1
# Copyright (C) 2005, 2006, 2007, 2009 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
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)
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
28
from datetime import datetime
30
from ntpath import (abspath as _nt_abspath,
32
normpath as _nt_normpath,
33
realpath as _nt_realpath,
34
splitdrive as _nt_splitdrive,
43
from tempfile import (
55
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
57
if sys.version_info < (2, 5):
58
import md5 as _mod_md5
60
import sha as _mod_sha
70
from bzrlib import symbol_versioning
73
# On win32, O_BINARY is used to indicate the file should
74
# be opened in binary mode, rather than text mode.
75
# On other platforms, O_BINARY doesn't exist, because
76
# they always open in binary mode, so it is okay to
77
# OR with 0 on those platforms
78
O_BINARY = getattr(os, 'O_BINARY', 0)
80
# O_NOINHERIT exists only on win32 too.
81
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
84
def get_unicode_argv():
86
user_encoding = get_user_encoding()
87
return [a.decode(user_encoding) for a in sys.argv[1:]]
88
except UnicodeDecodeError:
89
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
93
def make_readonly(filename):
94
"""Make a filename read-only."""
95
mod = os.lstat(filename).st_mode
96
if not stat.S_ISLNK(mod):
98
os.chmod(filename, mod)
101
def make_writable(filename):
102
mod = os.lstat(filename).st_mode
103
if not stat.S_ISLNK(mod):
105
os.chmod(filename, mod)
108
def minimum_path_selection(paths):
109
"""Return the smallset subset of paths which are outside paths.
111
:param paths: A container (and hence not None) of paths.
112
:return: A set of paths sufficient to include everything in paths via
113
is_inside, drawn from the paths parameter.
119
return path.split('/')
120
sorted_paths = sorted(list(paths), key=sort_key)
122
search_paths = [sorted_paths[0]]
123
for path in sorted_paths[1:]:
124
if not is_inside(search_paths[-1], path):
125
# This path is unique, add it
126
search_paths.append(path)
128
return set(search_paths)
135
"""Return a quoted filename filename
137
This previously used backslash quoting, but that works poorly on
139
# TODO: I'm not really sure this is the best format either.x
141
if _QUOTE_RE is None:
142
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
144
if _QUOTE_RE.search(f):
150
_directory_kind = 'directory'
153
"""Return the current umask"""
154
# Assume that people aren't messing with the umask while running
155
# XXX: This is not thread safe, but there is no way to get the
156
# umask without setting it
164
_directory_kind: "/",
166
'tree-reference': '+',
170
def kind_marker(kind):
172
return _kind_marker_map[kind]
174
raise errors.BzrError('invalid file kind %r' % kind)
177
lexists = getattr(os.path, 'lexists', None)
181
stat = getattr(os, 'lstat', os.stat)
185
if e.errno == errno.ENOENT:
188
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
191
def fancy_rename(old, new, rename_func, unlink_func):
192
"""A fancy rename, when you don't have atomic rename.
194
:param old: The old path, to rename from
195
:param new: The new path, to rename to
196
:param rename_func: The potentially non-atomic rename function
197
:param unlink_func: A way to delete the target file if the full rename succeeds
200
# sftp rename doesn't allow overwriting, so play tricks:
201
base = os.path.basename(new)
202
dirname = os.path.dirname(new)
203
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
204
tmp_name = pathjoin(dirname, tmp_name)
206
# Rename the file out of the way, but keep track if it didn't exist
207
# We don't want to grab just any exception
208
# something like EACCES should prevent us from continuing
209
# The downside is that the rename_func has to throw an exception
210
# with an errno = ENOENT, or NoSuchFile
213
rename_func(new, tmp_name)
214
except (errors.NoSuchFile,), e:
217
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
218
# function raises an IOError with errno is None when a rename fails.
219
# This then gets caught here.
220
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
223
if (getattr(e, 'errno', None) is None
224
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
232
# This may throw an exception, in which case success will
234
rename_func(old, new)
236
except (IOError, OSError), e:
237
# source and target may be aliases of each other (e.g. on a
238
# case-insensitive filesystem), so we may have accidentally renamed
239
# source by when we tried to rename target
240
if not (file_existed and e.errno in (None, errno.ENOENT)):
244
# If the file used to exist, rename it back into place
245
# otherwise just delete it from the tmp location
247
unlink_func(tmp_name)
249
rename_func(tmp_name, new)
252
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
253
# choke on a Unicode string containing a relative path if
254
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
256
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
257
def _posix_abspath(path):
258
# jam 20060426 rather than encoding to fsencoding
259
# copy posixpath.abspath, but use os.getcwdu instead
260
if not posixpath.isabs(path):
261
path = posixpath.join(getcwd(), path)
262
return posixpath.normpath(path)
265
def _posix_realpath(path):
266
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
269
def _win32_fixdrive(path):
270
"""Force drive letters to be consistent.
272
win32 is inconsistent whether it returns lower or upper case
273
and even if it was consistent the user might type the other
274
so we force it to uppercase
275
running python.exe under cmd.exe return capital C:\\
276
running win32 python inside a cygwin shell returns lowercase c:\\
278
drive, path = _nt_splitdrive(path)
279
return drive.upper() + path
282
def _win32_abspath(path):
283
# Real _nt_abspath doesn't have a problem with a unicode cwd
284
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
287
def _win98_abspath(path):
288
"""Return the absolute version of a path.
