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
25
from bzrlib.lazy_import import lazy_import
25
from .lazy_import import lazy_import
26
26
lazy_import(globals(), """
27
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,
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
43
from tempfile import (
42
from tempfile import mkdtemp
50
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
88
66
# On win32, O_BINARY is used to indicate the file should
89
67
# be opened in binary mode, rather than text mode.
96
74
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
99
def get_unicode_argv():
101
user_encoding = get_user_encoding()
102
return [a.decode(user_encoding) for a in sys.argv[1:]]
103
except UnicodeDecodeError:
104
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
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
108
86
def make_readonly(filename):
109
87
"""Make a filename read-only."""
110
88
mod = os.lstat(filename).st_mode
111
89
if not stat.S_ISLNK(mod):
113
os.chmod(filename, mod)
91
chmod_if_possible(filename, mod)
116
94
def make_writable(filename):
117
95
mod = os.lstat(filename).st_mode
118
96
if not stat.S_ISLNK(mod):
120
os.chmod(filename, 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" % (
123
120
def minimum_path_selection(paths):
232
234
file_existed = False
234
236
rename_func(new, tmp_name)
235
except (errors.NoSuchFile,), e:
237
except (errors.NoSuchFile,):
238
240
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
239
241
# function raises an IOError with errno is None when a rename fails.
240
242
# This then gets caught here.
241
243
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
245
except Exception as e:
244
246
if (getattr(e, 'errno', None) is None
245
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
247
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
248
250
file_existed = True
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)
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)
264
263
and old.lower() == new.lower()):
265
# source and target are the same file on a case-insensitive
266
# filesystem, so we don't generate an exception
264
# source and target are the same file on a case-insensitive
265
# filesystem, so we don't generate an exception
270
271
# If the file used to exist, rename it back into place
273
274
unlink_func(tmp_name)
275
276
rename_func(tmp_name, new)
276
if failure_exc is not None:
277
raise failure_exc[0], failure_exc[1], failure_exc[2]
280
279
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
281
280
# choke on a Unicode string containing a relative path if
282
281
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
284
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
285
283
def _posix_abspath(path):
286
284
# jam 20060426 rather than encoding to fsencoding
287
285
# copy posixpath.abspath, but use os.getcwdu instead
288
286
if not posixpath.isabs(path):
289
287
path = posixpath.join(getcwd(), path)
290
return posixpath.normpath(path)
288
return _posix_normpath(path)
293
291
def _posix_realpath(path):
294
292
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()
297
326
def _win32_fixdrive(path):
298
327
"""Force drive letters to be consistent.
303
332
running python.exe under cmd.exe return capital C:\\
304
333
running win32 python inside a cygwin shell returns lowercase c:\\
306
drive, path = _nt_splitdrive(path)
335
drive, path = ntpath.splitdrive(path)
307
336
return drive.upper() + path
310
339
def _win32_abspath(path):
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('\\', '/'))
340
# Real ntpath.abspath doesn't have a problem with a unicode cwd
341
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
342
344
def _win32_realpath(path):
343
# Real _nt_realpath doesn't have a problem with a unicode cwd
344
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
345
# Real ntpath.realpath doesn't have a problem with a unicode cwd
346
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
347
349
def _win32_pathjoin(*args):
348
return _nt_join(*args).replace('\\', '/')
350
return ntpath.join(*args).replace('\\', '/')
351
353
def _win32_normpath(path):
352
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
354
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
355
357
def _win32_getcwd():
356
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
358
return _win32_fixdrive(_getcwd().replace('\\', '/'))
359
361
def _win32_mkdtemp(*args, **kwargs):
360
362
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
371
365
def _win32_rename(old, new):
372
366
"""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)
399
383
def _mac_getcwd():
400
return unicodedata.normalize('NFC', os.getcwdu())
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)
