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
259
file_existed = False
234
261
rename_func(new, tmp_name)
235
except (errors.NoSuchFile,), e:
262
except (errors.NoSuchFile,) as e:
238
265
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
239
266
# function raises an IOError with errno is None when a rename fails.
240
267
# This then gets caught here.
241
268
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
270
except Exception as e:
244
271
if (getattr(e, 'errno', None) is None
245
272
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
248
275
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
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
270
296
# If the file used to exist, rename it back into place
273
299
unlink_func(tmp_name)
275
301
rename_func(tmp_name, new)
276
if failure_exc is not None:
277
raise failure_exc[0], failure_exc[1], failure_exc[2]
280
304
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
281
305
# choke on a Unicode string containing a relative path if
282
306
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
284
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
285
308
def _posix_abspath(path):
286
309
# jam 20060426 rather than encoding to fsencoding
287
310
# copy posixpath.abspath, but use os.getcwdu instead
288
311
if not posixpath.isabs(path):
289
312
path = posixpath.join(getcwd(), path)
290
return posixpath.normpath(path)
313
return _posix_normpath(path)
293
316
def _posix_realpath(path):
294
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))
297
375
def _win32_fixdrive(path):
298
376
"""Force drive letters to be consistent.
303
381
running python.exe under cmd.exe return capital C:\\
304
382
running win32 python inside a cygwin shell returns lowercase c:\\
306
drive, path = _nt_splitdrive(path)
384
drive, path = ntpath.splitdrive(path)
307
385
return drive.upper() + path
310
388
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('\\', '/'))
389
# Real ntpath.abspath doesn't have a problem with a unicode cwd
390
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
342
393
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('\\', '/'))
394
# Real ntpath.realpath doesn't have a problem with a unicode cwd
395
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
347
398
def _win32_pathjoin(*args):
348
return _nt_join(*args).replace('\\', '/')
399
return ntpath.join(*args).replace('\\', '/')
351
402
def _win32_normpath(path):
352
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
403
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
355
406
def _win32_getcwd():
356
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
407
return _win32_fixdrive(_getcwd().replace('\\', '/'))
359
410
def _win32_mkdtemp(*args, **kwargs):
360
411
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
414
def _win32_rename(old, new):
372
415
"""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
432
def _mac_getcwd():
400
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)
403
466
# Default is to just use the python builtins, but these can be rebound on
404
467
# particular platforms.
405
468
abspath = _posix_abspath
406
469
realpath = _posix_realpath
407
470
pathjoin = os.path.join
408
normpath = os.path.normpath
409
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
411
476
dirname = os.path.dirname
412
477
basename = os.path.basename
413
478
split = os.path.split
414
479
splitext = os.path.splitext
415
# These were already imported into local scope
480
# These were already lazily imported into local scope
416
481
# mkdtemp = tempfile.mkdtemp
417
482
# rmtree = shutil.rmtree
419
490
MIN_ABS_PATHLENGTH = 1
422
493
if sys.platform == 'win32':
423
if win32utils.winver == 'Windows 98':
424
abspath = _win98_abspath
426
abspath = _win32_abspath
494
abspath = _win32_abspath
427
495
realpath = _win32_realpath
428
496
pathjoin = _win32_pathjoin
429
497
normpath = _win32_normpath
430
498
getcwd = _win32_getcwd
431
499
mkdtemp = _win32_mkdtemp
432
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
434
510
MIN_ABS_PATHLENGTH = 3
471
550
On my standard US Windows XP, the preferred encoding is
472
551
cp1252, but the console is cp437
553
:param trace: If True trace the selected encoding via mutter().
