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
20
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
25
from .lazy_import import lazy_import
26
lazy_import(globals(), """
27
28
from datetime import datetime
30
from ntpath import (abspath as _nt_abspath,
32
normpath as _nt_normpath,
33
realpath as _nt_realpath,
34
splitdrive as _nt_splitdrive,
33
# We need to import both shutil and rmtree as we export the later on posix
34
# and need the former on windows
36
from shutil import rmtree
39
# We need to import both tempfile and mkdtemp as we export the later on posix
40
# and need the former on windows
42
from tempfile import mkdtemp
43
from tempfile import (
50
from breezy.i18n import gettext
55
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
57
if sys.version_info < (2, 5):
58
import md5 as _mod_md5
60
import sha as _mod_sha
70
from bzrlib import symbol_versioning
66
73
# On win32, O_BINARY is used to indicate the file should
67
74
# be opened in binary mode, rather than text mode.
68
75
# On other platforms, O_BINARY doesn't exist, because
69
76
# they always open in binary mode, so it is okay to
70
# OR with 0 on those platforms.
71
# O_NOINHERIT and O_TEXT exists only on win32 too.
77
# OR with 0 on those platforms
72
78
O_BINARY = getattr(os, 'O_BINARY', 0)
73
O_TEXT = getattr(os, 'O_TEXT', 0)
74
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
77
class UnsupportedTimezoneFormat(errors.BzrError):
79
_fmt = ('Unsupported timezone format "%(timezone)s", '
80
'options are "utc", "original", "local".')
82
def __init__(self, timezone):
83
self.timezone = timezone
81
def get_unicode_argv():
83
user_encoding = get_user_encoding()
84
return [a.decode(user_encoding) for a in sys.argv[1:]]
85
except UnicodeDecodeError:
86
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
86
90
def make_readonly(filename):
87
91
"""Make a filename read-only."""
88
92
mod = os.lstat(filename).st_mode
89
93
if not stat.S_ISLNK(mod):
91
chmod_if_possible(filename, mod)
95
os.chmod(filename, mod)
94
98
def make_writable(filename):
95
99
mod = os.lstat(filename).st_mode
96
100
if not stat.S_ISLNK(mod):
98
chmod_if_possible(filename, mod)
101
def chmod_if_possible(filename, mode):
102
# Set file mode if that can be safely done.
103
# Sometimes even on unix the filesystem won't allow it - see
104
# https://bugs.launchpad.net/bzr/+bug/606537
106
# It is probably faster to just do the chmod, rather than
107
# doing a stat, and then trying to compare
108
os.chmod(filename, mode)
109
except (IOError, OSError) as e:
110
# Permission/access denied seems to commonly happen on smbfs; there's
111
# probably no point warning about it.
112
# <https://bugs.launchpad.net/bzr/+bug/606537>
113
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
114
trace.mutter("ignore error on chmod of %r: %r" % (
102
os.chmod(filename, mod)
120
105
def minimum_path_selection(paths):
234
208
file_existed = False
236
210
rename_func(new, tmp_name)
237
except (errors.NoSuchFile,):
211
except (errors.NoSuchFile,), e:
240
214
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
241
215
# function raises an IOError with errno is None when a rename fails.
242
216
# This then gets caught here.
243
217
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
245
except Exception as e:
246
220
if (getattr(e, 'errno', None) is None
247
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
221
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
250
224
file_existed = True
254
# This may throw an exception, in which case success will
256
rename_func(old, new)
258
except (IOError, OSError) as e:
259
# source and target may be aliases of each other (e.g. on a
260
# case-insensitive filesystem), so we may have accidentally renamed
261
# source by when we tried to rename target
262
if (file_existed and e.errno in (None, errno.ENOENT)
263
and old.lower() == new.lower()):
264
# source and target are the same file on a case-insensitive
265
# filesystem, so we don't generate an exception
229
# This may throw an exception, in which case success will
231
rename_func(old, new)
233
except (IOError, OSError), e:
234
# source and target may be aliases of each other (e.g. on a
235
# case-insensitive filesystem), so we may have accidentally renamed
236
# source by when we tried to rename target
237
if not (file_existed and e.errno in (None, errno.ENOENT)):
271
241
# If the file used to exist, rename it back into place
280
250
# choke on a Unicode string containing a relative path if
281
251
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
253
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
283
254
def _posix_abspath(path):
284
255
# jam 20060426 rather than encoding to fsencoding
285
256
# copy posixpath.abspath, but use os.getcwdu instead
286
257
if not posixpath.isabs(path):
287
258
path = posixpath.join(getcwd(), path)
288
return _posix_normpath(path)
259
return posixpath.normpath(path)
291
262
def _posix_realpath(path):
292
263
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
295
def _posix_normpath(path):
296
path = posixpath.normpath(path)
297
# Bug 861008: posixpath.normpath() returns a path normalized according to
298
# the POSIX standard, which stipulates (for compatibility reasons) that two
299
# leading slashes must not be simplified to one, and only if there are 3 or
300
# more should they be simplified as one. So we treat the leading 2 slashes
301
# as a special case here by simply removing the first slash, as we consider
302
# that breaking POSIX compatibility for this obscure feature is acceptable.
303
# This is not a paranoid precaution, as we notably get paths like this when
304
# the repo is hosted at the root of the filesystem, i.e. in "/".
305
if path.startswith('//'):
310
def _posix_get_home_dir():
311
"""Get the home directory of the current user as a unicode path"""
312
path = posixpath.expanduser("~")
314
return path.decode(_fs_enc)
315
except AttributeError:
317
except UnicodeDecodeError:
318
raise errors.BadFilenameEncoding(path, _fs_enc)
321
def _posix_getuser_unicode():
322
"""Get username from environment or password database as unicode"""
323
return getpass.getuser()
326
266
def _win32_fixdrive(path):
327
267
"""Force drive letters to be consistent.
