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
25
from bzrlib.lazy_import import lazy_import
27
from .lazy_import import lazy_import
26
28
lazy_import(globals(), """
27
29
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
43
from tempfile import (
44
from tempfile import mkdtemp
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
73
# Cross platform wall-clock time functionality with decent resolution.
96
90
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
93
class UnsupportedTimezoneFormat(errors.BzrError):
95
_fmt = ('Unsupported timezone format "%(timezone)s", '
96
'options are "utc", "original", "local".')
98
def __init__(self, timezone):
99
self.timezone = timezone
99
102
def get_unicode_argv():
101
106
user_encoding = get_user_encoding()
102
107
return [a.decode(user_encoding) for a in sys.argv[1:]]
103
108
except UnicodeDecodeError:
104
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
109
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
110
"application locale.").format(a, user_encoding))
108
113
def make_readonly(filename):
109
114
"""Make a filename read-only."""
110
115
mod = os.lstat(filename).st_mode
111
116
if not stat.S_ISLNK(mod):
113
os.chmod(filename, mod)
118
chmod_if_possible(filename, mod)
116
121
def make_writable(filename):
117
122
mod = os.lstat(filename).st_mode
118
123
if not stat.S_ISLNK(mod):
120
os.chmod(filename, mod)
125
chmod_if_possible(filename, mod)
128
def chmod_if_possible(filename, mode):
129
# Set file mode if that can be safely done.
130
# Sometimes even on unix the filesystem won't allow it - see
131
# https://bugs.launchpad.net/bzr/+bug/606537
133
# It is probably faster to just do the chmod, rather than
134
# doing a stat, and then trying to compare
135
os.chmod(filename, mode)
136
except (IOError, OSError) as e:
137
# Permission/access denied seems to commonly happen on smbfs; there's
138
# probably no point warning about it.
139
# <https://bugs.launchpad.net/bzr/+bug/606537>
140
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
141
trace.mutter("ignore error on chmod of %r: %r" % (
123
147
def minimum_path_selection(paths):
232
260
file_existed = False
234
262
rename_func(new, tmp_name)
235
except (errors.NoSuchFile,), e:
263
except (errors.NoSuchFile,) as e:
238
266
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
239
267
# function raises an IOError with errno is None when a rename fails.
240
268
# This then gets caught here.
241
269
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
271
except Exception as e:
244
272
if (getattr(e, 'errno', None) is None
245
273
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
248
276
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)
264
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
280
# This may throw an exception, in which case success will
282
rename_func(old, new)
284
except (IOError, OSError) as e:
285
# source and target may be aliases of each other (e.g. on a
286
# case-insensitive filesystem), so we may have accidentally renamed
287
# source by when we tried to rename target
288
if (file_existed and e.errno in (None, errno.ENOENT)
289
and old.lower() == new.lower()):
290
# source and target are the same file on a case-insensitive
291
# filesystem, so we don't generate an exception
270
297
# If the file used to exist, rename it back into place
273
300
unlink_func(tmp_name)
275
302
rename_func(tmp_name, new)
276
if failure_exc is not None:
277
raise failure_exc[0], failure_exc[1], failure_exc[2]
280
305
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
281
306
# choke on a Unicode string containing a relative path if
282
307
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
284
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
285
309
def _posix_abspath(path):
286
310
# jam 20060426 rather than encoding to fsencoding
287
311
# copy posixpath.abspath, but use os.getcwdu instead
288
312
if not posixpath.isabs(path):
289
313
path = posixpath.join(getcwd(), path)
290
return posixpath.normpath(path)
314
return _posix_normpath(path)
293
317
def _posix_realpath(path):
294
318
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
321
def _posix_normpath(path):
322
path = posixpath.normpath(path)
323
# Bug 861008: posixpath.normpath() returns a path normalized according to
324
# the POSIX standard, which stipulates (for compatibility reasons) that two
325
# leading slashes must not be simplified to one, and only if there are 3 or
326
# more should they be simplified as one. So we treat the leading 2 slashes
327
# as a special case here by simply removing the first slash, as we consider
328
# that breaking POSIX compatibility for this obscure feature is acceptable.
329
# This is not a paranoid precaution, as we notably get paths like this when
330
# the repo is hosted at the root of the filesystem, i.e. in "/".
331
if path.startswith('//'):
336
def _posix_path_from_environ(key):
337
"""Get unicode path from `key` in environment or None if not present
339
Note that posix systems use arbitrary byte strings for filesystem objects,
340
so a path that raises BadFilenameEncoding here may still be accessible.
