1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005-2010 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, types
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
22
from errors import bailout
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 "
24
101
def make_readonly(filename):
25
102
"""Make a filename read-only."""
26
# TODO: probably needs to be fixed for windows
27
mod = os.stat(filename).st_mode
29
os.chmod(filename, mod)
103
mod = os.lstat(filename).st_mode
104
if not stat.S_ISLNK(mod):
106
os.chmod(filename, mod)
32
109
def make_writable(filename):
33
mod = os.stat(filename).st_mode
35
os.chmod(filename, mod)
38
_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)
40
"""Return shell-quoted filename"""
41
## We could be a bit more terse by using double-quotes etc
42
f = _QUOTE_RE.sub(r'\\\1', f)
49
mode = os.lstat(f)[ST_MODE]
57
bailout("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
# Slightly faster than using .get(, '') when the common case is that
187
lexists = getattr(os.path, 'lexists', None)
191
stat = getattr(os, 'lstat', os.stat)
195
if e.errno == errno.ENOENT:
198
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
201
def fancy_rename(old, new, rename_func, unlink_func):
202
"""A fancy rename, when you don't have atomic rename.
204
:param old: The old path, to rename from
205
:param new: The new path, to rename to
206
:param rename_func: The potentially non-atomic rename function
207
:param unlink_func: A way to delete the target file if the full rename
210
# sftp rename doesn't allow overwriting, so play tricks:
211
base = os.path.basename(new)
212
dirname = os.path.dirname(new)
213
# callers use different encodings for the paths so the following MUST
214
# respect that. We rely on python upcasting to unicode if new is unicode
215
# and keeping a str if not.
216
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
217
os.getpid(), rand_chars(10))
218
tmp_name = pathjoin(dirname, tmp_name)
220
# Rename the file out of the way, but keep track if it didn't exist
221
# We don't want to grab just any exception
222
# something like EACCES should prevent us from continuing
223
# The downside is that the rename_func has to throw an exception
224
# with an errno = ENOENT, or NoSuchFile
227
rename_func(new, tmp_name)
228
except (errors.NoSuchFile,), e:
231
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
232
# function raises an IOError with errno is None when a rename fails.
233
# This then gets caught here.
234
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
237
if (getattr(e, 'errno', None) is None
238
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
247
# This may throw an exception, in which case success will
249
rename_func(old, new)
251
except (IOError, OSError), e:
252
# source and target may be aliases of each other (e.g. on a
253
# case-insensitive filesystem), so we may have accidentally renamed
254
# source by when we tried to rename target
255
failure_exc = sys.exc_info()
256
if (file_existed and e.errno in (None, errno.ENOENT)
257
and old.lower() == new.lower()):
258
# source and target are the same file on a case-insensitive
259
# filesystem, so we don't generate an exception
263
# If the file used to exist, rename it back into place
264
# otherwise just delete it from the tmp location
266
unlink_func(tmp_name)
268
rename_func(tmp_name, new)
269
if failure_exc is not None:
270
raise failure_exc[0], failure_exc[1], failure_exc[2]
273
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
274
# choke on a Unicode string containing a relative path if
275
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
277
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
278
def _posix_abspath(path):
279
# jam 20060426 rather than encoding to fsencoding
280
# copy posixpath.abspath, but use os.getcwdu instead
281
if not posixpath.isabs(path):
282
path = posixpath.join(getcwd(), path)
283
return posixpath.normpath(path)
286
def _posix_realpath(path):
287
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
290
def _win32_fixdrive(path):
291
"""Force drive letters to be consistent.
293
win32 is inconsistent whether it returns lower or upper case
294
and even if it was consistent the user might type the other
295
so we force it to uppercase
296
running python.exe under cmd.exe return capital C:\\
297
running win32 python inside a cygwin shell returns lowercase c:\\
299
drive, path = _nt_splitdrive(path)
300
return drive.upper() + path
303
def _win32_abspath(path):
304
# Real _nt_abspath doesn't have a problem with a unicode cwd
305
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
308
def _win98_abspath(path):
309
"""Return the absolute version of a path.
310
Windows 98 safe implementation (python reimplementation
311
of Win32 API function GetFullPathNameW)
316
# \\HOST\path => //HOST/path
317
# //HOST/path => //HOST/path
318
# path => C:/cwd/path
321
# check for absolute path
322
drive = _nt_splitdrive(path)[0]
323
if drive == '' and path[:2] not in('//','\\\\'):
325
# we cannot simply os.path.join cwd and path
326
# because os.path.join('C:','/path') produce '/path'
327
# and this is incorrect
328
if path[:1] in ('/','\\'):
329
cwd = _nt_splitdrive(cwd)[0]
331
path = cwd + '\\' + path
332
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
335
def _win32_realpath(path):
336
# Real _nt_realpath doesn't have a problem with a unicode cwd
337
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
340
def _win32_pathjoin(*args):
341
return _nt_join(*args).replace('\\', '/')
344
def _win32_normpath(path):
345
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
349
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
352
def _win32_mkdtemp(*args, **kwargs):
353
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
356
def _win32_rename(old, new):
357
"""We expect to be able to atomically replace 'new' with old.
359
On win32, if new exists, it must be moved out of the way first,
363
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
365
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
366
# If we try to rename a non-existant file onto cwd, we get
367
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
368
# if the old path doesn't exist, sometimes we get EACCES
369
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
375
return unicodedata.normalize('NFC', os.getcwdu())
378
# Default is to just use the python builtins, but these can be rebound on
379
# particular platforms.
380
abspath = _posix_abspath
381
realpath = _posix_realpath
382
pathjoin = os.path.join
383
normpath = os.path.normpath
386
dirname = os.path.dirname
387
basename = os.path.basename
388
split = os.path.split
389
splitext = os.path.splitext
390
# These were already imported into local scope
391
# mkdtemp = tempfile.mkdtemp
392
# rmtree = shutil.rmtree
394
MIN_ABS_PATHLENGTH = 1
397
if sys.platform == 'win32':
398
if win32utils.winver == 'Windows 98':
399
abspath = _win98_abspath
401
abspath = _win32_abspath
402
realpath = _win32_realpath
403
pathjoin = _win32_pathjoin
404
normpath = _win32_normpath
405
getcwd = _win32_getcwd
406
mkdtemp = _win32_mkdtemp
407
rename = _win32_rename
409
MIN_ABS_PATHLENGTH = 3
411
def _win32_delete_readonly(function, path, excinfo):
412
"""Error handler for shutil.rmtree function [for win32]
413
Helps to remove files and dirs marked as read-only.
