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
30
from ntpath import (abspath as _nt_abspath,
32
normpath as _nt_normpath,
33
realpath as _nt_realpath,
34
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 (
44
from tempfile import mkdtemp
52
from breezy.i18n import gettext
57
from bzrlib.symbol_versioning import (
62
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
64
if sys.version_info < (2, 5):
65
import md5 as _mod_md5
67
import sha as _mod_sha
77
from bzrlib import symbol_versioning
80
73
# Cross platform wall-clock time functionality with decent resolution.
97
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
100
102
def get_unicode_argv():
102
106
user_encoding = get_user_encoding()
103
107
return [a.decode(user_encoding) for a in sys.argv[1:]]
104
108
except UnicodeDecodeError:
105
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))
109
113
def make_readonly(filename):
110
114
"""Make a filename read-only."""
111
115
mod = os.lstat(filename).st_mode
112
116
if not stat.S_ISLNK(mod):
114
os.chmod(filename, mod)
118
chmod_if_possible(filename, mod)
117
121
def make_writable(filename):
118
122
mod = os.lstat(filename).st_mode
119
123
if not stat.S_ISLNK(mod):
121
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" % (
124
147
def minimum_path_selection(paths):
233
259
file_existed = False
235
261
rename_func(new, tmp_name)
236
except (errors.NoSuchFile,), e:
262
except (errors.NoSuchFile,) as e:
239
265
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
240
266
# function raises an IOError with errno is None when a rename fails.
241
267
# This then gets caught here.
242
268
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
270
except Exception as e:
245
271
if (getattr(e, 'errno', None) is None
246
272
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
249
275
file_existed = True
255
# This may throw an exception, in which case success will
257
rename_func(old, new)
259
except (IOError, OSError), e:
260
# source and target may be aliases of each other (e.g. on a
261
# case-insensitive filesystem), so we may have accidentally renamed
262
# source by when we tried to rename target
263
failure_exc = sys.exc_info()
264
if (file_existed and e.errno in (None, errno.ENOENT)
265
and old.lower() == new.lower()):
266
# source and target are the same file on a case-insensitive
267
# filesystem, so we don't generate an exception
279
# This may throw an exception, in which case success will
281
rename_func(old, new)
283
except (IOError, OSError) as e:
284
# source and target may be aliases of each other (e.g. on a
285
# case-insensitive filesystem), so we may have accidentally renamed
286
# source by when we tried to rename target
287
if (file_existed and e.errno in (None, errno.ENOENT)
288
and old.lower() == new.lower()):
289
# source and target are the same file on a case-insensitive
290
# filesystem, so we don't generate an exception
271
296
# If the file used to exist, rename it back into place
274
299
unlink_func(tmp_name)
276
301
rename_func(tmp_name, new)
277
if failure_exc is not None:
278
raise failure_exc[0], failure_exc[1], failure_exc[2]
281
304
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
282
305
# choke on a Unicode string containing a relative path if
283
306
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
285
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
286
308
def _posix_abspath(path):
287
309
# jam 20060426 rather than encoding to fsencoding
288
310
# copy posixpath.abspath, but use os.getcwdu instead
289
311
if not posixpath.isabs(path):
290
312
path = posixpath.join(getcwd(), path)
291
return posixpath.normpath(path)
313
return _posix_normpath(path)
294
316
def _posix_realpath(path):
295
317
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
320
def _posix_normpath(path):
321
path = posixpath.normpath(path)
322
# Bug 861008: posixpath.normpath() returns a path normalized according to
323
# the POSIX standard, which stipulates (for compatibility reasons) that two
324
# leading slashes must not be simplified to one, and only if there are 3 or
325
# more should they be simplified as one. So we treat the leading 2 slashes
326
# as a special case here by simply removing the first slash, as we consider
327
# that breaking POSIX compatibility for this obscure feature is acceptable.
328
# This is not a paranoid precaution, as we notably get paths like this when
329
# the repo is hosted at the root of the filesystem, i.e. in "/".
330
if path.startswith('//'):
335
def _posix_path_from_environ(key):
336
"""Get unicode path from `key` in environment or None if not present
338
Note that posix systems use arbitrary byte strings for filesystem objects,
339
so a path that raises BadFilenameEncoding here may still be accessible.