403
414
# Default is to just use the python builtins, but these can be rebound on
404
415
# particular platforms.
405
416
abspath = _posix_abspath
406
417
realpath = _posix_realpath
407
418
pathjoin = os.path.join
408
normpath = os.path.normpath
409
rename = _wrapped_rename # overridden below on win32
419
normpath = _posix_normpath
420
_get_home_dir = _posix_get_home_dir
421
getuser_unicode = _posix_getuser_unicode
411
423
dirname = os.path.dirname
412
424
basename = os.path.basename
413
425
split = os.path.split
414
426
splitext = os.path.splitext
415
# These were already imported into local scope
427
# These were already lazily imported into local scope
416
428
# mkdtemp = tempfile.mkdtemp
417
429
# rmtree = shutil.rmtree
419
438
MIN_ABS_PATHLENGTH = 1
422
441
if sys.platform == 'win32':
423
if win32utils.winver == 'Windows 98':
424
abspath = _win98_abspath
426
abspath = _win32_abspath
442
abspath = _win32_abspath
427
443
realpath = _win32_realpath
428
444
pathjoin = _win32_pathjoin
429
445
normpath = _win32_normpath
430
446
getcwd = _win32_getcwd
431
447
mkdtemp = _win32_mkdtemp
432
rename = _win32_rename
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
434
458
MIN_ABS_PATHLENGTH = 3
471
494
On my standard US Windows XP, the preferred encoding is
472
495
cp1252, but the console is cp437
497
:param trace: If True trace the selected encoding via mutter().
474
from bzrlib.trace import mutter
499
from .trace import mutter
475
500
output_encoding = getattr(sys.stdout, 'encoding', None)
476
501
if not output_encoding:
477
502
input_encoding = getattr(sys.stdin, 'encoding', None)
478
503
if not input_encoding:
479
504
output_encoding = get_user_encoding()
480
mutter('encoding stdout as osutils.get_user_encoding() %r',
506
mutter('encoding stdout as osutils.get_user_encoding() %r',
483
509
output_encoding = input_encoding
484
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
511
mutter('encoding stdout as sys.stdin encoding %r',
486
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
515
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
487
516
if output_encoding == 'cp0':
488
517
# invalid encoding (cp0 means 'no codepage' on Windows)
489
518
output_encoding = get_user_encoding()
490
mutter('cp0 is invalid encoding.'
491
' encoding stdout as osutils.get_user_encoding() %r',
520
mutter('cp0 is invalid encoding.'
521
' encoding stdout as osutils.get_user_encoding() %r',
495
525
codecs.lookup(output_encoding)
496
526
except LookupError:
497
sys.stderr.write('bzr: warning:'
527
sys.stderr.write('brz: warning:'
498
528
' unknown terminal encoding %s.\n'
499
529
' Using encoding %s instead.\n'
500
530
% (output_encoding, get_user_encoding())
502
532
output_encoding = get_user_encoding()
504
534
return output_encoding
878
923
return '%d hours, %d minute%s %s' % (hours, minutes,
879
924
plural_minutes, direction)
882
928
"""Return size of given open file."""
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
929
return os.fstat(f.fileno())[stat.ST_SIZE]
932
# Alias os.urandom to support platforms (which?) without /dev/urandom and
933
# override if it doesn't work. Avoid checking on windows where there is
934
# significant initialisation cost that can be avoided for some bzr calls.
936
rand_bytes = os.urandom
938
if rand_bytes.__module__ != "nt":
896
rand_bytes = file('/dev/urandom', 'rb').read
897
# Otherwise, use this hack as a last resort
898
except (IOError, OSError):
941
except NotImplementedError:
899
942
# not well seeded, but better than nothing
900
943
def rand_bytes(n):
917
962
for raw_byte in rand_bytes(num):
918
s += ALNUM[ord(raw_byte) % 36]
963
s += ALNUM[raw_byte % 36]
922
## TODO: We could later have path objects that remember their list
923
## decomposition (might be too tricksy though.)
967
# TODO: We could later have path objects that remember their list
968
# decomposition (might be too tricksy though.)
925
970
def splitpath(p):
926
971
"""Turn string into list of parts."""