415
exception = excinfo[1]
416
if function in (os.remove, os.rmdir) \
417
and isinstance(exception, OSError) \
418
and exception.errno == errno.EACCES:
424
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
425
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
426
return shutil.rmtree(path, ignore_errors, onerror)
428
f = win32utils.get_unicode_argv # special function or None
432
elif sys.platform == 'darwin':
436
def get_terminal_encoding():
437
"""Find the best encoding for printing to the screen.
439
This attempts to check both sys.stdout and sys.stdin to see
440
what encoding they are in, and if that fails it falls back to
441
osutils.get_user_encoding().
442
The problem is that on Windows, locale.getpreferredencoding()
443
is not the same encoding as that used by the console:
444
http://mail.python.org/pipermail/python-list/2003-May/162357.html
446
On my standard US Windows XP, the preferred encoding is
447
cp1252, but the console is cp437
449
from bzrlib.trace import mutter
450
output_encoding = getattr(sys.stdout, 'encoding', None)
451
if not output_encoding:
452
input_encoding = getattr(sys.stdin, 'encoding', None)
453
if not input_encoding:
454
output_encoding = get_user_encoding()
455
mutter('encoding stdout as osutils.get_user_encoding() %r',
458
output_encoding = input_encoding
459
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
461
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
462
if output_encoding == 'cp0':
463
# invalid encoding (cp0 means 'no codepage' on Windows)
464
output_encoding = get_user_encoding()
465
mutter('cp0 is invalid encoding.'
466
' encoding stdout as osutils.get_user_encoding() %r',
470
codecs.lookup(output_encoding)
472
sys.stderr.write('bzr: warning:'
473
' unknown terminal encoding %s.\n'
474
' Using encoding %s instead.\n'
475
% (output_encoding, get_user_encoding())
477
output_encoding = get_user_encoding()
479
return output_encoding
482
def normalizepath(f):
483
if getattr(os.path, 'realpath', None) is not None:
487
[p,e] = os.path.split(f)
488
if e == "" or e == "." or e == "..":
491
return pathjoin(F(p), e)
197
706
def local_time_offset(t=None):
198
707
"""Return offset of local zone from GMT, either at present or at time t."""
199
# python2.3 localtime() can't take None
203
if time.localtime(t).tm_isdst and time.daylight:
206
return -time.timezone
209
def format_date(t, offset=0, timezone='original'):
210
## TODO: Perhaps a global option to use either universal or local time?
211
## Or perhaps just let people set $TZ?
214
assert isinstance(t, float)
710
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
711
return offset.days * 86400 + offset.seconds
713
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
714
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
717
def format_date(t, offset=0, timezone='original', date_fmt=None,
719
"""Return a formatted date string.
721
:param t: Seconds since the epoch.
722
:param offset: Timezone offset in seconds east of utc.
723
:param timezone: How to display the time: 'utc', 'original' for the
724
timezone specified by offset, or 'local' for the process's current
726
:param date_fmt: strftime format.
727
:param show_offset: Whether to append the timezone.
729
(date_fmt, tt, offset_str) = \
730
_format_date(t, offset, timezone, date_fmt, show_offset)
731
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
732
date_str = time.strftime(date_fmt, tt)
733
return date_str + offset_str
736
# Cache of formatted offset strings
740
def format_date_with_offset_in_original_timezone(t, offset=0,
741
_cache=_offset_cache):
742
"""Return a formatted date string in the original timezone.
744
This routine may be faster then format_date.
746
:param t: Seconds since the epoch.
747
:param offset: Timezone offset in seconds east of utc.
751
tt = time.gmtime(t + offset)
752
date_fmt = _default_format_by_weekday_num[tt[6]]
753
date_str = time.strftime(date_fmt, tt)
754
offset_str = _cache.get(offset, None)
755
if offset_str is None:
756
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
757
_cache[offset] = offset_str
758
return date_str + offset_str
761
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
763
"""Return an unicode date string formatted according to the current locale.
765
:param t: Seconds since the epoch.
766
:param offset: Timezone offset in seconds east of utc.
767
:param timezone: How to display the time: 'utc', 'original' for the
768
timezone specified by offset, or 'local' for the process's current
770
:param date_fmt: strftime format.
771
:param show_offset: Whether to append the timezone.
773
(date_fmt, tt, offset_str) = \
774
_format_date(t, offset, timezone, date_fmt, show_offset)
775
date_str = time.strftime(date_fmt, tt)
776
if not isinstance(date_str, unicode):
777
date_str = date_str.decode(get_user_encoding(), 'replace')
778
return date_str + offset_str
781
def _format_date(t, offset, timezone, date_fmt, show_offset):
216
782
if timezone == 'utc':
217
783
tt = time.gmtime(t)
219
785
elif timezone == 'original':
222
788
tt = time.gmtime(t + offset)
223
789
elif timezone == 'local':
224
790
tt = time.localtime(t)
225
791
offset = local_time_offset(t)
227
bailout("unsupported timezone format %r",
228
['options are "utc", "original", "local"'])
230
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
231
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
793
raise errors.UnsupportedTimezoneFormat(timezone)
795
date_fmt = "%a %Y-%m-%d %H:%M:%S"
797
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
800
return (date_fmt, tt, offset_str)
234
803
def compact_date(when):
235
804
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
807
def format_delta(delta):
808
"""Get a nice looking string for a time delta.
810
:param delta: The time difference in seconds, can be positive or negative.