332
272
running python.exe under cmd.exe return capital C:\\
333
273
running win32 python inside a cygwin shell returns lowercase c:\\
335
drive, path = ntpath.splitdrive(path)
275
drive, path = _nt_splitdrive(path)
336
276
return drive.upper() + path
339
279
def _win32_abspath(path):
340
# Real ntpath.abspath doesn't have a problem with a unicode cwd
341
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
280
# Real _nt_abspath doesn't have a problem with a unicode cwd
281
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
284
def _win98_abspath(path):
285
"""Return the absolute version of a path.
286
Windows 98 safe implementation (python reimplementation
287
of Win32 API function GetFullPathNameW)
292
# \\HOST\path => //HOST/path
293
# //HOST/path => //HOST/path
294
# path => C:/cwd/path
297
# check for absolute path
298
drive = _nt_splitdrive(path)[0]
299
if drive == '' and path[:2] not in('//','\\\\'):
301
# we cannot simply os.path.join cwd and path
302
# because os.path.join('C:','/path') produce '/path'
303
# and this is incorrect
304
if path[:1] in ('/','\\'):
305
cwd = _nt_splitdrive(cwd)[0]
307
path = cwd + '\\' + path
308
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
344
311
def _win32_realpath(path):
345
# Real ntpath.realpath doesn't have a problem with a unicode cwd
346
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
312
# Real _nt_realpath doesn't have a problem with a unicode cwd
313
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
349
316
def _win32_pathjoin(*args):
350
return ntpath.join(*args).replace('\\', '/')
317
return _nt_join(*args).replace('\\', '/')
353
320
def _win32_normpath(path):
354
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
321
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
357
324
def _win32_getcwd():
358
return _win32_fixdrive(_getcwd().replace('\\', '/'))
325
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
361
328
def _win32_mkdtemp(*args, **kwargs):
383
350
def _mac_getcwd():
384
return unicodedata.normalize('NFC', _getcwd())
387
def _rename_wrap_exception(rename_func):
388
"""Adds extra information to any exceptions that come from rename().
390
The exception has an updated message and 'old_filename' and 'new_filename'
394
def _rename_wrapper(old, new):
396
rename_func(old, new)
398
detailed_error = OSError(e.errno, e.strerror +
399
" [occurred when renaming '%s' to '%s']" %
401
detailed_error.old_filename = old
402
detailed_error.new_filename = new
405
return _rename_wrapper
411
# Default rename wraps os.rename()
412
rename = _rename_wrap_exception(os.rename)
351
return unicodedata.normalize('NFC', os.getcwdu())
414
354
# Default is to just use the python builtins, but these can be rebound on
415
355
# particular platforms.
416
356
abspath = _posix_abspath
417
357
realpath = _posix_realpath
418
358
pathjoin = os.path.join
419
normpath = _posix_normpath
420
_get_home_dir = _posix_get_home_dir
421
getuser_unicode = _posix_getuser_unicode
359
normpath = os.path.normpath
423
362
dirname = os.path.dirname
424
363
basename = os.path.basename
425
364
split = os.path.split
426
365
splitext = os.path.splitext
427
# These were already lazily imported into local scope
366
# These were already imported into local scope
428
367
# mkdtemp = tempfile.mkdtemp
429
368
# rmtree = shutil.rmtree
438
370
MIN_ABS_PATHLENGTH = 1
441
373
if sys.platform == 'win32':
442
abspath = _win32_abspath
374
if win32utils.winver == 'Windows 98':
375
abspath = _win98_abspath
377
abspath = _win32_abspath
443
378
realpath = _win32_realpath
444
379
pathjoin = _win32_pathjoin
445
380
normpath = _win32_normpath
446
381
getcwd = _win32_getcwd
447
382
mkdtemp = _win32_mkdtemp
448
rename = _rename_wrap_exception(_win32_rename)
450
from . import _walkdirs_win32
454
lstat = _walkdirs_win32.lstat
455
fstat = _walkdirs_win32.fstat
456
wrap_stat = _walkdirs_win32.wrap_stat
383
rename = _win32_rename
458
385
MIN_ABS_PATHLENGTH = 3
494
422
On my standard US Windows XP, the preferred encoding is
495
423
cp1252, but the console is cp437
497
:param trace: If True trace the selected encoding via mutter().
499
from .trace import mutter
425
from bzrlib.trace import mutter
500
426
output_encoding = getattr(sys.stdout, 'encoding', None)
501
427
if not output_encoding:
502
428
input_encoding = getattr(sys.stdin, 'encoding', None)
503
429
if not input_encoding:
504
430
output_encoding = get_user_encoding()
506
mutter('encoding stdout as osutils.get_user_encoding() %r',
431
mutter('encoding stdout as osutils.get_user_encoding() %r',
509
434
output_encoding = input_encoding
511
mutter('encoding stdout as sys.stdin encoding %r',
435
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
515
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
437
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
516
438
if output_encoding == 'cp0':
517
439
# invalid encoding (cp0 means 'no codepage' on Windows)
518
440
output_encoding = get_user_encoding()
520
mutter('cp0 is invalid encoding.'