927
# split on either delimiter because people might use either on
929
ps = re.split(r'[\\/]', p)
972
use_bytes = isinstance(p, bytes)
973
if os.path.sep == '\\':
974
# split on either delimiter because people might use either on
977
ps = re.split(b'[\\\\/]', p)
979
ps = re.split(r'[\\/]', p)
988
current_empty_dir = (b'.', b'')
991
current_empty_dir = ('.', '')
934
raise errors.BzrError("sorry, %r not allowed in path" % f)
935
elif (f == '.') or (f == ''):
996
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
997
elif f in current_empty_dir:
993
1054
def report_extension_load_failures():
994
1055
if not _extension_load_failures:
996
from bzrlib.config import GlobalConfig
997
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1057
if config.GlobalConfig().suppress_warning('missing_extensions'):
999
1059
# the warnings framework should by default show this only once
1000
from bzrlib.trace import warning
1060
from .trace import warning
1002
"bzr: warning: some compiled extensions could not be loaded; "
1003
"see <https://answers.launchpad.net/bzr/+faq/703>")
1062
"brz: warning: some compiled extensions could not be loaded; "
1063
"see ``brz help missing-extensions``")
1004
1064
# we no longer show the specific missing extensions here, because it makes
1005
1065
# the message too long and scary - see
1006
1066
# https://bugs.launchpad.net/bzr/+bug/430529
1010
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1011
except ImportError, e:
1070
from ._chunks_to_lines_pyx import chunks_to_lines
1071
except ImportError as e:
1012
1072
failed_to_load_extension(e)
1013
from bzrlib._chunks_to_lines_py import chunks_to_lines
1073
from ._chunks_to_lines_py import chunks_to_lines
1016
1076
def split_lines(s):
1017
1077
"""Split s into lines, but without removing the newline characters."""
1018
1078
# Trivially convert a fulltext into a 'chunked' representation, and let
1019
1079
# chunks_to_lines do the heavy lifting.
1020
if isinstance(s, str):
1080
if isinstance(s, bytes):
1021
1081
# chunks_to_lines only supports 8-bit strings
1022
1082
return chunks_to_lines([s])
1284
1362
return unicode_or_utf8_string.encode('utf-8')
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):
1365
def safe_revision_id(unicode_or_utf8_string):
1293
1366
"""Revision ids should now be utf8, but at one point they were unicode.
1295
1368
: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
1298
1370
:return: None or a utf8 revision id.
1300
1372
if (unicode_or_utf8_string is None
1301
or unicode_or_utf8_string.__class__ == str):
1373
or unicode_or_utf8_string.__class__ == bytes):
1302
1374
return 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):
1375
raise TypeError('Unicode revision ids are no longer supported. '
1376
'Revision id generators should be creating utf8 revision '
1380
def safe_file_id(unicode_or_utf8_string):
1314
1381
"""File ids should now be utf8, but at one point they were unicode.
1316
1383
This is the same as safe_utf8, except it uses the cached encode functions
1424
1499
None is returned if the width can't established precisely.
1427
- if BZR_COLUMNS is set, returns its value
1502
- if BRZ_COLUMNS is set, returns its value
1428
1503
- if there is no controlling terminal, returns None
1504
- query the OS, if the queried size has changed since the last query,
1429
1506
- if COLUMNS is set, returns its value,
1507
- if the OS has a value (even though it's never changed), return its value.
1431
1509
From there, we need to query the OS to get the size of the controlling
1512
On Unices we query the OS by:
1435
1513
- get termios.TIOCGWINSZ
1436
1514
- if an error occurs or a negative value is obtained, returns None
1516
On Windows we query the OS by:
1440
1517
- win32utils.get_console_size() decides,
1441
1518
- returns None on error (provided default value)
1520
# Note to implementors: if changing the rules for determining the width,
1521
# make sure you've considered the behaviour in these cases:
1522
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1523
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1525
# - (add more interesting cases here, if you find any)
1526
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1527
# but we don't want to register a signal handler because it is impossible
1528
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1529
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1530
# time so we can notice if the reported size has changed, which should have
1444
# If BZR_COLUMNS is set, take it, user is always right
1533
# If BRZ_COLUMNS is set, take it, user is always right
1534
# Except if they specified 0 in which case, impose no limit here
1446
return int(os.environ['BZR_COLUMNS'])
1536
width = int(os.environ['BRZ_COLUMNS'])
1447
1537
except (KeyError, ValueError):
1539
if width is not None:
1450
1545
isatty = getattr(sys.stdout, 'isatty', None)
1451
if isatty is None or not isatty():
1452
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1546
if isatty is None or not isatty():
1547
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
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))
1551
width, height = os_size = _terminal_size(None, None)
1552
global _first_terminal_size, _terminal_size_state
1553
if _terminal_size_state == 'no_data':
1554
_first_terminal_size = os_size
1555
_terminal_size_state = 'unchanged'
1556
elif (_terminal_size_state == 'unchanged' and
1557
_first_terminal_size != os_size):
1558
_terminal_size_state = 'changed'