811
positive indicates time in the past, negative indicates time in the
812
future. (usually time.time() - stored_time)
813
:return: String formatted to show approximate resolution
819
direction = 'in the future'
823
if seconds < 90: # print seconds up to 90 seconds
825
return '%d second %s' % (seconds, direction,)
827
return '%d seconds %s' % (seconds, direction)
829
minutes = int(seconds / 60)
830
seconds -= 60 * minutes
835
if minutes < 90: # print minutes, seconds up to 90 minutes
837
return '%d minute, %d second%s %s' % (
838
minutes, seconds, plural_seconds, direction)
840
return '%d minutes, %d second%s %s' % (
841
minutes, seconds, plural_seconds, direction)
843
hours = int(minutes / 60)
844
minutes -= 60 * hours
851
return '%d hour, %d minute%s %s' % (hours, minutes,
852
plural_minutes, direction)
853
return '%d hours, %d minute%s %s' % (hours, minutes,
854
plural_minutes, direction)
240
857
"""Return size of given open file."""
241
858
return os.fstat(f.fileno())[ST_SIZE]
244
if hasattr(os, 'urandom'): # python 2.4 and later
861
# Define rand_bytes based on platform.
863
# Python 2.4 and later have os.urandom,
864
# but it doesn't work on some arches
245
866
rand_bytes = os.urandom
247
# FIXME: No good on non-Linux
248
_rand_file = file('/dev/urandom', 'rb')
249
rand_bytes = _rand_file.read
867
except (NotImplementedError, AttributeError):
868
# If python doesn't have os.urandom, or it doesn't work,
869
# then try to first pull random data from /dev/urandom
871
rand_bytes = file('/dev/urandom', 'rb').read
872
# Otherwise, use this hack as a last resort
873
except (IOError, OSError):
874
# not well seeded, but better than nothing
879
s += chr(random.randint(0, 255))
884
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
886
"""Return a random string of num alphanumeric characters
888
The result only contains lowercase chars because it may be used on
889
case-insensitive filesystems.
892
for raw_byte in rand_bytes(num):
893
s += ALNUM[ord(raw_byte) % 36]
252
897
## TODO: We could later have path objects that remember their list
253
898
## decomposition (might be too tricksy though.)
255
900
def splitpath(p):
256
"""Turn string into list of parts.
901
"""Turn string into list of parts."""
902
# split on either delimiter because people might use either on
904
ps = re.split(r'[\\/]', p)
262
>>> splitpath('a/./b')
264
>>> splitpath('a/.b')
266
>>> splitpath('a/../b')
267
Traceback (most recent call last):
269
BzrError: ("sorry, '..' not allowed in path", [])
271
assert isinstance(p, types.StringTypes)
272
ps = [f for f in p.split('/') if f != '.']
275
bailout("sorry, %r not allowed in path" % f)
909
raise errors.BzrError("sorry, %r not allowed in path" % f)
910
elif (f == '.') or (f == ''):
279
assert isinstance(p, list)
281
919
if (f == '..') or (f is None) or (f == ''):
282
bailout("sorry, %r not allowed in path" % f)
286
def appendpath(p1, p2):
293
def extern_command(cmd, ignore_errors = False):
294
mutter('external command: %s' % `cmd`)
296
if not ignore_errors:
297
bailout('command failed')
920
raise errors.BzrError("sorry, %r not allowed in path" % f)
924
def parent_directories(filename):
925
"""Return the list of parent directories, deepest first.
927
For example, parent_directories("a/b/c") -> ["a/b", "a"].
930
parts = splitpath(dirname(filename))
932
parents.append(joinpath(parts))
937
_extension_load_failures = []
940
def failed_to_load_extension(exception):
941
"""Handle failing to load a binary extension.
943
This should be called from the ImportError block guarding the attempt to
944
import the native extension. If this function returns, the pure-Python
945
implementation should be loaded instead::
948
>>> import bzrlib._fictional_extension_pyx
949
>>> except ImportError, e:
950
>>> bzrlib.osutils.failed_to_load_extension(e)
951
>>> import bzrlib._fictional_extension_py
953
# NB: This docstring is just an example, not a doctest, because doctest
954
# currently can't cope with the use of lazy imports in this namespace --
957
# This currently doesn't report the failure at the time it occurs, because
958
# they tend to happen very early in startup when we can't check config
959
# files etc, and also we want to report all failures but not spam the user
961
from bzrlib import trace
962
exception_str = str(exception)
963
if exception_str not in _extension_load_failures:
964
trace.mutter("failed to load compiled extension: %s" % exception_str)
965
_extension_load_failures.append(exception_str)
968
def report_extension_load_failures():
969
if not _extension_load_failures:
971
from bzrlib.config import GlobalConfig
972
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
974
# the warnings framework should by default show this only once
975
from bzrlib.trace import warning
977
"bzr: warning: some compiled extensions could not be loaded; "
978
"see <https://answers.launchpad.net/bzr/+faq/703>")
979
# we no longer show the specific missing extensions here, because it makes
980
# the message too long and scary - see
981
# https://bugs.launchpad.net/bzr/+bug/430529
985
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
986
except ImportError, e:
987
failed_to_load_extension(e)
988
from bzrlib._chunks_to_lines_py import chunks_to_lines
992
"""Split s into lines, but without removing the newline characters."""
993
# Trivially convert a fulltext into a 'chunked' representation, and let
994
# chunks_to_lines do the heavy lifting.
995
if isinstance(s, str):
996
# chunks_to_lines only supports 8-bit strings
997
return chunks_to_lines([s])
999
return _split_lines(s)
1002
def _split_lines(s):
1003
"""Split s into lines, but without removing the newline characters.
1005
This supports Unicode or plain string objects.
1007
lines = s.split('\n')
1008
result = [line + '\n' for line in lines[:-1]]
1010
result.append(lines[-1])
1014
def hardlinks_good():
1015
return sys.platform not in ('win32', 'cygwin', 'darwin')
1018
def link_or_copy(src, dest):
1019
"""Hardlink a file, or copy it if it can't be hardlinked."""