474
from bzrlib.trace import mutter
555
from .trace import mutter
475
556
output_encoding = getattr(sys.stdout, 'encoding', None)
476
557
if not output_encoding:
477
558
input_encoding = getattr(sys.stdin, 'encoding', None)
478
559
if not input_encoding:
479
560
output_encoding = get_user_encoding()
480
mutter('encoding stdout as osutils.get_user_encoding() %r',
562
mutter('encoding stdout as osutils.get_user_encoding() %r',
483
565
output_encoding = input_encoding
484
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
567
mutter('encoding stdout as sys.stdin encoding %r',
486
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
571
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
487
572
if output_encoding == 'cp0':
488
573
# invalid encoding (cp0 means 'no codepage' on Windows)
489
574
output_encoding = get_user_encoding()
490
mutter('cp0 is invalid encoding.'
576
mutter('cp0 is invalid encoding.'
491
577
' encoding stdout as osutils.get_user_encoding() %r',
495
581
codecs.lookup(output_encoding)
496
582
except LookupError:
497
sys.stderr.write('bzr: warning:'
583
sys.stderr.write('brz: warning:'
498
584
' unknown terminal encoding %s.\n'
499
585
' Using encoding %s instead.\n'
500
586
% (output_encoding, get_user_encoding())
882
984
"""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
985
return os.fstat(f.fileno())[stat.ST_SIZE]
988
# Alias os.urandom to support platforms (which?) without /dev/urandom and
989
# override if it doesn't work. Avoid checking on windows where there is
990
# significant initialisation cost that can be avoided for some bzr calls.
992
rand_bytes = os.urandom
994
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):
997
except NotImplementedError:
899
998
# not well seeded, but better than nothing
900
999
def rand_bytes(n):
970
1075
implementation should be loaded instead::
973
>>> import bzrlib._fictional_extension_pyx
1078
>>> import breezy._fictional_extension_pyx
974
1079
>>> except ImportError, e:
975
>>> bzrlib.osutils.failed_to_load_extension(e)
976
>>> import bzrlib._fictional_extension_py
1080
>>> breezy.osutils.failed_to_load_extension(e)
1081
>>> import breezy._fictional_extension_py
978
1083
# NB: This docstring is just an example, not a doctest, because doctest
979
1084
# currently can't cope with the use of lazy imports in this namespace --
982
1087
# This currently doesn't report the failure at the time it occurs, because
983
1088
# they tend to happen very early in startup when we can't check config
984
1089
# files etc, and also we want to report all failures but not spam the user
985
1090
# with 10 warnings.
986
from bzrlib import trace
987
1091
exception_str = str(exception)
988
1092
if exception_str not in _extension_load_failures:
989
1093
trace.mutter("failed to load compiled extension: %s" % exception_str)
993
1097
def report_extension_load_failures():
994
1098
if not _extension_load_failures:
996
from bzrlib.config import GlobalConfig
997
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1100
if config.GlobalConfig().suppress_warning('missing_extensions'):
999
1102
# the warnings framework should by default show this only once
1000
from bzrlib.trace import warning
1103
from .trace import warning
1002
"bzr: warning: some compiled extensions could not be loaded; "
1003
"see <https://answers.launchpad.net/bzr/+faq/703>")
1105
"brz: warning: some compiled extensions could not be loaded; "
1106
"see ``brz help missing-extensions``")
1004
1107
# we no longer show the specific missing extensions here, because it makes
1005
1108
# the message too long and scary - see
1006
1109
# https://bugs.launchpad.net/bzr/+bug/430529
1010
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1011
except ImportError, e:
1113
from ._chunks_to_lines_pyx import chunks_to_lines
1114
except ImportError as e:
1012
1115
failed_to_load_extension(e)
1013
from bzrlib._chunks_to_lines_py import chunks_to_lines
1116
from ._chunks_to_lines_py import chunks_to_lines
1016
1119
def split_lines(s):
1017
1120
"""Split s into lines, but without removing the newline characters."""
