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
27
from .lazy_import import lazy_import
25
from bzrlib.lazy_import import lazy_import
28
26
lazy_import(globals(), """
29
27
from datetime import datetime
29
from ntpath import (abspath as _nt_abspath,
31
normpath as _nt_normpath,
32
realpath as _nt_realpath,
33
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
43
from tempfile import (
52
from breezy.i18n import gettext
56
from bzrlib.symbol_versioning import (
61
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
63
if sys.version_info < (2, 5):
64
import md5 as _mod_md5
66
import sha as _mod_sha
76
from bzrlib import symbol_versioning
79
# Cross platform wall-clock time functionality with decent resolution.
80
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
81
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
82
# synchronized with ``time.time()``, this is only meant to be used to find
83
# delta times by subtracting from another call to this function.
84
timer_func = time.time
85
if sys.platform == 'win32':
86
timer_func = time.clock
73
88
# On win32, O_BINARY is used to indicate the file should
74
89
# be opened in binary mode, rather than text mode.
81
96
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
84
class UnsupportedTimezoneFormat(errors.BzrError):
86
_fmt = ('Unsupported timezone format "%(timezone)s", '
87
'options are "utc", "original", "local".')
89
def __init__(self, timezone):
90
self.timezone = timezone
93
99
def get_unicode_argv():
97
101
user_encoding = get_user_encoding()
98
102
return [a.decode(user_encoding) for a in sys.argv[1:]]
99
103
except UnicodeDecodeError:
100
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
101
"application locale.").format(a, user_encoding))
104
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
104
108
def make_readonly(filename):
105
109
"""Make a filename read-only."""
106
110
mod = os.lstat(filename).st_mode
107
111
if not stat.S_ISLNK(mod):
109
chmod_if_possible(filename, mod)
113
os.chmod(filename, mod)
112
116
def make_writable(filename):
113
117
mod = os.lstat(filename).st_mode
114
118
if not stat.S_ISLNK(mod):
116
chmod_if_possible(filename, mod)
119
def chmod_if_possible(filename, mode):
120
# Set file mode if that can be safely done.
121
# Sometimes even on unix the filesystem won't allow it - see
122
# https://bugs.launchpad.net/bzr/+bug/606537
124
# It is probably faster to just do the chmod, rather than
125
# doing a stat, and then trying to compare
126
os.chmod(filename, mode)
127
except (IOError, OSError) as e:
128
# Permission/access denied seems to commonly happen on smbfs; there's
129
# probably no point warning about it.
130
# <https://bugs.launchpad.net/bzr/+bug/606537>
131
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
132
trace.mutter("ignore error on chmod of %r: %r" % (
120
os.chmod(filename, mod)
138
123
def minimum_path_selection(paths):
252
232
file_existed = False
254
234
rename_func(new, tmp_name)
255
except (errors.NoSuchFile,):
235
except (errors.NoSuchFile,), e:
258
238
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
259
239
# function raises an IOError with errno is None when a rename fails.
260
240
# This then gets caught here.
261
241
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
263
except Exception as e:
264
244
if (getattr(e, 'errno', None) is None
265
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
245
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
268
248
file_existed = True
272
# This may throw an exception, in which case success will
274
rename_func(old, new)
276
except (IOError, OSError) as e:
277
# source and target may be aliases of each other (e.g. on a
278
# case-insensitive filesystem), so we may have accidentally renamed
279
# source by when we tried to rename target
280
if (file_existed and e.errno in (None, errno.ENOENT)
254
# This may throw an exception, in which case success will
256
rename_func(old, new)
258
except (IOError, OSError), 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
failure_exc = sys.exc_info()
263
if (file_existed and e.errno in (None, errno.ENOENT)
281
264
and old.lower() == new.lower()):
282
# source and target are the same file on a case-insensitive
283
# filesystem, so we don't generate an exception
265
# source and target are the same file on a case-insensitive
266
# filesystem, so we don't generate an exception
289
270
# If the file used to exist, rename it back into place
292
273
unlink_func(tmp_name)
294
275
rename_func(tmp_name, new)
276
if failure_exc is not None:
277
raise failure_exc[0], failure_exc[1], failure_exc[2]
297
280
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
298
281
# choke on a Unicode string containing a relative path if
299
282
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
284
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
301
285
def _posix_abspath(path):
302
286
# jam 20060426 rather than encoding to fsencoding
303
287
# copy posixpath.abspath, but use os.getcwdu instead
304
288
if not posixpath.isabs(path):
305
289
path = posixpath.join(getcwd(), path)
306
return _posix_normpath(path)
290
return posixpath.normpath(path)
309
293
def _posix_realpath(path):
310
294
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
313
def _posix_normpath(path):
314
path = posixpath.normpath(path)
315
# Bug 861008: posixpath.normpath() returns a path normalized according to
316
# the POSIX standard, which stipulates (for compatibility reasons) that two
317
# leading slashes must not be simplified to one, and only if there are 3 or
318
# more should they be simplified as one. So we treat the leading 2 slashes
319
# as a special case here by simply removing the first slash, as we consider
320
# that breaking POSIX compatibility for this obscure feature is acceptable.
321
# This is not a paranoid precaution, as we notably get paths like this when
322
# the repo is hosted at the root of the filesystem, i.e. in "/".
323
if path.startswith('//'):
328
def _posix_path_from_environ(key):
329
"""Get unicode path from `key` in environment or None if not present
331
Note that posix systems use arbitrary byte strings for filesystem objects,
332
so a path that raises BadFilenameEncoding here may still be accessible.