341
val = os.environ.get(key, None)
342
if PY3 or val is None:
345
return val.decode(_fs_enc)
346
except UnicodeDecodeError:
347
# GZ 2011-12-12:Ideally want to include `key` in the exception message
348
raise errors.BadFilenameEncoding(val, _fs_enc)
351
def _posix_get_home_dir():
352
"""Get the home directory of the current user as a unicode path"""
353
path = posixpath.expanduser("~")
355
return path.decode(_fs_enc)
356
except AttributeError:
358
except UnicodeDecodeError:
359
raise errors.BadFilenameEncoding(path, _fs_enc)
362
def _posix_getuser_unicode():
363
"""Get username from environment or password database as unicode"""
364
name = getpass.getuser()
367
user_encoding = get_user_encoding()
369
return name.decode(user_encoding)
370
except UnicodeDecodeError:
371
raise errors.BzrError("Encoding of username %r is unsupported by %s "
372
"application locale." % (name, user_encoding))
298
375
def _win32_fixdrive(path):
299
376
"""Force drive letters to be consistent.
304
381
running python.exe under cmd.exe return capital C:\\
305
382
running win32 python inside a cygwin shell returns lowercase c:\\
307
drive, path = _nt_splitdrive(path)
384
drive, path = ntpath.splitdrive(path)
308
385
return drive.upper() + path
311
388
def _win32_abspath(path):
312
# Real _nt_abspath doesn't have a problem with a unicode cwd
313
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
316
def _win98_abspath(path):
317
"""Return the absolute version of a path.
318
Windows 98 safe implementation (python reimplementation
319
of Win32 API function GetFullPathNameW)
324
# \\HOST\path => //HOST/path
325
# //HOST/path => //HOST/path
326
# path => C:/cwd/path
329
# check for absolute path
330
drive = _nt_splitdrive(path)[0]
331
if drive == '' and path[:2] not in('//','\\\\'):
333
# we cannot simply os.path.join cwd and path
334
# because os.path.join('C:','/path') produce '/path'
335
# and this is incorrect
336
if path[:1] in ('/','\\'):
337
cwd = _nt_splitdrive(cwd)[0]
339
path = cwd + '\\' + path
340
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
389
# Real ntpath.abspath doesn't have a problem with a unicode cwd
390
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
343
393
def _win32_realpath(path):
344
# Real _nt_realpath doesn't have a problem with a unicode cwd
345
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
394
# Real ntpath.realpath doesn't have a problem with a unicode cwd
395
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
348
398
def _win32_pathjoin(*args):
349
return _nt_join(*args).replace('\\', '/')
399
return ntpath.join(*args).replace('\\', '/')
352
402
def _win32_normpath(path):
353
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
403
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
356
406
def _win32_getcwd():
357
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
407
return _win32_fixdrive(_getcwd().replace('\\', '/'))
360
410
def _win32_mkdtemp(*args, **kwargs):
361
411
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
364
def _add_rename_error_details(e, old, new):
365
new_e = OSError(e.errno, "failed to rename %s to %s: %s"
366
% (old, new, e.strerror))
368
new_e.to_filename = new
372
414
def _win32_rename(old, new):
373
415
"""We expect to be able to atomically replace 'new' with old.
390
def _wrapped_rename(old, new):
391
"""Rename a file or directory"""
394
except (IOError, OSError), e:
395
# this is eventually called by all rename-like functions, so should
397
raise _add_rename_error_details(e, old, new)
400
432
def _mac_getcwd():
401
return unicodedata.normalize('NFC', os.getcwdu())
433
return unicodedata.normalize('NFC', _getcwd())
436
def _rename_wrap_exception(rename_func):
437
"""Adds extra information to any exceptions that come from rename().
439
The exception has an updated message and 'old_filename' and 'new_filename'
443
def _rename_wrapper(old, new):
445
rename_func(old, new)
447
detailed_error = OSError(e.errno, e.strerror +
448
" [occurred when renaming '%s' to '%s']" %
450
detailed_error.old_filename = old
451
detailed_error.new_filename = new
454
return _rename_wrapper
457
if sys.version_info > (3,):
463
# Default rename wraps os.rename()
464
rename = _rename_wrap_exception(os.rename)