1020
if not hardlinks_good():
1021
shutil.copyfile(src, dest)
1025
except (OSError, IOError), e:
1026
if e.errno != errno.EXDEV:
1028
shutil.copyfile(src, dest)
1031
def delete_any(path):
1032
"""Delete a file, symlink or directory.
1034
Will delete even if readonly.
1037
_delete_file_or_dir(path)
1038
except (OSError, IOError), e:
1039
if e.errno in (errno.EPERM, errno.EACCES):
1040
# make writable and try again
1043
except (OSError, IOError):
1045
_delete_file_or_dir(path)
1050
def _delete_file_or_dir(path):
1051
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1052
# Forgiveness than Permission (EAFP) because:
1053
# - root can damage a solaris file system by using unlink,
1054
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1055
# EACCES, OSX: EPERM) when invoked on a directory.
1056
if isdir(path): # Takes care of symlinks
1063
if getattr(os, 'symlink', None) is not None:
1069
def has_hardlinks():
1070
if getattr(os, 'link', None) is not None:
1076
def host_os_dereferences_symlinks():
1077
return (has_symlinks()
1078
and sys.platform not in ('cygwin', 'win32'))
1081
def readlink(abspath):
1082
"""Return a string representing the path to which the symbolic link points.
1084
:param abspath: The link absolute unicode path.
1086
This his guaranteed to return the symbolic link in unicode in all python
1089
link = abspath.encode(_fs_enc)
1090
target = os.readlink(link)
1091
target = target.decode(_fs_enc)
1095
def contains_whitespace(s):
1096
"""True if there are any whitespace characters in s."""
1097
# string.whitespace can include '\xa0' in certain locales, because it is
1098
# considered "non-breaking-space" as part of ISO-8859-1. But it
1099
# 1) Isn't a breaking whitespace
1100
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1102
# 3) '\xa0' isn't unicode safe since it is >128.
1104
# This should *not* be a unicode set of characters in case the source
1105
# string is not a Unicode string. We can auto-up-cast the characters since
1106
# they are ascii, but we don't want to auto-up-cast the string in case it
1108
for ch in ' \t\n\r\v\f':
1115
def contains_linebreaks(s):
1116
"""True if there is any vertical whitespace in s."""
1124
def relpath(base, path):
1125
"""Return path relative to base, or raise exception.
1127
The path may be either an absolute path or a path relative to the
1128
current working directory.
1130
os.path.commonprefix (python2.4) has a bad bug that it works just
1131
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1132
avoids that problem.
1135
if len(base) < MIN_ABS_PATHLENGTH:
1136
# must have space for e.g. a drive letter
1137
raise ValueError('%r is too short to calculate a relative path'
1145
if len(head) <= len(base) and head != base:
1146
raise errors.PathNotChild(rp, base)
1149
head, tail = split(head)
1154
return pathjoin(*reversed(s))
1159
def _cicp_canonical_relpath(base, path):
1160
"""Return the canonical path relative to base.
1162
Like relpath, but on case-insensitive-case-preserving file-systems, this
1163
will return the relpath as stored on the file-system rather than in the
1164
case specified in the input string, for all existing portions of the path.
1166
This will cause O(N) behaviour if called for every path in a tree; if you
1167
have a number of paths to convert, you should use canonical_relpaths().
1169
# TODO: it should be possible to optimize this for Windows by using the
1170
# win32 API FindFiles function to look for the specified name - but using
1171
# os.listdir() still gives us the correct, platform agnostic semantics in
1174
rel = relpath(base, path)
1175
# '.' will have been turned into ''
1179
abs_base = abspath(base)
1181
_listdir = os.listdir
1183
# use an explicit iterator so we can easily consume the rest on early exit.
1184
bit_iter = iter(rel.split('/'))
1185
for bit in bit_iter:
1188
next_entries = _listdir(current)
1189
except OSError: # enoent, eperm, etc
1190
# We can't find this in the filesystem, so just append the
1192
current = pathjoin(current, bit, *list(bit_iter))
1194
for look in next_entries:
1195
if lbit == look.lower():
1196
current = pathjoin(current, look)
1199
# got to the end, nothing matched, so we just return the
1200
# non-existing bits as they were specified (the filename may be
1201
# the target of a move, for example).
1202
current = pathjoin(current, bit, *list(bit_iter))
1204
return current[len(abs_base):].lstrip('/')
1206
# XXX - TODO - we need better detection/integration of case-insensitive
1207
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1208
# filesystems), for example, so could probably benefit from the same basic
1209
# support there. For now though, only Windows and OSX get that support, and
1210
# they get it for *all* file-systems!
1211
if sys.platform in ('win32', 'darwin'):
1212
canonical_relpath = _cicp_canonical_relpath
1214
canonical_relpath = relpath
1216
def canonical_relpaths(base, paths):
1217
"""Create an iterable to canonicalize a sequence of relative paths.
1219
The intent is for this implementation to use a cache, vastly speeding
1220
up multiple transformations in the same directory.
1222
# but for now, we haven't optimized...
1223
return [canonical_relpath(base, p) for p in paths]
1225
def safe_unicode(unicode_or_utf8_string):
1226
"""Coerce unicode_or_utf8_string into unicode.
1228
If it is unicode, it is returned.
1229
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1230
wrapped in a BzrBadParameterNotUnicode exception.
1232
if isinstance(unicode_or_utf8_string, unicode):
1233
return unicode_or_utf8_string
1235
return unicode_or_utf8_string.decode('utf8')
1236
except UnicodeDecodeError:
1237
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1240
def safe_utf8(unicode_or_utf8_string):
1241
"""Coerce unicode_or_utf8_string to a utf8 string.
1243
If it is a str, it is returned.
1244
If it is Unicode, it is encoded into a utf-8 string.
1246
if isinstance(unicode_or_utf8_string, str):
1247
# TODO: jam 20070209 This is overkill, and probably has an impact on
1248
# performance if we are dealing with lots of apis that want a
1251
# Make sure it is a valid utf-8 string
1252
unicode_or_utf8_string.decode('utf-8')
1253
except UnicodeDecodeError:
1254
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1255
return unicode_or_utf8_string
1256
return unicode_or_utf8_string.encode('utf-8')
1259
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1260
' Revision id generators should be creating utf8'
1264
def safe_revision_id(unicode_or_utf8_string, warn=True):
1265
"""Revision ids should now be utf8, but at one point they were unicode.