521
' encoding stdout as osutils.get_user_encoding() %r',
441
mutter('cp0 is invalid encoding.'
442
' encoding stdout as osutils.get_user_encoding() %r',
525
446
codecs.lookup(output_encoding)
526
447
except LookupError:
527
sys.stderr.write('brz: warning:'
448
sys.stderr.write('bzr: warning:'
528
449
' unknown terminal encoding %s.\n'
529
450
' Using encoding %s instead.\n'
530
451
% (output_encoding, get_user_encoding())
532
453
output_encoding = get_user_encoding()
534
455
return output_encoding
797
701
:param show_offset: Whether to append the timezone.
799
703
(date_fmt, tt, offset_str) = \
800
_format_date(t, offset, timezone, date_fmt, show_offset)
704
_format_date(t, offset, timezone, date_fmt, show_offset)
801
705
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
802
706
date_str = time.strftime(date_fmt, tt)
803
707
return date_str + offset_str
806
# Cache of formatted offset strings
810
def format_date_with_offset_in_original_timezone(t, offset=0,
811
_cache=_offset_cache):
812
"""Return a formatted date string in the original timezone.
814
This routine may be faster then format_date.
816
:param t: Seconds since the epoch.
817
:param offset: Timezone offset in seconds east of utc.
821
tt = time.gmtime(t + offset)
822
date_fmt = _default_format_by_weekday_num[tt[6]]
823
date_str = time.strftime(date_fmt, tt)
824
offset_str = _cache.get(offset, None)
825
if offset_str is None:
826
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
827
_cache[offset] = offset_str
828
return date_str + offset_str
831
709
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
832
710
show_offset=True):
833
711
"""Return an unicode date string formatted according to the current locale.
1024
_extension_load_failures = []
1027
def failed_to_load_extension(exception):
1028
"""Handle failing to load a binary extension.
1030
This should be called from the ImportError block guarding the attempt to
1031
import the native extension. If this function returns, the pure-Python
1032
implementation should be loaded instead::
1035
>>> import breezy._fictional_extension_pyx
1036
>>> except ImportError, e:
1037
>>> breezy.osutils.failed_to_load_extension(e)
1038
>>> import breezy._fictional_extension_py
1040
# NB: This docstring is just an example, not a doctest, because doctest
1041
# currently can't cope with the use of lazy imports in this namespace --
1044
# This currently doesn't report the failure at the time it occurs, because
1045
# they tend to happen very early in startup when we can't check config
1046
# files etc, and also we want to report all failures but not spam the user
1048
exception_str = str(exception)
1049
if exception_str not in _extension_load_failures:
1050
trace.mutter("failed to load compiled extension: %s" % exception_str)
1051
_extension_load_failures.append(exception_str)
1054
def report_extension_load_failures():
1055
if not _extension_load_failures:
1057
if config.GlobalConfig().suppress_warning('missing_extensions'):
1059
# the warnings framework should by default show this only once
1060
from .trace import warning
1062
"brz: warning: some compiled extensions could not be loaded; "
1063
"see ``brz help missing-extensions``")
1064
# we no longer show the specific missing extensions here, because it makes
1065
# the message too long and scary - see
1066
# https://bugs.launchpad.net/bzr/+bug/430529
1070
from ._chunks_to_lines_pyx import chunks_to_lines
1071
except ImportError as e:
1072
failed_to_load_extension(e)
1073
from ._chunks_to_lines_py import chunks_to_lines
885
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
887
from bzrlib._chunks_to_lines_py import chunks_to_lines
1076
890
def split_lines(s):
1077
891
"""Split s into lines, but without removing the newline characters."""
1078
892
# Trivially convert a fulltext into a 'chunked' representation, and let
1079
893
# chunks_to_lines do the heavy lifting.
1080
if isinstance(s, bytes):
894
if isinstance(s, str):
1081
895
# chunks_to_lines only supports 8-bit strings
1082
896
return chunks_to_lines([s])
1362
1148
return unicode_or_utf8_string.encode('utf-8')
1151
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1152
' Revision id generators should be creating utf8'
1156
def safe_revision_id(unicode_or_utf8_string, warn=True):
1157
"""Revision ids should now be utf8, but at one point they were unicode.
1159
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1161
:param warn: Functions that are sanitizing user data can set warn=False
1162
:return: None or a utf8 revision id.
1164
if (unicode_or_utf8_string is None
1165
or unicode_or_utf8_string.__class__ == str):
1166
return unicode_or_utf8_string
1168
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1170
return cache_utf8.encode(unicode_or_utf8_string)
1173
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1174
' generators should be creating utf8 file ids.')
1177
def safe_file_id(unicode_or_utf8_string, warn=True):
1178
"""File ids should now be utf8, but at one point they were unicode.
1180
This is the same as safe_utf8, except it uses the cached encode functions
1181
to save a little bit of performance.
1183
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1185
:param warn: Functions that are sanitizing user data can set warn=False
1186
:return: None or a utf8 file id.
1188
if (unicode_or_utf8_string is None
1189
or unicode_or_utf8_string.__class__ == str):
1190
return unicode_or_utf8_string
1192
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1194
return cache_utf8.encode(unicode_or_utf8_string)
1365
1197
_platform_normalizes_filenames = False
1366
1198
if sys.platform == 'darwin':
1367
1199
_platform_normalizes_filenames = True
1410
1238
normalized_filename = _inaccessible_normalized_filename
1413
def set_signal_handler(signum, handler, restart_syscall=True):
1414
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1415
on platforms that support that.