404
466
# Default is to just use the python builtins, but these can be rebound on
405
467
# particular platforms.
406
468
abspath = _posix_abspath
407
469
realpath = _posix_realpath
408
470
pathjoin = os.path.join
409
normpath = os.path.normpath
410
rename = _wrapped_rename # overridden below on win32
471
normpath = _posix_normpath
472
path_from_environ = _posix_path_from_environ
473
_get_home_dir = _posix_get_home_dir
474
getuser_unicode = _posix_getuser_unicode
412
476
dirname = os.path.dirname
413
477
basename = os.path.basename
414
478
split = os.path.split
415
479
splitext = os.path.splitext
416
# These were already imported into local scope
480
# These were already lazily imported into local scope
417
481
# mkdtemp = tempfile.mkdtemp
418
482
# rmtree = shutil.rmtree
420
490
MIN_ABS_PATHLENGTH = 1
423
493
if sys.platform == 'win32':
424
if win32utils.winver == 'Windows 98':
425
abspath = _win98_abspath
427
abspath = _win32_abspath
494
abspath = _win32_abspath
428
495
realpath = _win32_realpath
429
496
pathjoin = _win32_pathjoin
430
497
normpath = _win32_normpath
431
498
getcwd = _win32_getcwd
432
499
mkdtemp = _win32_mkdtemp
433
rename = _win32_rename
500
rename = _rename_wrap_exception(_win32_rename)
502
from . import _walkdirs_win32
506
lstat = _walkdirs_win32.lstat
507
fstat = _walkdirs_win32.fstat
508
wrap_stat = _walkdirs_win32.wrap_stat
435
510
MIN_ABS_PATHLENGTH = 3
472
550
On my standard US Windows XP, the preferred encoding is
473
551
cp1252, but the console is cp437
553
:param trace: If True trace the selected encoding via mutter().
475
from bzrlib.trace import mutter
555
from .trace import mutter
476
556
output_encoding = getattr(sys.stdout, 'encoding', None)
477
557
if not output_encoding:
478
558
input_encoding = getattr(sys.stdin, 'encoding', None)
479
559
if not input_encoding:
480
560
output_encoding = get_user_encoding()
481
mutter('encoding stdout as osutils.get_user_encoding() %r',
562
mutter('encoding stdout as osutils.get_user_encoding() %r',
484
565
output_encoding = input_encoding
485
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
567
mutter('encoding stdout as sys.stdin encoding %r',
487
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
571
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
488
572
if output_encoding == 'cp0':
489
573
# invalid encoding (cp0 means 'no codepage' on Windows)
490
574
output_encoding = get_user_encoding()
491
mutter('cp0 is invalid encoding.'
576
mutter('cp0 is invalid encoding.'
492
577
' encoding stdout as osutils.get_user_encoding() %r',
496
581
codecs.lookup(output_encoding)
497
582
except LookupError:
498
sys.stderr.write('bzr: warning:'
583
sys.stderr.write('brz: warning:'
499
584
' unknown terminal encoding %s.\n'
500
585
' Using encoding %s instead.\n'
501
586
% (output_encoding, get_user_encoding())
883
980
"""Return size of given open file."""
884
return os.fstat(f.fileno())[ST_SIZE]
887
# Define rand_bytes based on platform.
889
# Python 2.4 and later have os.urandom,
890
# but it doesn't work on some arches
892
rand_bytes = os.urandom
893
except (NotImplementedError, AttributeError):
894
# If python doesn't have os.urandom, or it doesn't work,
895
# then try to first pull random data from /dev/urandom
981
return os.fstat(f.fileno())[stat.ST_SIZE]
984
# Alias os.urandom to support platforms (which?) without /dev/urandom and
985
# override if it doesn't work. Avoid checking on windows where there is
986
# significant initialisation cost that can be avoided for some bzr calls.
988
rand_bytes = os.urandom
990
if rand_bytes.__module__ != "nt":
897
rand_bytes = file('/dev/urandom', 'rb').read
898
# Otherwise, use this hack as a last resort
899
except (IOError, OSError):
993
except NotImplementedError:
900
994
# not well seeded, but better than nothing
901
995
def rand_bytes(n):
971
1071
implementation should be loaded instead::
974
>>> import bzrlib._fictional_extension_pyx
1074
>>> import breezy._fictional_extension_pyx
975
1075
>>> except ImportError, e:
976
>>> bzrlib.osutils.failed_to_load_extension(e)
977
>>> import bzrlib._fictional_extension_py
1076
>>> breezy.osutils.failed_to_load_extension(e)