1267
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1269
:param warn: Functions that are sanitizing user data can set warn=False
1270
:return: None or a utf8 revision id.
1272
if (unicode_or_utf8_string is None
1273
or unicode_or_utf8_string.__class__ == str):
1274
return unicode_or_utf8_string
1276
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1278
return cache_utf8.encode(unicode_or_utf8_string)
1281
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1282
' generators should be creating utf8 file ids.')
1285
def safe_file_id(unicode_or_utf8_string, warn=True):
1286
"""File ids should now be utf8, but at one point they were unicode.
1288
This is the same as safe_utf8, except it uses the cached encode functions
1289
to save a little bit of performance.
1291
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1293
:param warn: Functions that are sanitizing user data can set warn=False
1294
:return: None or a utf8 file id.
1296
if (unicode_or_utf8_string is None
1297
or unicode_or_utf8_string.__class__ == str):
1298
return unicode_or_utf8_string
1300
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1302
return cache_utf8.encode(unicode_or_utf8_string)
1305
_platform_normalizes_filenames = False
1306
if sys.platform == 'darwin':
1307
_platform_normalizes_filenames = True
1310
def normalizes_filenames():
1311
"""Return True if this platform normalizes unicode filenames.
1313
Mac OSX does, Windows/Linux do not.
1315
return _platform_normalizes_filenames
1318
def _accessible_normalized_filename(path):
1319
"""Get the unicode normalized path, and if you can access the file.
1321
On platforms where the system normalizes filenames (Mac OSX),
1322
you can access a file by any path which will normalize correctly.
1323
On platforms where the system does not normalize filenames
1324
(Windows, Linux), you have to access a file by its exact path.
1326
Internally, bzr only supports NFC normalization, since that is
1327
the standard for XML documents.
1329
So return the normalized path, and a flag indicating if the file
1330
can be accessed by that path.
1333
return unicodedata.normalize('NFC', unicode(path)), True
1336
def _inaccessible_normalized_filename(path):
1337
__doc__ = _accessible_normalized_filename.__doc__
1339
normalized = unicodedata.normalize('NFC', unicode(path))
1340
return normalized, normalized == path
1343
if _platform_normalizes_filenames:
1344
normalized_filename = _accessible_normalized_filename
1346
normalized_filename = _inaccessible_normalized_filename
1349
default_terminal_width = 80
1350
"""The default terminal width for ttys.
1352
This is defined so that higher levels can share a common fallback value when
1353
terminal_width() returns None.
1357
def terminal_width():
1358
"""Return terminal width.
1360
None is returned if the width can't established precisely.
1363
- if BZR_COLUMNS is set, returns its value
1364
- if there is no controlling terminal, returns None
1365
- if COLUMNS is set, returns its value,
1367
From there, we need to query the OS to get the size of the controlling
1371
- get termios.TIOCGWINSZ
1372
- if an error occurs or a negative value is obtained, returns None
1376
- win32utils.get_console_size() decides,
1377
- returns None on error (provided default value)
1380
# If BZR_COLUMNS is set, take it, user is always right
1382
return int(os.environ['BZR_COLUMNS'])
1383
except (KeyError, ValueError):
1386
isatty = getattr(sys.stdout, 'isatty', None)
1387
if isatty is None or not isatty():
1388
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1391
# If COLUMNS is set, take it, the terminal knows better (even inside a
1392
# given terminal, the application can decide to set COLUMNS to a lower
1393
# value (splitted screen) or a bigger value (scroll bars))
1395
return int(os.environ['COLUMNS'])
1396
except (KeyError, ValueError):
1399
width, height = _terminal_size(None, None)
1401
# Consider invalid values as meaning no width
1407
def _win32_terminal_size(width, height):
1408
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1409
return width, height
1412
def _ioctl_terminal_size(width, height):
1414
import struct, fcntl, termios
1415
s = struct.pack('HHHH', 0, 0, 0, 0)
1416
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1417
height, width = struct.unpack('HHHH', x)[0:2]
1418
except (IOError, AttributeError):
1420
return width, height
1422
_terminal_size = None
1423
"""Returns the terminal size as (width, height).
1425
:param width: Default value for width.
1426
:param height: Default value for height.
1428
This is defined specifically for each OS and query the size of the controlling
1429
terminal. If any error occurs, the provided default values should be returned.
1431
if sys.platform == 'win32':
1432
_terminal_size = _win32_terminal_size
1434
_terminal_size = _ioctl_terminal_size
1437
def _terminal_size_changed(signum, frame):
1438
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1439
width, height = _terminal_size(None, None)
1440
if width is not None:
1441
os.environ['COLUMNS'] = str(width)
1443
if sys.platform == 'win32':
1444
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from ReadConsoleInput but
1445
# I've no idea how to plug that in the current design -- vila 20091216
1448
signal.signal(signal.SIGWINCH, _terminal_size_changed)
1451
def supports_executable():
1452
return sys.platform != "win32"
1455
def supports_posix_readonly():
1456
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1458
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1459
directory controls creation/deletion, etc.
1461
And under win32, readonly means that the directory itself cannot be
1462
deleted. The contents of a readonly directory can be changed, unlike POSIX
1463
where files in readonly directories cannot be added, deleted or renamed.
1465
return sys.platform != "win32"
1468
def set_or_unset_env(env_variable, value):
1469
"""Modify the environment, setting or removing the env_variable.
1471
:param env_variable: The environment variable in question
1472
:param value: The value to set the environment to. If None, then
1473
the variable will be removed.
1474
:return: The original value of the environment variable.
1476
orig_val = os.environ.get(env_variable)
1478
if orig_val is not None:
1479
del os.environ[env_variable]
1481
if isinstance(value, unicode):
1482
value = value.encode(get_user_encoding())
1483
os.environ[env_variable] = value
1487
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1490
def check_legal_path(path):
1491
"""Check whether the supplied path is legal.