1417
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1418
automatically restart (by calling `signal.siginterrupt(signum,
1419
False)`). May be ignored if the feature is not available on this
1420
platform or Python version.
1424
siginterrupt = signal.siginterrupt
1426
# This python implementation doesn't provide signal support, hence no
1429
except AttributeError:
1430
# siginterrupt doesn't exist on this platform, or for this version
1432
def siginterrupt(signum, flag): return None
1434
def sig_handler(*args):
1435
# Python resets the siginterrupt flag when a signal is
1436
# received. <http://bugs.python.org/issue8354>
1437
# As a workaround for some cases, set it back the way we want it.
1438
siginterrupt(signum, False)
1439
# Now run the handler function passed to set_signal_handler.
1442
sig_handler = handler
1443
old_handler = signal.signal(signum, sig_handler)
1445
siginterrupt(signum, False)
1449
default_terminal_width = 80
1450
"""The default terminal width for ttys.
1452
This is defined so that higher levels can share a common fallback value when
1453
terminal_width() returns None.
1456
# Keep some state so that terminal_width can detect if _terminal_size has
1457
# returned a different size since the process started. See docstring and
1458
# comments of terminal_width for details.
1459
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1460
_terminal_size_state = 'no_data'
1461
_first_terminal_size = None
1464
1241
def terminal_width():
1465
"""Return terminal width.
1467
None is returned if the width can't established precisely.
1470
- if BRZ_COLUMNS is set, returns its value
1471
- if there is no controlling terminal, returns None
1472
- query the OS, if the queried size has changed since the last query,
1474
- if COLUMNS is set, returns its value,
1475
- if the OS has a value (even though it's never changed), return its value.
1477
From there, we need to query the OS to get the size of the controlling
1480
On Unices we query the OS by:
1481
- get termios.TIOCGWINSZ
1482
- if an error occurs or a negative value is obtained, returns None
1484
On Windows we query the OS by:
1485
- win32utils.get_console_size() decides,
1486
- returns None on error (provided default value)
1488
# Note to implementors: if changing the rules for determining the width,
1489
# make sure you've considered the behaviour in these cases:
1490
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1491
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1493
# - (add more interesting cases here, if you find any)
1494
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1495
# but we don't want to register a signal handler because it is impossible
1496
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1497
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1498
# time so we can notice if the reported size has changed, which should have
1501
# If BRZ_COLUMNS is set, take it, user is always right
1502
# Except if they specified 0 in which case, impose no limit here
1504
width = int(os.environ['BRZ_COLUMNS'])
1505
except (KeyError, ValueError):
1507
if width is not None:
1513
isatty = getattr(sys.stdout, 'isatty', None)
1514
if isatty is None or not isatty():
1515
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
1519
width, height = os_size = _terminal_size(None, None)
1520
global _first_terminal_size, _terminal_size_state
1521
if _terminal_size_state == 'no_data':
1522
_first_terminal_size = os_size
1523
_terminal_size_state = 'unchanged'
1524
elif (_terminal_size_state == 'unchanged' and
1525
_first_terminal_size != os_size):
1526
_terminal_size_state = 'changed'
1528
# If the OS claims to know how wide the terminal is, and this value has
1529
# ever changed, use that.
1530
if _terminal_size_state == 'changed':
1531
if width is not None and width > 0:
1534
# If COLUMNS is set, use it.
1536
return int(os.environ['COLUMNS'])
1537
except (KeyError, ValueError):
1540
# Finally, use an unchanged size from the OS, if we have one.
1541
if _terminal_size_state == 'unchanged':
1542
if width is not None and width > 0:
1545
# The width could not be determined.
1549
def _win32_terminal_size(width, height):
1550
width, height = win32utils.get_console_size(
1551
defaultx=width, defaulty=height)
1552
return width, height
1555
def _ioctl_terminal_size(width, height):
1242
"""Return estimated terminal width."""
1243
if sys.platform == 'win32':
1244
return win32utils.get_console_size()[0]
1247
import struct, fcntl, termios
1560
1248
s = struct.pack('HHHH', 0, 0, 0, 0)
1561
1249
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1562
height, width = struct.unpack('HHHH', x)[0:2]
1563
except (IOError, AttributeError):
1250
width = struct.unpack('HHHH', x)[1]
1565
return width, height
1568
_terminal_size = None
1569
"""Returns the terminal size as (width, height).
1571
:param width: Default value for width.
1572
:param height: Default value for height.
1574
This is defined specifically for each OS and query the size of the controlling
1575
terminal. If any error occurs, the provided default values should be returned.
1577
if sys.platform == 'win32':
1578
_terminal_size = _win32_terminal_size
1580
_terminal_size = _ioctl_terminal_size
1583
def supports_executable(path):
1584
"""Return if filesystem at path supports executable bit.
1586
:param path: Path for which to check the file system
1587
:return: boolean indicating whether executable bit can be stored/relied upon
1589
if sys.platform == 'win32':
1592
fs_type = get_fs_type(path)
1593
except errors.DependencyNotPresent as e:
1594
trace.mutter('Unable to get fs type for %r: %s', path, e)
1596
if fs_type in ('vfat', 'ntfs'):
1597
# filesystems known to not support executable bit
1602
def supports_symlinks(path):
1603
"""Return if the filesystem at path supports the creation of symbolic links.
