if l is None:
1652 1652 return ''.join(self.iter)
1653 1653
1654 1654 left = l
1655 1655 buf = []
1656 1656 queue = self._queue
1657 1657 while left > 0:
1658 1658 # refill the queue
1659 1659 if not queue:
1660 1660 target = 2**18
1661 1661 for chunk in self.iter:
1662 1662 queue.append(chunk)
1663 1663 target -= len(chunk)
1664 1664 if target <= 0:
1665 1665 break
1666 1666 if not queue:
1667 1667 break
1668 1668
1669 1669 # The easy way to do this would be to queue.popleft(), modify the
1670 1670 # chunk (if necessary), then queue.appendleft(). However, for cases
1671 1671 # where we read partial chunk content, this incurs 2 dequeue
1672 1672 # mutations and creates a new str for the remaining chunk in the
1673 1673 # queue. Our code below avoids this overhead.
1674 1674
1675 1675 chunk = queue[0]
1676 1676 chunkl = len(chunk)
1677 1677 offset = self._chunkoffset
1678 1678
1679 1679 # Use full chunk.
1680 1680 if offset == 0 and left >= chunkl:
1681 1681 left -= chunkl
1682 1682 queue.popleft()
1683 1683 buf.append(chunk)
1684 1684 # self._chunkoffset remains at 0.
1685 1685 continue
1686 1686
1687 1687 chunkremaining = chunkl - offset
1688 1688
1689 1689 # Use all of unconsumed part of chunk.
1690 1690 if left >= chunkremaining:
1691 1691 left -= chunkremaining
1692 1692 queue.popleft()
1693 1693 # offset == 0 is enabled by block above, so this won't merely
1694 1694 # copy via ``chunk[0:]``.
1695 1695 buf.append(chunk[offset:])
1696 1696 self._chunkoffset = 0
1697 1697
1698 1698 # Partial chunk needed.
1699 1699 else:
1700 1700 buf.append(chunk[offset:offset + left])
1701 1701 self._chunkoffset += left
1702 1702 left -= chunkremaining
1703 1703
1704 1704 return ''.join(buf)
1705 1705
1706 1706 def filechunkiter(f, size=131072, limit=None):
1707 1707 """Create a generator that produces the data in the file size
1708 1708 (default 131072) bytes at a time, up to optional limit (default is
1709 1709 to read all data). Chunks may be less than size bytes if the
1710 1710 chunk is the last chunk in the file, or the file is a socket or
1711 1711 some other type of file that sometimes reads less data than is
1712 1712 requested."""
1713 1713 assert size >= 0
1714 1714 assert limit is None or limit >= 0
1715 1715 while True:
1716 1716 if limit is None:
1717 1717 nbytes = size
1718 1718 else:
1719 1719 nbytes = min(limit, size)
1720 1720 s = nbytes and f.read(nbytes)
1721 1721 if not s:
1722 1722 break
1723 1723 if limit:
1724 1724 limit -= len(s)
1725 1725 yield s
1726 1726
1727 1727 def makedate(timestamp=None):
1728 1728 '''Return a unix timestamp (or the current time) as a (unixtime,
1729 1729 offset) tuple based off the local timezone.'''
1730 1730 if timestamp is None:
1731 1731 timestamp = time.time()
1732 1732 if timestamp < 0:
1733 1733 hint = _("check your clock")
1734 1734 raise Abort(_("negative timestamp: %d") % timestamp, hint=hint)
1735 1735 delta = (datetime.datetime.utcfromtimestamp(timestamp) -
1736 1736 datetime.datetime.fromtimestamp(timestamp))
1737 1737 tz = delta.days * 86400 + delta.seconds
1738 1738 return timestamp, tz
1739 1739
1740 1740 def datestr(date=None, format='%a %b %d %H:%M:%S %Y %1%2'):
1741 1741 """represent a (unixtime, offset) tuple as a localized time.
1742 1742 unixtime is seconds since the epoch, and offset is the time zone's
1743 1743 number of seconds away from UTC.
1744 1744
1745 1745 >>> datestr((0, 0))
1746 1746 'Thu Jan 01 00:00:00 1970 +0000'
1747 1747 >>> datestr((42, 0))
1748 1748 'Thu Jan 01 00:00:42 1970 +0000'
1749 1749 >>> datestr((-42, 0))
1750 1750 'Wed Dec 31 23:59:18 1969 +0000'
1751 1751 >>> datestr((0x7fffffff, 0))
1752 1752 'Tue Jan 19 03:14:07 2038 +0000'
1753 1753 >>> datestr((-0x80000000, 0))
1754 1754 'Fri Dec 13 20:45:52 1901 +0000'
1755 1755 """
1756 1756 t, tz = date or makedate()
1757 1757 if "%1" in format or "%2" in format or "%z" in format:
1758 1758 sign = (tz > 0) and "-" or "+"
1759 1759 minutes = abs(tz) // 60
1760 1760 q, r = divmod(minutes, 60)
1761 1761 format = format.replace("%z", "%1%2")
1762 1762 format = format.replace("%1", "%c%02d" % (sign, q))
1763 1763 format = format.replace("%2", "%02d" % r)
1764 1764 d = t - tz
1765 1765 if d > 0x7fffffff:
1766 1766 d = 0x7fffffff
1767 1767 elif d < -0x80000000:
1768 1768 d = -0x80000000
1769 1769 # Never use time.gmtime() and datetime.datetime.fromtimestamp()
1770 1770 # because they use the gmtime() system call which is buggy on Windows
1771 1771 # for negative values.
1772 1772 t = datetime.datetime(1970, 1, 1) + datetime.timedelta(seconds=d)
1773 1773 s = t.strftime(format)
1774 1774 return s
1775 1775
1776 1776 def shortdate(date=None):
1777 1777 """turn (timestamp, tzoff) tuple into iso 8631 date."""
1778 1778 return datestr(date, format='%Y-%m-%d')
1779 1779
1780 1780 def parsetimezone(s):
1781 1781 """find a trailing timezone, if any, in string, and return a
1782 1782 (offset, remainder) pair"""
1783 1783
1784 1784 if s.endswith("GMT") or s.endswith("UTC"):
1785 1785 return 0, s[:-3].rstrip()
1786 1786
1787 1787 # Unix-style timezones [+-]hhmm
1788 1788 if len(s) >= 5 and s[-5] in "+-" and s[-4:].isdigit():
1789 1789 sign = (s[-5] == "+") and 1 or -1
1790 1790 hours = int(s[-4:-2])
1791 1791 minutes = int(s[-2:])
1792 1792 return -sign * (hours * 60 + minutes) * 60, s[:-5].rstrip()
1793 1793
1794 1794 # ISO8601 trailing Z
1795 1795 if s.endswith("Z") and s[-2:-1].isdigit():
1796 1796 return 0, s[:-1]
1797 1797
1798 1798 # ISO8601-style [+-]hh:mm
1799 1799 if (len(s) >= 6 and s[-6] in "+-" and s[-3] == ":" and
1800 1800 s[-5:-3].isdigit() and s[-2:].isdigit()):
1801 1801 sign = (s[-6] == "+") and 1 or -1
1802 1802 hours = int(s[-5:-3])
1803 1803 minutes = int(s[-2:])
1804 1804 return -sign * (hours * 60 + minutes) * 60, s[:-6]
1805 1805
1806 1806 return None, s
1807 1807
1808 1808 def strdate(string, format, defaults=[]):
1809 1809 """parse a localized time string and return a (unixtime, offset) tuple.
1810 1810 if the string cannot be parsed, ValueError is raised."""
1811 1811 # NOTE: unixtime = localunixtime + offset
1812 1812 offset, date = parsetimezone(string)
1813 1813
1814 1814 # add missing elements from defaults
1815 1815 usenow = False # default to using biased defaults
1816 1816 for part in ("S", "M", "HI", "d", "mb", "yY"): # decreasing specificity
1817 1817 found = [True for p in part if ("%"+p) in format]
1818 1818 if not found:
1819 1819 date += "@" + defaults[part][usenow]
1820 1820 format += "@%" + part[0]
1821 1821 else:
1822 1822 # We've found a specific time element, less specific time
1823 1823 # elements are relative to today
1824 1824 usenow = True
1825 1825
1826 1826 timetuple = time.strptime(date, format)
1827 1827 localunixtime = int(calendar.timegm(timetuple))
1828 1828 if offset is None:
1829 1829 # local timezone
1830 1830 unixtime = int(time.mktime(timetuple))
1831 1831 offset = unixtime - localunixtime
1832 1832 else:
1833 1833 unixtime = localunixtime + offset
1834 1834 return unixtime, offset
1835 1835
1836 1836 def parsedate(date, formats=None, bias=None):
1837 1837 """parse a localized date/time and return a (unixtime, offset) tuple.