1492
This is only required on Windows, so we don't test on other platforms
1495
if sys.platform != "win32":
1497
if _validWin32PathRE.match(path) is None:
1498
raise errors.IllegalPath(path)
1501
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1503
def _is_error_enotdir(e):
1504
"""Check if this exception represents ENOTDIR.
1506
Unfortunately, python is very inconsistent about the exception
1507
here. The cases are:
1508
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1509
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1510
which is the windows error code.
1511
3) Windows, Python2.5 uses errno == EINVAL and
1512
winerror == ERROR_DIRECTORY
1514
:param e: An Exception object (expected to be OSError with an errno
1515
attribute, but we should be able to cope with anything)
1516
:return: True if this represents an ENOTDIR error. False otherwise.
1518
en = getattr(e, 'errno', None)
1519
if (en == errno.ENOTDIR
1520
or (sys.platform == 'win32'
1521
and (en == _WIN32_ERROR_DIRECTORY
1522
or (en == errno.EINVAL
1523
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1529
def walkdirs(top, prefix=""):
1530
"""Yield data about all the directories in a tree.
1532
This yields all the data about the contents of a directory at a time.
1533
After each directory has been yielded, if the caller has mutated the list
1534
to exclude some directories, they are then not descended into.
1536
The data yielded is of the form:
1537
((directory-relpath, directory-path-from-top),
1538
[(relpath, basename, kind, lstat, path-from-top), ...]),
1539
- directory-relpath is the relative path of the directory being returned
1540
with respect to top. prefix is prepended to this.
1541
- directory-path-from-root is the path including top for this directory.
1542
It is suitable for use with os functions.
1543
- relpath is the relative path within the subtree being walked.
1544
- basename is the basename of the path
1545
- kind is the kind of the file now. If unknown then the file is not
1546
present within the tree - but it may be recorded as versioned. See
1548
- lstat is the stat data *if* the file was statted.
1549
- planned, not implemented:
1550
path_from_tree_root is the path from the root of the tree.
1552
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1553
allows one to walk a subtree but get paths that are relative to a tree
1555
:return: an iterator over the dirs.
1557
#TODO there is a bit of a smell where the results of the directory-
1558
# summary in this, and the path from the root, may not agree
1559
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1560
# potentially confusing output. We should make this more robust - but
1561
# not at a speed cost. RBC 20060731
1563
_directory = _directory_kind
1564
_listdir = os.listdir
1565
_kind_from_mode = file_kind_from_stat_mode
1566
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1568
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1569
relroot, _, _, _, top = pending.pop()
1571
relprefix = relroot + u'/'
1574
top_slash = top + u'/'
1577
append = dirblock.append
1579
names = sorted(_listdir(top))
1581
if not _is_error_enotdir(e):
1585
abspath = top_slash + name
1586
statvalue = _lstat(abspath)
1587
kind = _kind_from_mode(statvalue.st_mode)
1588
append((relprefix + name, name, kind, statvalue, abspath))
1589
yield (relroot, top), dirblock
1591
# push the user specified dirs from dirblock
1592
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1595
class DirReader(object):
1596
"""An interface for reading directories."""
1598
def top_prefix_to_starting_dir(self, top, prefix=""):
1599
"""Converts top and prefix to a starting dir entry
1601
:param top: A utf8 path
1602
:param prefix: An optional utf8 path to prefix output relative paths
1604
:return: A tuple starting with prefix, and ending with the native
1607
raise NotImplementedError(self.top_prefix_to_starting_dir)
1609
def read_dir(self, prefix, top):
1610
"""Read a specific dir.
1612
:param prefix: A utf8 prefix to be preprended to the path basenames.
1613
:param top: A natively encoded path to read.
1614
:return: A list of the directories contents. Each item contains:
1615
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1617
raise NotImplementedError(self.read_dir)
1620
_selected_dir_reader = None
1623
def _walkdirs_utf8(top, prefix=""):
1624
"""Yield data about all the directories in a tree.
1626
This yields the same information as walkdirs() only each entry is yielded
1627
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1628
are returned as exact byte-strings.
1630
:return: yields a tuple of (dir_info, [file_info])
1631
dir_info is (utf8_relpath, path-from-top)
1632
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1633
if top is an absolute path, path-from-top is also an absolute path.
1634
path-from-top might be unicode or utf8, but it is the correct path to
1635
pass to os functions to affect the file in question. (such as os.lstat)
1637
global _selected_dir_reader
1638
if _selected_dir_reader is None:
1639
fs_encoding = _fs_enc.upper()
1640
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1641
# Win98 doesn't have unicode apis like FindFirstFileW
1642
# TODO: We possibly could support Win98 by falling back to the
1643
# original FindFirstFile, and using TCHAR instead of WCHAR,
1644
# but that gets a bit tricky, and requires custom compiling
1647
from bzrlib._walkdirs_win32 import Win32ReadDir
1648
_selected_dir_reader = Win32ReadDir()
1651
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1652
# ANSI_X3.4-1968 is a form of ASCII
1654
from bzrlib._readdir_pyx import UTF8DirReader
1655
_selected_dir_reader = UTF8DirReader()
1656
except ImportError, e:
1657
failed_to_load_extension(e)
1660
if _selected_dir_reader is None:
1661
# Fallback to the python version
1662
_selected_dir_reader = UnicodeDirReader()
1664
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1665
# But we don't actually uses 1-3 in pending, so set them to None
1666
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1667
read_dir = _selected_dir_reader.read_dir
1668
_directory = _directory_kind
1670
relroot, _, _, _, top = pending[-1].pop()
1673
dirblock = sorted(read_dir(relroot, top))
1674
yield (relroot, top), dirblock
1675
# push the user specified dirs from dirblock
1676
next = [d for d in reversed(dirblock) if d[2] == _directory]
1678
pending.append(next)
1681
class UnicodeDirReader(DirReader):
1682
"""A dir reader for non-utf8 file systems, which transcodes."""