334
val = os.environ.get(key, None)
335
if PY3 or val is None:
338
return val.decode(_fs_enc)
339
except UnicodeDecodeError:
340
# GZ 2011-12-12:Ideally want to include `key` in the exception message
341
raise errors.BadFilenameEncoding(val, _fs_enc)
344
def _posix_get_home_dir():
345
"""Get the home directory of the current user as a unicode path"""
346
path = posixpath.expanduser("~")
348
return path.decode(_fs_enc)
349
except AttributeError:
351
except UnicodeDecodeError:
352
raise errors.BadFilenameEncoding(path, _fs_enc)
355
def _posix_getuser_unicode():
356
"""Get username from environment or password database as unicode"""
357
name = getpass.getuser()
360
user_encoding = get_user_encoding()
362
return name.decode(user_encoding)
363
except UnicodeDecodeError:
364
raise errors.BzrError("Encoding of username %r is unsupported by %s "
365
"application locale." % (name, user_encoding))
368
297
def _win32_fixdrive(path):
369
298
"""Force drive letters to be consistent.
374
303
running python.exe under cmd.exe return capital C:\\
375
304
running win32 python inside a cygwin shell returns lowercase c:\\
377
drive, path = ntpath.splitdrive(path)
306
drive, path = _nt_splitdrive(path)
378
307
return drive.upper() + path
381
310
def _win32_abspath(path):
382
# Real ntpath.abspath doesn't have a problem with a unicode cwd
383
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
311
# Real _nt_abspath doesn't have a problem with a unicode cwd
312
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
315
def _win98_abspath(path):
316
"""Return the absolute version of a path.
317
Windows 98 safe implementation (python reimplementation
318
of Win32 API function GetFullPathNameW)
323
# \\HOST\path => //HOST/path
324
# //HOST/path => //HOST/path
325
# path => C:/cwd/path
328
# check for absolute path
329
drive = _nt_splitdrive(path)[0]
330
if drive == '' and path[:2] not in('//','\\\\'):
332
# we cannot simply os.path.join cwd and path
333
# because os.path.join('C:','/path') produce '/path'
334
# and this is incorrect
335
if path[:1] in ('/','\\'):
336
cwd = _nt_splitdrive(cwd)[0]
338
path = cwd + '\\' + path
339
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
386
342
def _win32_realpath(path):
387
# Real ntpath.realpath doesn't have a problem with a unicode cwd
388
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
343
# Real _nt_realpath doesn't have a problem with a unicode cwd
344
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
391
347
def _win32_pathjoin(*args):
392
return ntpath.join(*args).replace('\\', '/')
348
return _nt_join(*args).replace('\\', '/')
395
351
def _win32_normpath(path):
396
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
352
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
399
355
def _win32_getcwd():
400
return _win32_fixdrive(_getcwd().replace('\\', '/'))
356
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
403
359
def _win32_mkdtemp(*args, **kwargs):
404
360
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
363
def _add_rename_error_details(e, old, new):
364
new_e = OSError(e.errno, "failed to rename %s to %s: %s"
365
% (old, new, e.strerror))
367
new_e.to_filename = new
407
371
def _win32_rename(old, new):
408
372
"""We expect to be able to atomically replace 'new' with old.
389
def _wrapped_rename(old, new):
390
"""Rename a file or directory"""
393
except (IOError, OSError), e:
394
# this is eventually called by all rename-like functions, so should
396
raise _add_rename_error_details(e, old, new)
425
399
def _mac_getcwd():
426
return unicodedata.normalize('NFC', _getcwd())
429
def _rename_wrap_exception(rename_func):
430
"""Adds extra information to any exceptions that come from rename().
432
The exception has an updated message and 'old_filename' and 'new_filename'
436
def _rename_wrapper(old, new):
438
rename_func(old, new)
440
detailed_error = OSError(e.errno, e.strerror +
441
" [occurred when renaming '%s' to '%s']" %
443
detailed_error.old_filename = old
444
detailed_error.new_filename = new
447
return _rename_wrapper
450
if sys.version_info > (3,):
456
# Default rename wraps os.rename()
457
rename = _rename_wrap_exception(os.rename)
400
return unicodedata.normalize('NFC', os.getcwdu())