1077
>>> import breezy._fictional_extension_py
979
1079
# NB: This docstring is just an example, not a doctest, because doctest
980
1080
# currently can't cope with the use of lazy imports in this namespace --
983
1083
# This currently doesn't report the failure at the time it occurs, because
984
1084
# they tend to happen very early in startup when we can't check config
985
1085
# files etc, and also we want to report all failures but not spam the user
986
1086
# with 10 warnings.
987
from bzrlib import trace
988
1087
exception_str = str(exception)
989
1088
if exception_str not in _extension_load_failures:
990
1089
trace.mutter("failed to load compiled extension: %s" % exception_str)
994
1093
def report_extension_load_failures():
995
1094
if not _extension_load_failures:
997
from bzrlib.config import GlobalConfig
998
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1096
if config.GlobalConfig().suppress_warning('missing_extensions'):
1000
1098
# the warnings framework should by default show this only once
1001
from bzrlib.trace import warning
1099
from .trace import warning
1003
"bzr: warning: some compiled extensions could not be loaded; "
1004
"see <https://answers.launchpad.net/bzr/+faq/703>")
1101
"brz: warning: some compiled extensions could not be loaded; "
1102
"see ``brz help missing-extensions``")
1005
1103
# we no longer show the specific missing extensions here, because it makes
1006
1104
# the message too long and scary - see
1007
1105
# https://bugs.launchpad.net/bzr/+bug/430529
1011
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1012
except ImportError, e:
1109
from ._chunks_to_lines_pyx import chunks_to_lines
1110
except ImportError as e:
1013
1111
failed_to_load_extension(e)
1014
from bzrlib._chunks_to_lines_py import chunks_to_lines
1112
from ._chunks_to_lines_py import chunks_to_lines
1017
1115
def split_lines(s):
1018
1116
"""Split s into lines, but without removing the newline characters."""
1019
1117
# Trivially convert a fulltext into a 'chunked' representation, and let
1020
1118
# chunks_to_lines do the heavy lifting.
1021
if isinstance(s, str):
1119
if isinstance(s, bytes):
1022
1120
# chunks_to_lines only supports 8-bit strings
1023
1121
return chunks_to_lines([s])
1285
1400
return unicode_or_utf8_string.encode('utf-8')
1288
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1289
' Revision id generators should be creating utf8'
1293
def safe_revision_id(unicode_or_utf8_string, warn=True):
1403
def safe_revision_id(unicode_or_utf8_string):
1294
1404
"""Revision ids should now be utf8, but at one point they were unicode.
1296
1406
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1298
:param warn: Functions that are sanitizing user data can set warn=False
1299
1408
:return: None or a utf8 revision id.
1301
1410
if (unicode_or_utf8_string is None
1302
or unicode_or_utf8_string.__class__ == str):
1411
or unicode_or_utf8_string.__class__ == bytes):
1303
1412
return unicode_or_utf8_string
1305
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1307
return cache_utf8.encode(unicode_or_utf8_string)
1310
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1311
' generators should be creating utf8 file ids.')
1314
def safe_file_id(unicode_or_utf8_string, warn=True):
1413
raise TypeError('Unicode revision ids are no longer supported. '
1414
'Revision id generators should be creating utf8 revision '
1418
def safe_file_id(unicode_or_utf8_string):
1315
1419
"""File ids should now be utf8, but at one point they were unicode.
1317
1421
This is the same as safe_utf8, except it uses the cached encode functions
1359
1460
can be accessed by that path.
1362
return unicodedata.normalize('NFC', unicode(path)), True
1463
if isinstance(path, bytes):
1464
path = path.decode(sys.getfilesystemencoding())
1465
return unicodedata.normalize('NFC', path), True
1365
1468
def _inaccessible_normalized_filename(path):
1366
1469
__doc__ = _accessible_normalized_filename.__doc__
1368
normalized = unicodedata.normalize('NFC', unicode(path))
1471
if isinstance(path, bytes):
1472
path = path.decode(sys.getfilesystemencoding())
1473
normalized = unicodedata.normalize('NFC', path)
1369
1474
return normalized, normalized == path
1425
1536
None is returned if the width can't established precisely.
1428
- if BZR_COLUMNS is set, returns its value
1539
- if BRZ_COLUMNS is set, returns its value
1429
1540
- if there is no controlling terminal, returns None
1541
- query the OS, if the queried size has changed since the last query,
1430
1543
- if COLUMNS is set, returns its value,
1544
- if the OS has a value (even though it's never changed), return its value.