289
Windows 98 safe implementation (python reimplementation
290
of Win32 API function GetFullPathNameW)
295
# \\HOST\path => //HOST/path
296
# //HOST/path => //HOST/path
297
# path => C:/cwd/path
300
# check for absolute path
301
drive = _nt_splitdrive(path)[0]
302
if drive == '' and path[:2] not in('//','\\\\'):
304
# we cannot simply os.path.join cwd and path
305
# because os.path.join('C:','/path') produce '/path'
306
# and this is incorrect
307
if path[:1] in ('/','\\'):
308
cwd = _nt_splitdrive(cwd)[0]
310
path = cwd + '\\' + path
311
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
314
def _win32_realpath(path):
315
# Real _nt_realpath doesn't have a problem with a unicode cwd
316
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
319
def _win32_pathjoin(*args):
320
return _nt_join(*args).replace('\\', '/')
323
def _win32_normpath(path):
324
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
328
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
331
def _win32_mkdtemp(*args, **kwargs):
332
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
335
def _win32_rename(old, new):
336
"""We expect to be able to atomically replace 'new' with old.
338
On win32, if new exists, it must be moved out of the way first,
342
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
344
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
345
# If we try to rename a non-existant file onto cwd, we get
346
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
347
# if the old path doesn't exist, sometimes we get EACCES
348
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
354
return unicodedata.normalize('NFC', os.getcwdu())
357
# Default is to just use the python builtins, but these can be rebound on
358
# particular platforms.
359
abspath = _posix_abspath
360
realpath = _posix_realpath
361
pathjoin = os.path.join
362
normpath = os.path.normpath
365
dirname = os.path.dirname
366
basename = os.path.basename
367
split = os.path.split
368
splitext = os.path.splitext
369
# These were already imported into local scope
370
# mkdtemp = tempfile.mkdtemp
371
# rmtree = shutil.rmtree
373
MIN_ABS_PATHLENGTH = 1
376
if sys.platform == 'win32':
377
if win32utils.winver == 'Windows 98':
378
abspath = _win98_abspath
380
abspath = _win32_abspath
381
realpath = _win32_realpath
382
pathjoin = _win32_pathjoin
383
normpath = _win32_normpath
384
getcwd = _win32_getcwd
385
mkdtemp = _win32_mkdtemp
386
rename = _win32_rename
388
MIN_ABS_PATHLENGTH = 3
390
def _win32_delete_readonly(function, path, excinfo):
391
"""Error handler for shutil.rmtree function [for win32]
392
Helps to remove files and dirs marked as read-only.
394
exception = excinfo[1]
395
if function in (os.remove, os.rmdir) \
396
and isinstance(exception, OSError) \
397
and exception.errno == errno.EACCES:
403
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
404
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
405
return shutil.rmtree(path, ignore_errors, onerror)
407
f = win32utils.get_unicode_argv # special function or None
411
elif sys.platform == 'darwin':
415
def get_terminal_encoding():
416
"""Find the best encoding for printing to the screen.
418
This attempts to check both sys.stdout and sys.stdin to see
419
what encoding they are in, and if that fails it falls back to
420
osutils.get_user_encoding().
421
The problem is that on Windows, locale.getpreferredencoding()
422
is not the same encoding as that used by the console:
423
http://mail.python.org/pipermail/python-list/2003-May/162357.html
425
On my standard US Windows XP, the preferred encoding is
426
cp1252, but the console is cp437
428
from bzrlib.trace import mutter
429
output_encoding = getattr(sys.stdout, 'encoding', None)
430
if not output_encoding:
431
input_encoding = getattr(sys.stdin, 'encoding', None)
432
if not input_encoding:
433
output_encoding = get_user_encoding()
434
mutter('encoding stdout as osutils.get_user_encoding() %r',
437
output_encoding = input_encoding
438
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
440
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
441
if output_encoding == 'cp0':
442
# invalid encoding (cp0 means 'no codepage' on Windows)
443
output_encoding = get_user_encoding()
444
mutter('cp0 is invalid encoding.'
445
' encoding stdout as osutils.get_user_encoding() %r',
449
codecs.lookup(output_encoding)
451
sys.stderr.write('bzr: warning:'
452
' unknown terminal encoding %s.\n'
453
' Using encoding %s instead.\n'
454
% (output_encoding, get_user_encoding())
456
output_encoding = get_user_encoding()
458
return output_encoding
461
def normalizepath(f):
462
if getattr(os.path, 'realpath', None) is not None:
466
[p,e] = os.path.split(f)
467
if e == "" or e == "." or e == "..":
470
return pathjoin(F(p), e)
474
"""True if f is an accessible directory."""
476
return S_ISDIR(os.lstat(f)[ST_MODE])
482
"""True if f is a regular file."""
484
return S_ISREG(os.lstat(f)[ST_MODE])
489
"""True if f is a symlink."""
491
return S_ISLNK(os.lstat(f)[ST_MODE])
495
def is_inside(dir, fname):
496
"""True if fname is inside dir.
498
The parameters should typically be passed to osutils.normpath first, so
499
that . and .. and repeated slashes are eliminated, and the separators
500
are canonical for the platform.
502
The empty string as a dir name is taken as top-of-tree and matches
505
# XXX: Most callers of this can actually do something smarter by
506
# looking at the inventory
516
return fname.startswith(dir)
519
def is_inside_any(dir_list, fname):
520
"""True if fname is inside any of given dirs."""
521
for dirname in dir_list:
522
if is_inside(dirname, fname):
527
def is_inside_or_parent_of_any(dir_list, fname):
528
"""True if fname is a child or a parent of any of the given files."""
529
for dirname in dir_list:
530
if is_inside(dirname, fname) or is_inside(fname, dirname):
535
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
536
report_activity=None, direction='read'):
537
"""Copy contents of one file to another.
539
The read_length can either be -1 to read to end-of-file (EOF) or
540
it can specify the maximum number of bytes to read.
542
The buff_size represents the maximum size for each read operation
543
performed on from_file.
545
:param report_activity: Call this as bytes are read, see
546
Transport._report_activity
547
:param direction: Will be passed to report_activity
549
:return: The number of bytes copied.
