14
14
# along with this program; if not, write to the Free Software
15
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from __future__ import absolute_import
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(), """
27
from .lazy_import import lazy_import
28
lazy_import(globals(), """
29
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,
35
# We need to import both shutil and rmtree as we export the later on posix
36
# and need the former on windows
38
from shutil import rmtree
41
# We need to import both tempfile and mkdtemp as we export the later on posix
42
# and need the former on windows
44
from tempfile import mkdtemp
45
from tempfile import (
52
from breezy.i18n import gettext
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
73
75
# Cross platform wall-clock time functionality with decent resolution.
83
85
# be opened in binary mode, rather than text mode.
84
86
# On other platforms, O_BINARY doesn't exist, because
85
87
# they always open in binary mode, so it is okay to
86
# OR with 0 on those platforms.
87
# O_NOINHERIT and O_TEXT exists only on win32 too.
88
# OR with 0 on those platforms
88
89
O_BINARY = getattr(os, 'O_BINARY', 0)
89
O_TEXT = getattr(os, 'O_TEXT', 0)
90
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
93
class UnsupportedTimezoneFormat(errors.BzrError):
95
_fmt = ('Unsupported timezone format "%(timezone)s", '
96
'options are "utc", "original", "local".')
98
def __init__(self, timezone):
99
self.timezone = timezone
102
92
def get_unicode_argv():
106
94
user_encoding = get_user_encoding()
107
95
return [a.decode(user_encoding) for a in sys.argv[1:]]
108
96
except UnicodeDecodeError:
109
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
110
"application locale.").format(a, user_encoding))
97
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
113
101
def make_readonly(filename):
114
102
"""Make a filename read-only."""
115
103
mod = os.lstat(filename).st_mode
116
104
if not stat.S_ISLNK(mod):
118
chmod_if_possible(filename, mod)
106
os.chmod(filename, mod)
121
109
def make_writable(filename):
122
110
mod = os.lstat(filename).st_mode
123
111
if not stat.S_ISLNK(mod):
125
chmod_if_possible(filename, mod)
128
def chmod_if_possible(filename, mode):
129
# Set file mode if that can be safely done.
130
# Sometimes even on unix the filesystem won't allow it - see
131
# https://bugs.launchpad.net/bzr/+bug/606537
133
# It is probably faster to just do the chmod, rather than
134
# doing a stat, and then trying to compare
135
os.chmod(filename, mode)
136
except (IOError, OSError) as e:
137
# Permission/access denied seems to commonly happen on smbfs; there's
138
# probably no point warning about it.
139
# <https://bugs.launchpad.net/bzr/+bug/606537>
140
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
141
trace.mutter("ignore error on chmod of %r: %r" % (
113
os.chmod(filename, mod)
147
116
def minimum_path_selection(paths):
256
223
file_existed = False
258
225
rename_func(new, tmp_name)
259
except (errors.NoSuchFile,) as e:
226
except (errors.NoSuchFile,), e:
262
229
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
263
230
# function raises an IOError with errno is None when a rename fails.
264
231
# This then gets caught here.
265
232
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
267
except Exception as e:
268
235
if (getattr(e, 'errno', None) is None
269
236
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
272
239
file_existed = True
276
# This may throw an exception, in which case success will
278
rename_func(old, new)
280
except (IOError, OSError) as e:
281
# source and target may be aliases of each other (e.g. on a
282
# case-insensitive filesystem), so we may have accidentally renamed
283
# source by when we tried to rename target
284
if (file_existed and e.errno in (None, errno.ENOENT)
285
and old.lower() == new.lower()):
286
# source and target are the same file on a case-insensitive
287
# filesystem, so we don't generate an exception
245
# This may throw an exception, in which case success will
247
rename_func(old, new)
249
except (IOError, OSError), e:
250
# source and target may be aliases of each other (e.g. on a
251
# case-insensitive filesystem), so we may have accidentally renamed
252
# source by when we tried to rename target
253
failure_exc = sys.exc_info()
254
if (file_existed and e.errno in (None, errno.ENOENT)
255
and old.lower() == new.lower()):
256
# source and target are the same file on a case-insensitive
257
# filesystem, so we don't generate an exception
293
261
# If the file used to exist, rename it back into place
296
264
unlink_func(tmp_name)
298
266
rename_func(tmp_name, new)
267
if failure_exc is not None:
268
raise failure_exc[0], failure_exc[1], failure_exc[2]
301
271
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
302
272
# choke on a Unicode string containing a relative path if
303
273
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
275
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
305
276
def _posix_abspath(path):
306
277
# jam 20060426 rather than encoding to fsencoding
307
278
# copy posixpath.abspath, but use os.getcwdu instead
308
279
if not posixpath.isabs(path):
309
280
path = posixpath.join(getcwd(), path)
310
return _posix_normpath(path)
281
return posixpath.normpath(path)
313
284
def _posix_realpath(path):
314
285
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
317
def _posix_normpath(path):
318
path = posixpath.normpath(path)