459
403
# Default is to just use the python builtins, but these can be rebound on
460
404
# particular platforms.
461
405
abspath = _posix_abspath
462
406
realpath = _posix_realpath
463
407
pathjoin = os.path.join
464
normpath = _posix_normpath
465
path_from_environ = _posix_path_from_environ
466
_get_home_dir = _posix_get_home_dir
467
getuser_unicode = _posix_getuser_unicode
408
normpath = os.path.normpath
409
rename = _wrapped_rename # overridden below on win32
469
411
dirname = os.path.dirname
470
412
basename = os.path.basename
471
413
split = os.path.split
472
414
splitext = os.path.splitext
473
# These were already lazily imported into local scope
415
# These were already imported into local scope
474
416
# mkdtemp = tempfile.mkdtemp
475
417
# rmtree = shutil.rmtree
484
419
MIN_ABS_PATHLENGTH = 1
487
422
if sys.platform == 'win32':
488
abspath = _win32_abspath
423
if win32utils.winver == 'Windows 98':
424
abspath = _win98_abspath
426
abspath = _win32_abspath
489
427
realpath = _win32_realpath
490
428
pathjoin = _win32_pathjoin
491
429
normpath = _win32_normpath
492
430
getcwd = _win32_getcwd
493
431
mkdtemp = _win32_mkdtemp
494
rename = _rename_wrap_exception(_win32_rename)
496
from . import _walkdirs_win32
500
lstat = _walkdirs_win32.lstat
501
fstat = _walkdirs_win32.fstat
502
wrap_stat = _walkdirs_win32.wrap_stat
432
rename = _win32_rename
504
434
MIN_ABS_PATHLENGTH = 3
544
471
On my standard US Windows XP, the preferred encoding is
545
472
cp1252, but the console is cp437
547
:param trace: If True trace the selected encoding via mutter().
549
from .trace import mutter
474
from bzrlib.trace import mutter
550
475
output_encoding = getattr(sys.stdout, 'encoding', None)
551
476
if not output_encoding:
552
477
input_encoding = getattr(sys.stdin, 'encoding', None)
553
478
if not input_encoding:
554
479
output_encoding = get_user_encoding()
556
mutter('encoding stdout as osutils.get_user_encoding() %r',
480
mutter('encoding stdout as osutils.get_user_encoding() %r',
559
483
output_encoding = input_encoding
561
mutter('encoding stdout as sys.stdin encoding %r',
484
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
565
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
486
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
566
487
if output_encoding == 'cp0':
567
488
# invalid encoding (cp0 means 'no codepage' on Windows)
568
489
output_encoding = get_user_encoding()
570
mutter('cp0 is invalid encoding.'
571
' encoding stdout as osutils.get_user_encoding() %r',
490
mutter('cp0 is invalid encoding.'
491
' encoding stdout as osutils.get_user_encoding() %r',
575
495
codecs.lookup(output_encoding)
576
496
except LookupError:
577
sys.stderr.write('brz: warning:'
497
sys.stderr.write('bzr: warning:'
578
498
' unknown terminal encoding %s.\n'
579
499
' Using encoding %s instead.\n'
580
500
% (output_encoding, get_user_encoding())
582
502
output_encoding = get_user_encoding()
584
504
return output_encoding
977
878
return '%d hours, %d minute%s %s' % (hours, minutes,
978
879
plural_minutes, direction)
982
882
"""Return size of given open file."""
983
return os.fstat(f.fileno())[stat.ST_SIZE]
986
# Alias os.urandom to support platforms (which?) without /dev/urandom and
987
# override if it doesn't work. Avoid checking on windows where there is
988
# significant initialisation cost that can be avoided for some bzr calls.
990
rand_bytes = os.urandom
992
if rand_bytes.__module__ != "nt":
883
return os.fstat(f.fileno())[ST_SIZE]
886
# Define rand_bytes based on platform.
888
# Python 2.4 and later have os.urandom,
889
# but it doesn't work on some arches
891
rand_bytes = os.urandom
892
except (NotImplementedError, AttributeError):
893
# If python doesn't have os.urandom, or it doesn't work,
894
# then try to first pull random data from /dev/urandom
995
except NotImplementedError:
896
rand_bytes = file('/dev/urandom', 'rb').read
897
# Otherwise, use this hack as a last resort
898
except (IOError, OSError):
996
899
# not well seeded, but better than nothing
997
900
def rand_bytes(n):
1016
917
for raw_byte in rand_bytes(num):
1018
s += ALNUM[ord(raw_byte) % 36]
1020
s += ALNUM[raw_byte % 36]
918
s += ALNUM[ord(raw_byte) % 36]
1024
# TODO: We could later have path objects that remember their list
1025
# decomposition (might be too tricksy though.)
922
## TODO: We could later have path objects that remember their list
923
## decomposition (might be too tricksy though.)
1027
925
def splitpath(p):
1028
926
"""Turn string into list of parts."""
1029
use_bytes = isinstance(p, bytes)
1030
if os.path.sep == '\\':
1031
# split on either delimiter because people might use either on
1034
ps = re.split(b'[\\\\/]', p)
1036
ps = re.split(r'[\\/]', p)
1045
current_empty_dir = (b'.', b'')
1048
current_empty_dir = ('.', '')
927
# split on either delimiter because people might use either on
929
ps = re.split(r'[\\/]', p)
1053
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1054
elif f in current_empty_dir:
934
raise errors.BzrError("sorry, %r not allowed in path" % f)
935
elif (f == '.') or (f == ''):
1111
993
def report_extension_load_failures():
1112
994
if not _extension_load_failures:
1114
if config.GlobalConfig().suppress_warning('missing_extensions'):
996
from bzrlib.config import GlobalConfig
997
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1116
999
# the warnings framework should by default show this only once
1117
from .trace import warning
1000
from bzrlib.trace import warning
1119
"brz: warning: some compiled extensions could not be loaded; "
1120
"see ``brz help missing-extensions``")
1002
"bzr: warning: some compiled extensions could not be loaded; "
1003
"see <https://answers.launchpad.net/bzr/+faq/703>")
1121
1004
# we no longer show the specific missing extensions here, because it makes
1122
1005
# the message too long and scary - see
1123
1006
# https://bugs.launchpad.net/bzr/+bug/430529
1127
from ._chunks_to_lines_pyx import chunks_to_lines
1128
except ImportError as e:
1010
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1011
except ImportError, e:
1129
1012
failed_to_load_extension(e)
1130
from ._chunks_to_lines_py import chunks_to_lines
1013
from bzrlib._chunks_to_lines_py import chunks_to_lines
1133
1016
def split_lines(s):
1134
1017
"""Split s into lines, but without removing the newline characters."""