342
val = os.environ.get(key, None)
343
if PY3 or val is None:
346
return val.decode(_fs_enc)
347
except UnicodeDecodeError:
348
# GZ 2011-12-12:Ideally want to include `key` in the exception message
349
raise errors.BadFilenameEncoding(val, _fs_enc)
352
def _posix_get_home_dir():
353
"""Get the home directory of the current user as a unicode path"""
354
path = posixpath.expanduser("~")
356
return path.decode(_fs_enc)
357
except AttributeError:
359
except UnicodeDecodeError:
360
raise errors.BadFilenameEncoding(path, _fs_enc)
363
def _posix_getuser_unicode():
364
"""Get username from environment or password database as unicode"""
365
name = getpass.getuser()
368
user_encoding = get_user_encoding()
370
return name.decode(user_encoding)
371
except UnicodeDecodeError:
372
raise errors.BzrError("Encoding of username %r is unsupported by %s "
373
"application locale." % (name, user_encoding))
297
376
def _win32_fixdrive(path):
298
377
"""Force drive letters to be consistent.
303
382
running python.exe under cmd.exe return capital C:\\
304
383
running win32 python inside a cygwin shell returns lowercase c:\\
306
drive, path = _nt_splitdrive(path)
385
drive, path = ntpath.splitdrive(path)
307
386
return drive.upper() + path
310
389
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('\\', '/'))
390
# Real ntpath.abspath doesn't have a problem with a unicode cwd
391
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
342
394
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('\\', '/'))
395
# Real ntpath.realpath doesn't have a problem with a unicode cwd
396
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
347
399
def _win32_pathjoin(*args):
348
return _nt_join(*args).replace('\\', '/')
400
return ntpath.join(*args).replace('\\', '/')
351
403
def _win32_normpath(path):
352
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
404
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
355
407
def _win32_getcwd():
356
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
408
return _win32_fixdrive(_getcwd().replace('\\', '/'))
359
411
def _win32_mkdtemp(*args, **kwargs):
360
412
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
415
def _win32_rename(old, new):
372
416
"""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
433
def _mac_getcwd():
400
return unicodedata.normalize('NFC', os.getcwdu())
434
return unicodedata.normalize('NFC', _getcwd())
437
def _rename_wrap_exception(rename_func):
438
"""Adds extra information to any exceptions that come from rename().
440
The exception has an updated message and 'old_filename' and 'new_filename'
444
def _rename_wrapper(old, new):
446
rename_func(old, new)
448
detailed_error = OSError(e.errno, e.strerror +
449
" [occurred when renaming '%s' to '%s']" %
451
detailed_error.old_filename = old
452
detailed_error.new_filename = new
455
return _rename_wrapper
458
if sys.version_info > (3,):
464
# Default rename wraps os.rename()
465
rename = _rename_wrap_exception(os.rename)