1432
1546
From there, we need to query the OS to get the size of the controlling
1549
On Unices we query the OS by:
1436
1550
- get termios.TIOCGWINSZ
1437
1551
- if an error occurs or a negative value is obtained, returns None
1553
On Windows we query the OS by:
1441
1554
- win32utils.get_console_size() decides,
1442
1555
- returns None on error (provided default value)
1557
# Note to implementors: if changing the rules for determining the width,
1558
# make sure you've considered the behaviour in these cases:
1559
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1560
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1562
# - (add more interesting cases here, if you find any)
1563
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1564
# but we don't want to register a signal handler because it is impossible
1565
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1566
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1567
# time so we can notice if the reported size has changed, which should have
1445
# If BZR_COLUMNS is set, take it, user is always right
1570
# If BRZ_COLUMNS is set, take it, user is always right
1571
# Except if they specified 0 in which case, impose no limit here
1447
return int(os.environ['BZR_COLUMNS'])
1573
width = int(os.environ['BRZ_COLUMNS'])
1448
1574
except (KeyError, ValueError):
1576
if width is not None:
1451
1582
isatty = getattr(sys.stdout, 'isatty', None)
1452
if isatty is None or not isatty():
1453
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1583
if isatty is None or not isatty():
1584
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
1456
# If COLUMNS is set, take it, the terminal knows better (even inside a
1457
# given terminal, the application can decide to set COLUMNS to a lower
1458
# value (splitted screen) or a bigger value (scroll bars))
1588
width, height = os_size = _terminal_size(None, None)
1589
global _first_terminal_size, _terminal_size_state
1590
if _terminal_size_state == 'no_data':
1591
_first_terminal_size = os_size
1592
_terminal_size_state = 'unchanged'
1593
elif (_terminal_size_state == 'unchanged' and
1594
_first_terminal_size != os_size):
1595
_terminal_size_state = 'changed'
1597
# If the OS claims to know how wide the terminal is, and this value has
1598
# ever changed, use that.
1599
if _terminal_size_state == 'changed':
1600
if width is not None and width > 0:
1603
# If COLUMNS is set, use it.
1460
1605
return int(os.environ['COLUMNS'])
1461
1606
except (KeyError, ValueError):
1464
width, height = _terminal_size(None, None)
1466
# Consider invalid values as meaning no width
1609
# Finally, use an unchanged size from the OS, if we have one.
1610
if _terminal_size_state == 'unchanged':
1611
if width is not None and width > 0:
1614
# The width could not be determined.
1472
1618
def _win32_terminal_size(width, height):
1499
1645
_terminal_size = _ioctl_terminal_size
1502
def _terminal_size_changed(signum, frame):
1503
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1504
width, height = _terminal_size(None, None)
1505
if width is not None:
1506
os.environ['COLUMNS'] = str(width)
1509
_registered_sigwinch = False
1510
def watch_sigwinch():
1511
"""Register for SIGWINCH, once and only once.
1513
Do nothing if the signal module is not available.
1515
global _registered_sigwinch
1516
if not _registered_sigwinch:
1519
if getattr(signal, "SIGWINCH", None) is not None:
1520
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1522
# python doesn't provide signal support, nothing we can do about it
1524
_registered_sigwinch = True
1527
1648
def supports_executable():
1528
1649
return sys.platform != "win32"
1713
1834
global _selected_dir_reader
1714
1835
if _selected_dir_reader is None:
1715
fs_encoding = _fs_enc.upper()
1716
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1717
# Win98 doesn't have unicode apis like FindFirstFileW
1718
# TODO: We possibly could support Win98 by falling back to the
1719
# original FindFirstFile, and using TCHAR instead of WCHAR,
1720
# but that gets a bit tricky, and requires custom compiling
1836
if sys.platform == "win32":
1723
from bzrlib._walkdirs_win32 import Win32ReadDir
1838
from ._walkdirs_win32 import Win32ReadDir
1724
1839
_selected_dir_reader = Win32ReadDir()
1725
1840
except ImportError:
1727
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1728
# ANSI_X3.4-1968 is a form of ASCII
1842
elif _fs_enc in ('utf-8', 'ascii'):
1730
from bzrlib._readdir_pyx import UTF8DirReader
1844
from ._readdir_pyx import UTF8DirReader
1731
1845
_selected_dir_reader = UTF8DirReader()
1732
except ImportError, e:
1846
except ImportError as e:
1733
1847
failed_to_load_extension(e)
1887
2004
_cached_user_encoding = None
1890
def get_user_encoding(use_cache=True):
2007
def get_user_encoding():
1891
2008
"""Find out what the preferred user encoding is.
1893
2010
This is generally the encoding that is used for command line parameters
1894
2011
and file contents. This may be different from the terminal encoding
1895
2012
or the filesystem encoding.
1897
:param use_cache: Enable cache for detected encoding.