1135
1018
# Trivially convert a fulltext into a 'chunked' representation, and let
1136
1019
# chunks_to_lines do the heavy lifting.
1137
if isinstance(s, bytes):
1020
if isinstance(s, str):
1138
1021
# chunks_to_lines only supports 8-bit strings
1139
1022
return chunks_to_lines([s])
1420
1284
return unicode_or_utf8_string.encode('utf-8')
1423
def safe_revision_id(unicode_or_utf8_string):
1287
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1288
' Revision id generators should be creating utf8'
1292
def safe_revision_id(unicode_or_utf8_string, warn=True):
1424
1293
"""Revision ids should now be utf8, but at one point they were unicode.
1426
1295
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1297
:param warn: Functions that are sanitizing user data can set warn=False
1428
1298
:return: None or a utf8 revision id.
1430
1300
if (unicode_or_utf8_string is None
1431
or unicode_or_utf8_string.__class__ == bytes):
1301
or unicode_or_utf8_string.__class__ == str):
1432
1302
return unicode_or_utf8_string
1433
raise TypeError('Unicode revision ids are no longer supported. '
1434
'Revision id generators should be creating utf8 revision '
1438
def safe_file_id(unicode_or_utf8_string):
1304
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1306
return cache_utf8.encode(unicode_or_utf8_string)
1309
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1310
' generators should be creating utf8 file ids.')
1313
def safe_file_id(unicode_or_utf8_string, warn=True):
1439
1314
"""File ids should now be utf8, but at one point they were unicode.
1441
1316
This is the same as safe_utf8, except it uses the cached encode functions
1557
1424
None is returned if the width can't established precisely.
1560
- if BRZ_COLUMNS is set, returns its value
1427
- if BZR_COLUMNS is set, returns its value
1561
1428
- if there is no controlling terminal, returns None
1562
- query the OS, if the queried size has changed since the last query,
1564
1429
- if COLUMNS is set, returns its value,
1565
- if the OS has a value (even though it's never changed), return its value.
1567
1431
From there, we need to query the OS to get the size of the controlling
1570
On Unices we query the OS by:
1571
1435
- get termios.TIOCGWINSZ
1572
1436
- if an error occurs or a negative value is obtained, returns None
1574
On Windows we query the OS by:
1575
1440
- win32utils.get_console_size() decides,
1576
1441
- returns None on error (provided default value)
1578
# Note to implementors: if changing the rules for determining the width,
1579
# make sure you've considered the behaviour in these cases:
1580
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1581
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1583
# - (add more interesting cases here, if you find any)
1584
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1585
# but we don't want to register a signal handler because it is impossible
1586
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1587
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1588
# time so we can notice if the reported size has changed, which should have
1591
# If BRZ_COLUMNS is set, take it, user is always right
1592
# Except if they specified 0 in which case, impose no limit here
1444
# If BZR_COLUMNS is set, take it, user is always right
1594
width = int(os.environ['BRZ_COLUMNS'])
1446
return int(os.environ['BZR_COLUMNS'])
1595
1447
except (KeyError, ValueError):
1597
if width is not None:
1603
1450
isatty = getattr(sys.stdout, 'isatty', None)
1604
if isatty is None or not isatty():
1605
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
1451
if isatty is None or not isatty():
1452
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1609
width, height = os_size = _terminal_size(None, None)
1610
global _first_terminal_size, _terminal_size_state
1611
if _terminal_size_state == 'no_data':
1612
_first_terminal_size = os_size
1613
_terminal_size_state = 'unchanged'
1614
elif (_terminal_size_state == 'unchanged' and
1615
_first_terminal_size != os_size):
1616
_terminal_size_state = 'changed'
1618
# If the OS claims to know how wide the terminal is, and this value has
1619
# ever changed, use that.
1620
if _terminal_size_state == 'changed':
1621
if width is not None and width > 0:
1624
# If COLUMNS is set, use it.
1455
# If COLUMNS is set, take it, the terminal knows better (even inside a
1456
# given terminal, the application can decide to set COLUMNS to a lower
1457
# value (splitted screen) or a bigger value (scroll bars))
1626
1459
return int(os.environ['COLUMNS'])
1627
1460
except (KeyError, ValueError):
1630
# Finally, use an unchanged size from the OS, if we have one.
1631
if _terminal_size_state == 'unchanged':
1632
if width is not None and width > 0:
1463
width, height = _terminal_size(None, None)
1465
# Consider invalid values as meaning no width
1635
# The width could not be determined.
1639
1471
def _win32_terminal_size(width, height):
1640
width, height = win32utils.get_console_size(
1641
defaultx=width, defaulty=height)
1472
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1642
1473
return width, height
1645
1476
def _ioctl_terminal_size(width, height):
1478
import struct, fcntl, termios
1650
1479
s = struct.pack('HHHH', 0, 0, 0, 0)
1651
1480
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1652
1481
height, width = struct.unpack('HHHH', x)[0:2]
1670
1498
_terminal_size = _ioctl_terminal_size
1673
def supports_executable(path):
1674
"""Return if filesystem at path supports executable bit.