1560
# If the OS claims to know how wide the terminal is, and this value has
1561
# ever changed, use that.
1562
if _terminal_size_state == 'changed':
1563
if width is not None and width > 0:
1566
# If COLUMNS is set, use it.
1459
1568
return int(os.environ['COLUMNS'])
1460
1569
except (KeyError, ValueError):
1463
width, height = _terminal_size(None, None)
1465
# Consider invalid values as meaning no width
1572
# Finally, use an unchanged size from the OS, if we have one.
1573
if _terminal_size_state == 'unchanged':
1574
if width is not None and width > 0:
1577
# The width could not be determined.
1471
1581
def _win32_terminal_size(width, height):
1472
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1582
width, height = win32utils.get_console_size(
1583
defaultx=width, defaulty=height)
1473
1584
return width, height
1476
1587
def _ioctl_terminal_size(width, height):
1478
import struct, fcntl, termios
1479
1592
s = struct.pack('HHHH', 0, 0, 0, 0)
1480
1593
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1481
1594
height, width = struct.unpack('HHHH', x)[0:2]
1498
1612
_terminal_size = _ioctl_terminal_size
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"
1615
def supports_executable(path):
1616
"""Return if filesystem at path supports executable bit.
1618
:param path: Path for which to check the file system
1619
:return: boolean indicating whether executable bit can be stored/relied upon
1621
if sys.platform == 'win32':
1624
fs_type = get_fs_type(path)
1625
except errors.DependencyNotPresent as e:
1626
trace.mutter('Unable to get fs type for %r: %s', path, e)
1628
if fs_type in ('vfat', 'ntfs'):
1629
# filesystems known to not support executable bit
1634
def supports_symlinks(path):
1635
"""Return if the filesystem at path supports the creation of symbolic links.
1638
if not has_symlinks():
1641
fs_type = get_fs_type(path)
1642
except errors.DependencyNotPresent as e:
1643
trace.mutter('Unable to get fs type for %r: %s', path, e)
1645
if fs_type in ('vfat', 'ntfs'):
1646
# filesystems known to not support symlinks
1530
1651
def supports_posix_readonly():
1712
1837
global _selected_dir_reader
1713
1838
if _selected_dir_reader is None:
1714
fs_encoding = _fs_enc.upper()
1715
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1716
# Win98 doesn't have unicode apis like FindFirstFileW
1717
# TODO: We possibly could support Win98 by falling back to the
1718
# original FindFirstFile, and using TCHAR instead of WCHAR,
1719
# but that gets a bit tricky, and requires custom compiling
1839
if sys.platform == "win32":
1722
from bzrlib._walkdirs_win32 import Win32ReadDir
1841
from ._walkdirs_win32 import Win32ReadDir
1723
1842
_selected_dir_reader = Win32ReadDir()
1724
1843
except ImportError:
1726
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1727
# ANSI_X3.4-1968 is a form of ASCII
1845
elif _fs_enc in ('utf-8', 'ascii'):
1729
from bzrlib._readdir_pyx import UTF8DirReader
1847
from ._readdir_pyx import UTF8DirReader
1730
1848
_selected_dir_reader = UTF8DirReader()
1731
except ImportError, e:
1849
except ImportError as e:
1732
1850
failed_to_load_extension(e)
1779
1897
See DirReader.read_dir for details.