1838 1838
1839 1839 The date may be a "unixtime offset" string or in one of the specified
1840 1840 formats. If the date already is a (unixtime, offset) tuple, it is returned.
1841 1841
1842 1842 >>> parsedate(' today ') == parsedate(\
1843 1843 datetime.date.today().strftime('%b %d'))
1844 1844 True
1845 1845 >>> parsedate( 'yesterday ') == parsedate((datetime.date.today() -\
1846 1846 datetime.timedelta(days=1)\
1847 1847 ).strftime('%b %d'))
1848 1848 True
1849 1849 >>> now, tz = makedate()
1850 1850 >>> strnow, strtz = parsedate('now')
1851 1851 >>> (strnow - now) < 1
1852 1852 True
1853 1853 >>> tz == strtz
1854 1854 True
1855 1855 """
1856 1856 if bias is None:
1857 1857 bias = {}
1858 1858 if not date:
1859 1859 return 0, 0
1860 1860 if isinstance(date, tuple) and len(date) == 2:
1861 1861 return date
1862 1862 if not formats:
1863 1863 formats = defaultdateformats
1864 1864 date = date.strip()
1865 1865
1866 1866 if date == 'now' or date == _('now'):
1867 1867 return makedate()
1868 1868 if date == 'today' or date == _('today'):
1869 1869 date = datetime.date.today().strftime('%b %d')
1870 1870 elif date == 'yesterday' or date == _('yesterday'):
1871 1871 date = (datetime.date.today() -
1872 1872 datetime.timedelta(days=1)).strftime('%b %d')
1873 1873
1874 1874 try:
1875 1875 when, offset = map(int, date.split(' '))
1876 1876 except ValueError:
1877 1877 # fill out defaults
1878 1878 now = makedate()
1879 1879 defaults = {}
1880 1880 for part in ("d", "mb", "yY", "HI", "M", "S"):
1881 1881 # this piece is for rounding the specific end of unknowns
1882 1882 b = bias.get(part)
1883 1883 if b is None:
1884 1884 if part[0] in "HMS":
1885 1885 b = "00"
1886 1886 else:
1887 1887 b = "0"
1888 1888
1889 1889 # this piece is for matching the generic end to today's date
1890 1890 n = datestr(now, "%" + part[0])
1891 1891
1892 1892 defaults[part] = (b, n)
1893 1893
1894 1894 for format in formats:
1895 1895 try:
1896 1896 when, offset = strdate(date, format, defaults)
1897 1897 except (ValueError, OverflowError):
1898 1898 pass
1899 1899 else:
1900 1900 break
1901 1901 else:
1902 1902 raise Abort(_('invalid date: %r') % date)
1903 1903 # validate explicit (probably user-specified) date and
1904 1904 # time zone offset. values must fit in signed 32 bits for
1905 1905 # current 32-bit linux runtimes. timezones go from UTC-12
1906 1906 # to UTC+14
1907 1907 if when < -0x80000000 or when > 0x7fffffff:
1908 1908 raise Abort(_('date exceeds 32 bits: %d') % when)
1909 1909 if offset < -50400 or offset > 43200:
1910 1910 raise Abort(_('impossible time zone offset: %d') % offset)
1911 1911 return when, offset
1912 1912
1913 1913 def matchdate(date):
1914 1914 """Return a function that matches a given date match specifier
1915 1915
1916 1916 Formats include:
1917 1917
1918 1918 '{date}' match a given date to the accuracy provided
1919 1919
1920 1920 '<{date}' on or before a given date
1921 1921
1922 1922 '>{date}' on or after a given date
1923 1923
1924 1924 >>> p1 = parsedate("10:29:59")
1925 1925 >>> p2 = parsedate("10:30:00")
1926 1926 >>> p3 = parsedate("10:30:59")
1927 1927 >>> p4 = parsedate("10:31:00")
1928 1928 >>> p5 = parsedate("Sep 15 10:30:00 1999")
1929 1929 >>> f = matchdate("10:30")
1930 1930 >>> f(p1[0])
1931 1931 False
1932 1932 >>> f(p2[0])
1933 1933 True
1934 1934 >>> f(p3[0])
1935 1935 True
1936 1936 >>> f(p4[0])
1937 1937 False
1938 1938 >>> f(p5[0])
1939 1939 False
1940 1940 """
1941 1941
1942 1942 def lower(date):
1943 1943 d = {'mb': "1", 'd': "1"}
1944 1944 return parsedate(date, extendeddateformats, d)[0]
1945 1945
1946 1946 def upper(date):
1947 1947 d = {'mb': "12", 'HI': "23", 'M': "59", 'S': "59"}
1948 1948 for days in ("31", "30", "29"):
1949 1949 try:
1950 1950 d["d"] = days
1951 1951 return parsedate(date, extendeddateformats, d)[0]
1952 1952 except Abort:
1953 1953 pass
1954 1954 d["d"] = "28"
1955 1955 return parsedate(date, extendeddateformats, d)[0]
1956 1956
1957 1957 date = date.strip()
1958 1958
1959 1959 if not date:
1960 1960 raise Abort(_("dates cannot consist entirely of whitespace"))
1961 1961 elif date[0] == "<":
1962 1962 if not date[1:]:
1963 1963 raise Abort(_("invalid day spec, use '<DATE'"))
1964 1964 when = upper(date[1:])
1965 1965 return lambda x: x <= when
1966 1966 elif date[0] == ">":
1967 1967 if not date[1:]:
1968 1968 raise Abort(_("invalid day spec, use '>DATE'"))
1969 1969 when = lower(date[1:])
1970 1970 return lambda x: x >= when
1971 1971 elif date[0] == "-":
1972 1972 try:
1973 1973 days = int(date[1:])
1974 1974 except ValueError:
1975 1975 raise Abort(_("invalid day spec: %s") % date[1:])
1976 1976 if days < 0:
1977 1977 raise Abort(_("%s must be nonnegative (see 'hg help dates')")
1978 1978 % date[1:])
1979 1979 when = makedate()[0] - days * 3600 * 24
1980 1980 return lambda x: x >= when
1981 1981 elif " to " in date:
1982 1982 a, b = date.split(" to ")
1983 1983 start, stop = lower(a), upper(b)
1984 1984 return lambda x: x >= start and x <= stop
1985 1985 else:
1986 1986 start, stop = lower(date), upper(date)
1987 1987 return lambda x: x >= start and x <= stop
1988 1988
1989 1989 def stringmatcher(pattern):
1990 1990 """
1991 1991 accepts a string, possibly starting with 're:' or 'literal:' prefix.
1992 1992 returns the matcher name, pattern, and matcher function.
1993 1993 missing or unknown prefixes are treated as literal matches.
1994 1994
1995 1995 helper for tests:
1996 1996 >>> def test(pattern, *tests):
1997 1997 ... kind, pattern, matcher = stringmatcher(pattern)
1998 1998 ... return (kind, pattern, [bool(matcher(t)) for t in tests])
1999 1999
2000 2000 exact matching (no prefix):
2001 2001 >>> test('abcdefg', 'abc', 'def', 'abcdefg')
2002 2002 ('literal', 'abcdefg', [False, False, True])
2003 2003
2004 2004 regex matching ('re:' prefix)
2005 2005 >>> test('re:a.+b', 'nomatch', 'fooadef', 'fooadefbar')
2006 2006 ('re', 'a.+b', [False, False, True])
2007 2007
2008 2008 force exact matches ('literal:' prefix)
2009 2009 >>> test('literal:re:foobar', 'foobar', 're:foobar')
2010 2010 ('literal', 're:foobar', [False, True])
2011 2011
2012 2012 unknown prefixes are ignored and treated as literals
2013 2013 >>> test('foo:bar', 'foo', 'bar', 'foo:bar')
2014 2014 ('literal', 'foo:bar', [False, False, True])
2015 2015 """
2016 2016 if pattern.startswith('re:'):
2017 2017 pattern = pattern[3:]
2018 2018 try:
2019 2019 regex = remod.compile(pattern)
2020 2020 except remod.error as e:
2021 2021 raise error.ParseError(_('invalid regular expression: %s')
2022 2022 % e)
2023 2023 return 're', pattern, regex.search
2024 2024 elif pattern.startswith('literal:'):
2025 2025 pattern = pattern[8:]
2026 2026 return 'literal', pattern, pattern.__eq__
2027 2027
2028 2028 def shortuser(user):
2029 2029 """Return a short representation of a user name or email address."""