1676
:param path: Path for which to check the file system
1677
:return: boolean indicating whether executable bit can be stored/relied upon
1679
if sys.platform == 'win32':
1682
fs_type = get_fs_type(path)
1683
except errors.DependencyNotPresent as e:
1684
trace.mutter('Unable to get fs type for %r: %s', path, e)
1686
if fs_type in ('vfat', 'ntfs'):
1687
# filesystems known to not support executable bit
1692
def supports_symlinks(path):
1693
"""Return if the filesystem at path supports the creation of symbolic links.
1696
if not has_symlinks():
1699
fs_type = get_fs_type(path)
1700
except errors.DependencyNotPresent as e:
1701
trace.mutter('Unable to get fs type for %r: %s', path, e)
1703
if fs_type in ('vfat', 'ntfs'):
1704
# filesystems known to not support symlinks
1501
def _terminal_size_changed(signum, frame):
1502
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1503
width, height = _terminal_size(None, None)
1504
if width is not None:
1505
os.environ['COLUMNS'] = str(width)
1508
_registered_sigwinch = False
1509
def watch_sigwinch():
1510
"""Register for SIGWINCH, once and only once.
1512
Do nothing if the signal module is not available.
1514
global _registered_sigwinch
1515
if not _registered_sigwinch:
1518
if getattr(signal, "SIGWINCH", None) is not None:
1519
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1521
# python doesn't provide signal support, nothing we can do about it
1523
_registered_sigwinch = True
1526
def supports_executable():
1527
return sys.platform != "win32"
1709
1530
def supports_posix_readonly():
1952
1779
See DirReader.read_dir for details.
1954
1781
_utf8_encode = self._utf8_encode
1956
def _fs_decode(s): return s.decode(_fs_enc)
1958
def _fs_encode(s): return s.encode(_fs_enc)
1959
1782
_lstat = os.lstat
1960
1783
_listdir = os.listdir
1961
1784
_kind_from_mode = file_kind_from_stat_mode
1964
relprefix = prefix + b'/'
1787
relprefix = prefix + '/'
1967
top_slash = top + '/'
1790
top_slash = top + u'/'
1970
1793
append = dirblock.append
1971
for name_native in _listdir(top.encode('utf-8')):
1794
for name in sorted(_listdir(top)):
1973
name = _fs_decode(name_native)
1796
name_utf8 = _utf8_encode(name)[0]
1974
1797
except UnicodeDecodeError:
1975
1798
raise errors.BadFilenameEncoding(
1976
relprefix + name_native, _fs_enc)
1977
name_utf8 = _utf8_encode(name)[0]
1799
_utf8_encode(relprefix)[0] + name, _fs_enc)
1978
1800
abspath = top_slash + name
1979
1801
statvalue = _lstat(abspath)
1980
1802
kind = _kind_from_mode(statvalue.st_mode)
1981
1803
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1982
return sorted(dirblock)
1985
1807
def copy_tree(from_path, to_path, handlers={}):
2055
1874
This can be used to sort paths in the same way that walkdirs does.
2057
return (dirname(path), path)
1876
return (dirname(path) , path)
2060
1879
def compare_paths_prefix_order(path_a, path_b):
2061
1880
"""Compare path_a and path_b to generate the same order walkdirs uses."""
2062
1881
key_a = path_prefix_key(path_a)
2063
1882
key_b = path_prefix_key(path_b)
2064
return (key_a > key_b) - (key_a < key_b)
1883
return cmp(key_a, key_b)
2067
1886
_cached_user_encoding = None
2070
def get_user_encoding():
1889
def get_user_encoding(use_cache=True):
2071
1890
"""Find out what the preferred user encoding is.
2073
1892
This is generally the encoding that is used for command line parameters
2074
1893
and file contents. This may be different from the terminal encoding
2075
1894
or the filesystem encoding.
1896
:param use_cache: Enable cache for detected encoding.
1897
(This parameter is turned on by default,
1898
and required only for selftesting)
2077
1900
:return: A string defining the preferred user encoding
2079
1902
global _cached_user_encoding
2080
if _cached_user_encoding is not None:
1903
if _cached_user_encoding is not None and use_cache:
2081
1904
return _cached_user_encoding
2083
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2084
# Use the existing locale settings and call nl_langinfo directly
2085
# rather than going through getpreferredencoding. This avoids
2086
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2087
# possibility of the setlocale call throwing an error.
2088
user_encoding = locale.nl_langinfo(locale.CODESET)
1906
if sys.platform == 'darwin':
1907
# python locale.getpreferredencoding() always return
1908
# 'mac-roman' on darwin. That's a lie.
1909
sys.platform = 'posix'
1911
if os.environ.get('LANG', None) is None:
1912
# If LANG is not set, we end up with 'ascii', which is bad
1913
# ('mac-roman' is more than ascii), so we set a default which
1914
# will give us UTF-8 (which appears to work in all cases on
1915
# OSX). Users are still free to override LANG of course, as
1916
# long as it give us something meaningful. This work-around
1917
# *may* not be needed with python 3k and/or OSX 10.5, but will
1918
# work with them too -- vila 20080908
1919
os.environ['LANG'] = 'en_US.UTF-8'
1922
sys.platform = 'darwin'