319
# Bug 861008: posixpath.normpath() returns a path normalized according to
320
# the POSIX standard, which stipulates (for compatibility reasons) that two
321
# leading slashes must not be simplified to one, and only if there are 3 or
322
# more should they be simplified as one. So we treat the leading 2 slashes
323
# as a special case here by simply removing the first slash, as we consider
324
# that breaking POSIX compatibility for this obscure feature is acceptable.
325
# This is not a paranoid precaution, as we notably get paths like this when
326
# the repo is hosted at the root of the filesystem, i.e. in "/".
327
if path.startswith('//'):
332
def _posix_path_from_environ(key):
333
"""Get unicode path from `key` in environment or None if not present
335
Note that posix systems use arbitrary byte strings for filesystem objects,
336
so a path that raises BadFilenameEncoding here may still be accessible.
338
val = os.environ.get(key, None)
339
if PY3 or val is None:
342
return val.decode(_fs_enc)
343
except UnicodeDecodeError:
344
# GZ 2011-12-12:Ideally want to include `key` in the exception message
345
raise errors.BadFilenameEncoding(val, _fs_enc)
348
def _posix_get_home_dir():
349
"""Get the home directory of the current user as a unicode path"""
350
path = posixpath.expanduser("~")
352
return path.decode(_fs_enc)
353
except AttributeError:
355
except UnicodeDecodeError:
356
raise errors.BadFilenameEncoding(path, _fs_enc)
359
def _posix_getuser_unicode():
360
"""Get username from environment or password database as unicode"""
361
name = getpass.getuser()
364
user_encoding = get_user_encoding()
366
return name.decode(user_encoding)
367
except UnicodeDecodeError:
368
raise errors.BzrError("Encoding of username %r is unsupported by %s "
369
"application locale." % (name, user_encoding))
372
288
def _win32_fixdrive(path):
373
289
"""Force drive letters to be consistent.
378
294
running python.exe under cmd.exe return capital C:\\
379
295
running win32 python inside a cygwin shell returns lowercase c:\\
381
drive, path = ntpath.splitdrive(path)
297
drive, path = _nt_splitdrive(path)
382
298
return drive.upper() + path
385
301
def _win32_abspath(path):
386
# Real ntpath.abspath doesn't have a problem with a unicode cwd
387
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
302
# Real _nt_abspath doesn't have a problem with a unicode cwd
303
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
306
def _win98_abspath(path):
307
"""Return the absolute version of a path.
308
Windows 98 safe implementation (python reimplementation
309
of Win32 API function GetFullPathNameW)
314
# \\HOST\path => //HOST/path
315
# //HOST/path => //HOST/path
316
# path => C:/cwd/path
319
# check for absolute path
320
drive = _nt_splitdrive(path)[0]
321
if drive == '' and path[:2] not in('//','\\\\'):
323
# we cannot simply os.path.join cwd and path
324
# because os.path.join('C:','/path') produce '/path'
325
# and this is incorrect
326
if path[:1] in ('/','\\'):
327
cwd = _nt_splitdrive(cwd)[0]
329
path = cwd + '\\' + path
330
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
390
333
def _win32_realpath(path):
391
# Real ntpath.realpath doesn't have a problem with a unicode cwd
392
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
334
# Real _nt_realpath doesn't have a problem with a unicode cwd
335
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
395
338
def _win32_pathjoin(*args):
396
return ntpath.join(*args).replace('\\', '/')
339
return _nt_join(*args).replace('\\', '/')
399
342
def _win32_normpath(path):
400
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
343
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
403
346
def _win32_getcwd():
404
return _win32_fixdrive(_getcwd().replace('\\', '/'))
347
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
407
350
def _win32_mkdtemp(*args, **kwargs):
429
372
def _mac_getcwd():
430
return unicodedata.normalize('NFC', _getcwd())
433
def _rename_wrap_exception(rename_func):
434
"""Adds extra information to any exceptions that come from rename().
436
The exception has an updated message and 'old_filename' and 'new_filename'
440
def _rename_wrapper(old, new):
442
rename_func(old, new)
444
detailed_error = OSError(e.errno, e.strerror +
445
" [occurred when renaming '%s' to '%s']" %
447
detailed_error.old_filename = old
448
detailed_error.new_filename = new
451
return _rename_wrapper
454
if sys.version_info > (3,):
460
# Default rename wraps os.rename()
461
rename = _rename_wrap_exception(os.rename)
373
return unicodedata.normalize('NFC', os.getcwdu())