403
467
# Default is to just use the python builtins, but these can be rebound on
404
468
# particular platforms.
405
469
abspath = _posix_abspath
406
470
realpath = _posix_realpath
407
471
pathjoin = os.path.join
408
normpath = os.path.normpath
409
rename = _wrapped_rename # overridden below on win32
472
normpath = _posix_normpath
473
path_from_environ = _posix_path_from_environ
474
_get_home_dir = _posix_get_home_dir
475
getuser_unicode = _posix_getuser_unicode
411
477
dirname = os.path.dirname
412
478
basename = os.path.basename
413
479
split = os.path.split
414
480
splitext = os.path.splitext
415
# These were already imported into local scope
481
# These were already lazily imported into local scope
416
482
# mkdtemp = tempfile.mkdtemp
417
483
# rmtree = shutil.rmtree
419
492
MIN_ABS_PATHLENGTH = 1
422
495
if sys.platform == 'win32':
423
if win32utils.winver == 'Windows 98':
424
abspath = _win98_abspath
426
abspath = _win32_abspath
496
abspath = _win32_abspath
427
497
realpath = _win32_realpath
428
498
pathjoin = _win32_pathjoin
429
499
normpath = _win32_normpath
430
500
getcwd = _win32_getcwd
431
501
mkdtemp = _win32_mkdtemp
432
rename = _win32_rename
502
rename = _rename_wrap_exception(_win32_rename)
504
from . import _walkdirs_win32
508
lstat = _walkdirs_win32.lstat
509
fstat = _walkdirs_win32.fstat
510
wrap_stat = _walkdirs_win32.wrap_stat
434
512
MIN_ABS_PATHLENGTH = 3
471
552
On my standard US Windows XP, the preferred encoding is
472
553
cp1252, but the console is cp437
555
:param trace: If True trace the selected encoding via mutter().
474
from bzrlib.trace import mutter
557
from .trace import mutter
475
558
output_encoding = getattr(sys.stdout, 'encoding', None)
476
559
if not output_encoding:
477
560
input_encoding = getattr(sys.stdin, 'encoding', None)
478
561
if not input_encoding:
479
562
output_encoding = get_user_encoding()
480
mutter('encoding stdout as osutils.get_user_encoding() %r',
564
mutter('encoding stdout as osutils.get_user_encoding() %r',
483
567
output_encoding = input_encoding
484
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
569
mutter('encoding stdout as sys.stdin encoding %r',
486
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
573
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
487
574
if output_encoding == 'cp0':
488
575
# invalid encoding (cp0 means 'no codepage' on Windows)
489
576
output_encoding = get_user_encoding()
490
mutter('cp0 is invalid encoding.'
578
mutter('cp0 is invalid encoding.'
491
579
' encoding stdout as osutils.get_user_encoding() %r',
495
583
codecs.lookup(output_encoding)
496
584
except LookupError:
497
sys.stderr.write('bzr: warning:'
585
sys.stderr.write('brz: warning:'
498
586
' unknown terminal encoding %s.\n'
499
587
' Using encoding %s instead.\n'
500
588
% (output_encoding, get_user_encoding())
882
988
"""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
989
return os.fstat(f.fileno())[stat.ST_SIZE]
992
# Alias os.urandom to support platforms (which?) without /dev/urandom and
993
# override if it doesn't work. Avoid checking on windows where there is
994
# significant initialisation cost that can be avoided for some bzr calls.
996
rand_bytes = os.urandom
998
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):
1001
except NotImplementedError:
899
1002
# not well seeded, but better than nothing
900
1003
def rand_bytes(n):
970
1079
implementation should be loaded instead::
973
>>> import bzrlib._fictional_extension_pyx
1082
>>> import breezy._fictional_extension_pyx
974
1083
>>> except ImportError, e:
975
>>> bzrlib.osutils.failed_to_load_extension(e)
976
>>> import bzrlib._fictional_extension_py
1084
>>> breezy.osutils.failed_to_load_extension(e)