2030 2030 f = user.find('@')
2031 2031 if f >= 0:
2032 2032 user = user[:f]
2033 2033 f = user.find('<')
2034 2034 if f >= 0:
2035 2035 user = user[f + 1:]
2036 2036 f = user.find(' ')
2037 2037 if f >= 0:
2038 2038 user = user[:f]
2039 2039 f = user.find('.')
2040 2040 if f >= 0:
2041 2041 user = user[:f]
2042 2042 return user
2043 2043
2044 2044 def emailuser(user):
2045 2045 """Return the user portion of an email address."""
2046 2046 f = user.find('@')
2047 2047 if f >= 0:
2048 2048 user = user[:f]
2049 2049 f = user.find('<')
2050 2050 if f >= 0:
2051 2051 user = user[f + 1:]
2052 2052 return user
2053 2053
2054 2054 def email(author):
2055 2055 '''get email of author.'''
2056 2056 r = author.find('>')
2057 2057 if r == -1:
2058 2058 r = None
2059 2059 return author[author.find('<') + 1:r]
2060 2060
2061 2061 def ellipsis(text, maxlength=400):
2062 2062 """Trim string to at most maxlength (default: 400) columns in display."""
2063 2063 return encoding.trim(text, maxlength, ellipsis='...')
2064 2064
2065 2065 def unitcountfn(*unittable):
2066 2066 '''return a function that renders a readable count of some quantity'''
2067 2067
2068 2068 def go(count):
2069 2069 for multiplier, divisor, format in unittable:
2070 2070 if count >= divisor * multiplier:
2071 2071 return format % (count / float(divisor))
2072 2072 return unittable[-1][2] % count
2073 2073
2074 2074 return go
2075 2075
2076 2076 bytecount = unitcountfn(
2077 2077 (100, 1 << 30, _('%.0f GB')),
2078 2078 (10, 1 << 30, _('%.1f GB')),
2079 2079 (1, 1 << 30, _('%.2f GB')),
2080 2080 (100, 1 << 20, _('%.0f MB')),
2081 2081 (10, 1 << 20, _('%.1f MB')),
2082 2082 (1, 1 << 20, _('%.2f MB')),
2083 2083 (100, 1 << 10, _('%.0f KB')),
2084 2084 (10, 1 << 10, _('%.1f KB')),
2085 2085 (1, 1 << 10, _('%.2f KB')),
2086 2086 (1, 1, _('%.0f bytes')),
2087 2087 )
2088 2088
2089 2089 def uirepr(s):
2090 2090 # Avoid double backslash in Windows path repr()
2091 2091 return repr(s).replace('\\\\', '\\')
2092 2092
2093 2093 # delay import of textwrap
2094 2094 def MBTextWrapper(**kwargs):
2095 2095 class tw(textwrap.TextWrapper):
2096 2096 """
2097 2097 Extend TextWrapper for width-awareness.
2098 2098
2099 2099 Neither number of 'bytes' in any encoding nor 'characters' is
2100 2100 appropriate to calculate terminal columns for specified string.
2101 2101
2102 2102 Original TextWrapper implementation uses built-in 'len()' directly,
2103 2103 so overriding is needed to use width information of each characters.
2104 2104
2105 2105 In addition, characters classified into 'ambiguous' width are
2106 2106 treated as wide in East Asian area, but as narrow in other.
2107 2107
2108 2108 This requires use decision to determine width of such characters.
2109 2109 """
2110 2110 def _cutdown(self, ucstr, space_left):
2111 2111 l = 0
2112 2112 colwidth = encoding.ucolwidth
2113 2113 for i in xrange(len(ucstr)):
2114 2114 l += colwidth(ucstr[i])
2115 2115 if space_left < l:
2116 2116 return (ucstr[:i], ucstr[i:])
2117 2117 return ucstr, ''
2118 2118
2119 2119 # overriding of base class
2120 2120 def _handle_long_word(self, reversed_chunks, cur_line, cur_len, width):
2121 2121 space_left = max(width - cur_len, 1)
2122 2122
2123 2123 if self.break_long_words:
2124 2124 cut, res = self._cutdown(reversed_chunks[-1], space_left)
2125 2125 cur_line.append(cut)
2126 2126 reversed_chunks[-1] = res
2127 2127 elif not cur_line:
2128 2128 cur_line.append(reversed_chunks.pop())
2129 2129
2130 2130 # this overriding code is imported from TextWrapper of Python 2.6
2131 2131 # to calculate columns of string by 'encoding.ucolwidth()'
2132 2132 def _wrap_chunks(self, chunks):
2133 2133 colwidth = encoding.ucolwidth
2134 2134
2135 2135 lines = []
2136 2136 if self.width <= 0:
2137 2137 raise ValueError("invalid width %r (must be > 0)" % self.width)
2138 2138
2139 2139 # Arrange in reverse order so items can be efficiently popped
2140 2140 # from a stack of chucks.
2141 2141 chunks.reverse()
2142 2142
2143 2143 while chunks:
2144 2144
2145 2145 # Start the list of chunks that will make up the current line.
2146 2146 # cur_len is just the length of all the chunks in cur_line.
2147 2147 cur_line = []
2148 2148 cur_len = 0
2149 2149
2150 2150 # Figure out which static string will prefix this line.
2151 2151 if lines:
2152 2152 indent = self.subsequent_indent
2153 2153 else:
2154 2154 indent = self.initial_indent
2155 2155
2156 2156 # Maximum width for this line.
2157 2157 width = self.width - len(indent)
2158 2158
2159 2159 # First chunk on line is whitespace -- drop it, unless this
2160 2160 # is the very beginning of the text (i.e. no lines started yet).
2161 2161 if self.drop_whitespace and chunks[-1].strip() == '' and lines:
2162 2162 del chunks[-1]
2163 2163
2164 2164 while chunks:
2165 2165 l = colwidth(chunks[-1])
2166 2166
2167 2167 # Can at least squeeze this chunk onto the current line.
2168 2168 if cur_len + l <= width:
2169 2169 cur_line.append(chunks.pop())
2170 2170 cur_len += l
2171 2171
2172 2172 # Nope, this line is full.
2173 2173 else:
2174 2174 break
2175 2175
2176 2176 # The current line is full, and the next chunk is too big to
2177 2177 # fit on *any* line (not just this one).
2178 2178 if chunks and colwidth(chunks[-1]) > width:
2179 2179 self._handle_long_word(chunks, cur_line, cur_len, width)
2180 2180
2181 2181 # If the last chunk on this line is all whitespace, drop it.
2182 2182 if (self.drop_whitespace and
2183 2183 cur_line and cur_line[-1].strip() == ''):
2184 2184 del cur_line[-1]
2185 2185
2186 2186 # Convert current line back to a string and store it in list
2187 2187 # of all lines (return value).
2188 2188 if cur_line:
2189 2189 lines.append(indent + ''.join(cur_line))
2190 2190
2191 2191 return lines
2192 2192
2193 2193 global MBTextWrapper
2194 2194 MBTextWrapper = tw
2195 2195 return tw(**kwargs)
2196 2196
2197 2197 def wrap(line, width, initindent='', hangindent=''):
2198 2198 maxindent = max(len(hangindent), len(initindent))
2199 2199 if width <= maxindent:
2200 2200 # adjust for weird terminal size
2201 2201 width = max(78, maxindent + 1)
2202 2202 line = line.decode(encoding.encoding, encoding.encodingmode)
2203 2203 initindent = initindent.decode(encoding.encoding, encoding.encodingmode)
2204 2204 hangindent = hangindent.decode(encoding.encoding, encoding.encodingmode)
2205 2205 wrapper = MBTextWrapper(width=width,
2206 2206 initial_indent=initindent,
2207 2207 subsequent_indent=hangindent)
2208 2208 return wrapper.fill(line).encode(encoding.encoding)
2209 2209
2210 2210 if (pyplatform.python_implementation() == 'CPython' and
2211 2211 sys.version_info < (3, 0)):
2212 2212 # There is an issue in CPython that some IO methods do not handle EINTR
2213 2213 # correctly. The following table shows what CPython version (and functions)
2214 2214 # are affected (buggy: has the EINTR bug, okay: otherwise):
2215 2215 #
2216 2216 # | < 2.7.4 | 2.7.4 to 2.7.12 | >= 3.0
2217 2217 # --------------------------------------------------
2218 2218 # fp.__iter__ | buggy | buggy | okay
2219 2219 # fp.read* | buggy | okay [1] | okay
2220 2220 #
2221 2221 # [1]: fixed by changeset 67dc99a989cd in the cpython hg repo.
