1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007, 2009 Canonical Ltd
5
3
# This program is free software; you can redistribute it and/or modify
6
4
# it under the terms of the GNU General Public License as published by
7
5
# the Free Software Foundation; either version 2 of the License, or
8
6
# (at your option) any later version.
10
8
# This program is distributed in the hope that it will be useful,
11
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
11
# GNU General Public License for more details.
15
13
# You should have received a copy of the GNU General Public License
16
14
# along with this program; if not, write to the Free Software
17
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
import os, types, re, time, errno, sys
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
22
from errors import bailout, BzrError
23
from trace import mutter
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
45
from tempfile import (
57
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
59
if sys.version_info < (2, 5):
60
import md5 as _mod_md5
62
import sha as _mod_sha
72
from bzrlib import symbol_versioning
75
# Cross platform wall-clock time functionality with decent resolution.
76
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
77
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
78
# synchronized with ``time.time()``, this is only meant to be used to find
79
# delta times by subtracting from another call to this function.
80
timer_func = time.time
81
if sys.platform == 'win32':
82
timer_func = time.clock
84
# On win32, O_BINARY is used to indicate the file should
85
# be opened in binary mode, rather than text mode.
86
# On other platforms, O_BINARY doesn't exist, because
87
# they always open in binary mode, so it is okay to
88
# OR with 0 on those platforms
89
O_BINARY = getattr(os, 'O_BINARY', 0)
92
def get_unicode_argv():
94
user_encoding = get_user_encoding()
95
return [a.decode(user_encoding) for a in sys.argv[1:]]
96
except UnicodeDecodeError:
97
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
26
101
def make_readonly(filename):
27
102
"""Make a filename read-only."""
28
# TODO: probably needs to be fixed for windows
29
mod = os.stat(filename).st_mode
31
os.chmod(filename, mod)
103
mod = os.lstat(filename).st_mode
104
if not stat.S_ISLNK(mod):
106
os.chmod(filename, mod)
34
109
def make_writable(filename):
35
mod = os.stat(filename).st_mode
37
os.chmod(filename, mod)
40
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/_~-])')
110
mod = os.lstat(filename).st_mode
111
if not stat.S_ISLNK(mod):
113
os.chmod(filename, mod)
116
def minimum_path_selection(paths):
117
"""Return the smallset subset of paths which are outside paths.
119
:param paths: A container (and hence not None) of paths.
120
:return: A set of paths sufficient to include everything in paths via
121
is_inside, drawn from the paths parameter.
127
return path.split('/')
128
sorted_paths = sorted(list(paths), key=sort_key)
130
search_paths = [sorted_paths[0]]
131
for path in sorted_paths[1:]:
132
if not is_inside(search_paths[-1], path):
133
# This path is unique, add it
134
search_paths.append(path)
136
return set(search_paths)
42
"""Return shell-quoted filename"""
43
## We could be a bit more terse by using double-quotes etc
44
f = _QUOTE_RE.sub(r'\\\1', f)
51
mode = os.lstat(f)[ST_MODE]
59
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
143
"""Return a quoted filename filename
145
This previously used backslash quoting, but that works poorly on
147
# TODO: I'm not really sure this is the best format either.x
149
if _QUOTE_RE is None:
150
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
152
if _QUOTE_RE.search(f):
158
_directory_kind = 'directory'
161
"""Return the current umask"""
162
# Assume that people aren't messing with the umask while running
163
# XXX: This is not thread safe, but there is no way to get the
164
# umask without setting it
172
_directory_kind: "/",
174
'tree-reference': '+',
178
def kind_marker(kind):
180
return _kind_marker_map[kind]
182
raise errors.BzrError('invalid file kind %r' % kind)
185
lexists = getattr(os.path, 'lexists', None)
189
stat = getattr(os, 'lstat', os.stat)
193
if e.errno == errno.ENOENT:
196
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
199
def fancy_rename(old, new, rename_func, unlink_func):
200
"""A fancy rename, when you don't have atomic rename.
202
:param old: The old path, to rename from
203
:param new: The new path, to rename to
204
:param rename_func: The potentially non-atomic rename function
205
:param unlink_func: A way to delete the target file if the full rename
208
# sftp rename doesn't allow overwriting, so play tricks:
209
base = os.path.basename(new)
210
dirname = os.path.dirname(new)
211
# callers use different encodings for the paths so the following MUST
212
# respect that. We rely on python upcasting to unicode if new is unicode
213
# and keeping a str if not.
214
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
215
os.getpid(), rand_chars(10))
216
tmp_name = pathjoin(dirname, tmp_name)
218
# Rename the file out of the way, but keep track if it didn't exist
219
# We don't want to grab just any exception
220
# something like EACCES should prevent us from continuing
221
# The downside is that the rename_func has to throw an exception
222
# with an errno = ENOENT, or NoSuchFile
225
rename_func(new, tmp_name)
226
except (errors.NoSuchFile,), e:
229
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
230
# function raises an IOError with errno is None when a rename fails.
231
# This then gets caught here.
232
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
235
if (getattr(e, 'errno', None) is None
236
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
245
# This may throw an exception, in which case success will
247
rename_func(old, new)
249
except (IOError, OSError), e:
250
# source and target may be aliases of each other (e.g. on a
251
# case-insensitive filesystem), so we may have accidentally renamed
252
# source by when we tried to rename target
253
failure_exc = sys.exc_info()
254
if (file_existed and e.errno in (None, errno.ENOENT)
255
and old.lower() == new.lower()):
256
# source and target are the same file on a case-insensitive
257
# filesystem, so we don't generate an exception
261
# If the file used to exist, rename it back into place
262
# otherwise just delete it from the tmp location
264
unlink_func(tmp_name)
266
rename_func(tmp_name, new)
267
if failure_exc is not None:
268
raise failure_exc[0], failure_exc[1], failure_exc[2]
271
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
272
# choke on a Unicode string containing a relative path if
273
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
275
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
276
def _posix_abspath(path):
277
# jam 20060426 rather than encoding to fsencoding
278
# copy posixpath.abspath, but use os.getcwdu instead
279
if not posixpath.isabs(path):
280
path = posixpath.join(getcwd(), path)
281
return posixpath.normpath(path)
284
def _posix_realpath(path):
285
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
288
def _win32_fixdrive(path):
289
"""Force drive letters to be consistent.