1085
>>> import breezy._fictional_extension_py
978
1087
# NB: This docstring is just an example, not a doctest, because doctest
979
1088
# currently can't cope with the use of lazy imports in this namespace --
982
1091
# This currently doesn't report the failure at the time it occurs, because
983
1092
# they tend to happen very early in startup when we can't check config
984
1093
# files etc, and also we want to report all failures but not spam the user
985
1094
# with 10 warnings.
986
from bzrlib import trace
987
1095
exception_str = str(exception)
988
1096
if exception_str not in _extension_load_failures:
989
1097
trace.mutter("failed to load compiled extension: %s" % exception_str)
993
1101
def report_extension_load_failures():
994
1102
if not _extension_load_failures:
996
from bzrlib.config import GlobalConfig
997
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1104
if config.GlobalConfig().suppress_warning('missing_extensions'):
999
1106
# the warnings framework should by default show this only once
1000
from bzrlib.trace import warning
1107
from .trace import warning
1002
"bzr: warning: some compiled extensions could not be loaded; "
1003
"see <https://answers.launchpad.net/bzr/+faq/703>")
1109
"brz: warning: some compiled extensions could not be loaded; "
1110
"see ``brz help missing-extensions``")
1004
1111
# we no longer show the specific missing extensions here, because it makes
1005
1112
# the message too long and scary - see
1006
1113
# https://bugs.launchpad.net/bzr/+bug/430529
1010
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1011
except ImportError, e:
1117
from ._chunks_to_lines_pyx import chunks_to_lines
1118
except ImportError as e:
1012
1119
failed_to_load_extension(e)
1013
from bzrlib._chunks_to_lines_py import chunks_to_lines
1120
from ._chunks_to_lines_py import chunks_to_lines
1016
1123
def split_lines(s):
1017
1124
"""Split s into lines, but without removing the newline characters."""
1018
1125
# Trivially convert a fulltext into a 'chunked' representation, and let
1019
1126
# chunks_to_lines do the heavy lifting.
1020
if isinstance(s, str):
1127
if isinstance(s, bytes):
1021
1128
# chunks_to_lines only supports 8-bit strings
1022
1129
return chunks_to_lines([s])
1049
1157
os.link(src, dest)
1050
except (OSError, IOError), e:
1158
except (OSError, IOError) as e:
1051
1159
if e.errno != errno.EXDEV:
1053
1161
shutil.copyfile(src, dest)
1056
1164
def delete_any(path):
1057
"""Delete a file, symlink or directory.
1165
"""Delete a file, symlink or directory.
1059
1167
Will delete even if readonly.
1062
1170
_delete_file_or_dir(path)
1063
except (OSError, IOError), e:
1171
except (OSError, IOError) as e:
1064
1172
if e.errno in (errno.EPERM, errno.EACCES):
1065
1173
# make writable and try again
1250
1358
# but for now, we haven't optimized...
1251
1359
return [canonical_relpath(base, p) for p in paths]
1362
def decode_filename(filename):
1363
"""Decode the filename using the filesystem encoding
1365
If it is unicode, it is returned.
1366
Otherwise it is decoded from the the filesystem's encoding. If decoding
1367
fails, a errors.BadFilenameEncoding exception is raised.
1369
if isinstance(filename, text_type):
1372
return filename.decode(_fs_enc)
1373
except UnicodeDecodeError:
1374
raise errors.BadFilenameEncoding(filename, _fs_enc)
1253
1377
def safe_unicode(unicode_or_utf8_string):
1254
1378
"""Coerce unicode_or_utf8_string into unicode.
1284
1408
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):
1411
def safe_revision_id(unicode_or_utf8_string):
1293
1412
"""Revision ids should now be utf8, but at one point they were unicode.
1295
1414
: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
1416
:return: None or a utf8 revision id.
1300
1418
if (unicode_or_utf8_string is None
1301
or unicode_or_utf8_string.__class__ == str):
1419
or unicode_or_utf8_string.__class__ == bytes):
1302
1420
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):
1421
raise TypeError('Unicode revision ids are no longer supported. '
1422
'Revision id generators should be creating utf8 revision '
1426
def safe_file_id(unicode_or_utf8_string):
1314
1427
"""File ids should now be utf8, but at one point they were unicode.
1316
1429
This is the same as safe_utf8, except it uses the cached encode functions
1358
1468
can be accessed by that path.
1361
return unicodedata.normalize('NFC', unicode(path)), True
1471
if isinstance(path, bytes):
1472
path = path.decode(sys.getfilesystemencoding())
1473
return unicodedata.normalize('NFC', path), True
1364
1476
def _inaccessible_normalized_filename(path):
1365
1477
__doc__ = _accessible_normalized_filename.__doc__
1367
normalized = unicodedata.normalize('NFC', unicode(path))
1479
if isinstance(path, bytes):
1480
path = path.decode(sys.getfilesystemencoding())
1481
normalized = unicodedata.normalize('NFC', path)
1368
1482
return normalized, normalized == path
1424
1544
None is returned if the width can't established precisely.