2222 2222 #
2223 2223 # Here we workaround the EINTR issue for fileobj.__iter__. Other methods
2224 2224 # like "read*" are ignored for now, as Python < 2.7.4 is a minority.
2225 2225 #
2226 2226 # Although we can workaround the EINTR issue for fp.__iter__, it is slower:
2227 2227 # "for x in fp" is 4x faster than "for x in iter(fp.readline, '')" in
2228 2228 # CPython 2, because CPython 2 maintains an internal readahead buffer for
2229 2229 # fp.__iter__ but not other fp.read* methods.
2230 2230 #
2231 2231 # On modern systems like Linux, the "read" syscall cannot be interrupted
2232 2232 # when reading "fast" files like on-disk files. So the EINTR issue only
2233 2233 # affects things like pipes, sockets, ttys etc. We treat "normal" (S_ISREG)
2234 2234 # files approximately as "fast" files and use the fast (unsafe) code path,
2235 2235 # to minimize the performance impact.
2236 2236 if sys.version_info >= (2, 7, 4):
2237 2237 # fp.readline deals with EINTR correctly, use it as a workaround.
2238 2238 def _safeiterfile(fp):
2239 2239 return iter(fp.readline, '')
2240 2240 else:
2241 2241 # fp.read* are broken too, manually deal with EINTR in a stupid way.
2242 2242 # note: this may block longer than necessary because of bufsize.
2243 2243 def _safeiterfile(fp, bufsize=4096):
2244 2244 fd = fp.fileno()
2245 2245 line = ''
2246 2246 while True:
2247 2247 try:
2248 2248 buf = os.read(fd, bufsize)
2249 2249 except OSError as ex:
2250 2250 # os.read only raises EINTR before any data is read
2251 2251 if ex.errno == errno.EINTR:
2252 2252 continue
2253 2253 else:
2254 2254 raise
2255 2255 line += buf
2256 2256 if '\n' in buf:
2257 2257 splitted = line.splitlines(True)
2258 2258 line = ''
2259 2259 for l in splitted:
2260 2260 if l[-1] == '\n':
2261 2261 yield l
2262 2262 else:
2263 2263 line = l
2264 2264 if not buf:
2265 2265 break
2266 2266 if line:
2267 2267 yield line
2268 2268
2269 2269 def iterfile(fp):
2270 2270 fastpath = True
2271 2271 if type(fp) is file:
2272 2272 fastpath = stat.S_ISREG(os.fstat(fp.fileno()).st_mode)
2273 2273 if fastpath:
2274 2274 return fp
2275 2275 else:
2276 2276 return _safeiterfile(fp)
2277 2277 else:
2278 2278 # PyPy and CPython 3 do not have the EINTR issue thus no workaround needed.
2279 2279 def iterfile(fp):
2280 2280 return fp
2281 2281
2282 2282 def iterlines(iterator):
2283 2283 for chunk in iterator:
2284 2284 for line in chunk.splitlines():
2285 2285 yield line
2286 2286
2287 2287 def expandpath(path):
2288 2288 return os.path.expanduser(os.path.expandvars(path))
2289 2289
2290 2290 def hgcmd():
2291 2291 """Return the command used to execute current hg
2292 2292
2293 2293 This is different from hgexecutable() because on Windows we want
2294 2294 to avoid things opening new shell windows like batch files, so we
2295 2295 get either the python call or current executable.
2296 2296 """
2297 2297 if mainfrozen():
2298 2298 if getattr(sys, 'frozen', None) == 'macosx_app':
2299 2299 # Env variable set by py2app
2300 2300 return [encoding.environ['EXECUTABLEPATH']]
2301 2301 else:
2302 2302 return [sys.executable]
2303 2303 return gethgcmd()
2304 2304
2305 2305 def rundetached(args, condfn):
2306 2306 """Execute the argument list in a detached process.
2307 2307
2308 2308 condfn is a callable which is called repeatedly and should return
2309 2309 True once the child process is known to have started successfully.
2310 2310 At this point, the child process PID is returned. If the child
2311 2311 process fails to start or finishes before condfn() evaluates to
2312 2312 True, return -1.
2313 2313 """
2314 2314 # Windows case is easier because the child process is either
2315 2315 # successfully starting and validating the condition or exiting
2316 2316 # on failure. We just poll on its PID. On Unix, if the child
2317 2317 # process fails to start, it will be left in a zombie state until
2318 2318 # the parent wait on it, which we cannot do since we expect a long
2319 2319 # running process on success. Instead we listen for SIGCHLD telling
2320 2320 # us our child process terminated.
2321 2321 terminated = set()
2322 2322 def handler(signum, frame):
2323 2323 terminated.add(os.wait())
2324 2324 prevhandler = None
2325 2325 SIGCHLD = getattr(signal, 'SIGCHLD', None)
2326 2326 if SIGCHLD is not None:
2327 2327 prevhandler = signal.signal(SIGCHLD, handler)
2328 2328 try:
2329 2329 pid = spawndetached(args)
2330 2330 while not condfn():
2331 2331 if ((pid in terminated or not testpid(pid))
2332 2332 and not condfn()):
2333 2333 return -1
2334 2334 time.sleep(0.1)
2335 2335 return pid
2336 2336 finally:
2337 2337 if prevhandler is not None:
2338 2338 signal.signal(signal.SIGCHLD, prevhandler)
2339 2339
2340 2340 def interpolate(prefix, mapping, s, fn=None, escape_prefix=False):
2341 2341 """Return the result of interpolating items in the mapping into string s.
2342 2342
2343 2343 prefix is a single character string, or a two character string with
2344 2344 a backslash as the first character if the prefix needs to be escaped in
2345 2345 a regular expression.
2346 2346
2347 2347 fn is an optional function that will be applied to the replacement text
2348 2348 just before replacement.
2349 2349
2350 2350 escape_prefix is an optional flag that allows using doubled prefix for
2351 2351 its escaping.
2352 2352 """
2353 2353 fn = fn or (lambda s: s)
2354 2354 patterns = '|'.join(mapping.keys())
2355 2355 if escape_prefix:
2356 2356 patterns += '|' + prefix
2357 2357 if len(prefix) > 1:
2358 2358 prefix_char = prefix[1:]
2359 2359 else:
2360 2360 prefix_char = prefix
2361 2361 mapping[prefix_char] = prefix_char
2362 2362 r = remod.compile(r'%s(%s)' % (prefix, patterns))
2363 2363 return r.sub(lambda x: fn(mapping[x.group()[1:]]), s)
2364 2364
2365 2365 def getport(port):
2366 2366 """Return the port for a given network service.
2367 2367
2368 2368 If port is an integer, it's returned as is. If it's a string, it's
2369 2369 looked up using socket.getservbyname(). If there's no matching
2370 2370 service, error.Abort is raised.
2371 2371 """
2372 2372 try:
2373 2373 return int(port)
2374 2374 except ValueError:
2375 2375 pass
2376 2376
2377 2377 try:
2378 2378 return socket.getservbyname(port)
2379 2379 except socket.error:
2380 2380 raise Abort(_("no port number associated with service '%s'") % port)
2381 2381
2382 2382 _booleans = {'1': True, 'yes': True, 'true': True, 'on': True, 'always': True,
2383 2383 '0': False, 'no': False, 'false': False, 'off': False,
2384 2384 'never': False}
2385 2385
2386 2386 def parsebool(s):
2387 2387 """Parse s into a boolean.
2388 2388
2389 2389 If s is not a valid boolean, returns None.
2390 2390 """
2391 2391 return _booleans.get(s.lower(), None)
2392 2392
2393 2393 _hextochr = dict((a + b, chr(int(a + b, 16)))
2394 2394 for a in string.hexdigits for b in string.hexdigits)
2395 2395
2396 2396 class url(object):
2397 2397 r"""Reliable URL parser.
2398 2398
2399 2399 This parses URLs and provides attributes for the following
2400 2400 components:
2401 2401
2402 2402 <scheme>://<user>:<passwd>@<host>:<port>/<path>?<query>#<fragment>
2403 2403
2404 2404 Missing components are set to None. The only exception is
2405 2405 fragment, which is set to '' if present but empty.
2406 2406
2407 2407 If parsefragment is False, fragment is included in query. If
2408 2408 parsequery is False, query is included in path. If both are
2409 2409 False, both fragment and query are included in path.
2410 2410
2411 2411 See http://www.ietf.org/rfc/rfc2396.txt for more information.
2412 2412
2413 2413 Note that for backward compatibility reasons, bundle URLs do not
2414 2414 take host names. That means 'bundle://../' has a path of '../'.