1781
1899
_utf8_encode = self._utf8_encode
1783
_listdir = os.listdir
1784
_kind_from_mode = file_kind_from_stat_mode
1901
def _fs_decode(s): return s.decode(_fs_enc)
1903
def _fs_encode(s): return s.encode(_fs_enc)
1787
relprefix = prefix + '/'
1906
relprefix = prefix + b'/'
1790
top_slash = top + u'/'
1909
top_slash = top + '/'
1793
1912
append = dirblock.append
1794
for name in sorted(_listdir(top)):
1913
for entry in scandir(safe_utf8(top)):
1796
name_utf8 = _utf8_encode(name)[0]
1915
name = _fs_decode(entry.name)
1797
1916
except UnicodeDecodeError:
1798
1917
raise errors.BadFilenameEncoding(
1799
_utf8_encode(relprefix)[0] + name, _fs_enc)
1918
relprefix + entry.name, _fs_enc)
1800
1919
abspath = top_slash + name
1801
statvalue = _lstat(abspath)
1802
kind = _kind_from_mode(statvalue.st_mode)
1920
name_utf8 = _utf8_encode(name)[0]
1921
statvalue = entry.stat(follow_symlinks=False)
1922
kind = file_kind_from_stat_mode(statvalue.st_mode)
1803
1923
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1924
return sorted(dirblock)
1807
1927
def copy_tree(from_path, to_path, handlers={}):
1874
1997
This can be used to sort paths in the same way that walkdirs does.
1876
return (dirname(path) , path)
1999
return (dirname(path), path)
1879
2002
def compare_paths_prefix_order(path_a, path_b):
1880
2003
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1881
2004
key_a = path_prefix_key(path_a)
1882
2005
key_b = path_prefix_key(path_b)
1883
return cmp(key_a, key_b)
2006
return (key_a > key_b) - (key_a < key_b)
1886
2009
_cached_user_encoding = None
1889
def get_user_encoding(use_cache=True):
2012
def get_user_encoding():
1890
2013
"""Find out what the preferred user encoding is.
1892
2015
This is generally the encoding that is used for command line parameters
1893
2016
and file contents. This may be different from the terminal encoding
1894
2017
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)
1900
2019
:return: A string defining the preferred user encoding
1902
2021
global _cached_user_encoding
1903
if _cached_user_encoding is not None and use_cache:
2022
if _cached_user_encoding is not None:
1904
2023
return _cached_user_encoding
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'
2025
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2026
# Use the existing locale settings and call nl_langinfo directly
2027
# rather than going through getpreferredencoding. This avoids
2028
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2029
# possibility of the setlocale call throwing an error.
2030
user_encoding = locale.nl_langinfo(locale.CODESET)
2032
# GZ 2011-12-19: On windows could call GetACP directly instead.
2033
user_encoding = locale.getpreferredencoding(False)
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:'
2036
user_encoding = codecs.lookup(user_encoding).name
2038
if user_encoding not in ("", "cp0"):
2039
sys.stderr.write('brz: warning:'
1950
2040
' unknown encoding %s.'
1951
2041
' Continuing with ascii encoding.\n'
1952
2042
% user_encoding
1954
user_encoding = 'ascii'
1957
_cached_user_encoding = user_encoding
2044
user_encoding = 'ascii'
2046
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2047
if user_encoding == 'ascii':
2048
if sys.platform == 'darwin':
2049
# OSX is special-cased in Python to have a UTF-8 filesystem
2050
# encoding and previously had LANG set here if not present.
2051
user_encoding = 'utf-8'
2052
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2053
# for some other posix platforms as well.
2055
_cached_user_encoding = user_encoding
1959
2056
return user_encoding
2059
def get_diff_header_encoding():
2060
return get_terminal_encoding()
1962
2063
def get_host_name():
1963
2064
"""Return the current unicode host name.
1979
2079
# data at once.
1980
2080
MAX_SOCKET_CHUNK = 64 * 1024
2082
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2083
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2084
_eno = getattr(errno, _eno, None)
2085
if _eno is not None:
2086
_end_of_stream_errors.append(_eno)
1982
2090
def read_bytes_from_socket(sock, report_activity=None,
1983
max_read_size=MAX_SOCKET_CHUNK):
2091
max_read_size=MAX_SOCKET_CHUNK):
1984
2092
"""Read up to max_read_size of bytes from sock and notify of progress.
1986
2094
Translates "Connection reset by peer" into file-like EOF (return an
1987
2095
empty string rather than raise an error), and repeats the recv if
1988
2096
interrupted by a signal.