1684
__slots__ = ['_utf8_encode']
1687
self._utf8_encode = codecs.getencoder('utf8')
1689
def top_prefix_to_starting_dir(self, top, prefix=""):
1690
"""See DirReader.top_prefix_to_starting_dir."""
1691
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1693
def read_dir(self, prefix, top):
1694
"""Read a single directory from a non-utf8 file system.
1696
top, and the abspath element in the output are unicode, all other paths
1697
are utf8. Local disk IO is done via unicode calls to listdir etc.
1699
This is currently the fallback code path when the filesystem encoding is
1700
not UTF-8. It may be better to implement an alternative so that we can
1701
safely handle paths that are not properly decodable in the current
1704
See DirReader.read_dir for details.
1706
_utf8_encode = self._utf8_encode
1708
_listdir = os.listdir
1709
_kind_from_mode = file_kind_from_stat_mode
1712
relprefix = prefix + '/'
1715
top_slash = top + u'/'
1718
append = dirblock.append
1719
for name in sorted(_listdir(top)):
1721
name_utf8 = _utf8_encode(name)[0]
1722
except UnicodeDecodeError:
1723
raise errors.BadFilenameEncoding(
1724
_utf8_encode(relprefix)[0] + name, _fs_enc)
1725
abspath = top_slash + name
1726
statvalue = _lstat(abspath)
1727
kind = _kind_from_mode(statvalue.st_mode)
1728
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1732
def copy_tree(from_path, to_path, handlers={}):
1733
"""Copy all of the entries in from_path into to_path.
1735
:param from_path: The base directory to copy.
1736
:param to_path: The target directory. If it does not exist, it will
1738
:param handlers: A dictionary of functions, which takes a source and
1739
destinations for files, directories, etc.
1740
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1741
'file', 'directory', and 'symlink' should always exist.
1742
If they are missing, they will be replaced with 'os.mkdir()',
1743
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1745
# Now, just copy the existing cached tree to the new location
1746
# We use a cheap trick here.
1747
# Absolute paths are prefixed with the first parameter
1748
# relative paths are prefixed with the second.
1749
# So we can get both the source and target returned
1750
# without any extra work.
1752
def copy_dir(source, dest):
1755
def copy_link(source, dest):
1756
"""Copy the contents of a symlink"""
1757
link_to = os.readlink(source)
1758
os.symlink(link_to, dest)
1760
real_handlers = {'file':shutil.copy2,
1761
'symlink':copy_link,
1762
'directory':copy_dir,
1764
real_handlers.update(handlers)
1766
if not os.path.exists(to_path):
1767
real_handlers['directory'](from_path, to_path)
1769
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1770
for relpath, name, kind, st, abspath in entries:
1771
real_handlers[kind](abspath, relpath)
1774
def path_prefix_key(path):
1775
"""Generate a prefix-order path key for path.
1777
This can be used to sort paths in the same way that walkdirs does.
1779
return (dirname(path) , path)
1782
def compare_paths_prefix_order(path_a, path_b):
1783
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1784
key_a = path_prefix_key(path_a)
1785
key_b = path_prefix_key(path_b)
1786
return cmp(key_a, key_b)
1789
_cached_user_encoding = None
1792
def get_user_encoding(use_cache=True):
1793
"""Find out what the preferred user encoding is.
1795
This is generally the encoding that is used for command line parameters
1796
and file contents. This may be different from the terminal encoding
1797
or the filesystem encoding.
1799
:param use_cache: Enable cache for detected encoding.
1800
(This parameter is turned on by default,
1801
and required only for selftesting)
1803
:return: A string defining the preferred user encoding
1805
global _cached_user_encoding
1806
if _cached_user_encoding is not None and use_cache:
1807
return _cached_user_encoding
1809
if sys.platform == 'darwin':
1810
# python locale.getpreferredencoding() always return
1811
# 'mac-roman' on darwin. That's a lie.
1812
sys.platform = 'posix'
1814
if os.environ.get('LANG', None) is None:
1815
# If LANG is not set, we end up with 'ascii', which is bad
1816
# ('mac-roman' is more than ascii), so we set a default which
1817
# will give us UTF-8 (which appears to work in all cases on
1818
# OSX). Users are still free to override LANG of course, as
1819
# long as it give us something meaningful. This work-around
1820
# *may* not be needed with python 3k and/or OSX 10.5, but will
1821
# work with them too -- vila 20080908
1822
os.environ['LANG'] = 'en_US.UTF-8'
1825
sys.platform = 'darwin'
1830
user_encoding = locale.getpreferredencoding()
1831
except locale.Error, e:
1832
sys.stderr.write('bzr: warning: %s\n'
1833
' Could not determine what text encoding to use.\n'
1834
' This error usually means your Python interpreter\n'
1835
' doesn\'t support the locale set by $LANG (%s)\n'
1836
" Continuing with ascii encoding.\n"
1837
% (e, os.environ.get('LANG')))
1838
user_encoding = 'ascii'
1840
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1841
# treat that as ASCII, and not support printing unicode characters to the
1844
# For python scripts run under vim, we get '', so also treat that as ASCII
1845
if user_encoding in (None, 'cp0', ''):
1846
user_encoding = 'ascii'
1850
codecs.lookup(user_encoding)
1852
sys.stderr.write('bzr: warning:'
1853
' unknown encoding %s.'
1854
' Continuing with ascii encoding.\n'
1857
user_encoding = 'ascii'
1860
_cached_user_encoding = user_encoding
1862
return user_encoding
1865
def get_host_name():
1866
"""Return the current unicode host name.
1868
This is meant to be used in place of socket.gethostname() because that
1869
behaves inconsistently on different platforms.
1871
if sys.platform == "win32":
1873
return win32utils.get_host_name()
1876
return socket.gethostname().decode(get_user_encoding())
1879
def recv_all(socket, bytes):
1880
"""Receive an exact number of bytes.
1882
Regular Socket.recv() may return less than the requested number of bytes,
1883
dependning on what's in the OS buffer. MSG_WAITALL is not available
1884
on all platforms, but this should work everywhere. This will return
1885
less than the requested amount if the remote end closes.