291
win32 is inconsistent whether it returns lower or upper case
292
and even if it was consistent the user might type the other
293
so we force it to uppercase
294
running python.exe under cmd.exe return capital C:\\
295
running win32 python inside a cygwin shell returns lowercase c:\\
297
drive, path = _nt_splitdrive(path)
298
return drive.upper() + path
301
def _win32_abspath(path):
302
# Real _nt_abspath doesn't have a problem with a unicode cwd
303
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
306
def _win98_abspath(path):
307
"""Return the absolute version of a path.
308
Windows 98 safe implementation (python reimplementation
309
of Win32 API function GetFullPathNameW)
314
# \\HOST\path => //HOST/path
315
# //HOST/path => //HOST/path
316
# path => C:/cwd/path
319
# check for absolute path
320
drive = _nt_splitdrive(path)[0]
321
if drive == '' and path[:2] not in('//','\\\\'):
323
# we cannot simply os.path.join cwd and path
324
# because os.path.join('C:','/path') produce '/path'
325
# and this is incorrect
326
if path[:1] in ('/','\\'):
327
cwd = _nt_splitdrive(cwd)[0]
329
path = cwd + '\\' + path
330
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
333
def _win32_realpath(path):
334
# Real _nt_realpath doesn't have a problem with a unicode cwd
335
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
338
def _win32_pathjoin(*args):
339
return _nt_join(*args).replace('\\', '/')
342
def _win32_normpath(path):
343
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
347
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
350
def _win32_mkdtemp(*args, **kwargs):
351
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
354
def _win32_rename(old, new):
355
"""We expect to be able to atomically replace 'new' with old.
357
On win32, if new exists, it must be moved out of the way first,
361
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
363
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
364
# If we try to rename a non-existant file onto cwd, we get
365
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
366
# if the old path doesn't exist, sometimes we get EACCES
367
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
373
return unicodedata.normalize('NFC', os.getcwdu())
376
# Default is to just use the python builtins, but these can be rebound on
377
# particular platforms.
378
abspath = _posix_abspath
379
realpath = _posix_realpath
380
pathjoin = os.path.join
381
normpath = os.path.normpath
384
dirname = os.path.dirname
385
basename = os.path.basename
386
split = os.path.split
387
splitext = os.path.splitext
388
# These were already imported into local scope
389
# mkdtemp = tempfile.mkdtemp
390
# rmtree = shutil.rmtree
392
MIN_ABS_PATHLENGTH = 1
395
if sys.platform == 'win32':
396
if win32utils.winver == 'Windows 98':
397
abspath = _win98_abspath
399
abspath = _win32_abspath
400
realpath = _win32_realpath
401
pathjoin = _win32_pathjoin
402
normpath = _win32_normpath
403
getcwd = _win32_getcwd
404
mkdtemp = _win32_mkdtemp
405
rename = _win32_rename
407
MIN_ABS_PATHLENGTH = 3
409
def _win32_delete_readonly(function, path, excinfo):
410
"""Error handler for shutil.rmtree function [for win32]
411
Helps to remove files and dirs marked as read-only.
413
exception = excinfo[1]
414
if function in (os.remove, os.rmdir) \
415
and isinstance(exception, OSError) \
416
and exception.errno == errno.EACCES:
422
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
423
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
424
return shutil.rmtree(path, ignore_errors, onerror)
426
f = win32utils.get_unicode_argv # special function or None
430
elif sys.platform == 'darwin':
434
def get_terminal_encoding():
435
"""Find the best encoding for printing to the screen.
437
This attempts to check both sys.stdout and sys.stdin to see
438
what encoding they are in, and if that fails it falls back to
439
osutils.get_user_encoding().
440
The problem is that on Windows, locale.getpreferredencoding()
441
is not the same encoding as that used by the console:
442
http://mail.python.org/pipermail/python-list/2003-May/162357.html
444
On my standard US Windows XP, the preferred encoding is
445
cp1252, but the console is cp437
447
from bzrlib.trace import mutter
448
output_encoding = getattr(sys.stdout, 'encoding', None)
449
if not output_encoding:
450
input_encoding = getattr(sys.stdin, 'encoding', None)
451
if not input_encoding:
452
output_encoding = get_user_encoding()
453
mutter('encoding stdout as osutils.get_user_encoding() %r',
456
output_encoding = input_encoding
457
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
459
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
460
if output_encoding == 'cp0':
461
# invalid encoding (cp0 means 'no codepage' on Windows)
462
output_encoding = get_user_encoding()
463
mutter('cp0 is invalid encoding.'
464
' encoding stdout as osutils.get_user_encoding() %r',
468
codecs.lookup(output_encoding)
470
sys.stderr.write('bzr: warning:'
471
' unknown terminal encoding %s.\n'
472
' Using encoding %s instead.\n'
473
% (output_encoding, get_user_encoding())
475
output_encoding = get_user_encoding()
477
return output_encoding
480
def normalizepath(f):
481
if getattr(os.path, 'realpath', None) is not None:
485
[p,e] = os.path.split(f)
486
if e == "" or e == "." or e == "..":
489
return pathjoin(F(p), e)
241
704
def local_time_offset(t=None):
242
705
"""Return offset of local zone from GMT, either at present or at time t."""
243
# python2.3 localtime() can't take None
247
if time.localtime(t).tm_isdst and time.daylight:
250
return -time.timezone
253
def format_date(t, offset=0, timezone='original'):
254
## TODO: Perhaps a global option to use either universal or local time?
255
## Or perhaps just let people set $TZ?
256
assert isinstance(t, float)
708
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
709
return offset.days * 86400 + offset.seconds
711
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
712
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
715
def format_date(t, offset=0, timezone='original', date_fmt=None,
717
"""Return a formatted date string.
719
:param t: Seconds since the epoch.
720
:param offset: Timezone offset in seconds east of utc.
721
:param timezone: How to display the time: 'utc', 'original' for the
722
timezone specified by offset, or 'local' for the process's current
724
:param date_fmt: strftime format.
725
:param show_offset: Whether to append the timezone.
727
(date_fmt, tt, offset_str) = \
728
_format_date(t, offset, timezone, date_fmt, show_offset)
729
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
730
date_str = time.strftime(date_fmt, tt)
731
return date_str + offset_str
734
# Cache of formatted offset strings
738
def format_date_with_offset_in_original_timezone(t, offset=0,
739
_cache=_offset_cache):
740
"""Return a formatted date string in the original timezone.
742
This routine may be faster then format_date.
744
:param t: Seconds since the epoch.