1992
bytes = sock.recv(max_read_size)
1993
except socket.error, e:
2100
data = sock.recv(max_read_size)
2101
except socket.error as e:
1994
2102
eno = e.args[0]
1995
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2103
if eno in _end_of_stream_errors:
1996
2104
# The connection was closed by the other side. Callers expect
1997
2105
# an empty string to signal end-of-stream.
1999
2107
elif eno == errno.EINTR:
2000
2108
# Retry the interrupted recv.
2004
2112
if report_activity is not None:
2005
report_activity(len(bytes), 'read')
2113
report_activity(len(data), 'read')
2009
2117
def recv_all(socket, count):
2017
2125
This isn't optimized and is intended mostly for use in testing.
2020
2128
while len(b) < count:
2021
2129
new = read_bytes_from_socket(socket, None, count - len(b))
2028
2136
def send_all(sock, bytes, report_activity=None):
2029
2137
"""Send all bytes on a socket.
2031
2139
Breaks large blocks in smaller chunks to avoid buffering limitations on
2032
2140
some platforms, and catches EINTR which may be thrown if the send is
2033
2141
interrupted by a signal.
2035
2143
This is preferred to socket.sendall(), because it avoids portability bugs
2036
2144
and provides activity reporting.
2038
2146
:param report_activity: Call this as bytes are read, see
2039
2147
Transport._report_activity
2042
2150
byte_count = len(bytes)
2151
view = memoryview(bytes)
2043
2152
while sent_total < byte_count:
2045
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2046
except socket.error, e:
2154
sent = sock.send(view[sent_total:sent_total + MAX_SOCKET_CHUNK])
2155
except (socket.error, IOError) as e:
2156
if e.args[0] in _end_of_stream_errors:
2157
raise errors.ConnectionReset(
2158
"Error trying to write to socket", e)
2047
2159
if e.args[0] != errno.EINTR:
2163
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2050
2165
sent_total += sent
2051
report_activity(sent, 'write')
2166
if report_activity is not None:
2167
report_activity(sent, 'write')
2170
def connect_socket(address):
2171
# Slight variation of the socket.create_connection() function (provided by
2172
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2173
# provide it for previous python versions. Also, we don't use the timeout
2174
# parameter (provided by the python implementation) so we don't implement
2176
err = socket.error('getaddrinfo returns an empty list')
2177
host, port = address
2178
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2179
af, socktype, proto, canonname, sa = res
2182
sock = socket.socket(af, socktype, proto)
2186
except socket.error as e:
2188
# 'err' is now the most recent error
2189
if sock is not None:
2054
2194
def dereference_path(path):
2082
2222
If and when pkg_resources becomes a standard library, this routine
2083
2223
can delegate to it.
2085
# Check package name is within bzrlib
2086
if package == "bzrlib":
2225
# Check package name is within breezy
2226
if package == "breezy":
2087
2227
resource_relpath = resource_name
2088
elif package.startswith("bzrlib."):
2089
package = package[len("bzrlib."):].replace('.', os.sep)
2228
elif package.startswith("breezy."):
2229
package = package[len("breezy."):].replace('.', os.sep)
2090
2230
resource_relpath = pathjoin(package, resource_name)
2092
raise errors.BzrError('resource package %s not in bzrlib' % package)
2232
raise errors.BzrError('resource package %s not in breezy' % package)
2094
2234
# Map the resource to a file and read its contents
2095
base = dirname(bzrlib.__file__)
2235
base = dirname(breezy.__file__)
2096
2236
if getattr(sys, 'frozen', None): # bzr.exe
2097
2237
base = abspath(pathjoin(base, '..', '..'))