463
376
# Default is to just use the python builtins, but these can be rebound on
464
377
# particular platforms.
465
378
abspath = _posix_abspath
466
379
realpath = _posix_realpath
467
380
pathjoin = os.path.join
468
normpath = _posix_normpath
469
path_from_environ = _posix_path_from_environ
470
_get_home_dir = _posix_get_home_dir
471
getuser_unicode = _posix_getuser_unicode
381
normpath = os.path.normpath
473
384
dirname = os.path.dirname
474
385
basename = os.path.basename
475
386
split = os.path.split
476
387
splitext = os.path.splitext
477
# These were already lazily imported into local scope
388
# These were already imported into local scope
478
389
# mkdtemp = tempfile.mkdtemp
479
390
# rmtree = shutil.rmtree
487
392
MIN_ABS_PATHLENGTH = 1
490
395
if sys.platform == 'win32':
491
abspath = _win32_abspath
396
if win32utils.winver == 'Windows 98':
397
abspath = _win98_abspath
399
abspath = _win32_abspath
492
400
realpath = _win32_realpath
493
401
pathjoin = _win32_pathjoin
494
402
normpath = _win32_normpath
495
403
getcwd = _win32_getcwd
496
404
mkdtemp = _win32_mkdtemp
497
rename = _rename_wrap_exception(_win32_rename)
499
from . import _walkdirs_win32
503
lstat = _walkdirs_win32.lstat
504
fstat = _walkdirs_win32.fstat
505
wrap_stat = _walkdirs_win32.wrap_stat
405
rename = _win32_rename
507
407
MIN_ABS_PATHLENGTH = 3
547
444
On my standard US Windows XP, the preferred encoding is
548
445
cp1252, but the console is cp437
550
:param trace: If True trace the selected encoding via mutter().
552
from .trace import mutter
447
from bzrlib.trace import mutter
553
448
output_encoding = getattr(sys.stdout, 'encoding', None)
554
449
if not output_encoding:
555
450
input_encoding = getattr(sys.stdin, 'encoding', None)
556
451
if not input_encoding:
557
452
output_encoding = get_user_encoding()
559
mutter('encoding stdout as osutils.get_user_encoding() %r',
453
mutter('encoding stdout as osutils.get_user_encoding() %r',
562
456
output_encoding = input_encoding
564
mutter('encoding stdout as sys.stdin encoding %r',
457
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
568
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
459
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
569
460
if output_encoding == 'cp0':
570
461
# invalid encoding (cp0 means 'no codepage' on Windows)
571
462
output_encoding = get_user_encoding()
573
mutter('cp0 is invalid encoding.'
463
mutter('cp0 is invalid encoding.'
574
464
' encoding stdout as osutils.get_user_encoding() %r',
578
468
codecs.lookup(output_encoding)
579
469
except LookupError:
580
sys.stderr.write('brz: warning:'
470
sys.stderr.write('bzr: warning:'
581
471
' unknown terminal encoding %s.\n'
582
472
' Using encoding %s instead.\n'
583
473
% (output_encoding, get_user_encoding())
977
855
"""Return size of given open file."""
978
return os.fstat(f.fileno())[stat.ST_SIZE]
981
# Alias os.urandom to support platforms (which?) without /dev/urandom and
982
# override if it doesn't work. Avoid checking on windows where there is
983
# significant initialisation cost that can be avoided for some bzr calls.
985
rand_bytes = os.urandom
987
if rand_bytes.__module__ != "nt":
856
return os.fstat(f.fileno())[ST_SIZE]
859
# Define rand_bytes based on platform.
861
# Python 2.4 and later have os.urandom,
862
# but it doesn't work on some arches
864
rand_bytes = os.urandom
865
except (NotImplementedError, AttributeError):
866
# If python doesn't have os.urandom, or it doesn't work,
867
# then try to first pull random data from /dev/urandom
990
except NotImplementedError:
869
rand_bytes = file('/dev/urandom', 'rb').read
870
# Otherwise, use this hack as a last resort
871
except (IOError, OSError):
991
872
# not well seeded, but better than nothing
992
873
def rand_bytes(n):
1065
943
implementation should be loaded instead::
1068
>>> import breezy._fictional_extension_pyx
946
>>> import bzrlib._fictional_extension_pyx
1069
947
>>> except ImportError, e:
1070
>>> breezy.osutils.failed_to_load_extension(e)
1071
>>> import breezy._fictional_extension_py
948
>>> bzrlib.osutils.failed_to_load_extension(e)