553
# read specified number of bytes
555
while read_length > 0:
556
num_bytes_to_read = min(read_length, buff_size)
558
block = from_file.read(num_bytes_to_read)
562
if report_activity is not None:
563
report_activity(len(block), direction)
566
actual_bytes_read = len(block)
567
read_length -= actual_bytes_read
568
length += actual_bytes_read
572
block = from_file.read(buff_size)
576
if report_activity is not None:
577
report_activity(len(block), direction)
583
def pump_string_file(bytes, file_handle, segment_size=None):
584
"""Write bytes to file_handle in many smaller writes.
586
:param bytes: The string to write.
587
:param file_handle: The file to write to.
589
# Write data in chunks rather than all at once, because very large
590
# writes fail on some platforms (e.g. Windows with SMB mounted
593
segment_size = 5242880 # 5MB
594
segments = range(len(bytes) / segment_size + 1)
595
write = file_handle.write
596
for segment_index in segments:
597
segment = buffer(bytes, segment_index * segment_size, segment_size)
601
def file_iterator(input_file, readsize=32768):
603
b = input_file.read(readsize)
610
"""Calculate the hexdigest of an open file.
612
The file cursor should be already at the start.
624
def size_sha_file(f):
625
"""Calculate the size and hexdigest of an open file.
627
The file cursor should be already at the start and
628
the caller is responsible for closing the file afterwards.
639
return size, s.hexdigest()
642
def sha_file_by_name(fname):
643
"""Calculate the SHA1 of a file by reading the full text"""
645
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
648
b = os.read(f, 1<<16)
656
def sha_strings(strings, _factory=sha):
657
"""Return the sha-1 of concatenation of strings"""
659
map(s.update, strings)
663
def sha_string(f, _factory=sha):
664
return _factory(f).hexdigest()
667
def fingerprint_file(f):
669
return {'size': len(b),
670
'sha1': sha(b).hexdigest()}
673
def compare_files(a, b):
674
"""Returns true if equal in contents"""
685
def local_time_offset(t=None):
686
"""Return offset of local zone from GMT, either at present or at time t."""
689
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
690
return offset.days * 86400 + offset.seconds
692
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
694
def format_date(t, offset=0, timezone='original', date_fmt=None,
696
"""Return a formatted date string.
698
:param t: Seconds since the epoch.
699
:param offset: Timezone offset in seconds east of utc.
700
:param timezone: How to display the time: 'utc', 'original' for the
701
timezone specified by offset, or 'local' for the process's current
703
:param date_fmt: strftime format.
704
:param show_offset: Whether to append the timezone.
706
(date_fmt, tt, offset_str) = \
707
_format_date(t, offset, timezone, date_fmt, show_offset)
708
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
709
date_str = time.strftime(date_fmt, tt)
710
return date_str + offset_str
712
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
714
"""Return an unicode date string formatted according to the current locale.
716
:param t: Seconds since the epoch.
717
:param offset: Timezone offset in seconds east of utc.
718
:param timezone: How to display the time: 'utc', 'original' for the
719
timezone specified by offset, or 'local' for the process's current
721
:param date_fmt: strftime format.
722
:param show_offset: Whether to append the timezone.
724
(date_fmt, tt, offset_str) = \
725
_format_date(t, offset, timezone, date_fmt, show_offset)
726
date_str = time.strftime(date_fmt, tt)
727
if not isinstance(date_str, unicode):
728
date_str = date_str.decode(get_user_encoding(), 'replace')
729
return date_str + offset_str
731
def _format_date(t, offset, timezone, date_fmt, show_offset):
732
if timezone == 'utc':
735
elif timezone == 'original':
738
tt = time.gmtime(t + offset)
739
elif timezone == 'local':
740
tt = time.localtime(t)
741
offset = local_time_offset(t)
743
raise errors.UnsupportedTimezoneFormat(timezone)
745
date_fmt = "%a %Y-%m-%d %H:%M:%S"
747
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
750
return (date_fmt, tt, offset_str)
753
def compact_date(when):
754
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
757
def format_delta(delta):
758
"""Get a nice looking string for a time delta.
760
:param delta: The time difference in seconds, can be positive or negative.
761
positive indicates time in the past, negative indicates time in the
762
future. (usually time.time() - stored_time)
763
:return: String formatted to show approximate resolution
769
direction = 'in the future'
773
if seconds < 90: # print seconds up to 90 seconds
775
return '%d second %s' % (seconds, direction,)
777
return '%d seconds %s' % (seconds, direction)
779
minutes = int(seconds / 60)
780
seconds -= 60 * minutes
785
if minutes < 90: # print minutes, seconds up to 90 minutes
787
return '%d minute, %d second%s %s' % (
788
minutes, seconds, plural_seconds, direction)
790
return '%d minutes, %d second%s %s' % (
791
minutes, seconds, plural_seconds, direction)
793
hours = int(minutes / 60)
794
minutes -= 60 * hours
801
return '%d hour, %d minute%s %s' % (hours, minutes,
802
plural_minutes, direction)
803
return '%d hours, %d minute%s %s' % (hours, minutes,
804
plural_minutes, direction)
807
"""Return size of given open file."""
808
return os.fstat(f.fileno())[ST_SIZE]
811
# Define rand_bytes based on platform.
813
# Python 2.4 and later have os.urandom,
814
# but it doesn't work on some arches
816
rand_bytes = os.urandom
817
except (NotImplementedError, AttributeError):
818
# If python doesn't have os.urandom, or it doesn't work,
819
# then try to first pull random data from /dev/urandom
821
rand_bytes = file('/dev/urandom', 'rb').read
822
# Otherwise, use this hack as a last resort
823
except (IOError, OSError):
824
# not well seeded, but better than nothing
829
s += chr(random.randint(0, 255))
834
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
836
"""Return a random string of num alphanumeric characters
838
The result only contains lowercase chars because it may be used on
839
case-insensitive filesystems.
842
for raw_byte in rand_bytes(num):
843
s += ALNUM[ord(raw_byte) % 36]
847
## TODO: We could later have path objects that remember their list
848
## decomposition (might be too tricksy though.)
851
"""Turn string into list of parts."""