1898
(This parameter is turned on by default,
1899
and required only for selftesting)
1901
2014
:return: A string defining the preferred user encoding
1903
2016
global _cached_user_encoding
1904
if _cached_user_encoding is not None and use_cache:
2017
if _cached_user_encoding is not None:
1905
2018
return _cached_user_encoding
1907
if sys.platform == 'darwin':
1908
# python locale.getpreferredencoding() always return
1909
# 'mac-roman' on darwin. That's a lie.
1910
sys.platform = 'posix'
1912
if os.environ.get('LANG', None) is None:
1913
# If LANG is not set, we end up with 'ascii', which is bad
1914
# ('mac-roman' is more than ascii), so we set a default which
1915
# will give us UTF-8 (which appears to work in all cases on
1916
# OSX). Users are still free to override LANG of course, as
1917
# long as it give us something meaningful. This work-around
1918
# *may* not be needed with python 3k and/or OSX 10.5, but will
1919
# work with them too -- vila 20080908
1920
os.environ['LANG'] = 'en_US.UTF-8'
1923
sys.platform = 'darwin'
2020
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2021
# Use the existing locale settings and call nl_langinfo directly
2022
# rather than going through getpreferredencoding. This avoids
2023
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2024
# possibility of the setlocale call throwing an error.
2025
user_encoding = locale.nl_langinfo(locale.CODESET)
2027
# GZ 2011-12-19: On windows could call GetACP directly instead.
2028
user_encoding = locale.getpreferredencoding(False)
1928
user_encoding = locale.getpreferredencoding()
1929
except locale.Error, e:
1930
sys.stderr.write('bzr: warning: %s\n'
1931
' Could not determine what text encoding to use.\n'
1932
' This error usually means your Python interpreter\n'
1933
' doesn\'t support the locale set by $LANG (%s)\n'
1934
" Continuing with ascii encoding.\n"
1935
% (e, os.environ.get('LANG')))
1936
user_encoding = 'ascii'
1938
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1939
# treat that as ASCII, and not support printing unicode characters to the
1942
# For python scripts run under vim, we get '', so also treat that as ASCII
1943
if user_encoding in (None, 'cp0', ''):
1944
user_encoding = 'ascii'
1948
codecs.lookup(user_encoding)
1950
sys.stderr.write('bzr: warning:'
2031
user_encoding = codecs.lookup(user_encoding).name
2033
if user_encoding not in ("", "cp0"):
2034
sys.stderr.write('brz: warning:'
1951
2035
' unknown encoding %s.'
1952
2036
' Continuing with ascii encoding.\n'
1953
2037
% user_encoding
1955
user_encoding = 'ascii'
1958
_cached_user_encoding = user_encoding
2039
user_encoding = 'ascii'
2041
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2042
if user_encoding == 'ascii':
2043
if sys.platform == 'darwin':
2044
# OSX is special-cased in Python to have a UTF-8 filesystem
2045
# encoding and previously had LANG set here if not present.
2046
user_encoding = 'utf-8'
2047
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2048
# for some other posix platforms as well.
2050
_cached_user_encoding = user_encoding
1960
2051
return user_encoding
2054
def get_diff_header_encoding():
2055
return get_terminal_encoding()
1963
2058
def get_host_name():
1964
2059
"""Return the current unicode host name.
2029
2133
def send_all(sock, bytes, report_activity=None):
2030
2134
"""Send all bytes on a socket.
2032
2136
Breaks large blocks in smaller chunks to avoid buffering limitations on
2033
2137
some platforms, and catches EINTR which may be thrown if the send is
2034
2138
interrupted by a signal.
2036
2140
This is preferred to socket.sendall(), because it avoids portability bugs
2037
2141
and provides activity reporting.
2039
2143
:param report_activity: Call this as bytes are read, see
2040
2144
Transport._report_activity
2043
2147
byte_count = len(bytes)
2148
view = memoryview(bytes)
2044
2149
while sent_total < byte_count:
2046
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2047
except socket.error, e:
2151
sent = sock.send(view[sent_total:sent_total+MAX_SOCKET_CHUNK])
2152
except (socket.error, IOError) as e:
2153
if e.args[0] in _end_of_stream_errors:
2154
raise errors.ConnectionReset(
2155
"Error trying to write to socket", e)
2048
2156
if e.args[0] != errno.EINTR:
2160
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2051
2162
sent_total += sent
2052
report_activity(sent, 'write')
2163
if report_activity is not None:
2164
report_activity(sent, 'write')
2167
def connect_socket(address):
2168
# Slight variation of the socket.create_connection() function (provided by
2169
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2170
# provide it for previous python versions. Also, we don't use the timeout
2171
# parameter (provided by the python implementation) so we don't implement
2173
err = socket.error('getaddrinfo returns an empty list')
2174
host, port = address
2175
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2176
af, socktype, proto, canonname, sa = res
2179
sock = socket.socket(af, socktype, proto)
2183
except socket.error as e:
2185
# 'err' is now the most recent error
2186
if sock is not None:
2055
2191
def dereference_path(path):
2083
2219
If and when pkg_resources becomes a standard library, this routine
2084
2220
can delegate to it.