2415 2415
2416 2416 Examples:
2417 2417
2418 2418 >>> url('http://www.ietf.org/rfc/rfc2396.txt')
2419 2419 <url scheme: 'http', host: 'www.ietf.org', path: 'rfc/rfc2396.txt'>
2420 2420 >>> url('ssh://[::1]:2200//home/joe/repo')
2421 2421 <url scheme: 'ssh', host: '[::1]', port: '2200', path: '/home/joe/repo'>
2422 2422 >>> url('file:///home/joe/repo')
2423 2423 <url scheme: 'file', path: '/home/joe/repo'>
2424 2424 >>> url('file:///c:/temp/foo/')
2425 2425 <url scheme: 'file', path: 'c:/temp/foo/'>
2426 2426 >>> url('bundle:foo')
2427 2427 <url scheme: 'bundle', path: 'foo'>
2428 2428 >>> url('bundle://../foo')
2429 2429 <url scheme: 'bundle', path: '../foo'>
2430 2430 >>> url(r'c:\foo\bar')
2431 2431 <url path: 'c:\\foo\\bar'>
2432 2432 >>> url(r'\\blah\blah\blah')
2433 2433 <url path: '\\\\blah\\blah\\blah'>
2434 2434 >>> url(r'\\blah\blah\blah#baz')
2435 2435 <url path: '\\\\blah\\blah\\blah', fragment: 'baz'>
2436 2436 >>> url(r'file:///C:\users\me')
2437 2437 <url scheme: 'file', path: 'C:\\users\\me'>
2438 2438
2439 2439 Authentication credentials:
2440 2440
2441 2441 >>> url('ssh://joe:xyz@x/repo')
2442 2442 <url scheme: 'ssh', user: 'joe', passwd: 'xyz', host: 'x', path: 'repo'>
2443 2443 >>> url('ssh://joe@x/repo')
2444 2444 <url scheme: 'ssh', user: 'joe', host: 'x', path: 'repo'>
2445 2445
2446 2446 Query strings and fragments:
2447 2447
2448 2448 >>> url('http://host/a?b#c')
2449 2449 <url scheme: 'http', host: 'host', path: 'a', query: 'b', fragment: 'c'>
2450 2450 >>> url('http://host/a?b#c', parsequery=False, parsefragment=False)
2451 2451 <url scheme: 'http', host: 'host', path: 'a?b#c'>
2452 2452
2453 2453 Empty path:
2454 2454
2455 2455 >>> url('')
2456 2456 <url path: ''>
2457 2457 >>> url('#a')
2458 2458 <url path: '', fragment: 'a'>
2459 2459 >>> url('http://host/')
2460 2460 <url scheme: 'http', host: 'host', path: ''>
2461 2461 >>> url('http://host/#a')
2462 2462 <url scheme: 'http', host: 'host', path: '', fragment: 'a'>
2463 2463
2464 2464 Only scheme:
2465 2465
2466 2466 >>> url('http:')
2467 2467 <url scheme: 'http'>
2468 2468 """
2469 2469
2470 2470 _safechars = "!~*'()+"
2471 2471 _safepchars = "/!~*'()+:\\"
2472 2472 _matchscheme = remod.compile('^[a-zA-Z0-9+.\\-]+:').match
2473 2473
2474 2474 def __init__(self, path, parsequery=True, parsefragment=True):
2475 2475 # We slowly chomp away at path until we have only the path left
2476 2476 self.scheme = self.user = self.passwd = self.host = None
2477 2477 self.port = self.path = self.query = self.fragment = None
2478 2478 self._localpath = True
2479 2479 self._hostport = ''
2480 2480 self._origpath = path
2481 2481
2482 2482 if parsefragment and '#' in path:
2483 2483 path, self.fragment = path.split('#', 1)
2484 2484
2485 2485 # special case for Windows drive letters and UNC paths
2486 2486 if hasdriveletter(path) or path.startswith('\\\\'):
2487 2487 self.path = path
2488 2488 return
2489 2489
2490 2490 # For compatibility reasons, we can't handle bundle paths as
2491 2491 # normal URLS
2492 2492 if path.startswith('bundle:'):
2493 2493 self.scheme = 'bundle'
2494 2494 path = path[7:]
2495 2495 if path.startswith('//'):
2496 2496 path = path[2:]
2497 2497 self.path = path
2498 2498 return
2499 2499
2500 2500 if self._matchscheme(path):
2501 2501 parts = path.split(':', 1)
2502 2502 if parts[0]:
2503 2503 self.scheme, path = parts
2504 2504 self._localpath = False
2505 2505
2506 2506 if not path:
2507 2507 path = None
2508 2508 if self._localpath:
2509 2509 self.path = ''
2510 2510 return
2511 2511 else:
2512 2512 if self._localpath:
2513 2513 self.path = path
2514 2514 return
2515 2515
2516 2516 if parsequery and '?' in path:
2517 2517 path, self.query = path.split('?', 1)
2518 2518 if not path:
2519 2519 path = None
2520 2520 if not self.query:
2521 2521 self.query = None
2522 2522
2523 2523 # // is required to specify a host/authority
2524 2524 if path and path.startswith('//'):
2525 2525 parts = path[2:].split('/', 1)
2526 2526 if len(parts) > 1:
2527 2527 self.host, path = parts
2528 2528 else:
2529 2529 self.host = parts[0]
2530 2530 path = None
2531 2531 if not self.host:
2532 2532 self.host = None
2533 2533 # path of file:///d is /d
2534 2534 # path of file:///d:/ is d:/, not /d:/
2535 2535 if path and not hasdriveletter(path):
2536 2536 path = '/' + path
2537 2537
2538 2538 if self.host and '@' in self.host:
2539 2539 self.user, self.host = self.host.rsplit('@', 1)
2540 2540 if ':' in self.user:
2541 2541 self.user, self.passwd = self.user.split(':', 1)
2542 2542 if not self.host:
2543 2543 self.host = None
2544 2544
2545 2545 # Don't split on colons in IPv6 addresses without ports
2546 2546 if (self.host and ':' in self.host and
2547 2547 not (self.host.startswith('[') and self.host.endswith(']'))):
2548 2548 self._hostport = self.host
2549 2549 self.host, self.port = self.host.rsplit(':', 1)
2550 2550 if not self.host:
2551 2551 self.host = None
2552 2552
2553 2553 if (self.host and self.scheme == 'file' and
2554 2554 self.host not in ('localhost', '127.0.0.1', '[::1]')):
2555 2555 raise Abort(_('file:// URLs can only refer to localhost'))
2556 2556
2557 2557 self.path = path
2558 2558
2559 2559 # leave the query string escaped
2560 2560 for a in ('user', 'passwd', 'host', 'port',
2561 2561 'path', 'fragment'):
2562 2562 v = getattr(self, a)
2563 2563 if v is not None:
2564 2564 setattr(self, a, pycompat.urlunquote(v))
2565 2565
2566 2566 def __repr__(self):
2567 2567 attrs = []
2568 2568 for a in ('scheme', 'user', 'passwd', 'host', 'port', 'path',
2569 2569 'query', 'fragment'):
2570 2570 v = getattr(self, a)
2571 2571 if v is not None:
2572 2572 attrs.append('%s: %r' % (a, v))
2573 2573 return '<url %s>' % ', '.join(attrs)
2574 2574
2575 2575 def __str__(self):
2576 2576 r"""Join the URL's components back into a URL string.