852
# split on either delimiter because people might use either on
854
ps = re.split(r'[\\/]', p)
859
raise errors.BzrError("sorry, %r not allowed in path" % f)
860
elif (f == '.') or (f == ''):
869
if (f == '..') or (f is None) or (f == ''):
870
raise errors.BzrError("sorry, %r not allowed in path" % f)
874
def parent_directories(filename):
875
"""Return the list of parent directories, deepest first.
877
For example, parent_directories("a/b/c") -> ["a/b", "a"].
880
parts = splitpath(dirname(filename))
882
parents.append(joinpath(parts))
888
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
890
from bzrlib._chunks_to_lines_py import chunks_to_lines
894
"""Split s into lines, but without removing the newline characters."""
895
# Trivially convert a fulltext into a 'chunked' representation, and let
896
# chunks_to_lines do the heavy lifting.
897
if isinstance(s, str):
898
# chunks_to_lines only supports 8-bit strings
899
return chunks_to_lines([s])
901
return _split_lines(s)
905
"""Split s into lines, but without removing the newline characters.
907
This supports Unicode or plain string objects.
909
lines = s.split('\n')
910
result = [line + '\n' for line in lines[:-1]]
912
result.append(lines[-1])
916
def hardlinks_good():
917
return sys.platform not in ('win32', 'cygwin', 'darwin')
920
def link_or_copy(src, dest):
921
"""Hardlink a file, or copy it if it can't be hardlinked."""
922
if not hardlinks_good():
923
shutil.copyfile(src, dest)
927
except (OSError, IOError), e:
928
if e.errno != errno.EXDEV:
930
shutil.copyfile(src, dest)
933
def delete_any(path):
934
"""Delete a file, symlink or directory.
936
Will delete even if readonly.
939
_delete_file_or_dir(path)
940
except (OSError, IOError), e:
941
if e.errno in (errno.EPERM, errno.EACCES):
942
# make writable and try again
945
except (OSError, IOError):
947
_delete_file_or_dir(path)
952
def _delete_file_or_dir(path):
953
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
954
# Forgiveness than Permission (EAFP) because:
955
# - root can damage a solaris file system by using unlink,
956
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
957
# EACCES, OSX: EPERM) when invoked on a directory.
958
if isdir(path): # Takes care of symlinks
965
if getattr(os, 'symlink', None) is not None:
972
if getattr(os, 'link', None) is not None:
978
def host_os_dereferences_symlinks():
979
return (has_symlinks()
980
and sys.platform not in ('cygwin', 'win32'))
983
def readlink(abspath):
984
"""Return a string representing the path to which the symbolic link points.
986
:param abspath: The link absolute unicode path.
988
This his guaranteed to return the symbolic link in unicode in all python
991
link = abspath.encode(_fs_enc)
992
target = os.readlink(link)
993
target = target.decode(_fs_enc)
997
def contains_whitespace(s):
998
"""True if there are any whitespace characters in s."""
999
# string.whitespace can include '\xa0' in certain locales, because it is
1000
# considered "non-breaking-space" as part of ISO-8859-1. But it
1001
# 1) Isn't a breaking whitespace
1002
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1004
# 3) '\xa0' isn't unicode safe since it is >128.
1006
# This should *not* be a unicode set of characters in case the source
1007
# string is not a Unicode string. We can auto-up-cast the characters since
1008
# they are ascii, but we don't want to auto-up-cast the string in case it
1010
for ch in ' \t\n\r\v\f':
1017
def contains_linebreaks(s):
1018
"""True if there is any vertical whitespace in s."""
1026
def relpath(base, path):
1027
"""Return path relative to base, or raise exception.
1029
The path may be either an absolute path or a path relative to the
1030
current working directory.
1032
os.path.commonprefix (python2.4) has a bad bug that it works just
1033
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1034
avoids that problem.
1037
if len(base) < MIN_ABS_PATHLENGTH:
1038
# must have space for e.g. a drive letter
1039
raise ValueError('%r is too short to calculate a relative path'
1047
if len(head) <= len(base) and head != base:
1048
raise errors.PathNotChild(rp, base)
1051
head, tail = split(head)
1056
return pathjoin(*reversed(s))
1061
def _cicp_canonical_relpath(base, path):
1062
"""Return the canonical path relative to base.
1064
Like relpath, but on case-insensitive-case-preserving file-systems, this
1065
will return the relpath as stored on the file-system rather than in the
1066
case specified in the input string, for all existing portions of the path.
1068
This will cause O(N) behaviour if called for every path in a tree; if you
1069
have a number of paths to convert, you should use canonical_relpaths().
1071
# TODO: it should be possible to optimize this for Windows by using the
1072
# win32 API FindFiles function to look for the specified name - but using
1073
# os.listdir() still gives us the correct, platform agnostic semantics in
1076
rel = relpath(base, path)
1077
# '.' will have been turned into ''
1081
abs_base = abspath(base)
1083
_listdir = os.listdir
1085
# use an explicit iterator so we can easily consume the rest on early exit.
1086
bit_iter = iter(rel.split('/'))
1087
for bit in bit_iter:
1090
next_entries = _listdir(current)
1091
except OSError: # enoent, eperm, etc
1092
# We can't find this in the filesystem, so just append the
1094
current = pathjoin(current, bit, *list(bit_iter))
1096
for look in next_entries:
1097
if lbit == look.lower():
1098
current = pathjoin(current, look)
1101
# got to the end, nothing matched, so we just return the
1102
# non-existing bits as they were specified (the filename may be
1103
# the target of a move, for example).
1104
current = pathjoin(current, bit, *list(bit_iter))
1106
return current[len(abs_base):].lstrip('/')
1108
# XXX - TODO - we need better detection/integration of case-insensitive
1109
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1110
# filesystems), for example, so could probably benefit from the same basic
1111
# support there. For now though, only Windows and OSX get that support, and
1112
# they get it for *all* file-systems!
1113
if sys.platform in ('win32', 'darwin'):
1114
canonical_relpath = _cicp_canonical_relpath
1116
canonical_relpath = relpath
1118
def canonical_relpaths(base, paths):
1119
"""Create an iterable to canonicalize a sequence of relative paths.