2086
# Check package name is within bzrlib
2087
if package == "bzrlib":
2222
# Check package name is within breezy
2223
if package == "breezy":
2088
2224
resource_relpath = resource_name
2089
elif package.startswith("bzrlib."):
2090
package = package[len("bzrlib."):].replace('.', os.sep)
2225
elif package.startswith("breezy."):
2226
package = package[len("breezy."):].replace('.', os.sep)
2091
2227
resource_relpath = pathjoin(package, resource_name)
2093
raise errors.BzrError('resource package %s not in bzrlib' % package)
2229
raise errors.BzrError('resource package %s not in breezy' % package)
2095
2231
# Map the resource to a file and read its contents
2096
base = dirname(bzrlib.__file__)
2232
base = dirname(breezy.__file__)
2097
2233
if getattr(sys, 'frozen', None): # bzr.exe
2098
2234
base = abspath(pathjoin(base, '..', '..'))
2099
filename = pathjoin(base, resource_relpath)
2100
return open(filename, 'rU').read()
2235
with open(pathjoin(base, resource_relpath), "rt") as f:
2103
2238
def file_kind_from_stat_mode_thunk(mode):
2104
2239
global file_kind_from_stat_mode
2105
2240
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2107
from bzrlib._readdir_pyx import UTF8DirReader
2242
from ._readdir_pyx import UTF8DirReader
2108
2243
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2109
except ImportError, e:
2244
except ImportError as e:
2110
2245
# This is one time where we won't warn that an extension failed to
2111
2246
# load. The extension is never available on Windows anyway.
2112
from bzrlib._readdir_py import (
2247
from ._readdir_py import (
2113
2248
_kind_from_mode as file_kind_from_stat_mode
2115
2250
return file_kind_from_stat_mode(mode)
2116
2251
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2119
def file_kind(f, _lstat=os.lstat):
2253
def file_stat(f, _lstat=os.lstat):
2121
return file_kind_from_stat_mode(_lstat(f).st_mode)
2257
except OSError as e:
2123
2258
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2124
2259
raise errors.NoSuchFile(f)
2262
def file_kind(f, _lstat=os.lstat):
2263
stat_value = file_stat(f, _lstat)
2264
return file_kind_from_stat_mode(stat_value.st_mode)
2128
2266
def until_no_eintr(f, *a, **kw):
2129
2267
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2131
2269
WARNING: you must be certain that it is safe to retry the call repeatedly
2132
2270
if EINTR does occur. This is typically only true for low-level operations
2133
2271
like os.read. If in any doubt, don't use this.
2135
2273
Keep in mind that this is not a complete solution to EINTR. There is
2136
2274
probably code in the Python standard library and other dependencies that
2137
2275
may encounter EINTR if a signal arrives (and there is signal handler for
2138
that signal). So this function can reduce the impact for IO that bzrlib
2276
that signal). So this function can reduce the impact for IO that breezy
2139
2277
directly controls, but it is not a complete solution.
2141
2279
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2144
2282
return f(*a, **kw)
2145
except (IOError, OSError), e:
2283
except (IOError, OSError) as e:
2146
2284
if e.errno == errno.EINTR:
2151
def re_compile_checked(re_string, flags=0, where=""):
2152
"""Return a compiled re, or raise a sensible error.
2154
This should only be used when compiling user-supplied REs.
2156
:param re_string: Text form of regular expression.
2157
:param flags: eg re.IGNORECASE
2158
:param where: Message explaining to the user the context where
2159
it occurred, eg 'log search filter'.