745
:param offset: Timezone offset in seconds east of utc.
749
tt = time.gmtime(t + offset)
750
date_fmt = _default_format_by_weekday_num[tt[6]]
751
date_str = time.strftime(date_fmt, tt)
752
offset_str = _cache.get(offset, None)
753
if offset_str is None:
754
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
755
_cache[offset] = offset_str
756
return date_str + offset_str
759
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
761
"""Return an unicode date string formatted according to the current locale.
763
:param t: Seconds since the epoch.
764
:param offset: Timezone offset in seconds east of utc.
765
:param timezone: How to display the time: 'utc', 'original' for the
766
timezone specified by offset, or 'local' for the process's current
768
:param date_fmt: strftime format.
769
:param show_offset: Whether to append the timezone.
771
(date_fmt, tt, offset_str) = \
772
_format_date(t, offset, timezone, date_fmt, show_offset)
773
date_str = time.strftime(date_fmt, tt)
774
if not isinstance(date_str, unicode):
775
date_str = date_str.decode(get_user_encoding(), 'replace')
776
return date_str + offset_str
779
def _format_date(t, offset, timezone, date_fmt, show_offset):
258
780
if timezone == 'utc':
259
781
tt = time.gmtime(t)
261
783
elif timezone == 'original':
264
786
tt = time.gmtime(t + offset)
265
787
elif timezone == 'local':
266
788
tt = time.localtime(t)
267
789
offset = local_time_offset(t)
269
bailout("unsupported timezone format %r",
270
['options are "utc", "original", "local"'])
272
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
273
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
791
raise errors.UnsupportedTimezoneFormat(timezone)
793
date_fmt = "%a %Y-%m-%d %H:%M:%S"
795
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
798
return (date_fmt, tt, offset_str)
276
801
def compact_date(when):
277
802
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
805
def format_delta(delta):
806
"""Get a nice looking string for a time delta.
808
:param delta: The time difference in seconds, can be positive or negative.
809
positive indicates time in the past, negative indicates time in the
810
future. (usually time.time() - stored_time)
811
:return: String formatted to show approximate resolution
817
direction = 'in the future'
821
if seconds < 90: # print seconds up to 90 seconds
823
return '%d second %s' % (seconds, direction,)
825
return '%d seconds %s' % (seconds, direction)
827
minutes = int(seconds / 60)
828
seconds -= 60 * minutes
833
if minutes < 90: # print minutes, seconds up to 90 minutes
835
return '%d minute, %d second%s %s' % (
836
minutes, seconds, plural_seconds, direction)
838
return '%d minutes, %d second%s %s' % (
839
minutes, seconds, plural_seconds, direction)
841
hours = int(minutes / 60)
842
minutes -= 60 * hours
849
return '%d hour, %d minute%s %s' % (hours, minutes,
850
plural_minutes, direction)
851
return '%d hours, %d minute%s %s' % (hours, minutes,
852
plural_minutes, direction)
282
855
"""Return size of given open file."""
283
856
return os.fstat(f.fileno())[ST_SIZE]
286
if hasattr(os, 'urandom'): # python 2.4 and later
859
# Define rand_bytes based on platform.
861
# Python 2.4 and later have os.urandom,
862
# but it doesn't work on some arches
287
864
rand_bytes = os.urandom
288
elif sys.platform == 'linux2':
289
rand_bytes = file('/dev/urandom', 'rb').read
291
# not well seeded, but better than nothing
296
s += chr(random.randint(0, 255))
865
except (NotImplementedError, AttributeError):
866
# If python doesn't have os.urandom, or it doesn't work,
867
# then try to first pull random data from /dev/urandom
869
rand_bytes = file('/dev/urandom', 'rb').read
870
# Otherwise, use this hack as a last resort
871
except (IOError, OSError):
872
# not well seeded, but better than nothing
877
s += chr(random.randint(0, 255))
882
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
884
"""Return a random string of num alphanumeric characters
886
The result only contains lowercase chars because it may be used on
887
case-insensitive filesystems.
890
for raw_byte in rand_bytes(num):
891
s += ALNUM[ord(raw_byte) % 36]
301
895
## TODO: We could later have path objects that remember their list
302
896
## decomposition (might be too tricksy though.)
304
898
def splitpath(p):
305
"""Turn string into list of parts.
311
>>> splitpath('a/./b')
313
>>> splitpath('a/.b')
315
>>> splitpath('a/../b')
316
Traceback (most recent call last):
318
BzrError: ("sorry, '..' not allowed in path", [])
320
assert isinstance(p, types.StringTypes)
899
"""Turn string into list of parts."""
322
900
# split on either delimiter because people might use either on
324
902
ps = re.split(r'[\\/]', p)
329
bailout("sorry, %r not allowed in path" % f)
907
raise errors.BzrError("sorry, %r not allowed in path" % f)
330
908
elif (f == '.') or (f == ''):
337
assert isinstance(p, list)
339
if (f == '..') or (f == None) or (f == ''):
340
bailout("sorry, %r not allowed in path" % f)
341
return os.path.join(*p)
344
def appendpath(p1, p2):
348
return os.path.join(p1, p2)
351
def extern_command(cmd, ignore_errors = False):
352
mutter('external command: %s' % `cmd`)
354
if not ignore_errors:
355
bailout('command failed')
917
if (f == '..') or (f is None) or (f == ''):
918
raise errors.BzrError("sorry, %r not allowed in path" % f)
922
def parent_directories(filename):
923
"""Return the list of parent directories, deepest first.
925
For example, parent_directories("a/b/c") -> ["a/b", "a"].
928
parts = splitpath(dirname(filename))
930
parents.append(joinpath(parts))
935
_extension_load_failures = []
938
def failed_to_load_extension(exception):
939
"""Handle failing to load a binary extension.