1121
The intent is for this implementation to use a cache, vastly speeding
1122
up multiple transformations in the same directory.
1124
# but for now, we haven't optimized...
1125
return [canonical_relpath(base, p) for p in paths]
1127
def safe_unicode(unicode_or_utf8_string):
1128
"""Coerce unicode_or_utf8_string into unicode.
1130
If it is unicode, it is returned.
1131
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1132
wrapped in a BzrBadParameterNotUnicode exception.
1134
if isinstance(unicode_or_utf8_string, unicode):
1135
return unicode_or_utf8_string
1137
return unicode_or_utf8_string.decode('utf8')
1138
except UnicodeDecodeError:
1139
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1142
def safe_utf8(unicode_or_utf8_string):
1143
"""Coerce unicode_or_utf8_string to a utf8 string.
1145
If it is a str, it is returned.
1146
If it is Unicode, it is encoded into a utf-8 string.
1148
if isinstance(unicode_or_utf8_string, str):
1149
# TODO: jam 20070209 This is overkill, and probably has an impact on
1150
# performance if we are dealing with lots of apis that want a
1153
# Make sure it is a valid utf-8 string
1154
unicode_or_utf8_string.decode('utf-8')
1155
except UnicodeDecodeError:
1156
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1157
return unicode_or_utf8_string
1158
return unicode_or_utf8_string.encode('utf-8')
1161
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1162
' Revision id generators should be creating utf8'
1166
def safe_revision_id(unicode_or_utf8_string, warn=True):
1167
"""Revision ids should now be utf8, but at one point they were unicode.
1169
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1171
:param warn: Functions that are sanitizing user data can set warn=False
1172
:return: None or a utf8 revision id.
1174
if (unicode_or_utf8_string is None
1175
or unicode_or_utf8_string.__class__ == str):
1176
return unicode_or_utf8_string
1178
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1180
return cache_utf8.encode(unicode_or_utf8_string)
1183
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1184
' generators should be creating utf8 file ids.')
1187
def safe_file_id(unicode_or_utf8_string, warn=True):
1188
"""File ids should now be utf8, but at one point they were unicode.
1190
This is the same as safe_utf8, except it uses the cached encode functions
1191
to save a little bit of performance.
1193
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1195
:param warn: Functions that are sanitizing user data can set warn=False
1196
:return: None or a utf8 file id.
1198
if (unicode_or_utf8_string is None
1199
or unicode_or_utf8_string.__class__ == str):
1200
return unicode_or_utf8_string
1202
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1204
return cache_utf8.encode(unicode_or_utf8_string)
1207
_platform_normalizes_filenames = False
1208
if sys.platform == 'darwin':
1209
_platform_normalizes_filenames = True
1212
def normalizes_filenames():
1213
"""Return True if this platform normalizes unicode filenames.
1215
Mac OSX does, Windows/Linux do not.
1217
return _platform_normalizes_filenames
1220
def _accessible_normalized_filename(path):
1221
"""Get the unicode normalized path, and if you can access the file.
1223
On platforms where the system normalizes filenames (Mac OSX),
1224
you can access a file by any path which will normalize correctly.
1225
On platforms where the system does not normalize filenames
1226
(Windows, Linux), you have to access a file by its exact path.
1228
Internally, bzr only supports NFC normalization, since that is
1229
the standard for XML documents.
1231
So return the normalized path, and a flag indicating if the file
1232
can be accessed by that path.
1235
return unicodedata.normalize('NFC', unicode(path)), True
1238
def _inaccessible_normalized_filename(path):
1239
__doc__ = _accessible_normalized_filename.__doc__
1241
normalized = unicodedata.normalize('NFC', unicode(path))
1242
return normalized, normalized == path
1245
if _platform_normalizes_filenames:
1246
normalized_filename = _accessible_normalized_filename
1248
normalized_filename = _inaccessible_normalized_filename
1251
def terminal_width():
1252
"""Return estimated terminal width."""
1253
if sys.platform == 'win32':
1254
return win32utils.get_console_size()[0]
1257
import struct, fcntl, termios
1258
s = struct.pack('HHHH', 0, 0, 0, 0)
1259
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1260
width = struct.unpack('HHHH', x)[1]
1265
width = int(os.environ['COLUMNS'])
1274
def supports_executable():
1275
return sys.platform != "win32"
1278
def supports_posix_readonly():
1279
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1281
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1282
directory controls creation/deletion, etc.
1284
And under win32, readonly means that the directory itself cannot be
1285
deleted. The contents of a readonly directory can be changed, unlike POSIX
1286
where files in readonly directories cannot be added, deleted or renamed.
1288
return sys.platform != "win32"
1291
def set_or_unset_env(env_variable, value):
1292
"""Modify the environment, setting or removing the env_variable.
1294
:param env_variable: The environment variable in question
1295
:param value: The value to set the environment to. If None, then
1296
the variable will be removed.
1297
:return: The original value of the environment variable.
1299
orig_val = os.environ.get(env_variable)
1301
if orig_val is not None:
1302
del os.environ[env_variable]
1304
if isinstance(value, unicode):
1305
value = value.encode(get_user_encoding())
1306
os.environ[env_variable] = value
1310
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1313
def check_legal_path(path):
1314
"""Check whether the supplied path is legal.
1315
This is only required on Windows, so we don't test on other platforms
1318
if sys.platform != "win32":
1320
if _validWin32PathRE.match(path) is None:
1321
raise errors.IllegalPath(path)
1324
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1326
def _is_error_enotdir(e):
1327
"""Check if this exception represents ENOTDIR.