1606
if not has_symlinks():
1609
fs_type = get_fs_type(path)
1610
except errors.DependencyNotPresent as e:
1611
trace.mutter('Unable to get fs type for %r: %s', path, e)
1613
if fs_type in ('vfat', 'ntfs'):
1614
# filesystems known to not support symlinks
1255
width = int(os.environ['COLUMNS'])
1264
def supports_executable():
1265
return sys.platform != "win32"
1619
1268
def supports_posix_readonly():
1934
1583
real_handlers[kind](abspath, relpath)
1937
def copy_ownership_from_path(dst, src=None):
1938
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1940
If src is None, the containing directory is used as source. If chown
1941
fails, the error is ignored and a warning is printed.
1943
chown = getattr(os, 'chown', None)
1948
src = os.path.dirname(dst)
1954
chown(dst, s.st_uid, s.st_gid)
1957
'Unable to copy ownership from "%s" to "%s". '
1958
'You may want to set it manually.', src, dst)
1959
trace.log_exception_quietly()
1962
1586
def path_prefix_key(path):
1963
1587
"""Generate a prefix-order path key for path.
1965
1589
This can be used to sort paths in the same way that walkdirs does.
1967
return (dirname(path), path)
1591
return (dirname(path) , path)
1970
1594
def compare_paths_prefix_order(path_a, path_b):
1971
1595
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1972
1596
key_a = path_prefix_key(path_a)
1973
1597
key_b = path_prefix_key(path_b)
1974
return (key_a > key_b) - (key_a < key_b)
1598
return cmp(key_a, key_b)
1977
1601
_cached_user_encoding = None
1980
def get_user_encoding():
1604
def get_user_encoding(use_cache=True):
1981
1605
"""Find out what the preferred user encoding is.
1983
1607
This is generally the encoding that is used for command line parameters
1984
1608
and file contents. This may be different from the terminal encoding
1985
1609
or the filesystem encoding.
1611
:param use_cache: Enable cache for detected encoding.
1612
(This parameter is turned on by default,
1613
and required only for selftesting)
1987
1615
:return: A string defining the preferred user encoding
1989
1617
global _cached_user_encoding
1990
if _cached_user_encoding is not None:
1618
if _cached_user_encoding is not None and use_cache:
1991
1619
return _cached_user_encoding
1993
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
1994
# Use the existing locale settings and call nl_langinfo directly
1995
# rather than going through getpreferredencoding. This avoids
1996
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
1997
# possibility of the setlocale call throwing an error.
1998
user_encoding = locale.nl_langinfo(locale.CODESET)
1621
if sys.platform == 'darwin':
1622
# python locale.getpreferredencoding() always return
1623
# 'mac-roman' on darwin. That's a lie.
1624
sys.platform = 'posix'
1626
if os.environ.get('LANG', None) is None:
1627
# If LANG is not set, we end up with 'ascii', which is bad
1628
# ('mac-roman' is more than ascii), so we set a default which
1629
# will give us UTF-8 (which appears to work in all cases on
1630
# OSX). Users are still free to override LANG of course, as
1631
# long as it give us something meaningful. This work-around
1632
# *may* not be needed with python 3k and/or OSX 10.5, but will
1633
# work with them too -- vila 20080908
1634
os.environ['LANG'] = 'en_US.UTF-8'
1637
sys.platform = 'darwin'
2000
# GZ 2011-12-19: On windows could call GetACP directly instead.
2001
user_encoding = locale.getpreferredencoding(False)
2004
user_encoding = codecs.lookup(user_encoding).name
2006
if user_encoding not in ("", "cp0"):
2007
sys.stderr.write('brz: warning:'
1642
user_encoding = locale.getpreferredencoding()
1643
except locale.Error, e:
1644
sys.stderr.write('bzr: warning: %s\n'
1645
' Could not determine what text encoding to use.\n'
1646
' This error usually means your Python interpreter\n'
1647
' doesn\'t support the locale set by $LANG (%s)\n'
1648
" Continuing with ascii encoding.\n"
1649
% (e, os.environ.get('LANG')))
1650
user_encoding = 'ascii'
1652
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1653
# treat that as ASCII, and not support printing unicode characters to the
1656
# For python scripts run under vim, we get '', so also treat that as ASCII
1657
if user_encoding in (None, 'cp0', ''):
1658
user_encoding = 'ascii'
1662
codecs.lookup(user_encoding)
1664
sys.stderr.write('bzr: warning:'
2008
1665
' unknown encoding %s.'
2009
1666
' Continuing with ascii encoding.\n'
2010
1667
% user_encoding
2012
user_encoding = 'ascii'
2014
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2015
if user_encoding == 'ascii':
2016
if sys.platform == 'darwin':
2017
# OSX is special-cased in Python to have a UTF-8 filesystem
2018
# encoding and previously had LANG set here if not present.
2019
user_encoding = 'utf-8'
2020
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2021
# for some other posix platforms as well.
2023
_cached_user_encoding = user_encoding
1669
user_encoding = 'ascii'
1672
_cached_user_encoding = user_encoding
2024
1674
return user_encoding
2027
def get_diff_header_encoding():
2028
return get_terminal_encoding()
2031
1677
def get_host_name():
2032
1678
"""Return the current unicode host name.
2035
1681
behaves inconsistently on different platforms.