949
>>> import bzrlib._fictional_extension_py
1073
951
# NB: This docstring is just an example, not a doctest, because doctest
1074
952
# currently can't cope with the use of lazy imports in this namespace --
1077
955
# This currently doesn't report the failure at the time it occurs, because
1078
956
# they tend to happen very early in startup when we can't check config
1079
957
# files etc, and also we want to report all failures but not spam the user
1080
958
# with 10 warnings.
959
from bzrlib import trace
1081
960
exception_str = str(exception)
1082
961
if exception_str not in _extension_load_failures:
1083
962
trace.mutter("failed to load compiled extension: %s" % exception_str)
1087
966
def report_extension_load_failures():
1088
967
if not _extension_load_failures:
1090
if config.GlobalConfig().suppress_warning('missing_extensions'):
969
from bzrlib.config import GlobalConfig
970
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1092
972
# the warnings framework should by default show this only once
1093
from .trace import warning
973
from bzrlib.trace import warning
1095
"brz: warning: some compiled extensions could not be loaded; "
1096
"see ``brz help missing-extensions``")
975
"bzr: warning: some compiled extensions could not be loaded; "
976
"see <https://answers.launchpad.net/bzr/+faq/703>")
1097
977
# we no longer show the specific missing extensions here, because it makes
1098
978
# the message too long and scary - see
1099
979
# https://bugs.launchpad.net/bzr/+bug/430529
1103
from ._chunks_to_lines_pyx import chunks_to_lines
1104
except ImportError as e:
983
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
984
except ImportError, e:
1105
985
failed_to_load_extension(e)
1106
from ._chunks_to_lines_py import chunks_to_lines
986
from bzrlib._chunks_to_lines_py import chunks_to_lines
1109
989
def split_lines(s):
1394
1254
return unicode_or_utf8_string.encode('utf-8')
1397
def safe_revision_id(unicode_or_utf8_string):
1257
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1258
' Revision id generators should be creating utf8'
1262
def safe_revision_id(unicode_or_utf8_string, warn=True):
1398
1263
"""Revision ids should now be utf8, but at one point they were unicode.
1400
1265
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1267
:param warn: Functions that are sanitizing user data can set warn=False
1402
1268
:return: None or a utf8 revision id.
1404
1270
if (unicode_or_utf8_string is None
1405
or unicode_or_utf8_string.__class__ == bytes):
1271
or unicode_or_utf8_string.__class__ == str):
1406
1272
return unicode_or_utf8_string
1407
raise TypeError('Unicode revision ids are no longer supported. '
1408
'Revision id generators should be creating utf8 revision '
1412
def safe_file_id(unicode_or_utf8_string):
1274
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1276
return cache_utf8.encode(unicode_or_utf8_string)
1279
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1280
' generators should be creating utf8 file ids.')
1283
def safe_file_id(unicode_or_utf8_string, warn=True):
1413
1284
"""File ids should now be utf8, but at one point they were unicode.
1415
1286
This is the same as safe_utf8, except it uses the cached encode functions
1474
1344
normalized_filename = _inaccessible_normalized_filename
1477
def set_signal_handler(signum, handler, restart_syscall=True):
1478
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1479
on platforms that support that.
1481
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1482
automatically restart (by calling `signal.siginterrupt(signum,
1483
False)`). May be ignored if the feature is not available on this
1484
platform or Python version.
1488
siginterrupt = signal.siginterrupt
1490
# This python implementation doesn't provide signal support, hence no
1493
except AttributeError:
1494
# siginterrupt doesn't exist on this platform, or for this version
1496
siginterrupt = lambda signum, flag: None
1498
def sig_handler(*args):
1499
# Python resets the siginterrupt flag when a signal is
1500
# received. <http://bugs.python.org/issue8354>
1501
# As a workaround for some cases, set it back the way we want it.
1502
siginterrupt(signum, False)
1503
# Now run the handler function passed to set_signal_handler.
1506
sig_handler = handler
1507
old_handler = signal.signal(signum, sig_handler)
1509
siginterrupt(signum, False)
1513
1347
default_terminal_width = 80
1514
1348
"""The default terminal width for ttys.
1530
1358
None is returned if the width can't established precisely.
1533
- if BRZ_COLUMNS is set, returns its value
1361
- if BZR_COLUMNS is set, returns its value
1534
1362
- if there is no controlling terminal, returns None
1535
- query the OS, if the queried size has changed since the last query,
1537
1363
- if COLUMNS is set, returns its value,
1538
- if the OS has a value (even though it's never changed), return its value.
1540
1365
From there, we need to query the OS to get the size of the controlling
1543
On Unices we query the OS by:
1544
1369
- get termios.TIOCGWINSZ
1545
1370
- if an error occurs or a negative value is obtained, returns None
1547
On Windows we query the OS by:
1548
1374
- win32utils.get_console_size() decides,
1549
1375
- returns None on error (provided default value)
1551
# Note to implementors: if changing the rules for determining the width,
1552
# make sure you've considered the behaviour in these cases:
1553
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1554
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1556
# - (add more interesting cases here, if you find any)
1557
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1558
# but we don't want to register a signal handler because it is impossible
1559
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1560
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1561
# time so we can notice if the reported size has changed, which should have
1564
# If BRZ_COLUMNS is set, take it, user is always right
1565
# Except if they specified 0 in which case, impose no limit here
1378
# If BZR_COLUMNS is set, take it, user is always right
1567
width = int(os.environ['BRZ_COLUMNS'])
1380
return int(os.environ['BZR_COLUMNS'])
1568
1381
except (KeyError, ValueError):
1570
if width is not None:
1576
1384
isatty = getattr(sys.stdout, 'isatty', None)
1577
if isatty is None or not isatty():
1578
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
1385
if isatty is None or not isatty():
1386
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1582
width, height = os_size = _terminal_size(None, None)
1583
global _first_terminal_size, _terminal_size_state
1584
if _terminal_size_state == 'no_data':
1585
_first_terminal_size = os_size
1586
_terminal_size_state = 'unchanged'
1587
elif (_terminal_size_state == 'unchanged' and
1588
_first_terminal_size != os_size):
1589
_terminal_size_state = 'changed'