1329
Unfortunately, python is very inconsistent about the exception
1330
here. The cases are:
1331
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1332
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1333
which is the windows error code.
1334
3) Windows, Python2.5 uses errno == EINVAL and
1335
winerror == ERROR_DIRECTORY
1337
:param e: An Exception object (expected to be OSError with an errno
1338
attribute, but we should be able to cope with anything)
1339
:return: True if this represents an ENOTDIR error. False otherwise.
1341
en = getattr(e, 'errno', None)
1342
if (en == errno.ENOTDIR
1343
or (sys.platform == 'win32'
1344
and (en == _WIN32_ERROR_DIRECTORY
1345
or (en == errno.EINVAL
1346
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1352
def walkdirs(top, prefix=""):
1353
"""Yield data about all the directories in a tree.
1355
This yields all the data about the contents of a directory at a time.
1356
After each directory has been yielded, if the caller has mutated the list
1357
to exclude some directories, they are then not descended into.
1359
The data yielded is of the form:
1360
((directory-relpath, directory-path-from-top),
1361
[(relpath, basename, kind, lstat, path-from-top), ...]),
1362
- directory-relpath is the relative path of the directory being returned
1363
with respect to top. prefix is prepended to this.
1364
- directory-path-from-root is the path including top for this directory.
1365
It is suitable for use with os functions.
1366
- relpath is the relative path within the subtree being walked.
1367
- basename is the basename of the path
1368
- kind is the kind of the file now. If unknown then the file is not
1369
present within the tree - but it may be recorded as versioned. See
1371
- lstat is the stat data *if* the file was statted.
1372
- planned, not implemented:
1373
path_from_tree_root is the path from the root of the tree.
1375
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1376
allows one to walk a subtree but get paths that are relative to a tree
1378
:return: an iterator over the dirs.
1380
#TODO there is a bit of a smell where the results of the directory-
1381
# summary in this, and the path from the root, may not agree
1382
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1383
# potentially confusing output. We should make this more robust - but
1384
# not at a speed cost. RBC 20060731
1386
_directory = _directory_kind
1387
_listdir = os.listdir
1388
_kind_from_mode = file_kind_from_stat_mode
1389
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1391
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1392
relroot, _, _, _, top = pending.pop()
1394
relprefix = relroot + u'/'
1397
top_slash = top + u'/'
1400
append = dirblock.append
1402
names = sorted(_listdir(top))
1404
if not _is_error_enotdir(e):
1408
abspath = top_slash + name
1409
statvalue = _lstat(abspath)
1410
kind = _kind_from_mode(statvalue.st_mode)
1411
append((relprefix + name, name, kind, statvalue, abspath))
1412
yield (relroot, top), dirblock
1414
# push the user specified dirs from dirblock
1415
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1418
class DirReader(object):
1419
"""An interface for reading directories."""
1421
def top_prefix_to_starting_dir(self, top, prefix=""):
1422
"""Converts top and prefix to a starting dir entry
1424
:param top: A utf8 path
1425
:param prefix: An optional utf8 path to prefix output relative paths
1427
:return: A tuple starting with prefix, and ending with the native
1430
raise NotImplementedError(self.top_prefix_to_starting_dir)
1432
def read_dir(self, prefix, top):
1433
"""Read a specific dir.
1435
:param prefix: A utf8 prefix to be preprended to the path basenames.
1436
:param top: A natively encoded path to read.
1437
:return: A list of the directories contents. Each item contains:
1438
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1440
raise NotImplementedError(self.read_dir)
1443
_selected_dir_reader = None
1446
def _walkdirs_utf8(top, prefix=""):
1447
"""Yield data about all the directories in a tree.
1449
This yields the same information as walkdirs() only each entry is yielded
1450
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1451
are returned as exact byte-strings.
1453
:return: yields a tuple of (dir_info, [file_info])
1454
dir_info is (utf8_relpath, path-from-top)
1455
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1456
if top is an absolute path, path-from-top is also an absolute path.
1457
path-from-top might be unicode or utf8, but it is the correct path to
1458
pass to os functions to affect the file in question. (such as os.lstat)
1460
global _selected_dir_reader
1461
if _selected_dir_reader is None:
1462
fs_encoding = _fs_enc.upper()
1463
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1464
# Win98 doesn't have unicode apis like FindFirstFileW
1465
# TODO: We possibly could support Win98 by falling back to the
1466
# original FindFirstFile, and using TCHAR instead of WCHAR,
1467
# but that gets a bit tricky, and requires custom compiling
1470
from bzrlib._walkdirs_win32 import Win32ReadDir
1471
_selected_dir_reader = Win32ReadDir()
1474
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1475
# ANSI_X3.4-1968 is a form of ASCII
1477
from bzrlib._readdir_pyx import UTF8DirReader
1478
_selected_dir_reader = UTF8DirReader()
1482
if _selected_dir_reader is None:
1483
# Fallback to the python version
1484
_selected_dir_reader = UnicodeDirReader()
1486
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1487
# But we don't actually uses 1-3 in pending, so set them to None
1488
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1489
read_dir = _selected_dir_reader.read_dir
1490
_directory = _directory_kind
1492
relroot, _, _, _, top = pending[-1].pop()
1495
dirblock = sorted(read_dir(relroot, top))
1496
yield (relroot, top), dirblock
1497
# push the user specified dirs from dirblock
1498
next = [d for d in reversed(dirblock) if d[2] == _directory]
1500
pending.append(next)
1503
class UnicodeDirReader(DirReader):
1504
"""A dir reader for non-utf8 file systems, which transcodes."""
1506
__slots__ = ['_utf8_encode']
1509
self._utf8_encode = codecs.getencoder('utf8')
1511
def top_prefix_to_starting_dir(self, top, prefix=""):
1512
"""See DirReader.top_prefix_to_starting_dir."""
1513
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1515
def read_dir(self, prefix, top):
1516
"""Read a single directory from a non-utf8 file system.