2037
1683
if sys.platform == "win32":
2038
1685
return win32utils.get_host_name()
2041
return socket.gethostname()
2044
# We must not read/write any more than 64k at a time from/to a socket so we
2045
# don't risk "no buffer space available" errors on some platforms. Windows in
2046
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2048
MAX_SOCKET_CHUNK = 64 * 1024
2050
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2051
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2052
_eno = getattr(errno, _eno, None)
2053
if _eno is not None:
2054
_end_of_stream_errors.append(_eno)
2058
def read_bytes_from_socket(sock, report_activity=None,
2059
max_read_size=MAX_SOCKET_CHUNK):
2060
"""Read up to max_read_size of bytes from sock and notify of progress.
2062
Translates "Connection reset by peer" into file-like EOF (return an
2063
empty string rather than raise an error), and repeats the recv if
2064
interrupted by a signal.
2068
data = sock.recv(max_read_size)
2069
except socket.error as e:
2071
if eno in _end_of_stream_errors:
2072
# The connection was closed by the other side. Callers expect
2073
# an empty string to signal end-of-stream.
2075
elif eno == errno.EINTR:
2076
# Retry the interrupted recv.
2080
if report_activity is not None:
2081
report_activity(len(data), 'read')
2085
def recv_all(socket, count):
1688
return socket.gethostname().decode(get_user_encoding())
1691
def recv_all(socket, bytes):
2086
1692
"""Receive an exact number of bytes.
2088
1694
Regular Socket.recv() may return less than the requested number of bytes,
2089
depending on what's in the OS buffer. MSG_WAITALL is not available
1695
dependning on what's in the OS buffer. MSG_WAITALL is not available
2090
1696
on all platforms, but this should work everywhere. This will return
2091
1697
less than the requested amount if the remote end closes.
2093
1699
This isn't optimized and is intended mostly for use in testing.
2096
while len(b) < count:
2097
new = read_bytes_from_socket(socket, None, count - len(b))
1702
while len(b) < bytes:
1703
new = until_no_eintr(socket.recv, bytes - len(b))
2104
def send_all(sock, bytes, report_activity=None):
1710
def send_all(socket, bytes, report_activity=None):
2105
1711
"""Send all bytes on a socket.
2107
Breaks large blocks in smaller chunks to avoid buffering limitations on
2108
some platforms, and catches EINTR which may be thrown if the send is
2109
interrupted by a signal.
2111
This is preferred to socket.sendall(), because it avoids portability bugs
2112
and provides activity reporting.
1713
Regular socket.sendall() can give socket error 10053 on Windows. This
1714
implementation sends no more than 64k at a time, which avoids this problem.
2114
1716
:param report_activity: Call this as bytes are read, see
2115
1717
Transport._report_activity
2118
byte_count = len(bytes)
2119
view = memoryview(bytes)
2120
while sent_total < byte_count:
2122
sent = sock.send(view[sent_total:sent_total + MAX_SOCKET_CHUNK])
2123
except (socket.error, IOError) as e:
2124
if e.args[0] in _end_of_stream_errors:
2125
raise errors.ConnectionReset(
2126
"Error trying to write to socket", e)
2127
if e.args[0] != errno.EINTR:
2131
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2134
if report_activity is not None:
2135
report_activity(sent, 'write')
2138
def connect_socket(address):
2139
# Slight variation of the socket.create_connection() function (provided by
2140
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2141
# provide it for previous python versions. Also, we don't use the timeout
2142
# parameter (provided by the python implementation) so we don't implement
2144
err = socket.error('getaddrinfo returns an empty list')
2145
host, port = address
2146
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2147
af, socktype, proto, canonname, sa = res
2150
sock = socket.socket(af, socktype, proto)
2154
except socket.error as e:
2156
# 'err' is now the most recent error
2157
if sock is not None:
1720
for pos in xrange(0, len(bytes), chunk_size):
1721
block = bytes[pos:pos+chunk_size]
1722
if report_activity is not None:
1723
report_activity(len(block), 'write')
1724
until_no_eintr(socket.sendall, block)
2162
1727
def dereference_path(path):
2190
1755
If and when pkg_resources becomes a standard library, this routine
2191
1756
can delegate to it.
2193
# Check package name is within breezy
2194
if package == "breezy":
1758
# Check package name is within bzrlib
1759
if package == "bzrlib":
2195
1760
resource_relpath = resource_name
2196
elif package.startswith("breezy."):
2197
package = package[len("breezy."):].replace('.', os.sep)
1761
elif package.startswith("bzrlib."):
1762
package = package[len("bzrlib."):].replace('.', os.sep)
2198
1763
resource_relpath = pathjoin(package, resource_name)
2200
raise errors.BzrError('resource package %s not in breezy' % package)
1765
raise errors.BzrError('resource package %s not in bzrlib' % package)
2202
1767
# Map the resource to a file and read its contents
2203
base = dirname(breezy.__file__)
1768
base = dirname(bzrlib.__file__)
2204
1769
if getattr(sys, 'frozen', None): # bzr.exe
2205
1770
base = abspath(pathjoin(base, '..', '..'))