1887
This isn't optimized and is intended mostly for use in testing.
1890
while len(b) < bytes:
1891
new = until_no_eintr(socket.recv, bytes - len(b))
1898
def send_all(socket, bytes, report_activity=None):
1899
"""Send all bytes on a socket.
1901
Regular socket.sendall() can give socket error 10053 on Windows. This
1902
implementation sends no more than 64k at a time, which avoids this problem.
1904
:param report_activity: Call this as bytes are read, see
1905
Transport._report_activity
1908
for pos in xrange(0, len(bytes), chunk_size):
1909
block = bytes[pos:pos+chunk_size]
1910
if report_activity is not None:
1911
report_activity(len(block), 'write')
1912
until_no_eintr(socket.sendall, block)
1915
def dereference_path(path):
1916
"""Determine the real path to a file.
1918
All parent elements are dereferenced. But the file itself is not
1920
:param path: The original path. May be absolute or relative.
1921
:return: the real path *to* the file
1923
parent, base = os.path.split(path)
1924
# The pathjoin for '.' is a workaround for Python bug #1213894.
1925
# (initial path components aren't dereferenced)
1926
return pathjoin(realpath(pathjoin('.', parent)), base)
1929
def supports_mapi():
1930
"""Return True if we can use MAPI to launch a mail client."""
1931
return sys.platform == "win32"
1934
def resource_string(package, resource_name):
1935
"""Load a resource from a package and return it as a string.
1937
Note: Only packages that start with bzrlib are currently supported.
1939
This is designed to be a lightweight implementation of resource
1940
loading in a way which is API compatible with the same API from
1942
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1943
If and when pkg_resources becomes a standard library, this routine
1946
# Check package name is within bzrlib
1947
if package == "bzrlib":
1948
resource_relpath = resource_name
1949
elif package.startswith("bzrlib."):
1950
package = package[len("bzrlib."):].replace('.', os.sep)
1951
resource_relpath = pathjoin(package, resource_name)
1953
raise errors.BzrError('resource package %s not in bzrlib' % package)
1955
# Map the resource to a file and read its contents
1956
base = dirname(bzrlib.__file__)
1957
if getattr(sys, 'frozen', None): # bzr.exe
1958
base = abspath(pathjoin(base, '..', '..'))
1959
filename = pathjoin(base, resource_relpath)
1960
return open(filename, 'rU').read()
1963
def file_kind_from_stat_mode_thunk(mode):
1964
global file_kind_from_stat_mode
1965
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1967
from bzrlib._readdir_pyx import UTF8DirReader
1968
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1969
except ImportError, e:
1970
# This is one time where we won't warn that an extension failed to
1971
# load. The extension is never available on Windows anyway.
1972
from bzrlib._readdir_py import (
1973
_kind_from_mode as file_kind_from_stat_mode
1975
return file_kind_from_stat_mode(mode)
1976
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1979
def file_kind(f, _lstat=os.lstat):
1981
return file_kind_from_stat_mode(_lstat(f).st_mode)
1983
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1984
raise errors.NoSuchFile(f)
1988
def until_no_eintr(f, *a, **kw):
1989
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1990
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1994
except (IOError, OSError), e:
1995
if e.errno == errno.EINTR:
1999
def re_compile_checked(re_string, flags=0, where=""):
2000
"""Return a compiled re, or raise a sensible error.
2002
This should only be used when compiling user-supplied REs.
2004
:param re_string: Text form of regular expression.
2005
:param flags: eg re.IGNORECASE
2006
:param where: Message explaining to the user the context where
2007
it occurred, eg 'log search filter'.
2009
# from https://bugs.launchpad.net/bzr/+bug/251352
2011
re_obj = re.compile(re_string, flags)
2016
where = ' in ' + where
2017
# despite the name 'error' is a type
2018
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2019
% (where, re_string, e))
2022
if sys.platform == "win32":
2025
return msvcrt.getch()
2030
fd = sys.stdin.fileno()
2031
settings = termios.tcgetattr(fd)
2034
ch = sys.stdin.read(1)
2036
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2040
if sys.platform == 'linux2':
2041
def _local_concurrency():
2043
prefix = 'processor'
2044
for line in file('/proc/cpuinfo', 'rb'):
2045
if line.startswith(prefix):
2046
concurrency = int(line[line.find(':')+1:]) + 1
2048
elif sys.platform == 'darwin':
2049
def _local_concurrency():
2050
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2051
stdout=subprocess.PIPE).communicate()[0]
2052
elif sys.platform[0:7] == 'freebsd':
2053
def _local_concurrency():
2054
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2055
stdout=subprocess.PIPE).communicate()[0]
2056
elif sys.platform == 'sunos5':
2057
def _local_concurrency():
2058
return subprocess.Popen(['psrinfo', '-p',],
2059
stdout=subprocess.PIPE).communicate()[0]
2060
elif sys.platform == "win32":
2061
def _local_concurrency():
2062
# This appears to return the number of cores.
2063
return os.environ.get('NUMBER_OF_PROCESSORS')
2065
def _local_concurrency():
2070
_cached_local_concurrency = None
2072
def local_concurrency(use_cache=True):
2073
"""Return how many processes can be run concurrently.
2075
Rely on platform specific implementations and default to 1 (one) if
2076
anything goes wrong.
2078
global _cached_local_concurrency
2080
if _cached_local_concurrency is not None and use_cache:
2081
return _cached_local_concurrency
2083
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2084
if concurrency is None:
2086
concurrency = _local_concurrency()
2087
except (OSError, IOError):
2090
concurrency = int(concurrency)
2091
except (TypeError, ValueError):
2094
_cached_concurrency = concurrency
2098
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2099
"""A stream writer that doesn't decode str arguments."""
2101
def __init__(self, encode, stream, errors='strict'):
2102
codecs.StreamWriter.__init__(self, stream, errors)
2103
self.encode = encode
2105
def write(self, object):
2106
if type(object) is str:
2107
self.stream.write(object)
2109
data, _ = self.encode(object, self.errors)
2110
self.stream.write(data)
added
removed
bzrlib/osutils.py