1591
# If the OS claims to know how wide the terminal is, and this value has
1592
# ever changed, use that.
1593
if _terminal_size_state == 'changed':
1594
if width is not None and width > 0:
1597
# If COLUMNS is set, use it.
1389
# If COLUMNS is set, take it, the terminal knows better (even inside a
1390
# given terminal, the application can decide to set COLUMNS to a lower
1391
# value (splitted screen) or a bigger value (scroll bars))
1599
1393
return int(os.environ['COLUMNS'])
1600
1394
except (KeyError, ValueError):
1603
# Finally, use an unchanged size from the OS, if we have one.
1604
if _terminal_size_state == 'unchanged':
1605
if width is not None and width > 0:
1397
width, height = _terminal_size(None, None)
1399
# Consider invalid values as meaning no width
1608
# The width could not be determined.
1612
1405
def _win32_terminal_size(width, height):
1828
1635
global _selected_dir_reader
1829
1636
if _selected_dir_reader is None:
1830
if sys.platform == "win32":
1637
fs_encoding = _fs_enc.upper()
1638
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1639
# Win98 doesn't have unicode apis like FindFirstFileW
1640
# TODO: We possibly could support Win98 by falling back to the
1641
# original FindFirstFile, and using TCHAR instead of WCHAR,
1642
# but that gets a bit tricky, and requires custom compiling
1832
from ._walkdirs_win32 import Win32ReadDir
1645
from bzrlib._walkdirs_win32 import Win32ReadDir
1833
1646
_selected_dir_reader = Win32ReadDir()
1834
1647
except ImportError:
1836
elif _fs_enc in ('utf-8', 'ascii'):
1649
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1650
# ANSI_X3.4-1968 is a form of ASCII
1838
from ._readdir_pyx import UTF8DirReader
1652
from bzrlib._readdir_pyx import UTF8DirReader
1839
1653
_selected_dir_reader = UTF8DirReader()
1840
except ImportError as e:
1654
except ImportError, e:
1841
1655
failed_to_load_extension(e)
1955
1769
real_handlers[kind](abspath, relpath)
1958
def copy_ownership_from_path(dst, src=None):
1959
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1961
If src is None, the containing directory is used as source. If chown
1962
fails, the error is ignored and a warning is printed.
1964
chown = getattr(os, 'chown', None)
1969
src = os.path.dirname(dst)
1975
chown(dst, s.st_uid, s.st_gid)
1976
except OSError as e:
1978
'Unable to copy ownership from "%s" to "%s". '
1979
'You may want to set it manually.', src, dst)
1980
trace.log_exception_quietly()
1983
1772
def path_prefix_key(path):
1984
1773
"""Generate a prefix-order path key for path.
1986
1775
This can be used to sort paths in the same way that walkdirs does.
1988
return (dirname(path), path)
1777
return (dirname(path) , path)
1991
1780
def compare_paths_prefix_order(path_a, path_b):
1998
1787
_cached_user_encoding = None
2001
def get_user_encoding():
1790
def get_user_encoding(use_cache=True):
2002
1791
"""Find out what the preferred user encoding is.
2004
1793
This is generally the encoding that is used for command line parameters
2005
1794
and file contents. This may be different from the terminal encoding
2006
1795
or the filesystem encoding.
1797
:param use_cache: Enable cache for detected encoding.
1798
(This parameter is turned on by default,
1799
and required only for selftesting)
2008
1801
:return: A string defining the preferred user encoding
2010
1803
global _cached_user_encoding
2011
if _cached_user_encoding is not None:
1804
if _cached_user_encoding is not None and use_cache:
2012
1805
return _cached_user_encoding
2014
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2015
# Use the existing locale settings and call nl_langinfo directly
2016
# rather than going through getpreferredencoding. This avoids
2017
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2018
# possibility of the setlocale call throwing an error.
2019
user_encoding = locale.nl_langinfo(locale.CODESET)
1807
if sys.platform == 'darwin':
1808
# python locale.getpreferredencoding() always return
1809
# 'mac-roman' on darwin. That's a lie.
1810
sys.platform = 'posix'
1812
if os.environ.get('LANG', None) is None:
1813
# If LANG is not set, we end up with 'ascii', which is bad
1814
# ('mac-roman' is more than ascii), so we set a default which
1815
# will give us UTF-8 (which appears to work in all cases on
1816
# OSX). Users are still free to override LANG of course, as
1817
# long as it give us something meaningful. This work-around
1818
# *may* not be needed with python 3k and/or OSX 10.5, but will
1819
# work with them too -- vila 20080908
1820
os.environ['LANG'] = 'en_US.UTF-8'
1823
sys.platform = 'darwin'