941
This should be called from the ImportError block guarding the attempt to
942
import the native extension. If this function returns, the pure-Python
943
implementation should be loaded instead::
946
>>> import bzrlib._fictional_extension_pyx
947
>>> except ImportError, e:
948
>>> bzrlib.osutils.failed_to_load_extension(e)
949
>>> import bzrlib._fictional_extension_py
951
# NB: This docstring is just an example, not a doctest, because doctest
952
# currently can't cope with the use of lazy imports in this namespace --
955
# This currently doesn't report the failure at the time it occurs, because
956
# they tend to happen very early in startup when we can't check config
957
# files etc, and also we want to report all failures but not spam the user
959
from bzrlib import trace
960
exception_str = str(exception)
961
if exception_str not in _extension_load_failures:
962
trace.mutter("failed to load compiled extension: %s" % exception_str)
963
_extension_load_failures.append(exception_str)
966
def report_extension_load_failures():
967
if not _extension_load_failures:
969
from bzrlib.config import GlobalConfig
970
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
972
# the warnings framework should by default show this only once
973
from bzrlib.trace import warning
975
"bzr: warning: some compiled extensions could not be loaded; "
976
"see <https://answers.launchpad.net/bzr/+faq/703>")
977
# we no longer show the specific missing extensions here, because it makes
978
# the message too long and scary - see
979
# https://bugs.launchpad.net/bzr/+bug/430529
983
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
984
except ImportError, e:
985
failed_to_load_extension(e)
986
from bzrlib._chunks_to_lines_py import chunks_to_lines
990
"""Split s into lines, but without removing the newline characters."""
991
# Trivially convert a fulltext into a 'chunked' representation, and let
992
# chunks_to_lines do the heavy lifting.
993
if isinstance(s, str):
994
# chunks_to_lines only supports 8-bit strings
995
return chunks_to_lines([s])
997
return _split_lines(s)
1000
def _split_lines(s):
1001
"""Split s into lines, but without removing the newline characters.
1003
This supports Unicode or plain string objects.
1005
lines = s.split('\n')
1006
result = [line + '\n' for line in lines[:-1]]
1008
result.append(lines[-1])
1012
def hardlinks_good():
1013
return sys.platform not in ('win32', 'cygwin', 'darwin')
1016
def link_or_copy(src, dest):
1017
"""Hardlink a file, or copy it if it can't be hardlinked."""
1018
if not hardlinks_good():
1019
shutil.copyfile(src, dest)
1023
except (OSError, IOError), e:
1024
if e.errno != errno.EXDEV:
1026
shutil.copyfile(src, dest)
1029
def delete_any(path):
1030
"""Delete a file, symlink or directory.
1032
Will delete even if readonly.
1035
_delete_file_or_dir(path)
1036
except (OSError, IOError), e:
1037
if e.errno in (errno.EPERM, errno.EACCES):
1038
# make writable and try again
1041
except (OSError, IOError):
1043
_delete_file_or_dir(path)
1048
def _delete_file_or_dir(path):
1049
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1050
# Forgiveness than Permission (EAFP) because:
1051
# - root can damage a solaris file system by using unlink,
1052
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1053
# EACCES, OSX: EPERM) when invoked on a directory.
1054
if isdir(path): # Takes care of symlinks
1061
if getattr(os, 'symlink', None) is not None:
1067
def has_hardlinks():
1068
if getattr(os, 'link', None) is not None:
1074
def host_os_dereferences_symlinks():
1075
return (has_symlinks()
1076
and sys.platform not in ('cygwin', 'win32'))
1079
def readlink(abspath):
1080
"""Return a string representing the path to which the symbolic link points.
1082
:param abspath: The link absolute unicode path.
1084
This his guaranteed to return the symbolic link in unicode in all python
1087
link = abspath.encode(_fs_enc)
1088
target = os.readlink(link)
1089
target = target.decode(_fs_enc)
1093
def contains_whitespace(s):
1094
"""True if there are any whitespace characters in s."""
1095
# string.whitespace can include '\xa0' in certain locales, because it is
1096
# considered "non-breaking-space" as part of ISO-8859-1. But it
1097
# 1) Isn't a breaking whitespace
1098
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1100
# 3) '\xa0' isn't unicode safe since it is >128.
1102
# This should *not* be a unicode set of characters in case the source
1103
# string is not a Unicode string. We can auto-up-cast the characters since
1104
# they are ascii, but we don't want to auto-up-cast the string in case it
1106
for ch in ' \t\n\r\v\f':
1113
def contains_linebreaks(s):
1114
"""True if there is any vertical whitespace in s."""
1122
def relpath(base, path):
1123
"""Return path relative to base, or raise exception.
1125
The path may be either an absolute path or a path relative to the
1126
current working directory.
1128
os.path.commonprefix (python2.4) has a bad bug that it works just
1129
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1130
avoids that problem.
1133
if len(base) < MIN_ABS_PATHLENGTH:
1134
# must have space for e.g. a drive letter
1135
raise ValueError('%r is too short to calculate a relative path'
1143
if len(head) <= len(base) and head != base:
1144
raise errors.PathNotChild(rp, base)
1147
head, tail = split(head)
1152
return pathjoin(*reversed(s))
1157
def _cicp_canonical_relpath(base, path):
1158
"""Return the canonical path relative to base.
1160
Like relpath, but on case-insensitive-case-preserving file-systems, this
1161
will return the relpath as stored on the file-system rather than in the
1162
case specified in the input string, for all existing portions of the path.
1164
This will cause O(N) behaviour if called for every path in a tree; if you
1165
have a number of paths to convert, you should use canonical_relpaths().
1167
# TODO: it should be possible to optimize this for Windows by using the
1168
# win32 API FindFiles function to look for the specified name - but using
1169
# os.listdir() still gives us the correct, platform agnostic semantics in
1172
rel = relpath(base, path)
1173
# '.' will have been turned into ''
1177
abs_base = abspath(base)
1179
_listdir = os.listdir
1181
# use an explicit iterator so we can easily consume the rest on early exit.
1182
bit_iter = iter(rel.split('/'))
1183
for bit in bit_iter:
1186
next_entries = _listdir(current)
1187
except OSError: # enoent, eperm, etc
1188
# We can't find this in the filesystem, so just append the
1190
current = pathjoin(current, bit, *list(bit_iter))
1192
for look in next_entries:
1193
if lbit == look.lower():
1194
current = pathjoin(current, look)
1197
# got to the end, nothing matched, so we just return the
1198
# non-existing bits as they were specified (the filename may be
1199
# the target of a move, for example).
1200
current = pathjoin(current, bit, *list(bit_iter))
1202
return current[len(abs_base):].lstrip('/')
1204
# XXX - TODO - we need better detection/integration of case-insensitive
1205
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1206
# filesystems), for example, so could probably benefit from the same basic
1207
# support there. For now though, only Windows and OSX get that support, and
1208
# they get it for *all* file-systems!
1209
if sys.platform in ('win32', 'darwin'):
1210
canonical_relpath = _cicp_canonical_relpath
1212
canonical_relpath = relpath
1214
def canonical_relpaths(base, paths):
1215
"""Create an iterable to canonicalize a sequence of relative paths.