1018
1121
# Trivially convert a fulltext into a 'chunked' representation, and let
1019
1122
# chunks_to_lines do the heavy lifting.
1020
if isinstance(s, str):
1123
if isinstance(s, bytes):
1021
1124
# chunks_to_lines only supports 8-bit strings
1022
1125
return chunks_to_lines([s])
1049
1153
os.link(src, dest)
1050
except (OSError, IOError), e:
1154
except (OSError, IOError) as e:
1051
1155
if e.errno != errno.EXDEV:
1053
1157
shutil.copyfile(src, dest)
1056
1160
def delete_any(path):
1057
"""Delete a file, symlink or directory.
1161
"""Delete a file, symlink or directory.
1059
1163
Will delete even if readonly.
1062
1166
_delete_file_or_dir(path)
1063
except (OSError, IOError), e:
1167
except (OSError, IOError) as e:
1064
1168
if e.errno in (errno.EPERM, errno.EACCES):
1065
1169
# make writable and try again
1250
1354
# but for now, we haven't optimized...
1251
1355
return [canonical_relpath(base, p) for p in paths]
1358
def decode_filename(filename):
1359
"""Decode the filename using the filesystem encoding
1361
If it is unicode, it is returned.
1362
Otherwise it is decoded from the the filesystem's encoding. If decoding
1363
fails, a errors.BadFilenameEncoding exception is raised.
1365
if isinstance(filename, text_type):
1368
return filename.decode(_fs_enc)
1369
except UnicodeDecodeError:
1370
raise errors.BadFilenameEncoding(filename, _fs_enc)
1253
1373
def safe_unicode(unicode_or_utf8_string):
1254
1374
"""Coerce unicode_or_utf8_string into unicode.
1284
1404
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):
1407
def safe_revision_id(unicode_or_utf8_string):
1293
1408
"""Revision ids should now be utf8, but at one point they were unicode.
1295
1410
: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
1412
:return: None or a utf8 revision id.
1300
1414
if (unicode_or_utf8_string is None
1301
or unicode_or_utf8_string.__class__ == str):
1415
or unicode_or_utf8_string.__class__ == bytes):
1302
1416
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):
1417
raise TypeError('Unicode revision ids are no longer supported. '
1418
'Revision id generators should be creating utf8 revision '
1422
def safe_file_id(unicode_or_utf8_string):
1314
1423
"""File ids should now be utf8, but at one point they were unicode.
1316
1425
This is the same as safe_utf8, except it uses the cached encode functions
1358
1464
can be accessed by that path.
1361
return unicodedata.normalize('NFC', unicode(path)), True
1467
if isinstance(path, bytes):
1468
path = path.decode(sys.getfilesystemencoding())
1469
return unicodedata.normalize('NFC', path), True
1364
1472
def _inaccessible_normalized_filename(path):
1365
1473
__doc__ = _accessible_normalized_filename.__doc__
1367
normalized = unicodedata.normalize('NFC', unicode(path))
1475
if isinstance(path, bytes):
1476
path = path.decode(sys.getfilesystemencoding())
1477
normalized = unicodedata.normalize('NFC', path)
1368
1478
return normalized, normalized == path
1424
1540
None is returned if the width can't established precisely.
1427
- if BZR_COLUMNS is set, returns its value
1543
- if BRZ_COLUMNS is set, returns its value
1428
1544
- if there is no controlling terminal, returns None
1545
- query the OS, if the queried size has changed since the last query,
1429
1547
- if COLUMNS is set, returns its value,
1548
- if the OS has a value (even though it's never changed), return its value.
1431
1550
From there, we need to query the OS to get the size of the controlling
1553
On Unices we query the OS by:
1435
1554
- get termios.TIOCGWINSZ
1436
1555
- if an error occurs or a negative value is obtained, returns None
1557
On Windows we query the OS by:
1440
1558
- win32utils.get_console_size() decides,
1441
1559
- returns None on error (provided default value)
1561
# Note to implementors: if changing the rules for determining the width,
1562
# make sure you've considered the behaviour in these cases:
1563
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1564
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1566
# - (add more interesting cases here, if you find any)
1567
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1568
# but we don't want to register a signal handler because it is impossible
1569
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1570
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1571
# 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
1574
# If BRZ_COLUMNS is set, take it, user is always right
1575
# Except if they specified 0 in which case, impose no limit here
1446
return int(os.environ['BZR_COLUMNS'])
1577
width = int(os.environ['BRZ_COLUMNS'])
1447
1578
except (KeyError, ValueError):
1580
if width is not None:
1450
1586
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.