2021
# GZ 2011-12-19: On windows could call GetACP directly instead.
2022
user_encoding = locale.getpreferredencoding(False)
2025
user_encoding = codecs.lookup(user_encoding).name
2027
if user_encoding not in ("", "cp0"):
2028
sys.stderr.write('brz: warning:'
1828
user_encoding = locale.getpreferredencoding()
1829
except locale.Error, e:
1830
sys.stderr.write('bzr: warning: %s\n'
1831
' Could not determine what text encoding to use.\n'
1832
' This error usually means your Python interpreter\n'
1833
' doesn\'t support the locale set by $LANG (%s)\n'
1834
" Continuing with ascii encoding.\n"
1835
% (e, os.environ.get('LANG')))
1836
user_encoding = 'ascii'
1838
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1839
# treat that as ASCII, and not support printing unicode characters to the
1842
# For python scripts run under vim, we get '', so also treat that as ASCII
1843
if user_encoding in (None, 'cp0', ''):
1844
user_encoding = 'ascii'
1848
codecs.lookup(user_encoding)
1850
sys.stderr.write('bzr: warning:'
2029
1851
' unknown encoding %s.'
2030
1852
' Continuing with ascii encoding.\n'
2031
1853
% user_encoding
2033
user_encoding = 'ascii'
2035
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2036
if user_encoding == 'ascii':
2037
if sys.platform == 'darwin':
2038
# OSX is special-cased in Python to have a UTF-8 filesystem
2039
# encoding and previously had LANG set here if not present.
2040
user_encoding = 'utf-8'
2041
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2042
# for some other posix platforms as well.
2044
_cached_user_encoding = user_encoding
1855
user_encoding = 'ascii'
1858
_cached_user_encoding = user_encoding
2045
1860
return user_encoding
2048
def get_diff_header_encoding():
2049
return get_terminal_encoding()
2052
1863
def get_host_name():
2053
1864
"""Return the current unicode host name.
2056
1867
behaves inconsistently on different platforms.
2058
1869
if sys.platform == "win32":
2059
1871
return win32utils.get_host_name()
2063
return socket.gethostname()
2064
1874
return socket.gethostname().decode(get_user_encoding())
2067
# We must not read/write any more than 64k at a time from/to a socket so we
2068
# don't risk "no buffer space available" errors on some platforms. Windows in
2069
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2071
MAX_SOCKET_CHUNK = 64 * 1024
2073
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2074
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2075
_eno = getattr(errno, _eno, None)
2076
if _eno is not None:
2077
_end_of_stream_errors.append(_eno)
2081
def read_bytes_from_socket(sock, report_activity=None,
2082
max_read_size=MAX_SOCKET_CHUNK):
2083
"""Read up to max_read_size of bytes from sock and notify of progress.
2085
Translates "Connection reset by peer" into file-like EOF (return an
2086
empty string rather than raise an error), and repeats the recv if
2087
interrupted by a signal.
2091
data = sock.recv(max_read_size)
2092
except socket.error as e:
2094
if eno in _end_of_stream_errors:
2095
# The connection was closed by the other side. Callers expect
2096
# an empty string to signal end-of-stream.
2098
elif eno == errno.EINTR:
2099
# Retry the interrupted recv.
2103
if report_activity is not None:
2104
report_activity(len(data), 'read')
2108
def recv_all(socket, count):
1877
def recv_all(socket, bytes):
2109
1878
"""Receive an exact number of bytes.
2111
1880
Regular Socket.recv() may return less than the requested number of bytes,
2112
depending on what's in the OS buffer. MSG_WAITALL is not available
1881
dependning on what's in the OS buffer. MSG_WAITALL is not available
2113
1882
on all platforms, but this should work everywhere. This will return
2114
1883
less than the requested amount if the remote end closes.
2116
1885
This isn't optimized and is intended mostly for use in testing.
2119
while len(b) < count:
2120
new = read_bytes_from_socket(socket, None, count - len(b))
1888
while len(b) < bytes:
1889
new = until_no_eintr(socket.recv, bytes - len(b))
2127
def send_all(sock, bytes, report_activity=None):
1896
def send_all(socket, bytes, report_activity=None):
2128
1897
"""Send all bytes on a socket.
2130
Breaks large blocks in smaller chunks to avoid buffering limitations on
2131
some platforms, and catches EINTR which may be thrown if the send is
2132
interrupted by a signal.
2134
This is preferred to socket.sendall(), because it avoids portability bugs
2135
and provides activity reporting.
1899
Regular socket.sendall() can give socket error 10053 on Windows. This
1900
implementation sends no more than 64k at a time, which avoids this problem.