1217
The intent is for this implementation to use a cache, vastly speeding
1218
up multiple transformations in the same directory.
1220
# but for now, we haven't optimized...
1221
return [canonical_relpath(base, p) for p in paths]
1223
def safe_unicode(unicode_or_utf8_string):
1224
"""Coerce unicode_or_utf8_string into unicode.
1226
If it is unicode, it is returned.
1227
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1228
wrapped in a BzrBadParameterNotUnicode exception.
1230
if isinstance(unicode_or_utf8_string, unicode):
1231
return unicode_or_utf8_string
1233
return unicode_or_utf8_string.decode('utf8')
1234
except UnicodeDecodeError:
1235
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1238
def safe_utf8(unicode_or_utf8_string):
1239
"""Coerce unicode_or_utf8_string to a utf8 string.
1241
If it is a str, it is returned.
1242
If it is Unicode, it is encoded into a utf-8 string.
1244
if isinstance(unicode_or_utf8_string, str):
1245
# TODO: jam 20070209 This is overkill, and probably has an impact on
1246
# performance if we are dealing with lots of apis that want a
1249
# Make sure it is a valid utf-8 string
1250
unicode_or_utf8_string.decode('utf-8')
1251
except UnicodeDecodeError:
1252
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1253
return unicode_or_utf8_string
1254
return unicode_or_utf8_string.encode('utf-8')
1257
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1258
' Revision id generators should be creating utf8'
1262
def safe_revision_id(unicode_or_utf8_string, warn=True):
1263
"""Revision ids should now be utf8, but at one point they were unicode.
1265
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1267
:param warn: Functions that are sanitizing user data can set warn=False
1268
:return: None or a utf8 revision id.
1270
if (unicode_or_utf8_string is None
1271
or unicode_or_utf8_string.__class__ == str):
1272
return unicode_or_utf8_string
1274
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1276
return cache_utf8.encode(unicode_or_utf8_string)
1279
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1280
' generators should be creating utf8 file ids.')
1283
def safe_file_id(unicode_or_utf8_string, warn=True):
1284
"""File ids should now be utf8, but at one point they were unicode.
1286
This is the same as safe_utf8, except it uses the cached encode functions
1287
to save a little bit of performance.
1289
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1291
:param warn: Functions that are sanitizing user data can set warn=False
1292
:return: None or a utf8 file id.
1294
if (unicode_or_utf8_string is None
1295
or unicode_or_utf8_string.__class__ == str):
1296
return unicode_or_utf8_string
1298
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1300
return cache_utf8.encode(unicode_or_utf8_string)
1303
_platform_normalizes_filenames = False
1304
if sys.platform == 'darwin':
1305
_platform_normalizes_filenames = True
1308
def normalizes_filenames():
1309
"""Return True if this platform normalizes unicode filenames.
1311
Mac OSX does, Windows/Linux do not.
1313
return _platform_normalizes_filenames
1316
def _accessible_normalized_filename(path):
1317
"""Get the unicode normalized path, and if you can access the file.
1319
On platforms where the system normalizes filenames (Mac OSX),
1320
you can access a file by any path which will normalize correctly.
1321
On platforms where the system does not normalize filenames
1322
(Windows, Linux), you have to access a file by its exact path.
1324
Internally, bzr only supports NFC normalization, since that is
1325
the standard for XML documents.
1327
So return the normalized path, and a flag indicating if the file
1328
can be accessed by that path.
1331
return unicodedata.normalize('NFC', unicode(path)), True
1334
def _inaccessible_normalized_filename(path):
1335
__doc__ = _accessible_normalized_filename.__doc__
1337
normalized = unicodedata.normalize('NFC', unicode(path))
1338
return normalized, normalized == path
1341
if _platform_normalizes_filenames:
1342
normalized_filename = _accessible_normalized_filename
1344
normalized_filename = _inaccessible_normalized_filename
1347
default_terminal_width = 80
1348
"""The default terminal width for ttys.
1350
This is defined so that higher levels can share a common fallback value when
1351
terminal_width() returns None.
1355
def terminal_width():
1356
"""Return terminal width.
1358
None is returned if the width can't established precisely.
1361
- if BZR_COLUMNS is set, returns its value
1362
- if there is no controlling terminal, returns None
1363
- if COLUMNS is set, returns its value,
1365
From there, we need to query the OS to get the size of the controlling
1369
- get termios.TIOCGWINSZ
1370
- if an error occurs or a negative value is obtained, returns None
1374
- win32utils.get_console_size() decides,
1375
- returns None on error (provided default value)
1378
# If BZR_COLUMNS is set, take it, user is always right
1380
return int(os.environ['BZR_COLUMNS'])
1381
except (KeyError, ValueError):
1384
isatty = getattr(sys.stdout, 'isatty', None)
1385
if isatty is None or not isatty():
1386
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1389
# If COLUMNS is set, take it, the terminal knows better (even inside a
1390
# given terminal, the application can decide to set COLUMNS to a lower
1391
# value (splitted screen) or a bigger value (scroll bars))
1393
return int(os.environ['COLUMNS'])
1394
except (KeyError, ValueError):
1397
width, height = _terminal_size(None, None)
1399
# Consider invalid values as meaning no width
1405
def _win32_terminal_size(width, height):
1406
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1407
return width, height
1410
def _ioctl_terminal_size(width, height):
1412
import struct, fcntl, termios
1413
s = struct.pack('HHHH', 0, 0, 0, 0)
1414
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1415
height, width = struct.unpack('HHHH', x)[0:2]
1416
except (IOError, AttributeError):
1418
return width, height
1420
_terminal_size = None
1421
"""Returns the terminal size as (width, height).
1423
:param width: Default value for width.
1424
:param height: Default value for height.
1426
This is defined specifically for each OS and query the size of the controlling
1427
terminal. If any error occurs, the provided default values should be returned.
1429
if sys.platform == 'win32':
1430
_terminal_size = _win32_terminal_size
1432
_terminal_size = _ioctl_terminal_size
1435
def _terminal_size_changed(signum, frame):
1436
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1437
width, height = _terminal_size(None, None)
1438
if width is not None:
1439
os.environ['COLUMNS'] = str(width)
1441
if sys.platform == 'win32':
1442
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from ReadConsoleInput but
1443
# I've no idea how to plug that in the current design -- vila 20091216
1446
signal.signal(signal.SIGWINCH, _terminal_size_changed)
1449
def supports_executable():
1450
return sys.platform != "win32"
1453
def supports_posix_readonly():
1454
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1456
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1457
directory controls creation/deletion, etc.