1518
top, and the abspath element in the output are unicode, all other paths
1519
are utf8. Local disk IO is done via unicode calls to listdir etc.
1521
This is currently the fallback code path when the filesystem encoding is
1522
not UTF-8. It may be better to implement an alternative so that we can
1523
safely handle paths that are not properly decodable in the current
1526
See DirReader.read_dir for details.
1528
_utf8_encode = self._utf8_encode
1530
_listdir = os.listdir
1531
_kind_from_mode = file_kind_from_stat_mode
1534
relprefix = prefix + '/'
1537
top_slash = top + u'/'
1540
append = dirblock.append
1541
for name in sorted(_listdir(top)):
1543
name_utf8 = _utf8_encode(name)[0]
1544
except UnicodeDecodeError:
1545
raise errors.BadFilenameEncoding(
1546
_utf8_encode(relprefix)[0] + name, _fs_enc)
1547
abspath = top_slash + name
1548
statvalue = _lstat(abspath)
1549
kind = _kind_from_mode(statvalue.st_mode)
1550
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1554
def copy_tree(from_path, to_path, handlers={}):
1555
"""Copy all of the entries in from_path into to_path.
1557
:param from_path: The base directory to copy.
1558
:param to_path: The target directory. If it does not exist, it will
1560
:param handlers: A dictionary of functions, which takes a source and
1561
destinations for files, directories, etc.
1562
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1563
'file', 'directory', and 'symlink' should always exist.
1564
If they are missing, they will be replaced with 'os.mkdir()',
1565
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1567
# Now, just copy the existing cached tree to the new location
1568
# We use a cheap trick here.
1569
# Absolute paths are prefixed with the first parameter
1570
# relative paths are prefixed with the second.
1571
# So we can get both the source and target returned
1572
# without any extra work.
1574
def copy_dir(source, dest):
1577
def copy_link(source, dest):
1578
"""Copy the contents of a symlink"""
1579
link_to = os.readlink(source)
1580
os.symlink(link_to, dest)
1582
real_handlers = {'file':shutil.copy2,
1583
'symlink':copy_link,
1584
'directory':copy_dir,
1586
real_handlers.update(handlers)
1588
if not os.path.exists(to_path):
1589
real_handlers['directory'](from_path, to_path)
1591
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1592
for relpath, name, kind, st, abspath in entries:
1593
real_handlers[kind](abspath, relpath)
1596
def path_prefix_key(path):
1597
"""Generate a prefix-order path key for path.
1599
This can be used to sort paths in the same way that walkdirs does.
1601
return (dirname(path) , path)
1604
def compare_paths_prefix_order(path_a, path_b):
1605
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1606
key_a = path_prefix_key(path_a)
1607
key_b = path_prefix_key(path_b)
1608
return cmp(key_a, key_b)
1611
_cached_user_encoding = None
1614
def get_user_encoding(use_cache=True):
1615
"""Find out what the preferred user encoding is.
1617
This is generally the encoding that is used for command line parameters
1618
and file contents. This may be different from the terminal encoding
1619
or the filesystem encoding.
1621
:param use_cache: Enable cache for detected encoding.
1622
(This parameter is turned on by default,
1623
and required only for selftesting)
1625
:return: A string defining the preferred user encoding
1627
global _cached_user_encoding
1628
if _cached_user_encoding is not None and use_cache:
1629
return _cached_user_encoding
1631
if sys.platform == 'darwin':
1632
# python locale.getpreferredencoding() always return
1633
# 'mac-roman' on darwin. That's a lie.
1634
sys.platform = 'posix'
1636
if os.environ.get('LANG', None) is None:
1637
# If LANG is not set, we end up with 'ascii', which is bad
1638
# ('mac-roman' is more than ascii), so we set a default which
1639
# will give us UTF-8 (which appears to work in all cases on
1640
# OSX). Users are still free to override LANG of course, as
1641
# long as it give us something meaningful. This work-around
1642
# *may* not be needed with python 3k and/or OSX 10.5, but will
1643
# work with them too -- vila 20080908
1644
os.environ['LANG'] = 'en_US.UTF-8'
1647
sys.platform = 'darwin'
1652
user_encoding = locale.getpreferredencoding()
1653
except locale.Error, e:
1654
sys.stderr.write('bzr: warning: %s\n'
1655
' Could not determine what text encoding to use.\n'
1656
' This error usually means your Python interpreter\n'
1657
' doesn\'t support the locale set by $LANG (%s)\n'
1658
" Continuing with ascii encoding.\n"
1659
% (e, os.environ.get('LANG')))
1660
user_encoding = 'ascii'
1662
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1663
# treat that as ASCII, and not support printing unicode characters to the
1666
# For python scripts run under vim, we get '', so also treat that as ASCII
1667
if user_encoding in (None, 'cp0', ''):
1668
user_encoding = 'ascii'
1672
codecs.lookup(user_encoding)
1674
sys.stderr.write('bzr: warning:'
1675
' unknown encoding %s.'
1676
' Continuing with ascii encoding.\n'
1679
user_encoding = 'ascii'
1682
_cached_user_encoding = user_encoding
1684
return user_encoding
1687
def get_host_name():
1688
"""Return the current unicode host name.
1690
This is meant to be used in place of socket.gethostname() because that
1691
behaves inconsistently on different platforms.
1693
if sys.platform == "win32":
1695
return win32utils.get_host_name()
1698
return socket.gethostname().decode(get_user_encoding())
1701
def recv_all(socket, bytes):
1702
"""Receive an exact number of bytes.
1704
Regular Socket.recv() may return less than the requested number of bytes,
1705
dependning on what's in the OS buffer. MSG_WAITALL is not available
1706
on all platforms, but this should work everywhere. This will return
1707
less than the requested amount if the remote end closes.
1709
This isn't optimized and is intended mostly for use in testing.