2137
1902
:param report_activity: Call this as bytes are read, see
2138
1903
Transport._report_activity
2141
byte_count = len(bytes)
2142
view = memoryview(bytes)
2143
while sent_total < byte_count:
2145
sent = sock.send(view[sent_total:sent_total+MAX_SOCKET_CHUNK])
2146
except (socket.error, IOError) as e:
2147
if e.args[0] in _end_of_stream_errors:
2148
raise errors.ConnectionReset(
2149
"Error trying to write to socket", e)
2150
if e.args[0] != errno.EINTR:
2154
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2157
if report_activity is not None:
2158
report_activity(sent, 'write')
2161
def connect_socket(address):
2162
# Slight variation of the socket.create_connection() function (provided by
2163
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2164
# provide it for previous python versions. Also, we don't use the timeout
2165
# parameter (provided by the python implementation) so we don't implement
2167
err = socket.error('getaddrinfo returns an empty list')
2168
host, port = address
2169
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2170
af, socktype, proto, canonname, sa = res
2173
sock = socket.socket(af, socktype, proto)
2177
except socket.error as e:
2179
# 'err' is now the most recent error
2180
if sock is not None:
1906
for pos in xrange(0, len(bytes), chunk_size):
1907
block = bytes[pos:pos+chunk_size]
1908
if report_activity is not None:
1909
report_activity(len(block), 'write')
1910
until_no_eintr(socket.sendall, block)
2185
1913
def dereference_path(path):
2213
1941
If and when pkg_resources becomes a standard library, this routine
2214
1942
can delegate to it.
2216
# Check package name is within breezy
2217
if package == "breezy":
1944
# Check package name is within bzrlib
1945
if package == "bzrlib":
2218
1946
resource_relpath = resource_name
2219
elif package.startswith("breezy."):
2220
package = package[len("breezy."):].replace('.', os.sep)
1947
elif package.startswith("bzrlib."):
1948
package = package[len("bzrlib."):].replace('.', os.sep)
2221
1949
resource_relpath = pathjoin(package, resource_name)
2223
raise errors.BzrError('resource package %s not in breezy' % package)
1951
raise errors.BzrError('resource package %s not in bzrlib' % package)
2225
1953
# Map the resource to a file and read its contents
2226
base = dirname(breezy.__file__)
1954
base = dirname(bzrlib.__file__)
2227
1955
if getattr(sys, 'frozen', None): # bzr.exe
2228
1956
base = abspath(pathjoin(base, '..', '..'))
2229
with open(pathjoin(base, resource_relpath), "rU") as f:
1957
filename = pathjoin(base, resource_relpath)
1958
return open(filename, 'rU').read()
2232
1961
def file_kind_from_stat_mode_thunk(mode):
2233
1962
global file_kind_from_stat_mode
2234
1963
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2236
from ._readdir_pyx import UTF8DirReader
1965
from bzrlib._readdir_pyx import UTF8DirReader
2237
1966
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2238
except ImportError as e:
1967
except ImportError, e:
2239
1968
# This is one time where we won't warn that an extension failed to
2240
1969
# load. The extension is never available on Windows anyway.
2241
from ._readdir_py import (
1970
from bzrlib._readdir_py import (
2242
1971
_kind_from_mode as file_kind_from_stat_mode
2244
1973
return file_kind_from_stat_mode(mode)
2245
1974
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2247
def file_stat(f, _lstat=os.lstat):
1977
def file_kind(f, _lstat=os.lstat):
2251
except OSError as e:
1979
return file_kind_from_stat_mode(_lstat(f).st_mode)
2252
1981
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2253
1982
raise errors.NoSuchFile(f)
2256
def file_kind(f, _lstat=os.lstat):
2257
stat_value = file_stat(f, _lstat)
2258
return file_kind_from_stat_mode(stat_value.st_mode)
2260
1986
def until_no_eintr(f, *a, **kw):
2261
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2263
WARNING: you must be certain that it is safe to retry the call repeatedly
2264
if EINTR does occur. This is typically only true for low-level operations
2265
like os.read. If in any doubt, don't use this.
2267
Keep in mind that this is not a complete solution to EINTR. There is
2268
probably code in the Python standard library and other dependencies that
2269
may encounter EINTR if a signal arrives (and there is signal handler for
2270
that signal). So this function can reduce the impact for IO that breezy
2271
directly controls, but it is not a complete solution.
1987
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
2273
1988
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2276
1991
return f(*a, **kw)
2277
except (IOError, OSError) as e:
1992
except (IOError, OSError), e:
2278
1993
if e.errno == errno.EINTR:
1997
def re_compile_checked(re_string, flags=0, where=""):
1998
"""Return a compiled re, or raise a sensible error.
2000
This should only be used when compiling user-supplied REs.
2002
:param re_string: Text form of regular expression.
2003
:param flags: eg re.IGNORECASE
2004
:param where: Message explaining to the user the context where
2005
it occurred, eg 'log search filter'.