1459
And under win32, readonly means that the directory itself cannot be
1460
deleted. The contents of a readonly directory can be changed, unlike POSIX
1461
where files in readonly directories cannot be added, deleted or renamed.
1463
return sys.platform != "win32"
1466
def set_or_unset_env(env_variable, value):
1467
"""Modify the environment, setting or removing the env_variable.
1469
:param env_variable: The environment variable in question
1470
:param value: The value to set the environment to. If None, then
1471
the variable will be removed.
1472
:return: The original value of the environment variable.
1474
orig_val = os.environ.get(env_variable)
1476
if orig_val is not None:
1477
del os.environ[env_variable]
1479
if isinstance(value, unicode):
1480
value = value.encode(get_user_encoding())
1481
os.environ[env_variable] = value
1485
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1488
def check_legal_path(path):
1489
"""Check whether the supplied path is legal.
1490
This is only required on Windows, so we don't test on other platforms
1493
if sys.platform != "win32":
1495
if _validWin32PathRE.match(path) is None:
1496
raise errors.IllegalPath(path)
1499
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1501
def _is_error_enotdir(e):
1502
"""Check if this exception represents ENOTDIR.
1504
Unfortunately, python is very inconsistent about the exception
1505
here. The cases are:
1506
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1507
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1508
which is the windows error code.
1509
3) Windows, Python2.5 uses errno == EINVAL and
1510
winerror == ERROR_DIRECTORY
1512
:param e: An Exception object (expected to be OSError with an errno
1513
attribute, but we should be able to cope with anything)
1514
:return: True if this represents an ENOTDIR error. False otherwise.
1516
en = getattr(e, 'errno', None)
1517
if (en == errno.ENOTDIR
1518
or (sys.platform == 'win32'
1519
and (en == _WIN32_ERROR_DIRECTORY
1520
or (en == errno.EINVAL
1521
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1527
def walkdirs(top, prefix=""):
1528
"""Yield data about all the directories in a tree.
1530
This yields all the data about the contents of a directory at a time.
1531
After each directory has been yielded, if the caller has mutated the list
1532
to exclude some directories, they are then not descended into.
1534
The data yielded is of the form:
1535
((directory-relpath, directory-path-from-top),
1536
[(relpath, basename, kind, lstat, path-from-top), ...]),
1537
- directory-relpath is the relative path of the directory being returned
1538
with respect to top. prefix is prepended to this.
1539
- directory-path-from-root is the path including top for this directory.
1540
It is suitable for use with os functions.
1541
- relpath is the relative path within the subtree being walked.
1542
- basename is the basename of the path
1543
- kind is the kind of the file now. If unknown then the file is not
1544
present within the tree - but it may be recorded as versioned. See
1546
- lstat is the stat data *if* the file was statted.
1547
- planned, not implemented:
1548
path_from_tree_root is the path from the root of the tree.
1550
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1551
allows one to walk a subtree but get paths that are relative to a tree
1553
:return: an iterator over the dirs.
1555
#TODO there is a bit of a smell where the results of the directory-
1556
# summary in this, and the path from the root, may not agree
1557
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1558
# potentially confusing output. We should make this more robust - but
1559
# not at a speed cost. RBC 20060731
1561
_directory = _directory_kind
1562
_listdir = os.listdir
1563
_kind_from_mode = file_kind_from_stat_mode
1564
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1566
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1567
relroot, _, _, _, top = pending.pop()
1569
relprefix = relroot + u'/'
1572
top_slash = top + u'/'
1575
append = dirblock.append
1577
names = sorted(_listdir(top))
1579
if not _is_error_enotdir(e):
1583
abspath = top_slash + name
1584
statvalue = _lstat(abspath)
1585
kind = _kind_from_mode(statvalue.st_mode)
1586
append((relprefix + name, name, kind, statvalue, abspath))
1587
yield (relroot, top), dirblock
1589
# push the user specified dirs from dirblock
1590
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1593
class DirReader(object):
1594
"""An interface for reading directories."""
1596
def top_prefix_to_starting_dir(self, top, prefix=""):
1597
"""Converts top and prefix to a starting dir entry
1599
:param top: A utf8 path
1600
:param prefix: An optional utf8 path to prefix output relative paths
1602
:return: A tuple starting with prefix, and ending with the native
1605
raise NotImplementedError(self.top_prefix_to_starting_dir)
1607
def read_dir(self, prefix, top):
1608
"""Read a specific dir.
1610
:param prefix: A utf8 prefix to be preprended to the path basenames.
1611
:param top: A natively encoded path to read.
1612
:return: A list of the directories contents. Each item contains:
1613
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1615
raise NotImplementedError(self.read_dir)
1618
_selected_dir_reader = None
1621
def _walkdirs_utf8(top, prefix=""):
1622
"""Yield data about all the directories in a tree.
1624
This yields the same information as walkdirs() only each entry is yielded
1625
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1626
are returned as exact byte-strings.
1628
:return: yields a tuple of (dir_info, [file_info])
1629
dir_info is (utf8_relpath, path-from-top)
1630
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1631
if top is an absolute path, path-from-top is also an absolute path.
1632
path-from-top might be unicode or utf8, but it is the correct path to
1633
pass to os functions to affect the file in question. (such as os.lstat)
1635
global _selected_dir_reader
1636
if _selected_dir_reader is None:
1637
fs_encoding = _fs_enc.upper()
1638
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1639
# Win98 doesn't have unicode apis like FindFirstFileW
1640
# TODO: We possibly could support Win98 by falling back to the
1641
# original FindFirstFile, and using TCHAR instead of WCHAR,
1642
# but that gets a bit tricky, and requires custom compiling
1645
from bzrlib._walkdirs_win32 import Win32ReadDir
1646
_selected_dir_reader = Win32ReadDir()
1649
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1650
# ANSI_X3.4-1968 is a form of ASCII
1652
from bzrlib._readdir_pyx import UTF8DirReader
1653
_selected_dir_reader = UTF8DirReader()
1654
except ImportError, e:
1655
failed_to_load_extension(e)
1658
if _selected_dir_reader is None:
1659
# Fallback to the python version
1660
_selected_dir_reader = UnicodeDirReader()
1662
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1663
# But we don't actually uses 1-3 in pending, so set them to None
1664
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1665
read_dir = _selected_dir_reader.read_dir
1666
_directory = _directory_kind
1668
relroot, _, _, _, top = pending[-1].pop()
1671
dirblock = sorted(read_dir(relroot, top))
1672
yield (relroot, top), dirblock
1673
# push the user specified dirs from dirblock
1674
next = [d for d in reversed(dirblock) if d[2] == _directory]
1676
pending.append(next)
1679
class UnicodeDirReader(DirReader):
1680
"""A dir reader for non-utf8 file systems, which transcodes."""