2577 2577
2578 2578 Examples:
2579 2579
2580 2580 >>> str(url('http://user:pw@host:80/c:/bob?fo:oo#ba:ar'))
2581 2581 'http://user:pw@host:80/c:/bob?fo:oo#ba:ar'
2582 2582 >>> str(url('http://user:pw@host:80/?foo=bar&baz=42'))
2583 2583 'http://user:pw@host:80/?foo=bar&baz=42'
2584 2584 >>> str(url('http://user:pw@host:80/?foo=bar%3dbaz'))
2585 2585 'http://user:pw@host:80/?foo=bar%3dbaz'
2586 2586 >>> str(url('ssh://user:pw@[::1]:2200//home/joe#'))
2587 2587 'ssh://user:pw@[::1]:2200//home/joe#'
2588 2588 >>> str(url('http://localhost:80//'))
2589 2589 'http://localhost:80//'
2590 2590 >>> str(url('http://localhost:80/'))
2591 2591 'http://localhost:80/'
2592 2592 >>> str(url('http://localhost:80'))
2593 2593 'http://localhost:80/'
2594 2594 >>> str(url('bundle:foo'))
2595 2595 'bundle:foo'
2596 2596 >>> str(url('bundle://../foo'))
2597 2597 'bundle:../foo'
2598 2598 >>> str(url('path'))
2599 2599 'path'
2600 2600 >>> str(url('file:///tmp/foo/bar'))
2601 2601 'file:///tmp/foo/bar'
2602 2602 >>> str(url('file:///c:/tmp/foo/bar'))
2603 2603 'file:///c:/tmp/foo/bar'
2604 2604 >>> print url(r'bundle:foo\bar')
2605 2605 bundle:foo\bar
2606 2606 >>> print url(r'file:///D:\data\hg')
2607 2607 file:///D:\data\hg
2608 2608 """
2609 2609 if self._localpath:
2610 2610 s = self.path
2611 2611 if self.scheme == 'bundle':
2612 2612 s = 'bundle:' + s
2613 2613 if self.fragment:
2614 2614 s += '#' + self.fragment
2615 2615 return s
2616 2616
2617 2617 s = self.scheme + ':'
2618 2618 if self.user or self.passwd or self.host:
2619 2619 s += '//'
2620 2620 elif self.scheme and (not self.path or self.path.startswith('/')
2621 2621 or hasdriveletter(self.path)):
2622 2622 s += '//'
2623 2623 if hasdriveletter(self.path):
2624 2624 s += '/'
2625 2625 if self.user:
2626 2626 s += urlreq.quote(self.user, safe=self._safechars)
2627 2627 if self.passwd:
2628 2628 s += ':' + urlreq.quote(self.passwd, safe=self._safechars)
2629 2629 if self.user or self.passwd:
2630 2630 s += '@'
2631 2631 if self.host:
2632 2632 if not (self.host.startswith('[') and self.host.endswith(']')):
2633 2633 s += urlreq.quote(self.host)
2634 2634 else:
2635 2635 s += self.host
2636 2636 if self.port:
2637 2637 s += ':' + urlreq.quote(self.port)
2638 2638 if self.host:
2639 2639 s += '/'
2640 2640 if self.path:
2641 2641 # TODO: similar to the query string, we should not unescape the
2642 2642 # path when we store it, the path might contain '%2f' = '/',
2643 2643 # which we should *not* escape.
2644 2644 s += urlreq.quote(self.path, safe=self._safepchars)
2645 2645 if self.query:
2646 2646 # we store the query in escaped form.
2647 2647 s += '?' + self.query
2648 2648 if self.fragment is not None:
2649 2649 s += '#' + urlreq.quote(self.fragment, safe=self._safepchars)
2650 2650 return s
2651 2651
2652 2652 def authinfo(self):
2653 2653 user, passwd = self.user, self.passwd
2654 2654 try:
2655 2655 self.user, self.passwd = None, None
2656 2656 s = str(self)
2657 2657 finally:
2658 2658 self.user, self.passwd = user, passwd
2659 2659 if not self.user:
2660 2660 return (s, None)
2661 2661 # authinfo[1] is passed to urllib2 password manager, and its
2662 2662 # URIs must not contain credentials. The host is passed in the
2663 2663 # URIs list because Python < 2.4.3 uses only that to search for
2664 2664 # a password.
2665 2665 return (s, (None, (s, self.host),
2666 2666 self.user, self.passwd or ''))
2667 2667
2668 2668 def isabs(self):
2669 2669 if self.scheme and self.scheme != 'file':
2670 2670 return True # remote URL
2671 2671 if hasdriveletter(self.path):
2672 2672 return True # absolute for our purposes - can't be joined()
2673 2673 if self.path.startswith(r'\\'):
2674 2674 return True # Windows UNC path
2675 2675 if self.path.startswith('/'):
2676 2676 return True # POSIX-style
2677 2677 return False
2678 2678
2679 2679 def localpath(self):
2680 2680 if self.scheme == 'file' or self.scheme == 'bundle':
2681 2681 path = self.path or '/'
2682 2682 # For Windows, we need to promote hosts containing drive
2683 2683 # letters to paths with drive letters.
2684 2684 if hasdriveletter(self._hostport):
2685 2685 path = self._hostport + '/' + self.path
2686 2686 elif (self.host is not None and self.path
2687 2687 and not hasdriveletter(path)):
2688 2688 path = '/' + path
2689 2689 return path
2690 2690 return self._origpath
2691 2691
2692 2692 def islocal(self):
2693 2693 '''whether localpath will return something that posixfile can open'''
2694 2694 return (not self.scheme or self.scheme == 'file'
2695 2695 or self.scheme == 'bundle')
2696 2696
2697 2697 def hasscheme(path):
2698 2698 return bool(url(path).scheme)
2699 2699
2700 2700 def hasdriveletter(path):
2701 2701 return path and path[1:2] == ':' and path[0:1].isalpha()
2702 2702
2703 2703 def urllocalpath(path):
2704 2704 return url(path, parsequery=False, parsefragment=False).localpath()
2705 2705
2706 2706 def hidepassword(u):
2707 2707 '''hide user credential in a url string'''
2708 2708 u = url(u)
2709 2709 if u.passwd:
2710 2710 u.passwd = '***'
2711 2711 return str(u)
2712 2712
2713 2713 def removeauth(u):
2714 2714 '''remove all authentication information from a url string'''
2715 2715 u = url(u)
2716 2716 u.user = u.passwd = None
2717 2717 return str(u)
2718 2718
2719 2719 def isatty(fp):
2720 2720 try:
2721 2721 return fp.isatty()
2722 2722 except AttributeError:
2723 2723 return False
2724 2724
2725 2725 timecount = unitcountfn(
2726 2726 (1, 1e3, _('%.0f s')),
2727 2727 (100, 1, _('%.1f s')),
2728 2728 (10, 1, _('%.2f s')),
2729 2729 (1, 1, _('%.3f s')),
2730 2730 (100, 0.001, _('%.1f ms')),
2731 2731 (10, 0.001, _('%.2f ms')),
2732 2732 (1, 0.001, _('%.3f ms')),
2733 2733 (100, 0.000001, _('%.1f us')),
2734 2734 (10, 0.000001, _('%.2f us')),
2735 2735 (1, 0.000001, _('%.3f us')),
2736 2736 (100, 0.000000001, _('%.1f ns')),
2737 2737 (10, 0.000000001, _('%.2f ns')),
2738 2738 (1, 0.000000001, _('%.3f ns')),
2739 2739 )
2740 2740
2741 2741 _timenesting = [0]
2742 2742
2743 2743 def timed(func):
2744 2744 '''Report the execution time of a function call to stderr.
2745 2745
2746 2746 During development, use as a decorator when you need to measure
2747 2747 the cost of a function, e.g. as follows:
2748 2748
2749 2749 @util.timed
2750 2750 def foo(a, b, c):
2751 2751 pass
2752 2752 '''
2753 2753
2754 2754 def wrapper(*args, **kwargs):
2755 2755 start = time.time()
2756 2756 indent = 2
2757 2757 _timenesting[0] += indent
2758 2758 try:
2759 2759 return func(*args, **kwargs)
2760 2760 finally:
2761 2761 elapsed = time.time() - start
2762 2762 _timenesting[0] -= indent
2763 2763 stderr.write('%s%s: %s\n' %
2764 2764 (' ' * _timenesting[0], func.__name__,
2765 2765 timecount(elapsed)))
2766 2766 return wrapper
2767 2767
2768 2768 _sizeunits = (('m', 2**20), ('k', 2**10), ('g', 2**30),
2769 2769 ('kb', 2**10), ('mb', 2**20), ('gb', 2**30), ('b', 1))
2770 2770
2771 2771 def sizetoint(s):
2772 2772 '''Convert a space specifier to a byte count.
2773 2773
2774 2774 >>> sizetoint('30')
2775 2775 30
2776 2776 >>> sizetoint('2.2kb')
2777 2777 2252
2778 2778 >>> sizetoint('6M')
2779 2779 6291456
2780 2780 '''
2781 2781 t = s.strip().lower()
2782 2782 try:
2783 2783 for k, u in _sizeunits:
2784 2784 if t.endswith(k):
2785 2785 return int(float(t[:-len(k)]) * u)
2786 2786 return int(t)
2787 2787 except ValueError:
2788 2788 raise error.ParseError(_("couldn't parse size: %s") % s)
2789 2789
2790 2790 class hooks(object):
2791 2791 '''A collection of hook functions that can be used to extend a
2792 2792 function's behavior. Hooks are called in lexicographic order,
2793 2793 based on the names of their sources.'''