1427
- if BZR_COLUMNS is set, returns its value
1547
- if BRZ_COLUMNS is set, returns its value
1428
1548
- if there is no controlling terminal, returns None
1549
- query the OS, if the queried size has changed since the last query,
1429
1551
- if COLUMNS is set, returns its value,
1552
- if the OS has a value (even though it's never changed), return its value.
1431
1554
From there, we need to query the OS to get the size of the controlling
1557
On Unices we query the OS by:
1435
1558
- get termios.TIOCGWINSZ
1436
1559
- if an error occurs or a negative value is obtained, returns None
1561
On Windows we query the OS by:
1440
1562
- win32utils.get_console_size() decides,
1441
1563
- returns None on error (provided default value)
1565
# Note to implementors: if changing the rules for determining the width,
1566
# make sure you've considered the behaviour in these cases:
1567
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1568
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1570
# - (add more interesting cases here, if you find any)
1571
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1572
# but we don't want to register a signal handler because it is impossible
1573
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1574
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1575
# 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
1578
# If BRZ_COLUMNS is set, take it, user is always right
1579
# Except if they specified 0 in which case, impose no limit here
1446
return int(os.environ['BZR_COLUMNS'])
1581
width = int(os.environ['BRZ_COLUMNS'])
1447
1582
except (KeyError, ValueError):
1584
if width is not None:
1450
1590
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.
1591
if isatty is None or not isatty():
1592
# 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))
1596
width, height = os_size = _terminal_size(None, None)
1597
global _first_terminal_size, _terminal_size_state
1598
if _terminal_size_state == 'no_data':
1599
_first_terminal_size = os_size
1600
_terminal_size_state = 'unchanged'
1601
elif (_terminal_size_state == 'unchanged' and
1602
_first_terminal_size != os_size):
1603
_terminal_size_state = 'changed'
1605
# If the OS claims to know how wide the terminal is, and this value has
1606
# ever changed, use that.
1607
if _terminal_size_state == 'changed':
1608
if width is not None and width > 0:
1611
# If COLUMNS is set, use it.
1459
1613
return int(os.environ['COLUMNS'])
1460
1614
except (KeyError, ValueError):
1463
width, height = _terminal_size(None, None)
1465
# Consider invalid values as meaning no width
1617
# Finally, use an unchanged size from the OS, if we have one.
1618
if _terminal_size_state == 'unchanged':
1619
if width is not None and width > 0:
1622
# The width could not be determined.
1471
1626
def _win32_terminal_size(width, height):
1498
1653
_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
1656
def supports_executable():
1527
1657
return sys.platform != "win32"
1712
1842
global _selected_dir_reader
1713
1843
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
1844
if sys.platform == "win32":
1722
from bzrlib._walkdirs_win32 import Win32ReadDir
1846
from ._walkdirs_win32 import Win32ReadDir
1723
1847
_selected_dir_reader = Win32ReadDir()
1724
1848
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
1850
elif _fs_enc in ('utf-8', 'ascii'):
1729
from bzrlib._readdir_pyx import UTF8DirReader
1852
from ._readdir_pyx import UTF8DirReader
1730
1853
_selected_dir_reader = UTF8DirReader()
1731
except ImportError, e:
1854
except ImportError as e:
1732
1855
failed_to_load_extension(e)
1779
1902
See DirReader.read_dir for details.