1682
__slots__ = ['_utf8_encode']
1685
self._utf8_encode = codecs.getencoder('utf8')
1687
def top_prefix_to_starting_dir(self, top, prefix=""):
1688
"""See DirReader.top_prefix_to_starting_dir."""
1689
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1691
def read_dir(self, prefix, top):
1692
"""Read a single directory from a non-utf8 file system.
1694
top, and the abspath element in the output are unicode, all other paths
1695
are utf8. Local disk IO is done via unicode calls to listdir etc.
1697
This is currently the fallback code path when the filesystem encoding is
1698
not UTF-8. It may be better to implement an alternative so that we can
1699
safely handle paths that are not properly decodable in the current
1702
See DirReader.read_dir for details.
1704
_utf8_encode = self._utf8_encode
1706
_listdir = os.listdir
1707
_kind_from_mode = file_kind_from_stat_mode
1710
relprefix = prefix + '/'
1713
top_slash = top + u'/'
1716
append = dirblock.append
1717
for name in sorted(_listdir(top)):
1719
name_utf8 = _utf8_encode(name)[0]
1720
except UnicodeDecodeError:
1721
raise errors.BadFilenameEncoding(
1722
_utf8_encode(relprefix)[0] + name, _fs_enc)
1723
abspath = top_slash + name
1724
statvalue = _lstat(abspath)
1725
kind = _kind_from_mode(statvalue.st_mode)
1726
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1730
def copy_tree(from_path, to_path, handlers={}):
1731
"""Copy all of the entries in from_path into to_path.
1733
:param from_path: The base directory to copy.
1734
:param to_path: The target directory. If it does not exist, it will
1736
:param handlers: A dictionary of functions, which takes a source and
1737
destinations for files, directories, etc.
1738
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1739
'file', 'directory', and 'symlink' should always exist.
1740
If they are missing, they will be replaced with 'os.mkdir()',
1741
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1743
# Now, just copy the existing cached tree to the new location
1744
# We use a cheap trick here.
1745
# Absolute paths are prefixed with the first parameter
1746
# relative paths are prefixed with the second.
1747
# So we can get both the source and target returned
1748
# without any extra work.
1750
def copy_dir(source, dest):
1753
def copy_link(source, dest):
1754
"""Copy the contents of a symlink"""
1755
link_to = os.readlink(source)
1756
os.symlink(link_to, dest)
1758
real_handlers = {'file':shutil.copy2,
1759
'symlink':copy_link,
1760
'directory':copy_dir,
1762
real_handlers.update(handlers)
1764
if not os.path.exists(to_path):
1765
real_handlers['directory'](from_path, to_path)
1767
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1768
for relpath, name, kind, st, abspath in entries:
1769
real_handlers[kind](abspath, relpath)
1772
def path_prefix_key(path):
1773
"""Generate a prefix-order path key for path.
1775
This can be used to sort paths in the same way that walkdirs does.
1777
return (dirname(path) , path)
1780
def compare_paths_prefix_order(path_a, path_b):
1781
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1782
key_a = path_prefix_key(path_a)
1783
key_b = path_prefix_key(path_b)
1784
return cmp(key_a, key_b)
1787
_cached_user_encoding = None
1790
def get_user_encoding(use_cache=True):
1791
"""Find out what the preferred user encoding is.
1793
This is generally the encoding that is used for command line parameters
1794
and file contents. This may be different from the terminal encoding
1795
or the filesystem encoding.
1797
:param use_cache: Enable cache for detected encoding.
1798
(This parameter is turned on by default,
1799
and required only for selftesting)
1801
:return: A string defining the preferred user encoding
1803
global _cached_user_encoding
1804
if _cached_user_encoding is not None and use_cache:
1805
return _cached_user_encoding
1807
if sys.platform == 'darwin':
1808
# python locale.getpreferredencoding() always return
1809
# 'mac-roman' on darwin. That's a lie.
1810
sys.platform = 'posix'
1812
if os.environ.get('LANG', None) is None:
1813
# If LANG is not set, we end up with 'ascii', which is bad
1814
# ('mac-roman' is more than ascii), so we set a default which
1815
# will give us UTF-8 (which appears to work in all cases on
1816
# OSX). Users are still free to override LANG of course, as
1817
# long as it give us something meaningful. This work-around
1818
# *may* not be needed with python 3k and/or OSX 10.5, but will
1819
# work with them too -- vila 20080908
1820
os.environ['LANG'] = 'en_US.UTF-8'
1823
sys.platform = 'darwin'
1828
user_encoding = locale.getpreferredencoding()
1829
except locale.Error, e:
1830
sys.stderr.write('bzr: warning: %s\n'
1831
' Could not determine what text encoding to use.\n'
1832
' This error usually means your Python interpreter\n'
1833
' doesn\'t support the locale set by $LANG (%s)\n'
1834
" Continuing with ascii encoding.\n"
1835
% (e, os.environ.get('LANG')))
1836
user_encoding = 'ascii'
1838
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1839
# treat that as ASCII, and not support printing unicode characters to the
1842
# For python scripts run under vim, we get '', so also treat that as ASCII
1843
if user_encoding in (None, 'cp0', ''):
1844
user_encoding = 'ascii'
1848
codecs.lookup(user_encoding)
1850
sys.stderr.write('bzr: warning:'
1851
' unknown encoding %s.'
1852
' Continuing with ascii encoding.\n'
1855
user_encoding = 'ascii'
1858
_cached_user_encoding = user_encoding
1860
return user_encoding
1863
def get_host_name():
1864
"""Return the current unicode host name.
1866
This is meant to be used in place of socket.gethostname() because that
1867
behaves inconsistently on different platforms.
1869
if sys.platform == "win32":
1871
return win32utils.get_host_name()
1874
return socket.gethostname().decode(get_user_encoding())
1877
def recv_all(socket, bytes):
1878
"""Receive an exact number of bytes.