2090
# GZ 2011-12-19: On windows could call GetACP directly instead.
2091
user_encoding = locale.getpreferredencoding(False)
2094
user_encoding = codecs.lookup(user_encoding).name
2096
if user_encoding not in ("", "cp0"):
2097
sys.stderr.write('brz: warning:'
1927
user_encoding = locale.getpreferredencoding()
1928
except locale.Error, e:
1929
sys.stderr.write('bzr: warning: %s\n'
1930
' Could not determine what text encoding to use.\n'
1931
' This error usually means your Python interpreter\n'
1932
' doesn\'t support the locale set by $LANG (%s)\n'
1933
" Continuing with ascii encoding.\n"
1934
% (e, os.environ.get('LANG')))
1935
user_encoding = 'ascii'
1937
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1938
# treat that as ASCII, and not support printing unicode characters to the
1941
# For python scripts run under vim, we get '', so also treat that as ASCII
1942
if user_encoding in (None, 'cp0', ''):
1943
user_encoding = 'ascii'
1947
codecs.lookup(user_encoding)
1949
sys.stderr.write('bzr: warning:'
2098
1950
' unknown encoding %s.'
2099
1951
' Continuing with ascii encoding.\n'
2100
1952
% user_encoding
2102
user_encoding = 'ascii'
2104
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2105
if user_encoding == 'ascii':
2106
if sys.platform == 'darwin':
2107
# OSX is special-cased in Python to have a UTF-8 filesystem
2108
# encoding and previously had LANG set here if not present.
2109
user_encoding = 'utf-8'
2110
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2111
# for some other posix platforms as well.
2113
_cached_user_encoding = user_encoding
1954
user_encoding = 'ascii'
1957
_cached_user_encoding = user_encoding
2114
1959
return user_encoding
2117
def get_diff_header_encoding():
2118
return get_terminal_encoding()
2121
1962
def get_host_name():
2122
1963
"""Return the current unicode host name.
2139
1979
# data at once.
2140
1980
MAX_SOCKET_CHUNK = 64 * 1024
2142
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2143
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2144
_eno = getattr(errno, _eno, None)
2145
if _eno is not None:
2146
_end_of_stream_errors.append(_eno)
2150
1982
def read_bytes_from_socket(sock, report_activity=None,
2151
max_read_size=MAX_SOCKET_CHUNK):
1983
max_read_size=MAX_SOCKET_CHUNK):
2152
1984
"""Read up to max_read_size of bytes from sock and notify of progress.
2154
1986
Translates "Connection reset by peer" into file-like EOF (return an
2155
1987
empty string rather than raise an error), and repeats the recv if
2156
1988
interrupted by a signal.
2160
data = sock.recv(max_read_size)
2161
except socket.error as e:
1992
bytes = sock.recv(max_read_size)
1993
except socket.error, e:
2162
1994
eno = e.args[0]
2163
if eno in _end_of_stream_errors:
1995
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2164
1996
# The connection was closed by the other side. Callers expect
2165
1997
# an empty string to signal end-of-stream.
2167
1999
elif eno == errno.EINTR:
2168
2000
# Retry the interrupted recv.
2172
2004
if report_activity is not None:
2173
report_activity(len(data), 'read')
2005
report_activity(len(bytes), 'read')
2177
2009
def recv_all(socket, count):
2185
2017
This isn't optimized and is intended mostly for use in testing.
2188
2020
while len(b) < count:
2189
2021
new = read_bytes_from_socket(socket, None, count - len(b))
2196
2028
def send_all(sock, bytes, report_activity=None):
2197
2029
"""Send all bytes on a socket.
2199
2031
Breaks large blocks in smaller chunks to avoid buffering limitations on
2200
2032
some platforms, and catches EINTR which may be thrown if the send is
2201
2033
interrupted by a signal.
2203
2035
This is preferred to socket.sendall(), because it avoids portability bugs
2204
2036
and provides activity reporting.
2206
2038
:param report_activity: Call this as bytes are read, see
2207
2039
Transport._report_activity
2210
2042
byte_count = len(bytes)
2211
view = memoryview(bytes)
2212
2043
while sent_total < byte_count:
2214
sent = sock.send(view[sent_total:sent_total + MAX_SOCKET_CHUNK])
2215
except (socket.error, IOError) as e:
2216
if e.args[0] in _end_of_stream_errors:
2217
raise errors.ConnectionReset(
2218
"Error trying to write to socket", e)
2045
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2046
except socket.error, e:
2219
2047
if e.args[0] != errno.EINTR:
2223
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2225
2050
sent_total += sent
2226
if report_activity is not None:
2227
report_activity(sent, 'write')
2230
def connect_socket(address):
2231
# Slight variation of the socket.create_connection() function (provided by
2232
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2233
# provide it for previous python versions. Also, we don't use the timeout
2234
# parameter (provided by the python implementation) so we don't implement
2236
err = socket.error('getaddrinfo returns an empty list')
2237
host, port = address
2238
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2239
af, socktype, proto, canonname, sa = res
2242
sock = socket.socket(af, socktype, proto)
2246
except socket.error as e:
2248
# 'err' is now the most recent error
2249
if sock is not None:
2051
report_activity(sent, 'write')
2254
2054
def dereference_path(path):
2282
2082
If and when pkg_resources becomes a standard library, this routine
2283
2083
can delegate to it.