1781
1904
_utf8_encode = self._utf8_encode
1905
_fs_decode = lambda s: s.decode(_fs_enc)
1906
_fs_encode = lambda s: s.encode(_fs_enc)
1782
1907
_lstat = os.lstat
1783
1908
_listdir = os.listdir
1784
1909
_kind_from_mode = file_kind_from_stat_mode
1787
relprefix = prefix + '/'
1912
relprefix = prefix + b'/'
1790
top_slash = top + u'/'
1915
top_slash = top + '/'
1793
1918
append = dirblock.append
1794
for name in sorted(_listdir(top)):
1919
for name_native in _listdir(top.encode('utf-8')):
1796
name_utf8 = _utf8_encode(name)[0]
1921
name = _fs_decode(name_native)
1797
1922
except UnicodeDecodeError:
1798
1923
raise errors.BadFilenameEncoding(
1799
_utf8_encode(relprefix)[0] + name, _fs_enc)
1924
relprefix + name_native, _fs_enc)
1925
name_utf8 = _utf8_encode(name)[0]
1800
1926
abspath = top_slash + name
1801
1927
statvalue = _lstat(abspath)
1802
1928
kind = _kind_from_mode(statvalue.st_mode)
1803
1929
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1930
return sorted(dirblock)
1807
1933
def copy_tree(from_path, to_path, handlers={}):
1874
2003
This can be used to sort paths in the same way that walkdirs does.
1876
return (dirname(path) , path)
2005
return (dirname(path), path)
1879
2008
def compare_paths_prefix_order(path_a, path_b):
1880
2009
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1881
2010
key_a = path_prefix_key(path_a)
1882
2011
key_b = path_prefix_key(path_b)
1883
return cmp(key_a, key_b)
2012
return (key_a > key_b) - (key_a < key_b)
1886
2015
_cached_user_encoding = None
1889
def get_user_encoding(use_cache=True):
2018
def get_user_encoding():
1890
2019
"""Find out what the preferred user encoding is.
1892
2021
This is generally the encoding that is used for command line parameters
1893
2022
and file contents. This may be different from the terminal encoding
1894
2023
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
2025
:return: A string defining the preferred user encoding
1902
2027
global _cached_user_encoding
1903
if _cached_user_encoding is not None and use_cache:
2028
if _cached_user_encoding is not None:
1904
2029
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'
2031
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2032
# Use the existing locale settings and call nl_langinfo directly
2033
# rather than going through getpreferredencoding. This avoids
2034
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2035
# possibility of the setlocale call throwing an error.
2036
user_encoding = locale.nl_langinfo(locale.CODESET)
2038
# GZ 2011-12-19: On windows could call GetACP directly instead.
2039
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:'
2042
user_encoding = codecs.lookup(user_encoding).name
2044
if user_encoding not in ("", "cp0"):
2045
sys.stderr.write('brz: warning:'
1950
2046
' unknown encoding %s.'
1951
2047
' Continuing with ascii encoding.\n'
1952
2048
% user_encoding
1954
user_encoding = 'ascii'
1957
_cached_user_encoding = user_encoding
2050
user_encoding = 'ascii'
2052
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2053
if user_encoding == 'ascii':
2054
if sys.platform == 'darwin':
2055
# OSX is special-cased in Python to have a UTF-8 filesystem
2056
# encoding and previously had LANG set here if not present.
2057
user_encoding = 'utf-8'
2058
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2059
# for some other posix platforms as well.
2061
_cached_user_encoding = user_encoding
1959
2062
return user_encoding
2065
def get_diff_header_encoding():
2066
return get_terminal_encoding()
1962
2069
def get_host_name():
1963
2070
"""Return the current unicode host name.
1987
2103
empty string rather than raise an error), and repeats the recv if
1988
2104
interrupted by a signal.
1992
bytes = sock.recv(max_read_size)
1993
except socket.error, e:
2108
data = sock.recv(max_read_size)
2109
except socket.error as e:
1994
2110
eno = e.args[0]
1995
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2111
if eno in _end_of_stream_errors:
1996
2112
# The connection was closed by the other side. Callers expect
1997
2113
# an empty string to signal end-of-stream.
1999
2115
elif eno == errno.EINTR:
2000
2116
# Retry the interrupted recv.
2004
2120
if report_activity is not None:
2005
report_activity(len(bytes), 'read')
2121
report_activity(len(data), 'read')
2009
2125
def recv_all(socket, count):
2028
2144
def send_all(sock, bytes, report_activity=None):
2029
2145
"""Send all bytes on a socket.