1880
Regular Socket.recv() may return less than the requested number of bytes,
1881
dependning on what's in the OS buffer. MSG_WAITALL is not available
1882
on all platforms, but this should work everywhere. This will return
1883
less than the requested amount if the remote end closes.
1885
This isn't optimized and is intended mostly for use in testing.
1888
while len(b) < bytes:
1889
new = until_no_eintr(socket.recv, bytes - len(b))
1896
def send_all(socket, bytes, report_activity=None):
1897
"""Send all bytes on a socket.
1899
Regular socket.sendall() can give socket error 10053 on Windows. This
1900
implementation sends no more than 64k at a time, which avoids this problem.
1902
:param report_activity: Call this as bytes are read, see
1903
Transport._report_activity
1906
for pos in xrange(0, len(bytes), chunk_size):
1907
block = bytes[pos:pos+chunk_size]
1908
if report_activity is not None:
1909
report_activity(len(block), 'write')
1910
until_no_eintr(socket.sendall, block)
1913
def dereference_path(path):
1914
"""Determine the real path to a file.
1916
All parent elements are dereferenced. But the file itself is not
1918
:param path: The original path. May be absolute or relative.
1919
:return: the real path *to* the file
1921
parent, base = os.path.split(path)
1922
# The pathjoin for '.' is a workaround for Python bug #1213894.
1923
# (initial path components aren't dereferenced)
1924
return pathjoin(realpath(pathjoin('.', parent)), base)
1927
def supports_mapi():
1928
"""Return True if we can use MAPI to launch a mail client."""
1929
return sys.platform == "win32"
1932
def resource_string(package, resource_name):
1933
"""Load a resource from a package and return it as a string.
1935
Note: Only packages that start with bzrlib are currently supported.
1937
This is designed to be a lightweight implementation of resource
1938
loading in a way which is API compatible with the same API from
1940
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1941
If and when pkg_resources becomes a standard library, this routine
1944
# Check package name is within bzrlib
1945
if package == "bzrlib":
1946
resource_relpath = resource_name
1947
elif package.startswith("bzrlib."):
1948
package = package[len("bzrlib."):].replace('.', os.sep)
1949
resource_relpath = pathjoin(package, resource_name)
1951
raise errors.BzrError('resource package %s not in bzrlib' % package)
1953
# Map the resource to a file and read its contents
1954
base = dirname(bzrlib.__file__)
1955
if getattr(sys, 'frozen', None): # bzr.exe
1956
base = abspath(pathjoin(base, '..', '..'))
1957
filename = pathjoin(base, resource_relpath)
1958
return open(filename, 'rU').read()
1961
def file_kind_from_stat_mode_thunk(mode):
1962
global file_kind_from_stat_mode
1963
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1965
from bzrlib._readdir_pyx import UTF8DirReader
1966
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1967
except ImportError, e:
1968
# This is one time where we won't warn that an extension failed to
1969
# load. The extension is never available on Windows anyway.
1970
from bzrlib._readdir_py import (
1971
_kind_from_mode as file_kind_from_stat_mode
1973
return file_kind_from_stat_mode(mode)
1974
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1977
def file_kind(f, _lstat=os.lstat):
1979
return file_kind_from_stat_mode(_lstat(f).st_mode)
1981
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1982
raise errors.NoSuchFile(f)
1986
def until_no_eintr(f, *a, **kw):
1987
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1988
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1992
except (IOError, OSError), e:
1993
if e.errno == errno.EINTR:
1997
def re_compile_checked(re_string, flags=0, where=""):
1998
"""Return a compiled re, or raise a sensible error.
2000
This should only be used when compiling user-supplied REs.
2002
:param re_string: Text form of regular expression.
2003
:param flags: eg re.IGNORECASE
2004
:param where: Message explaining to the user the context where
2005
it occurred, eg 'log search filter'.
2007
# from https://bugs.launchpad.net/bzr/+bug/251352
2009
re_obj = re.compile(re_string, flags)
2014
where = ' in ' + where
2015
# despite the name 'error' is a type
2016
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2017
% (where, re_string, e))
2020
if sys.platform == "win32":
2023
return msvcrt.getch()
2028
fd = sys.stdin.fileno()
2029
settings = termios.tcgetattr(fd)
2032
ch = sys.stdin.read(1)
2034
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2038
if sys.platform == 'linux2':
2039
def _local_concurrency():
2041
prefix = 'processor'
2042
for line in file('/proc/cpuinfo', 'rb'):
2043
if line.startswith(prefix):
2044
concurrency = int(line[line.find(':')+1:]) + 1
2046
elif sys.platform == 'darwin':
2047
def _local_concurrency():
2048
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2049
stdout=subprocess.PIPE).communicate()[0]
2050
elif sys.platform[0:7] == 'freebsd':
2051
def _local_concurrency():
2052
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2053
stdout=subprocess.PIPE).communicate()[0]
2054
elif sys.platform == 'sunos5':
2055
def _local_concurrency():
2056
return subprocess.Popen(['psrinfo', '-p',],
2057
stdout=subprocess.PIPE).communicate()[0]
2058
elif sys.platform == "win32":
2059
def _local_concurrency():
2060
# This appears to return the number of cores.
2061
return os.environ.get('NUMBER_OF_PROCESSORS')
2063
def _local_concurrency():
2068
_cached_local_concurrency = None
2070
def local_concurrency(use_cache=True):
2071
"""Return how many processes can be run concurrently.
2073
Rely on platform specific implementations and default to 1 (one) if
2074
anything goes wrong.
2076
global _cached_local_concurrency
2078
if _cached_local_concurrency is not None and use_cache:
2079
return _cached_local_concurrency
2081
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2082
if concurrency is None:
2084
concurrency = _local_concurrency()
2085
except (OSError, IOError):
2088
concurrency = int(concurrency)
2089
except (TypeError, ValueError):
2092
_cached_concurrency = concurrency
2096
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2097
"""A stream writer that doesn't decode str arguments."""
2099
def __init__(self, encode, stream, errors='strict'):
2100
codecs.StreamWriter.__init__(self, stream, errors)
2101
self.encode = encode
2103
def write(self, object):
2104
if type(object) is str:
2105
self.stream.write(object)
2107
data, _ = self.encode(object, self.errors)
2108
self.stream.write(data)
added
removed
bzrlib/osutils.py