2206
with open(pathjoin(base, resource_relpath), "rt") as f:
1771
filename = pathjoin(base, resource_relpath)
1772
return open(filename, 'rU').read()
2210
1775
def file_kind_from_stat_mode_thunk(mode):
2211
1776
global file_kind_from_stat_mode
2212
1777
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2214
from ._readdir_pyx import UTF8DirReader
1779
from bzrlib._readdir_pyx import UTF8DirReader
2215
1780
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2216
1781
except ImportError:
2217
# This is one time where we won't warn that an extension failed to
2218
# load. The extension is never available on Windows anyway.
2219
from ._readdir_py import (
1782
from bzrlib._readdir_py import (
2220
1783
_kind_from_mode as file_kind_from_stat_mode
2222
1785
return file_kind_from_stat_mode(mode)
2225
1786
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2228
def file_stat(f, _lstat=os.lstat):
1789
def file_kind(f, _lstat=os.lstat):
2232
except OSError as e:
1791
return file_kind_from_stat_mode(_lstat(f).st_mode)
2233
1793
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2234
1794
raise errors.NoSuchFile(f)
2238
def file_kind(f, _lstat=os.lstat):
2239
stat_value = file_stat(f, _lstat)
2240
return file_kind_from_stat_mode(stat_value.st_mode)
2243
1798
def until_no_eintr(f, *a, **kw):
2244
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2246
WARNING: you must be certain that it is safe to retry the call repeatedly
2247
if EINTR does occur. This is typically only true for low-level operations
2248
like os.read. If in any doubt, don't use this.
2250
Keep in mind that this is not a complete solution to EINTR. There is
2251
probably code in the Python standard library and other dependencies that
2252
may encounter EINTR if a signal arrives (and there is signal handler for
2253
that signal). So this function can reduce the impact for IO that breezy
2254
directly controls, but it is not a complete solution.
1799
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
2256
1800
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2259
1803
return f(*a, **kw)
2260
except (IOError, OSError) as e:
1804
except (IOError, OSError), e:
2261
1805
if e.errno == errno.EINTR:
1809
def re_compile_checked(re_string, flags=0, where=""):
1810
"""Return a compiled re, or raise a sensible error.
1812
This should only be used when compiling user-supplied REs.
1814
:param re_string: Text form of regular expression.
1815
:param flags: eg re.IGNORECASE
1816
:param where: Message explaining to the user the context where
1817
it occurred, eg 'log search filter'.
1819
# from https://bugs.launchpad.net/bzr/+bug/251352
1821
re_obj = re.compile(re_string, flags)
1826
where = ' in ' + where
1827
# despite the name 'error' is a type
1828
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
1829
% (where, re_string, e))
2266
1832
if sys.platform == "win32":
2269
1835
return msvcrt.getch()
2274
1840
fd = sys.stdin.fileno()
2275
1841
settings = termios.tcgetattr(fd)
2280
1846
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2283
if sys.platform.startswith('linux'):
1850
if sys.platform == 'linux2':
2284
1851
def _local_concurrency():
2286
return os.sysconf('SC_NPROCESSORS_ONLN')
2287
except (ValueError, OSError, AttributeError):
1853
prefix = 'processor'
1854
for line in file('/proc/cpuinfo', 'rb'):
1855
if line.startswith(prefix):
1856
concurrency = int(line[line.find(':')+1:]) + 1
2289
1858
elif sys.platform == 'darwin':
2290
1859
def _local_concurrency():
2291
1860
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2292
1861
stdout=subprocess.PIPE).communicate()[0]
2293
elif "bsd" in sys.platform:
1862
elif sys.platform[0:7] == 'freebsd':
2294
1863
def _local_concurrency():
2295
1864
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2296
1865
stdout=subprocess.PIPE).communicate()[0]
2297
1866
elif sys.platform == 'sunos5':
2298
1867
def _local_concurrency():
2299
return subprocess.Popen(['psrinfo', '-p', ],
1868
return subprocess.Popen(['psrinfo', '-p',],
2300
1869
stdout=subprocess.PIPE).communicate()[0]
2301
1870
elif sys.platform == "win32":
2302
1871
def _local_concurrency():
2318
1886
anything goes wrong.
2320
1888
global _cached_local_concurrency
2322
1889
if _cached_local_concurrency is not None and use_cache:
2323
1890
return _cached_local_concurrency
2325
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2326
if concurrency is None:
2327
import multiprocessing
2329
concurrency = multiprocessing.cpu_count()
2330
except NotImplementedError:
2331
# multiprocessing.cpu_count() isn't implemented on all platforms
2333
concurrency = _local_concurrency()
2334
except (OSError, IOError):
1893
concurrency = _local_concurrency()
1894
except (OSError, IOError):
2337
1897
concurrency = int(concurrency)
2338
1898
except (TypeError, ValueError):
2339
1899
concurrency = 1
2341
_cached_local_concurrency = concurrency
1901
_cached_concurrency = concurrency
2342
1902
return concurrency
2345
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2346
"""A stream writer that doesn't decode str arguments."""
2348
def __init__(self, encode, stream, errors='strict'):
2349
codecs.StreamWriter.__init__(self, stream, errors)
2350
self.encode = encode
2352
def write(self, object):
2353
if isinstance(object, str):
2354
self.stream.write(object)
2356
data, _ = self.encode(object, self.errors)
2357
self.stream.write(data)
2360
if sys.platform == 'win32':
2361
def open_file(filename, mode='r', bufsize=-1):
2362
"""This function is used to override the ``open`` builtin.
2364
But it uses O_NOINHERIT flag so the file handle is not inherited by
2365
child processes. Deleting or renaming a closed file opened with this
2366
function is not blocking child processes.