2031
2147
Breaks large blocks in smaller chunks to avoid buffering limitations on
2032
2148
some platforms, and catches EINTR which may be thrown if the send is
2033
2149
interrupted by a signal.
2035
2151
This is preferred to socket.sendall(), because it avoids portability bugs
2036
2152
and provides activity reporting.
2038
2154
:param report_activity: Call this as bytes are read, see
2039
2155
Transport._report_activity
2042
2158
byte_count = len(bytes)
2159
view = memoryview(bytes)
2043
2160
while sent_total < byte_count:
2045
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2046
except socket.error, e:
2162
sent = sock.send(view[sent_total:sent_total+MAX_SOCKET_CHUNK])
2163
except (socket.error, IOError) as e:
2164
if e.args[0] in _end_of_stream_errors:
2165
raise errors.ConnectionReset(
2166
"Error trying to write to socket", e)
2047
2167
if e.args[0] != errno.EINTR:
2171
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2050
2173
sent_total += sent
2051
report_activity(sent, 'write')
2174
if report_activity is not None:
2175
report_activity(sent, 'write')
2178
def connect_socket(address):
2179
# Slight variation of the socket.create_connection() function (provided by
2180
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2181
# provide it for previous python versions. Also, we don't use the timeout
2182
# parameter (provided by the python implementation) so we don't implement
2184
err = socket.error('getaddrinfo returns an empty list')
2185
host, port = address
2186
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2187
af, socktype, proto, canonname, sa = res
2190
sock = socket.socket(af, socktype, proto)
2194
except socket.error as e:
2196
# 'err' is now the most recent error
2197
if sock is not None:
2054
2202
def dereference_path(path):
2082
2230
If and when pkg_resources becomes a standard library, this routine
2083
2231
can delegate to it.
2085
# Check package name is within bzrlib
2086
if package == "bzrlib":
2233
# Check package name is within breezy
2234
if package == "breezy":
2087
2235
resource_relpath = resource_name
2088
elif package.startswith("bzrlib."):
2089
package = package[len("bzrlib."):].replace('.', os.sep)
2236
elif package.startswith("breezy."):
2237
package = package[len("breezy."):].replace('.', os.sep)
2090
2238
resource_relpath = pathjoin(package, resource_name)
2092
raise errors.BzrError('resource package %s not in bzrlib' % package)
2240
raise errors.BzrError('resource package %s not in breezy' % package)
2094
2242
# Map the resource to a file and read its contents
2095
base = dirname(bzrlib.__file__)
2243
base = dirname(breezy.__file__)
2096
2244
if getattr(sys, 'frozen', None): # bzr.exe
2097
2245
base = abspath(pathjoin(base, '..', '..'))
2098
filename = pathjoin(base, resource_relpath)
2099
return open(filename, 'rU').read()
2246
with open(pathjoin(base, resource_relpath), "rt") as f:
2102
2249
def file_kind_from_stat_mode_thunk(mode):
2103
2250
global file_kind_from_stat_mode
2104
2251
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2106
from bzrlib._readdir_pyx import UTF8DirReader
2253
from ._readdir_pyx import UTF8DirReader
2107
2254
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2108
except ImportError, e:
2255
except ImportError as e:
2109
2256
# This is one time where we won't warn that an extension failed to
2110
2257
# load. The extension is never available on Windows anyway.
2111
from bzrlib._readdir_py import (
2258
from ._readdir_py import (
2112
2259
_kind_from_mode as file_kind_from_stat_mode
2114
2261
return file_kind_from_stat_mode(mode)
2115
2262
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2118
def file_kind(f, _lstat=os.lstat):
2264
def file_stat(f, _lstat=os.lstat):
2120
return file_kind_from_stat_mode(_lstat(f).st_mode)
2268
except OSError as e:
2122
2269
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2123
2270
raise errors.NoSuchFile(f)
2273
def file_kind(f, _lstat=os.lstat):
2274
stat_value = file_stat(f, _lstat)
2275
return file_kind_from_stat_mode(stat_value.st_mode)
2127
2277
def until_no_eintr(f, *a, **kw):
2128
2278
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2130
2280
WARNING: you must be certain that it is safe to retry the call repeatedly
2131
2281
if EINTR does occur. This is typically only true for low-level operations
2132
2282
like os.read. If in any doubt, don't use this.