2098
filename = pathjoin(base, resource_relpath)
2099
return open(filename, 'rU').read()
2238
with open(pathjoin(base, resource_relpath), "rt") as f:
2102
2242
def file_kind_from_stat_mode_thunk(mode):
2103
2243
global file_kind_from_stat_mode
2104
2244
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2106
from bzrlib._readdir_pyx import UTF8DirReader
2246
from ._readdir_pyx import UTF8DirReader
2107
2247
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2108
except ImportError, e:
2109
2249
# This is one time where we won't warn that an extension failed to
2110
2250
# load. The extension is never available on Windows anyway.
2111
from bzrlib._readdir_py import (
2251
from ._readdir_py import (
2112
2252
_kind_from_mode as file_kind_from_stat_mode
2114
2254
return file_kind_from_stat_mode(mode)
2115
2257
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2118
def file_kind(f, _lstat=os.lstat):
2260
def file_stat(f, _lstat=os.lstat):
2120
return file_kind_from_stat_mode(_lstat(f).st_mode)
2264
except OSError as e:
2122
2265
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2123
2266
raise errors.NoSuchFile(f)
2270
def file_kind(f, _lstat=os.lstat):
2271
stat_value = file_stat(f, _lstat)
2272
return file_kind_from_stat_mode(stat_value.st_mode)
2127
2275
def until_no_eintr(f, *a, **kw):
2128
2276
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2130
2278
WARNING: you must be certain that it is safe to retry the call repeatedly
2131
2279
if EINTR does occur. This is typically only true for low-level operations
2132
2280
like os.read. If in any doubt, don't use this.
2134
2282
Keep in mind that this is not a complete solution to EINTR. There is
2135
2283
probably code in the Python standard library and other dependencies that
2136
2284
may encounter EINTR if a signal arrives (and there is signal handler for
2137
that signal). So this function can reduce the impact for IO that bzrlib
2285
that signal). So this function can reduce the impact for IO that breezy
2138
2286
directly controls, but it is not a complete solution.
2140
2288
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2143
2291
return f(*a, **kw)
2144
except (IOError, OSError), e:
2292
except (IOError, OSError) as e:
2145
2293
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))
2173
2298
if sys.platform == "win32":
2176
2301
return msvcrt.getch()
2181
2306
fd = sys.stdin.fileno()
2182
2307
settings = termios.tcgetattr(fd)
2187
2312
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2191
if sys.platform == 'linux2':
2315
if sys.platform.startswith('linux'):
2192
2316
def _local_concurrency():
2194
prefix = 'processor'
2195
for line in file('/proc/cpuinfo', 'rb'):
2196
if line.startswith(prefix):
2197
concurrency = int(line[line.find(':')+1:]) + 1
2318
return os.sysconf('SC_NPROCESSORS_ONLN')
2319
except (ValueError, OSError, AttributeError):
2199
2321
elif sys.platform == 'darwin':
2200
2322
def _local_concurrency():
2201
2323
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2202
2324
stdout=subprocess.PIPE).communicate()[0]
2203
elif sys.platform[0:7] == 'freebsd':
2325
elif "bsd" in sys.platform:
2204
2326
def _local_concurrency():
2205
2327
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2206
2328
stdout=subprocess.PIPE).communicate()[0]
2207
2329
elif sys.platform == 'sunos5':
2208
2330
def _local_concurrency():
2209
return subprocess.Popen(['psrinfo', '-p',],
2331
return subprocess.Popen(['psrinfo', '-p', ],
2210
2332
stdout=subprocess.PIPE).communicate()[0]
2211
2333
elif sys.platform == "win32":
2212
2334
def _local_concurrency():
2302
2431
return os.fdopen(os.open(filename, flags), mode, bufsize)
2304
2433
open_file = open
2436
def available_backup_name(base, exists):
2437
"""Find a non-existing backup file name.
2439
This will *not* create anything, this only return a 'free' entry. This
2440
should be used for checking names in a directory below a locked
2441
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2442
Leap) and generally discouraged.
2444
:param base: The base name.
2446
:param exists: A callable returning True if the path parameter exists.
2449
name = "%s.~%d~" % (base, counter)
2452
name = "%s.~%d~" % (base, counter)
2456
def set_fd_cloexec(fd):
2457
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2458
support for this is not available.
2462
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2463
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2464
except (ImportError, AttributeError):
2465
# Either the fcntl module or specific constants are not present
2469
def find_executable_on_path(name):
2470
"""Finds an executable on the PATH.
2472
On Windows, this will try to append each extension in the PATHEXT
2473
environment variable to the name, if it cannot be found with the name
2476
:param name: The base name of the executable.