2794 2794
2795 2795 def __init__(self):
2796 2796 self._hooks = []
2797 2797
2798 2798 def add(self, source, hook):
2799 2799 self._hooks.append((source, hook))
2800 2800
2801 2801 def __call__(self, *args):
2802 2802 self._hooks.sort(key=lambda x: x[0])
2803 2803 results = []
2804 2804 for source, hook in self._hooks:
2805 2805 results.append(hook(*args))
2806 2806 return results
2807 2807
2808 2808 def getstackframes(skip=0, line=' %-*s in %s\n', fileline='%s:%s'):
2809 2809 '''Yields lines for a nicely formatted stacktrace.
2810 2810 Skips the 'skip' last entries.
2811 2811 Each file+linenumber is formatted according to fileline.
2812 2812 Each line is formatted according to line.
2813 2813 If line is None, it yields:
2814 2814 length of longest filepath+line number,
2815 2815 filepath+linenumber,
2816 2816 function
2817 2817
2818 2818 Not be used in production code but very convenient while developing.
2819 2819 '''
2820 2820 entries = [(fileline % (fn, ln), func)
2821 2821 for fn, ln, func, _text in traceback.extract_stack()[:-skip - 1]]
2822 2822 if entries:
2823 2823 fnmax = max(len(entry[0]) for entry in entries)
2824 2824 for fnln, func in entries:
2825 2825 if line is None:
2826 2826 yield (fnmax, fnln, func)
2827 2827 else:
2828 2828 yield line % (fnmax, fnln, func)
2829 2829
2830 2830 def debugstacktrace(msg='stacktrace', skip=0, f=stderr, otherf=stdout):
2831 2831 '''Writes a message to f (stderr) with a nicely formatted stacktrace.
2832 2832 Skips the 'skip' last entries. By default it will flush stdout first.
2833 2833 It can be used everywhere and intentionally does not require an ui object.
2834 2834 Not be used in production code but very convenient while developing.
2835 2835 '''
2836 2836 if otherf:
2837 2837 otherf.flush()
2838 2838 f.write('%s at:\n' % msg)
2839 2839 for line in getstackframes(skip + 1):
2840 2840 f.write(line)
2841 2841 f.flush()
2842 2842
2843 2843 class dirs(object):
2844 2844 '''a multiset of directory names from a dirstate or manifest'''
2845 2845
2846 2846 def __init__(self, map, skip=None):
2847 2847 self._dirs = {}
2848 2848 addpath = self.addpath
2849 2849 if safehasattr(map, 'iteritems') and skip is not None:
2850 2850 for f, s in map.iteritems():
2851 2851 if s[0] != skip:
2852 2852 addpath(f)
2853 2853 else:
2854 2854 for f in map:
2855 2855 addpath(f)
2856 2856
2857 2857 def addpath(self, path):
2858 2858 dirs = self._dirs
2859 2859 for base in finddirs(path):
2860 2860 if base in dirs:
2861 2861 dirs[base] += 1
2862 2862 return
2863 2863 dirs[base] = 1
2864 2864
2865 2865 def delpath(self, path):
2866 2866 dirs = self._dirs
2867 2867 for base in finddirs(path):
2868 2868 if dirs[base] > 1:
2869 2869 dirs[base] -= 1
2870 2870 return
2871 2871 del dirs[base]
2872 2872
2873 2873 def __iter__(self):
2874 2874 return self._dirs.iterkeys()
2875 2875
2876 2876 def __contains__(self, d):
2877 2877 return d in self._dirs
2878 2878
2879 2879 if safehasattr(parsers, 'dirs'):
2880 2880 dirs = parsers.dirs
2881 2881
2882 2882 def finddirs(path):
2883 2883 pos = path.rfind('/')
2884 2884 while pos != -1:
2885 2885 yield path[:pos]
2886 2886 pos = path.rfind('/', 0, pos)
2887 2887
2888 2888 class ctxmanager(object):
2889 2889 '''A context manager for use in 'with' blocks to allow multiple
2890 2890 contexts to be entered at once. This is both safer and more
2891 2891 flexible than contextlib.nested.
2892 2892
2893 2893 Once Mercurial supports Python 2.7+, this will become mostly
2894 2894 unnecessary.
2895 2895 '''
2896 2896
2897 2897 def __init__(self, *args):
2898 2898 '''Accepts a list of no-argument functions that return context
2899 2899 managers. These will be invoked at __call__ time.'''
2900 2900 self._pending = args
2901 2901 self._atexit = []
2902 2902
2903 2903 def __enter__(self):
2904 2904 return self
2905 2905
2906 2906 def enter(self):
2907 2907 '''Create and enter context managers in the order in which they were
2908 2908 passed to the constructor.'''
2909 2909 values = []
2910 2910 for func in self._pending:
2911 2911 obj = func()
2912 2912 values.append(obj.__enter__())
2913 2913 self._atexit.append(obj.__exit__)
2914 2914 del self._pending
2915 2915 return values
2916 2916
2917 2917 def atexit(self, func, *args, **kwargs):
2918 2918 '''Add a function to call when this context manager exits. The
2919 2919 ordering of multiple atexit calls is unspecified, save that
2920 2920 they will happen before any __exit__ functions.'''
2921 2921 def wrapper(exc_type, exc_val, exc_tb):
2922 2922 func(*args, **kwargs)
2923 2923 self._atexit.append(wrapper)
2924 2924 return func
2925 2925
2926 2926 def __exit__(self, exc_type, exc_val, exc_tb):
2927 2927 '''Context managers are exited in the reverse order from which
2928 2928 they were created.'''
2929 2929 received = exc_type is not None
2930 2930 suppressed = False
2931 2931 pending = None
2932 2932 self._atexit.reverse()
2933 2933 for exitfunc in self._atexit:
2934 2934 try:
2935 2935 if exitfunc(exc_type, exc_val, exc_tb):
2936 2936 suppressed = True
2937 2937 exc_type = None
2938 2938 exc_val = None
2939 2939 exc_tb = None
2940 2940 except BaseException:
2941 2941 pending = sys.exc_info()
2942 2942 exc_type, exc_val, exc_tb = pending = sys.exc_info()
2943 2943 del self._atexit
2944 2944 if pending:
2945 2945 raise exc_val
2946 2946 return received and suppressed
2947 2947
2948 2948 # compression code
2949 2949
2950 2950 class compressormanager(object):
2951 2951 """Holds registrations of various compression engines.
2952 2952
2953 2953 This class essentially abstracts the differences between compression
2954 2954 engines to allow new compression formats to be added easily, possibly from
2955 2955 extensions.
2956 2956
2957 2957 Compressors are registered against the global instance by calling its
2958 2958 ``register()`` method.
2959 2959 """
2960 2960 def __init__(self):
2961 2961 self._engines = {}
2962 2962 # Bundle spec human name to engine name.
2963 2963 self._bundlenames = {}
2964 2964 # Internal bundle identifier to engine name.
2965 2965 self._bundletypes = {}
2966 2966
2967 2967 def __getitem__(self, key):
2968 2968 return self._engines[key]
2969 2969
2970 2970 def __contains__(self, key):
2971 2971 return key in self._engines
2972 2972
2973 2973 def __iter__(self):
2974 2974 return iter(self._engines.keys())
2975 2975
2976 2976 def register(self, engine):
2977 2977 """Register a compression engine with the manager.
2978 2978
2979 2979 The argument must be a ``compressionengine`` instance.
2980 2980 """
2981 2981 if not isinstance(engine, compressionengine):
2982 2982 raise ValueError(_('argument must be a compressionengine'))
2983 2983
2984 2984 name = engine.name()
2985 2985
2986 2986 if name in self._engines:
2987 2987 raise error.Abort(_('compression engine %s already registered') %
2988 2988 name)
2989 2989
2990 2990 bundleinfo = engine.bundletype()
2991 2991 if bundleinfo:
2992 2992 bundlename, bundletype = bundleinfo
2993 2993
2994 2994 if bundlename in self._bundlenames:
2995 2995 raise error.Abort(_('bundle name %s already registered') %
2996 2996 bundlename)
2997 2997 if bundletype in self._bundletypes:
2998 2998 raise error.Abort(_('bundle type %s already registered by %s') %
2999 2999 (bundletype, self._bundletypes[bundletype]))
3000 3000
3001 3001 # No external facing name declared.
3002 3002 if bundlename:
3003 3003 self._bundlenames[bundlename] = name
3004 3004
3005 3005 self._bundletypes[bundletype] = name
3006 3006
3007 3007 self._engines[name] = engine
3008 3008
3009 3009 @property
3010 3010 def supportedbundlenames(self):
3011 3011 return set(self._bundlenames.keys())
3012 3012
3013 3013 @property
3014 3014 def supportedbundletypes(self):
3015 3015 return set(self._bundletypes.keys())
3016 3016
3017 3017 def forbundlename(self, bundlename):
3018 3018 """Obtain a compression engine registered to a bundle name.