2368
writing = 'w' in mode
2369
appending = 'a' in mode
2370
updating = '+' in mode
2371
binary = 'b' in mode
2374
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2375
# for flags for each modes.
2385
flags |= os.O_WRONLY
2386
flags |= os.O_CREAT | os.O_TRUNC
2391
flags |= os.O_WRONLY
2392
flags |= os.O_CREAT | os.O_APPEND
2397
flags |= os.O_RDONLY
2399
return os.fdopen(os.open(filename, flags), mode, bufsize)
2404
def available_backup_name(base, exists):
2405
"""Find a non-existing backup file name.
2407
This will *not* create anything, this only return a 'free' entry. This
2408
should be used for checking names in a directory below a locked
2409
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2410
Leap) and generally discouraged.
2412
:param base: The base name.
2414
:param exists: A callable returning True if the path parameter exists.
2417
name = "%s.~%d~" % (base, counter)
2420
name = "%s.~%d~" % (base, counter)
2424
def set_fd_cloexec(fd):
2425
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2426
support for this is not available.
2430
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2431
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2432
except (ImportError, AttributeError):
2433
# Either the fcntl module or specific constants are not present
2437
def find_executable_on_path(name):
2438
"""Finds an executable on the PATH.
2440
On Windows, this will try to append each extension in the PATHEXT
2441
environment variable to the name, if it cannot be found with the name
2444
:param name: The base name of the executable.
2445
:return: The path to the executable found or None.
2447
if sys.platform == 'win32':
2448
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2449
exts = [ext.lower() for ext in exts]
2450
base, ext = os.path.splitext(name)
2452
if ext.lower() not in exts:
2458
path = os.environ.get('PATH')
2459
if path is not None:
2460
path = path.split(os.pathsep)
2463
f = os.path.join(d, name) + ext
2464
if os.access(f, os.X_OK):
2466
if sys.platform == 'win32':
2467
app_path = win32utils.get_app_path(name)
2468
if app_path != name:
2473
def _posix_is_local_pid_dead(pid):
2474
"""True if pid doesn't correspond to live process on this machine"""
2476
# Special meaning of unix kill: just check if it's there.
2478
except OSError as e:
2479
if e.errno == errno.ESRCH:
2480
# On this machine, and really not found: as sure as we can be
2483
elif e.errno == errno.EPERM:
2484
# exists, though not ours
2487
trace.mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2488
# Don't really know.
2491
# Exists and our process: not dead.
2495
if sys.platform == "win32":
2496
is_local_pid_dead = win32utils.is_local_pid_dead
2498
is_local_pid_dead = _posix_is_local_pid_dead
2500
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2501
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2502
if getattr(errno, name, None) is not None]
2505
def fdatasync(fileno):
2506
"""Flush file contents to disk if possible.
2508
:param fileno: Integer OS file handle.
2509
:raises TransportNotPossible: If flushing to disk is not possible.
2511
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2515
except IOError as e:
2516
# See bug #1075108, on some platforms fdatasync exists, but can
2517
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2518
# and reduce the chance of corruption-on-powerloss situations. It
2519
# is not a mandatory call, so it is ok to suppress failures.
2520
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2521
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2525
def ensure_empty_directory_exists(path, exception_class):
2526
"""Make sure a local directory exists and is empty.
2528
If it does not exist, it is created. If it exists and is not empty, an
2529
instance of exception_class is raised.
2533
except OSError as e:
2534
if e.errno != errno.EEXIST:
2536
if os.listdir(path) != []:
2537
raise exception_class(path)
2540
def read_mtab(path):
2541
"""Read an fstab-style file and extract mountpoint+filesystem information.
2543
:param path: Path to read from
2544
:yield: Tuples with mountpoints (as bytestrings) and filesystem names
2546
with open(path, 'rb') as f:
2548
if line.startswith(b'#'):
2553
yield cols[1], cols[2].decode('ascii', 'replace')
2556
MTAB_PATH = '/etc/mtab'
2558
class FilesystemFinder(object):
2559
"""Find the filesystem for a particular path."""
2561
def __init__(self, mountpoints):
2564
self._mountpoints = sorted(mountpoints, key=key, reverse=True)
2568
"""Create a FilesystemFinder from an mtab-style file.
2570
Note that this will silenty ignore mtab if it doesn't exist or can not
2573
# TODO(jelmer): Use inotify to be notified when /etc/mtab changes and
2574
# we need to re-read it.
2576
return cls(read_mtab(MTAB_PATH))
2577
except EnvironmentError as e:
2578
trace.mutter('Unable to read mtab: %s', e)
2581
def find(self, path):
2582
"""Find the filesystem used by a particular path.
2584
:param path: Path to find (bytestring or text type)
2585
:return: Filesystem name (as text type) or None, if the filesystem is
2588
for mountpoint, filesystem in self._mountpoints:
2589
if is_inside(mountpoint, path):
2594
_FILESYSTEM_FINDER = None
2597
def get_fs_type(path):
2598
"""Return the filesystem type for the partition a path is in.
2600
:param path: Path to search filesystem type for
2601
:return: A FS type, as string. E.g. "ext2"
2603
global _FILESYSTEM_FINDER
2604
if _FILESYSTEM_FINDER is None:
2605
_FILESYSTEM_FINDER = FilesystemFinder.from_mtab()
2607
if not isinstance(path, bytes):
2608
path = path.encode(_fs_enc)
2610
return _FILESYSTEM_FINDER.find(path)
2613
perf_counter = time.perf_counter
added
removed
bzrlib/osutils.py