2477
:return: The path to the executable found or None.
2479
if sys.platform == 'win32':
2480
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2481
exts = [ext.lower() for ext in exts]
2482
base, ext = os.path.splitext(name)
2484
if ext.lower() not in exts:
2490
path = os.environ.get('PATH')
2491
if path is not None:
2492
path = path.split(os.pathsep)
2495
f = os.path.join(d, name) + ext
2496
if os.access(f, os.X_OK):
2498
if sys.platform == 'win32':
2499
app_path = win32utils.get_app_path(name)
2500
if app_path != name:
2505
def _posix_is_local_pid_dead(pid):
2506
"""True if pid doesn't correspond to live process on this machine"""
2508
# Special meaning of unix kill: just check if it's there.
2510
except OSError as e:
2511
if e.errno == errno.ESRCH:
2512
# On this machine, and really not found: as sure as we can be
2515
elif e.errno == errno.EPERM:
2516
# exists, though not ours
2519
trace.mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2520
# Don't really know.
2523
# Exists and our process: not dead.
2527
if sys.platform == "win32":
2528
is_local_pid_dead = win32utils.is_local_pid_dead
2530
is_local_pid_dead = _posix_is_local_pid_dead
2532
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2533
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2534
if getattr(errno, name, None) is not None]
2537
def fdatasync(fileno):
2538
"""Flush file contents to disk if possible.
2540
:param fileno: Integer OS file handle.
2541
:raises TransportNotPossible: If flushing to disk is not possible.
2543
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2547
except IOError as e:
2548
# See bug #1075108, on some platforms fdatasync exists, but can
2549
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2550
# and reduce the chance of corruption-on-powerloss situations. It
2551
# is not a mandatory call, so it is ok to suppress failures.
2552
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2553
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2557
def ensure_empty_directory_exists(path, exception_class):
2558
"""Make sure a local directory exists and is empty.
2560
If it does not exist, it is created. If it exists and is not empty, an
2561
instance of exception_class is raised.
2565
except OSError as e:
2566
if e.errno != errno.EEXIST:
2568
if os.listdir(path) != []:
2569
raise exception_class(path)
2572
def read_mtab(path):
2573
"""Read an fstab-style file and extract mountpoint+filesystem information.
2575
:param path: Path to read from
2576
:yield: Tuples with mountpoints (as bytestrings) and filesystem names
2578
with open(path, 'rb') as f:
2580
if line.startswith(b'#'):
2585
yield cols[1], cols[2].decode('ascii', 'replace')
2588
MTAB_PATH = '/etc/mtab'
2590
class FilesystemFinder(object):
2591
"""Find the filesystem for a particular path."""
2593
def __init__(self, mountpoints):
2596
self._mountpoints = sorted(mountpoints, key=key, reverse=True)
2600
"""Create a FilesystemFinder from an mtab-style file.
2602
Note that this will silenty ignore mtab if it doesn't exist or can not
2605
# TODO(jelmer): Use inotify to be notified when /etc/mtab changes and
2606
# we need to re-read it.
2608
return cls(read_mtab(MTAB_PATH))
2609
except EnvironmentError as e:
2610
trace.mutter('Unable to read mtab: %s', e)
2613
def find(self, path):
2614
"""Find the filesystem used by a particular path.
2616
:param path: Path to find (bytestring or text type)
2617
:return: Filesystem name (as text type) or None, if the filesystem is
2620
for mountpoint, filesystem in self._mountpoints:
2621
if is_inside(mountpoint, path):
2626
_FILESYSTEM_FINDER = None
2629
def get_fs_type(path):
2630
"""Return the filesystem type for the partition a path is in.
2632
:param path: Path to search filesystem type for
2633
:return: A FS type, as string. E.g. "ext2"
2635
global _FILESYSTEM_FINDER
2636
if _FILESYSTEM_FINDER is None:
2637
_FILESYSTEM_FINDER = FilesystemFinder.from_mtab()
2639
if not isinstance(path, bytes):
2640
path = path.encode(_fs_enc)
2642
return _FILESYSTEM_FINDER.find(path)
2645
perf_counter = time.perf_counter
added
removed
breezy/osutils.py