2007
# from https://bugs.launchpad.net/bzr/+bug/251352
2009
re_obj = re.compile(re_string, flags)
2014
where = ' in ' + where
2015
# despite the name 'error' is a type
2016
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2017
% (where, re_string, e))
2283
2020
if sys.platform == "win32":
2286
2023
return msvcrt.getch()
2291
2028
fd = sys.stdin.fileno()
2292
2029
settings = termios.tcgetattr(fd)
2297
2034
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2300
if sys.platform.startswith('linux'):
2038
if sys.platform == 'linux2':
2301
2039
def _local_concurrency():
2303
return os.sysconf('SC_NPROCESSORS_ONLN')
2304
except (ValueError, OSError, AttributeError):
2041
prefix = 'processor'
2042
for line in file('/proc/cpuinfo', 'rb'):
2043
if line.startswith(prefix):
2044
concurrency = int(line[line.find(':')+1:]) + 1
2306
2046
elif sys.platform == 'darwin':
2307
2047
def _local_concurrency():
2308
2048
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2309
2049
stdout=subprocess.PIPE).communicate()[0]
2310
elif "bsd" in sys.platform:
2050
elif sys.platform[0:7] == 'freebsd':
2311
2051
def _local_concurrency():
2312
2052
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2313
2053
stdout=subprocess.PIPE).communicate()[0]
2366
2101
self.encode = encode
2368
2103
def write(self, object):
2369
if isinstance(object, str):
2104
if type(object) is str:
2370
2105
self.stream.write(object)
2372
2107
data, _ = self.encode(object, self.errors)
2373
2108
self.stream.write(data)
2375
if sys.platform == 'win32':
2376
def open_file(filename, mode='r', bufsize=-1):
2377
"""This function is used to override the ``open`` builtin.
2379
But it uses O_NOINHERIT flag so the file handle is not inherited by
2380
child processes. Deleting or renaming a closed file opened with this
2381
function is not blocking child processes.
2383
writing = 'w' in mode
2384
appending = 'a' in mode
2385
updating = '+' in mode
2386
binary = 'b' in mode
2389
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2390
# for flags for each modes.
2400
flags |= os.O_WRONLY
2401
flags |= os.O_CREAT | os.O_TRUNC
2406
flags |= os.O_WRONLY
2407
flags |= os.O_CREAT | os.O_APPEND
2412
flags |= os.O_RDONLY
2414
return os.fdopen(os.open(filename, flags), mode, bufsize)
2419
def available_backup_name(base, exists):
2420
"""Find a non-existing backup file name.
2422
This will *not* create anything, this only return a 'free' entry. This
2423
should be used for checking names in a directory below a locked
2424
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2425
Leap) and generally discouraged.
2427
:param base: The base name.
2429
:param exists: A callable returning True if the path parameter exists.
2432
name = "%s.~%d~" % (base, counter)
2435
name = "%s.~%d~" % (base, counter)
2439
def set_fd_cloexec(fd):
2440
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2441
support for this is not available.
2445
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2446
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2447
except (ImportError, AttributeError):
2448
# Either the fcntl module or specific constants are not present
2452
def find_executable_on_path(name):
2453
"""Finds an executable on the PATH.
2455
On Windows, this will try to append each extension in the PATHEXT
2456
environment variable to the name, if it cannot be found with the name
2459
:param name: The base name of the executable.
2460
:return: The path to the executable found or None.
2462
if sys.platform == 'win32':
2463
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2464
exts = [ext.lower() for ext in exts]
2465
base, ext = os.path.splitext(name)
2467
if ext.lower() not in exts:
2473
path = os.environ.get('PATH')
2474
if path is not None:
2475
path = path.split(os.pathsep)
2478
f = os.path.join(d, name) + ext
2479
if os.access(f, os.X_OK):
2481
if sys.platform == 'win32':
2482
app_path = win32utils.get_app_path(name)
2483
if app_path != name:
2488
def _posix_is_local_pid_dead(pid):
2489
"""True if pid doesn't correspond to live process on this machine"""
2491
# Special meaning of unix kill: just check if it's there.
2493
except OSError as e:
2494
if e.errno == errno.ESRCH:
2495
# On this machine, and really not found: as sure as we can be
2498
elif e.errno == errno.EPERM:
2499
# exists, though not ours
2502
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2503
# Don't really know.
2506
# Exists and our process: not dead.
2509
if sys.platform == "win32":
2510
is_local_pid_dead = win32utils.is_local_pid_dead
2512
is_local_pid_dead = _posix_is_local_pid_dead
2514
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2515
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2516
if getattr(errno, name, None) is not None]
2519
def fdatasync(fileno):
2520
"""Flush file contents to disk if possible.
2522
:param fileno: Integer OS file handle.
2523
:raises TransportNotPossible: If flushing to disk is not possible.
2525
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2529
except IOError as e:
2530
# See bug #1075108, on some platforms fdatasync exists, but can
2531
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2532
# and reduce the chance of corruption-on-powerloss situations. It
2533
# is not a mandatory call, so it is ok to suppress failures.
2534
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2535
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2539
def ensure_empty_directory_exists(path, exception_class):
2540
"""Make sure a local directory exists and is empty.
2542
If it does not exist, it is created. If it exists and is not empty, an
2543
instance of exception_class is raised.
2547
except OSError as e:
2548
if e.errno != errno.EEXIST:
2550
if os.listdir(path) != []:
2551
raise exception_class(path)
2554
def is_environment_error(evalue):
2555
"""True if exception instance is due to a process environment issue
2557
This includes OSError and IOError, but also other errors that come from
2558
the operating system or core libraries but are not subclasses of those.
2560
if isinstance(evalue, (EnvironmentError, select.error)):
2562
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
added
removed
bzrlib/osutils.py