2285
# Check package name is within breezy
2286
if package == "breezy":
2085
# Check package name is within bzrlib
2086
if package == "bzrlib":
2287
2087
resource_relpath = resource_name
2288
elif package.startswith("breezy."):
2289
package = package[len("breezy."):].replace('.', os.sep)
2088
elif package.startswith("bzrlib."):
2089
package = package[len("bzrlib."):].replace('.', os.sep)
2290
2090
resource_relpath = pathjoin(package, resource_name)
2292
raise errors.BzrError('resource package %s not in breezy' % package)
2092
raise errors.BzrError('resource package %s not in bzrlib' % package)
2294
2094
# Map the resource to a file and read its contents
2295
base = dirname(breezy.__file__)
2095
base = dirname(bzrlib.__file__)
2296
2096
if getattr(sys, 'frozen', None): # bzr.exe
2297
2097
base = abspath(pathjoin(base, '..', '..'))
2298
with open(pathjoin(base, resource_relpath), "rt") as f:
2098
filename = pathjoin(base, resource_relpath)
2099
return open(filename, 'rU').read()
2302
2102
def file_kind_from_stat_mode_thunk(mode):
2303
2103
global file_kind_from_stat_mode
2304
2104
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2306
from ._readdir_pyx import UTF8DirReader
2106
from bzrlib._readdir_pyx import UTF8DirReader
2307
2107
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2108
except ImportError, e:
2309
2109
# This is one time where we won't warn that an extension failed to
2310
2110
# load. The extension is never available on Windows anyway.
2311
from ._readdir_py import (
2111
from bzrlib._readdir_py import (
2312
2112
_kind_from_mode as file_kind_from_stat_mode
2314
2114
return file_kind_from_stat_mode(mode)
2317
2115
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2320
def file_stat(f, _lstat=os.lstat):
2118
def file_kind(f, _lstat=os.lstat):
2324
except OSError as e:
2120
return file_kind_from_stat_mode(_lstat(f).st_mode)
2325
2122
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2326
2123
raise errors.NoSuchFile(f)
2330
def file_kind(f, _lstat=os.lstat):
2331
stat_value = file_stat(f, _lstat)
2332
return file_kind_from_stat_mode(stat_value.st_mode)
2335
2127
def until_no_eintr(f, *a, **kw):
2336
2128
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2338
2130
WARNING: you must be certain that it is safe to retry the call repeatedly
2339
2131
if EINTR does occur. This is typically only true for low-level operations
2340
2132
like os.read. If in any doubt, don't use this.
2342
2134
Keep in mind that this is not a complete solution to EINTR. There is
2343
2135
probably code in the Python standard library and other dependencies that
2344
2136
may encounter EINTR if a signal arrives (and there is signal handler for
2345
that signal). So this function can reduce the impact for IO that breezy
2137
that signal). So this function can reduce the impact for IO that bzrlib
2346
2138
directly controls, but it is not a complete solution.
2348
2140
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2351
2143
return f(*a, **kw)
2352
except (IOError, OSError) as e:
2144
except (IOError, OSError), e:
2353
2145
if e.errno == errno.EINTR:
2150
def re_compile_checked(re_string, flags=0, where=""):
2151
"""Return a compiled re, or raise a sensible error.
2153
This should only be used when compiling user-supplied REs.
2155
:param re_string: Text form of regular expression.
2156
:param flags: eg re.IGNORECASE
2157
:param where: Message explaining to the user the context where
2158
it occurred, eg 'log search filter'.
2160
# from https://bugs.launchpad.net/bzr/+bug/251352
2162
re_obj = re.compile(re_string, flags)
2167
where = ' in ' + where
2168
# despite the name 'error' is a type
2169
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2170
% (where, re_string, e))
2358
2173
if sys.platform == "win32":
2361
2176
return msvcrt.getch()
2366
2181
fd = sys.stdin.fileno()
2367
2182
settings = termios.tcgetattr(fd)
2372
2187
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2375
if sys.platform.startswith('linux'):
2191
if sys.platform == 'linux2':
2376
2192
def _local_concurrency():
2378
return os.sysconf('SC_NPROCESSORS_ONLN')
2379
except (ValueError, OSError, AttributeError):
2194
prefix = 'processor'
2195
for line in file('/proc/cpuinfo', 'rb'):
2196
if line.startswith(prefix):
2197
concurrency = int(line[line.find(':')+1:]) + 1
2381
2199
elif sys.platform == 'darwin':
2382
2200
def _local_concurrency():
2383
2201
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2384
2202
stdout=subprocess.PIPE).communicate()[0]
2385
elif "bsd" in sys.platform:
2203
elif sys.platform[0:7] == 'freebsd':
2386
2204
def _local_concurrency():
2387
2205
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2388
2206
stdout=subprocess.PIPE).communicate()[0]
2389
2207
elif sys.platform == 'sunos5':
2390
2208
def _local_concurrency():
2391
return subprocess.Popen(['psrinfo', '-p', ],
2209
return subprocess.Popen(['psrinfo', '-p',],
2392
2210
stdout=subprocess.PIPE).communicate()[0]
2393
2211
elif sys.platform == "win32":
2394
2212
def _local_concurrency():
2491
2302
return os.fdopen(os.open(filename, flags), mode, bufsize)
2493
2304
open_file = open
2496
def available_backup_name(base, exists):
2497
"""Find a non-existing backup file name.