3019 3019
3020 3020 Will raise KeyError if the bundle type isn't registered.
3021 3021
3022 3022 Will abort if the engine is known but not available.
3023 3023 """
3024 3024 engine = self._engines[self._bundlenames[bundlename]]
3025 3025 if not engine.available():
3026 3026 raise error.Abort(_('compression engine %s could not be loaded') %
3027 3027 engine.name())
3028 3028 return engine
3029 3029
3030 3030 def forbundletype(self, bundletype):
3031 3031 """Obtain a compression engine registered to a bundle type.
3032 3032
3033 3033 Will raise KeyError if the bundle type isn't registered.
3034 3034
3035 3035 Will abort if the engine is known but not available.
3036 3036 """
3037 3037 engine = self._engines[self._bundletypes[bundletype]]
3038 3038 if not engine.available():
3039 3039 raise error.Abort(_('compression engine %s could not be loaded') %
3040 3040 engine.name())
3041 3041 return engine
3042 3042
3043 3043 compengines = compressormanager()
3044 3044
3045 3045 class compressionengine(object):
3046 3046 """Base class for compression engines.
3047 3047
3048 3048 Compression engines must implement the interface defined by this class.
3049 3049 """
3050 3050 def name(self):
3051 3051 """Returns the name of the compression engine.
3052 3052
3053 3053 This is the key the engine is registered under.
3054 3054
3055 3055 This method must be implemented.
3056 3056 """
3057 3057 raise NotImplementedError()
3058 3058
3059 3059 def available(self):
3060 3060 """Whether the compression engine is available.
3061 3061
3062 3062 The intent of this method is to allow optional compression engines
3063 3063 that may not be available in all installations (such as engines relying
3064 3064 on C extensions that may not be present).
3065 3065 """
3066 3066 return True
3067 3067
3068 3068 def bundletype(self):
3069 3069 """Describes bundle identifiers for this engine.
3070 3070
3071 3071 If this compression engine isn't supported for bundles, returns None.
3072 3072
3073 3073 If this engine can be used for bundles, returns a 2-tuple of strings of
3074 3074 the user-facing "bundle spec" compression name and an internal
3075 3075 identifier used to denote the compression format within bundles. To
3076 3076 exclude the name from external usage, set the first element to ``None``.
3077 3077
3078 3078 If bundle compression is supported, the class must also implement
3079 3079 ``compressstream`` and `decompressorreader``.
3080 3080 """
3081 3081 return None
3082 3082
3083 3083 def compressstream(self, it, opts=None):
3084 3084 """Compress an iterator of chunks.
3085 3085
3086 3086 The method receives an iterator (ideally a generator) of chunks of
3087 3087 bytes to be compressed. It returns an iterator (ideally a generator)
3088 3088 of bytes of chunks representing the compressed output.
3089 3089
3090 3090 Optionally accepts an argument defining how to perform compression.
3091 3091 Each engine treats this argument differently.
3092 3092 """
3093 3093 raise NotImplementedError()
3094 3094
3095 3095 def decompressorreader(self, fh):
3096 3096 """Perform decompression on a file object.
3097 3097
3098 3098 Argument is an object with a ``read(size)`` method that returns
3099 3099 compressed data. Return value is an object with a ``read(size)`` that
3100 3100 returns uncompressed data.
3101 3101 """
3102 3102 raise NotImplementedError()
3103 3103
3104 3104 class _zlibengine(compressionengine):
3105 3105 def name(self):
3106 3106 return 'zlib'
3107 3107
3108 3108 def bundletype(self):
3109 3109 return 'gzip', 'GZ'
3110 3110
3111 3111 def compressstream(self, it, opts=None):
3112 3112 opts = opts or {}
3113 3113
3114 3114 z = zlib.compressobj(opts.get('level', -1))
3115 3115 for chunk in it:
3116 3116 data = z.compress(chunk)
3117 3117 # Not all calls to compress emit data. It is cheaper to inspect
3118 3118 # here than to feed empty chunks through generator.
3119 3119 if data:
3120 3120 yield data
3121 3121
3122 3122 yield z.flush()
3123 3123
3124 3124 def decompressorreader(self, fh):
3125 3125 def gen():
3126 3126 d = zlib.decompressobj()
3127 3127 for chunk in filechunkiter(fh):
3128 3128 while chunk:
3129 3129 # Limit output size to limit memory.
3130 3130 yield d.decompress(chunk, 2 ** 18)
3131 3131 chunk = d.unconsumed_tail
3132 3132
3133 3133 return chunkbuffer(gen())
3134 3134
3135 3135 compengines.register(_zlibengine())
3136 3136
3137 3137 class _bz2engine(compressionengine):
3138 3138 def name(self):
3139 3139 return 'bz2'
3140 3140
3141 3141 def bundletype(self):
3142 3142 return 'bzip2', 'BZ'
3143 3143
3144 3144 def compressstream(self, it, opts=None):
3145 3145 opts = opts or {}
3146 3146 z = bz2.BZ2Compressor(opts.get('level', 9))
3147 3147 for chunk in it:
3148 3148 data = z.compress(chunk)
3149 3149 if data:
3150 3150 yield data
3151 3151
3152 3152 yield z.flush()
3153 3153
3154 3154 def decompressorreader(self, fh):
3155 3155 def gen():
3156 3156 d = bz2.BZ2Decompressor()
3157 3157 for chunk in filechunkiter(fh):
3158 3158 yield d.decompress(chunk)
3159 3159
3160 3160 return chunkbuffer(gen())
3161 3161
3162 3162 compengines.register(_bz2engine())
3163 3163
3164 3164 class _truncatedbz2engine(compressionengine):
3165 3165 def name(self):
3166 3166 return 'bz2truncated'
3167 3167
3168 3168 def bundletype(self):
3169 3169 return None, '_truncatedBZ'
3170 3170
3171 3171 # We don't implement compressstream because it is hackily handled elsewhere.
3172 3172
3173 3173 def decompressorreader(self, fh):
3174 3174 def gen():
3175 3175 # The input stream doesn't have the 'BZ' header. So add it back.
3176 3176 d = bz2.BZ2Decompressor()
3177 3177 d.decompress('BZ')
3178 3178 for chunk in filechunkiter(fh):
3179 3179 yield d.decompress(chunk)
3180 3180
3181 3181 return chunkbuffer(gen())
3182 3182
3183 3183 compengines.register(_truncatedbz2engine())
3184 3184
3185 3185 class _noopengine(compressionengine):
3186 3186 def name(self):
3187 3187 return 'none'
3188 3188
3189 3189 def bundletype(self):
3190 3190 return 'none', 'UN'
3191 3191
3192 3192 def compressstream(self, it, opts=None):
3193 3193 return it
3194 3194
3195 3195 def decompressorreader(self, fh):
3196 3196 return fh
3197 3197
3198 3198 compengines.register(_noopengine())
3199 3199
3200 3200 class _zstdengine(compressionengine):
3201 3201 def name(self):
3202 3202 return 'zstd'
3203 3203
3204 3204 @propertycache
3205 3205 def _module(self):
3206 3206 # Not all installs have the zstd module available. So defer importing
3207 3207 # until first access.
3208 3208 try:
3209 3209 from . import zstd
3210 3210 # Force delayed import.
3211 3211 zstd.__version__
3212 3212 return zstd
3213 3213 except ImportError:
3214 3214 return None
3215 3215
3216 3216 def available(self):
3217 3217 return bool(self._module)
3218 3218
3219 3219 def bundletype(self):
3220 3220 return 'zstd', 'ZS'
3221 3221
3222 3222 def compressstream(self, it, opts=None):
3223 3223 opts = opts or {}
3224 3224 # zstd level 3 is almost always significantly faster than zlib
3225 3225 # while providing no worse compression. It strikes a good balance
3226 3226 # between speed and compression.
3227 3227 level = opts.get('level', 3)
3228 3228
3229 3229 zstd = self._module
3230 3230 z = zstd.ZstdCompressor(level=level).compressobj()
3231 3231 for chunk in it:
3232 3232 data = z.compress(chunk)
3233 3233 if data:
3234 3234 yield data
3235 3235
3236 3236 yield z.flush()
3237 3237
3238 3238 def decompressorreader(self, fh):
3239 3239 zstd = self._module
3240 3240 dctx = zstd.ZstdDecompressor()
3241 3241 return chunkbuffer(dctx.read_from(fh))
3242 3242
3243 3243 compengines.register(_zstdengine())
3244 3244
3245 3245 # convenient shortcut
3246 3246 dst = debugstacktrace
General Comments 0
You need to be logged in to leave comments. Login now