2134
2284
Keep in mind that this is not a complete solution to EINTR. There is
2135
2285
probably code in the Python standard library and other dependencies that
2136
2286
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
2287
that signal). So this function can reduce the impact for IO that breezy
2138
2288
directly controls, but it is not a complete solution.
2140
2290
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2143
2293
return f(*a, **kw)
2144
except (IOError, OSError), e:
2294
except (IOError, OSError) as e:
2145
2295
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
2300
if sys.platform == "win32":
2176
2303
return msvcrt.getch()
2181
2308
fd = sys.stdin.fileno()
2182
2309
settings = termios.tcgetattr(fd)
2187
2314
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2191
if sys.platform == 'linux2':
2317
if sys.platform.startswith('linux'):
2192
2318
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
2320
return os.sysconf('SC_NPROCESSORS_ONLN')
2321
except (ValueError, OSError, AttributeError):
2199
2323
elif sys.platform == 'darwin':
2200
2324
def _local_concurrency():
2201
2325
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2202
2326
stdout=subprocess.PIPE).communicate()[0]
2203
elif sys.platform[0:7] == 'freebsd':
2327
elif "bsd" in sys.platform:
2204
2328
def _local_concurrency():
2205
2329
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2206
2330
stdout=subprocess.PIPE).communicate()[0]
2302
2432
return os.fdopen(os.open(filename, flags), mode, bufsize)
2304
2434
open_file = open
2437
def available_backup_name(base, exists):
2438
"""Find a non-existing backup file name.
2440
This will *not* create anything, this only return a 'free' entry. This
2441
should be used for checking names in a directory below a locked
2442
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2443
Leap) and generally discouraged.
2445
:param base: The base name.
2447
:param exists: A callable returning True if the path parameter exists.
2450
name = "%s.~%d~" % (base, counter)
2453
name = "%s.~%d~" % (base, counter)
2457
def set_fd_cloexec(fd):
2458
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2459
support for this is not available.
2463
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2464
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2465
except (ImportError, AttributeError):
2466
# Either the fcntl module or specific constants are not present
2470
def find_executable_on_path(name):
2471
"""Finds an executable on the PATH.
2473
On Windows, this will try to append each extension in the PATHEXT
2474
environment variable to the name, if it cannot be found with the name
2477
:param name: The base name of the executable.
2478
:return: The path to the executable found or None.
2480
if sys.platform == 'win32':
2481
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2482
exts = [ext.lower() for ext in exts]
2483
base, ext = os.path.splitext(name)
2485
if ext.lower() not in exts:
2491
path = os.environ.get('PATH')
2492
if path is not None:
2493
path = path.split(os.pathsep)
2496
f = os.path.join(d, name) + ext
2497
if os.access(f, os.X_OK):
2499
if sys.platform == 'win32':
2500
app_path = win32utils.get_app_path(name)
2501
if app_path != name:
2506
def _posix_is_local_pid_dead(pid):
2507
"""True if pid doesn't correspond to live process on this machine"""
2509
# Special meaning of unix kill: just check if it's there.
2511
except OSError as e:
2512
if e.errno == errno.ESRCH:
2513
# On this machine, and really not found: as sure as we can be
2516
elif e.errno == errno.EPERM:
2517
# exists, though not ours
2520
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2521
# Don't really know.
2524
# 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 is_environment_error(evalue):
2573
"""True if exception instance is due to a process environment issue
2575
This includes OSError and IOError, but also other errors that come from
2576
the operating system or core libraries but are not subclasses of those.
2578
if isinstance(evalue, (EnvironmentError, select.error)):
2580
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
added
removed
breezy/osutils.py