2161
# from https://bugs.launchpad.net/bzr/+bug/251352
2163
re_obj = re.compile(re_string, flags)
2168
where = ' in ' + where
2169
# despite the name 'error' is a type
2170
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2171
% (where, re_string, e))
2174
2289
if sys.platform == "win32":
2177
2292
return msvcrt.getch()
2182
2297
fd = sys.stdin.fileno()
2183
2298
settings = termios.tcgetattr(fd)
2188
2303
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2192
if sys.platform == 'linux2':
2306
if sys.platform.startswith('linux'):
2193
2307
def _local_concurrency():
2195
prefix = 'processor'
2196
for line in file('/proc/cpuinfo', 'rb'):
2197
if line.startswith(prefix):
2198
concurrency = int(line[line.find(':')+1:]) + 1
2309
return os.sysconf('SC_NPROCESSORS_ONLN')
2310
except (ValueError, OSError, AttributeError):
2200
2312
elif sys.platform == 'darwin':
2201
2313
def _local_concurrency():
2202
2314
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2203
2315
stdout=subprocess.PIPE).communicate()[0]
2204
elif sys.platform[0:7] == 'freebsd':
2316
elif "bsd" in sys.platform:
2205
2317
def _local_concurrency():
2206
2318
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2207
2319
stdout=subprocess.PIPE).communicate()[0]
2305
2422
open_file = open
2308
def getuser_unicode():
2309
"""Return the username as unicode.
2312
user_encoding = get_user_encoding()
2313
username = getpass.getuser().decode(user_encoding)
2314
except UnicodeDecodeError:
2315
raise errors.BzrError("Can't decode username as %s." % \
2425
def available_backup_name(base, exists):
2426
"""Find a non-existing backup file name.
2428
This will *not* create anything, this only return a 'free' entry. This
2429
should be used for checking names in a directory below a locked
2430
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2431
Leap) and generally discouraged.
2433
:param base: The base name.
2435
:param exists: A callable returning True if the path parameter exists.
2438
name = "%s.~%d~" % (base, counter)
2441
name = "%s.~%d~" % (base, counter)
2445
def set_fd_cloexec(fd):
2446
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2447
support for this is not available.
2451
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2452
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2453
except (ImportError, AttributeError):
2454
# Either the fcntl module or specific constants are not present
2458
def find_executable_on_path(name):
2459
"""Finds an executable on the PATH.
2461
On Windows, this will try to append each extension in the PATHEXT
2462
environment variable to the name, if it cannot be found with the name
2465
:param name: The base name of the executable.
2466
:return: The path to the executable found or None.
2468
if sys.platform == 'win32':
2469
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2470
exts = [ext.lower() for ext in exts]
2471
base, ext = os.path.splitext(name)
2473
if ext.lower() not in exts:
2479
path = os.environ.get('PATH')
2480
if path is not None:
2481
path = path.split(os.pathsep)
2484
f = os.path.join(d, name) + ext
2485
if os.access(f, os.X_OK):
2487
if sys.platform == 'win32':
2488
app_path = win32utils.get_app_path(name)
2489
if app_path != name:
2494
def _posix_is_local_pid_dead(pid):
2495
"""True if pid doesn't correspond to live process on this machine"""
2497
# Special meaning of unix kill: just check if it's there.
2499
except OSError as e:
2500
if e.errno == errno.ESRCH:
2501
# On this machine, and really not found: as sure as we can be
2504
elif e.errno == errno.EPERM:
2505
# exists, though not ours
2508
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2509
# Don't really know.
2512
# Exists and our process: not dead.
2515
if sys.platform == "win32":
2516
is_local_pid_dead = win32utils.is_local_pid_dead
2518
is_local_pid_dead = _posix_is_local_pid_dead
2520
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2521
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2522
if getattr(errno, name, None) is not None]
2525
def fdatasync(fileno):
2526
"""Flush file contents to disk if possible.
2528
:param fileno: Integer OS file handle.
2529
:raises TransportNotPossible: If flushing to disk is not possible.
2531
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2535
except IOError as e:
2536
# See bug #1075108, on some platforms fdatasync exists, but can
2537
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2538
# and reduce the chance of corruption-on-powerloss situations. It
2539
# is not a mandatory call, so it is ok to suppress failures.
2540
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2541
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2545
def ensure_empty_directory_exists(path, exception_class):
2546
"""Make sure a local directory exists and is empty.
2548
If it does not exist, it is created. If it exists and is not empty, an
2549
instance of exception_class is raised.
2553
except OSError as e:
2554
if e.errno != errno.EEXIST:
2556
if os.listdir(path) != []:
2557
raise exception_class(path)
2560
def is_environment_error(evalue):
2561
"""True if exception instance is due to a process environment issue
2563
This includes OSError and IOError, but also other errors that come from
2564
the operating system or core libraries but are not subclasses of those.
2566
if isinstance(evalue, (EnvironmentError, select.error)):
2568
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):