2499
This will *not* create anything, this only return a 'free' entry. This
2500
should be used for checking names in a directory below a locked
2501
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2502
Leap) and generally discouraged.
2504
:param base: The base name.
2506
:param exists: A callable returning True if the path parameter exists.
2509
name = "%s.~%d~" % (base, counter)
2512
name = "%s.~%d~" % (base, counter)
2516
def set_fd_cloexec(fd):
2517
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2518
support for this is not available.
2522
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2523
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2524
except (ImportError, AttributeError):
2525
# Either the fcntl module or specific constants are not present
2529
def find_executable_on_path(name):
2530
"""Finds an executable on the PATH.
2532
On Windows, this will try to append each extension in the PATHEXT
2533
environment variable to the name, if it cannot be found with the name
2536
:param name: The base name of the executable.
2537
:return: The path to the executable found or None.
2539
if sys.platform == 'win32':
2540
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2541
exts = [ext.lower() for ext in exts]
2542
base, ext = os.path.splitext(name)
2544
if ext.lower() not in exts:
2550
path = os.environ.get('PATH')
2551
if path is not None:
2552
path = path.split(os.pathsep)
2555
f = os.path.join(d, name) + ext
2556
if os.access(f, os.X_OK):
2558
if sys.platform == 'win32':
2559
app_path = win32utils.get_app_path(name)
2560
if app_path != name:
2565
def _posix_is_local_pid_dead(pid):
2566
"""True if pid doesn't correspond to live process on this machine"""
2568
# Special meaning of unix kill: just check if it's there.
2570
except OSError as e:
2571
if e.errno == errno.ESRCH:
2572
# On this machine, and really not found: as sure as we can be
2575
elif e.errno == errno.EPERM:
2576
# exists, though not ours
2579
trace.mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2580
# Don't really know.
2583
# Exists and our process: not dead.
2587
if sys.platform == "win32":
2588
is_local_pid_dead = win32utils.is_local_pid_dead
2590
is_local_pid_dead = _posix_is_local_pid_dead
2592
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2593
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2594
if getattr(errno, name, None) is not None]
2597
def fdatasync(fileno):
2598
"""Flush file contents to disk if possible.
2600
:param fileno: Integer OS file handle.
2601
:raises TransportNotPossible: If flushing to disk is not possible.
2603
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2607
except IOError as e:
2608
# See bug #1075108, on some platforms fdatasync exists, but can
2609
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2610
# and reduce the chance of corruption-on-powerloss situations. It
2611
# is not a mandatory call, so it is ok to suppress failures.
2612
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2613
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2617
def ensure_empty_directory_exists(path, exception_class):
2618
"""Make sure a local directory exists and is empty.
2620
If it does not exist, it is created. If it exists and is not empty, an
2621
instance of exception_class is raised.
2625
except OSError as e:
2626
if e.errno != errno.EEXIST:
2628
if os.listdir(path) != []:
2629
raise exception_class(path)
2632
def is_environment_error(evalue):
2633
"""True if exception instance is due to a process environment issue
2635
This includes OSError and IOError, but also other errors that come from
2636
the operating system or core libraries but are not subclasses of those.
2638
if isinstance(evalue, (EnvironmentError, select.error)):
2640
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
2645
def read_mtab(path):
2646
"""Read an fstab-style file and extract mountpoint+filesystem information.
2648
:param path: Path to read from
2649
:yield: Tuples with mountpoints (as bytestrings) and filesystem names
2651
with open(path, 'rb') as f:
2653
if line.startswith(b'#'):
2658
yield cols[1], cols[2].decode('ascii', 'replace')
2661
MTAB_PATH = '/etc/mtab'
2663
class FilesystemFinder(object):
2664
"""Find the filesystem for a particular path."""
2666
def __init__(self, mountpoints):
2669
self._mountpoints = sorted(mountpoints, key=key, reverse=True)
2673
"""Create a FilesystemFinder from an mtab-style file.
2675
Note that this will silenty ignore mtab if it doesn't exist or can not
2678
# TODO(jelmer): Use inotify to be notified when /etc/mtab changes and
2679
# we need to re-read it.
2681
return cls(read_mtab(MTAB_PATH))
2682
except EnvironmentError as e:
2683
trace.mutter('Unable to read mtab: %s', e)
2686
def find(self, path):
2687
"""Find the filesystem used by a particular path.
2689
:param path: Path to find (bytestring or text type)
2690
:return: Filesystem name (as text type) or None, if the filesystem is
2693
for mountpoint, filesystem in self._mountpoints:
2694
if is_inside(mountpoint, path):
2699
_FILESYSTEM_FINDER = None
2702
def get_fs_type(path):
2703
"""Return the filesystem type for the partition a path is in.
2705
:param path: Path to search filesystem type for
2706
:return: A FS type, as string. E.g. "ext2"
2708
global _FILESYSTEM_FINDER
2709
if _FILESYSTEM_FINDER is None:
2710
_FILESYSTEM_FINDER = FilesystemFinder.from_mtab()
2712
if not isinstance(path, bytes):
2713
path = path.encode(_fs_enc)
2715
return _FILESYSTEM_FINDER.find(path)
2719
perf_counter = time.perf_counter
2721
perf_counter = time.clock
added
removed
bzrlib/osutils.py