1712
while len(b) < bytes:
1713
new = until_no_eintr(socket.recv, bytes - len(b))
1720
def send_all(socket, bytes, report_activity=None):
1721
"""Send all bytes on a socket.
1723
Regular socket.sendall() can give socket error 10053 on Windows. This
1724
implementation sends no more than 64k at a time, which avoids this problem.
1726
:param report_activity: Call this as bytes are read, see
1727
Transport._report_activity
1730
for pos in xrange(0, len(bytes), chunk_size):
1731
block = bytes[pos:pos+chunk_size]
1732
if report_activity is not None:
1733
report_activity(len(block), 'write')
1734
until_no_eintr(socket.sendall, block)
1737
def dereference_path(path):
1738
"""Determine the real path to a file.
1740
All parent elements are dereferenced. But the file itself is not
1742
:param path: The original path. May be absolute or relative.
1743
:return: the real path *to* the file
1745
parent, base = os.path.split(path)
1746
# The pathjoin for '.' is a workaround for Python bug #1213894.
1747
# (initial path components aren't dereferenced)
1748
return pathjoin(realpath(pathjoin('.', parent)), base)
1751
def supports_mapi():
1752
"""Return True if we can use MAPI to launch a mail client."""
1753
return sys.platform == "win32"
1756
def resource_string(package, resource_name):
1757
"""Load a resource from a package and return it as a string.
1759
Note: Only packages that start with bzrlib are currently supported.
1761
This is designed to be a lightweight implementation of resource
1762
loading in a way which is API compatible with the same API from
1764
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1765
If and when pkg_resources becomes a standard library, this routine
1768
# Check package name is within bzrlib
1769
if package == "bzrlib":
1770
resource_relpath = resource_name
1771
elif package.startswith("bzrlib."):
1772
package = package[len("bzrlib."):].replace('.', os.sep)
1773
resource_relpath = pathjoin(package, resource_name)
1775
raise errors.BzrError('resource package %s not in bzrlib' % package)
1777
# Map the resource to a file and read its contents
1778
base = dirname(bzrlib.__file__)
1779
if getattr(sys, 'frozen', None): # bzr.exe
1780
base = abspath(pathjoin(base, '..', '..'))
1781
filename = pathjoin(base, resource_relpath)
1782
return open(filename, 'rU').read()
1785
def file_kind_from_stat_mode_thunk(mode):
1786
global file_kind_from_stat_mode
1787
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1789
from bzrlib._readdir_pyx import UTF8DirReader
1790
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1792
from bzrlib._readdir_py import (
1793
_kind_from_mode as file_kind_from_stat_mode
1795
return file_kind_from_stat_mode(mode)
1796
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1799
def file_kind(f, _lstat=os.lstat):
1801
return file_kind_from_stat_mode(_lstat(f).st_mode)
1803
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1804
raise errors.NoSuchFile(f)
1808
def until_no_eintr(f, *a, **kw):
1809
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1810
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1814
except (IOError, OSError), e:
1815
if e.errno == errno.EINTR:
1819
def re_compile_checked(re_string, flags=0, where=""):
1820
"""Return a compiled re, or raise a sensible error.
1822
This should only be used when compiling user-supplied REs.
1824
:param re_string: Text form of regular expression.
1825
:param flags: eg re.IGNORECASE
1826
:param where: Message explaining to the user the context where
1827
it occurred, eg 'log search filter'.
1829
# from https://bugs.launchpad.net/bzr/+bug/251352
1831
re_obj = re.compile(re_string, flags)
1836
where = ' in ' + where
1837
# despite the name 'error' is a type
1838
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
1839
% (where, re_string, e))
1842
if sys.platform == "win32":
1845
return msvcrt.getch()
1850
fd = sys.stdin.fileno()
1851
settings = termios.tcgetattr(fd)
1854
ch = sys.stdin.read(1)
1856
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
1860
if sys.platform == 'linux2':
1861
def _local_concurrency():
1863
prefix = 'processor'
1864
for line in file('/proc/cpuinfo', 'rb'):
1865
if line.startswith(prefix):
1866
concurrency = int(line[line.find(':')+1:]) + 1
1868
elif sys.platform == 'darwin':
1869
def _local_concurrency():
1870
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
1871
stdout=subprocess.PIPE).communicate()[0]
1872
elif sys.platform[0:7] == 'freebsd':
1873
def _local_concurrency():
1874
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
1875
stdout=subprocess.PIPE).communicate()[0]
1876
elif sys.platform == 'sunos5':
1877
def _local_concurrency():
1878
return subprocess.Popen(['psrinfo', '-p',],
1879
stdout=subprocess.PIPE).communicate()[0]
1880
elif sys.platform == "win32":
1881
def _local_concurrency():
1882
# This appears to return the number of cores.
1883
return os.environ.get('NUMBER_OF_PROCESSORS')
1885
def _local_concurrency():
1890
_cached_local_concurrency = None
1892
def local_concurrency(use_cache=True):
1893
"""Return how many processes can be run concurrently.
1895
Rely on platform specific implementations and default to 1 (one) if
1896
anything goes wrong.
1898
global _cached_local_concurrency
1899
if _cached_local_concurrency is not None and use_cache:
1900
return _cached_local_concurrency
1903
concurrency = _local_concurrency()
1904
except (OSError, IOError):
1907
concurrency = int(concurrency)
1908
except (TypeError, ValueError):
1911
_cached_concurrency = concurrency
1914
if sys.platform == 'win32':
1915
def open(filename, mode='r', bufsize=-1):
1916
writing = 'w' in mode
1917
updating = '+' in mode
1918
binary = 'b' in mode
1920
flags = os.O_NOINHERIT
1924
flags |= os.O_WRONLY
1926
flags |= os.O_RDONLY
1927
flags |= os.O_BINARY if binary else os.O_TEXT
1929
return os.fdopen(os.open(filename, flags), mode, bufsize)
1932
open = __builtin__.open
added
removed
bzrlib/osutils.py