1587
if isatty is None or not isatty():
1588
# 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))
1592
width, height = os_size = _terminal_size(None, None)
1593
global _first_terminal_size, _terminal_size_state
1594
if _terminal_size_state == 'no_data':
1595
_first_terminal_size = os_size
1596
_terminal_size_state = 'unchanged'
1597
elif (_terminal_size_state == 'unchanged' and
1598
_first_terminal_size != os_size):
1599
_terminal_size_state = 'changed'
1601
# If the OS claims to know how wide the terminal is, and this value has
1602
# ever changed, use that.
1603
if _terminal_size_state == 'changed':
1604
if width is not None and width > 0:
1607
# If COLUMNS is set, use it.
1459
1609
return int(os.environ['COLUMNS'])
1460
1610
except (KeyError, ValueError):
1463
width, height = _terminal_size(None, None)
1465
# Consider invalid values as meaning no width
1613
# Finally, use an unchanged size from the OS, if we have one.
1614
if _terminal_size_state == 'unchanged':
1615
if width is not None and width > 0:
1618
# The width could not be determined.
1471
1622
def _win32_terminal_size(width, height):
1498
1649
_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
1652
def supports_executable():
1527
1653
return sys.platform != "win32"
1712
1838
global _selected_dir_reader
1713
1839
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
1840
if sys.platform == "win32":
1722
from bzrlib._walkdirs_win32 import Win32ReadDir
1842
from ._walkdirs_win32 import Win32ReadDir
1723
1843
_selected_dir_reader = Win32ReadDir()
1724
1844
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
1846
elif _fs_enc in ('utf-8', 'ascii'):
1729
from bzrlib._readdir_pyx import UTF8DirReader
1848
from ._readdir_pyx import UTF8DirReader
1730
1849
_selected_dir_reader = UTF8DirReader()
1731
except ImportError, e:
1850
except ImportError as e:
1732
1851
failed_to_load_extension(e)
1779
1898
See DirReader.read_dir for details.
1781
1900
_utf8_encode = self._utf8_encode
1901
_fs_decode = lambda s: s.decode(_fs_enc)
1902
_fs_encode = lambda s: s.encode(_fs_enc)
1782
1903
_lstat = os.lstat
1783
1904
_listdir = os.listdir
1784
1905
_kind_from_mode = file_kind_from_stat_mode
1787
relprefix = prefix + '/'
1908
relprefix = prefix + b'/'
1790
top_slash = top + u'/'
1911
top_slash = top + '/'
1793
1914
append = dirblock.append
1794
for name in sorted(_listdir(top)):
1915
for name_native in _listdir(top.encode('utf-8')):
1796
name_utf8 = _utf8_encode(name)[0]
1917
name = _fs_decode(name_native)
1797
1918
except UnicodeDecodeError:
1798
1919
raise errors.BadFilenameEncoding(
1799
_utf8_encode(relprefix)[0] + name, _fs_enc)
1920
relprefix + name_native, _fs_enc)
1921
name_utf8 = _utf8_encode(name)[0]
1800
1922
abspath = top_slash + name
1801
1923
statvalue = _lstat(abspath)
1802
1924
kind = _kind_from_mode(statvalue.st_mode)
1803
1925
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1926
return sorted(dirblock)
1807
1929
def copy_tree(from_path, to_path, handlers={}):
1874
1999
This can be used to sort paths in the same way that walkdirs does.
1876
return (dirname(path) , path)
2001
return (dirname(path), path)
1879
2004
def compare_paths_prefix_order(path_a, path_b):
1880
2005
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1881
2006
key_a = path_prefix_key(path_a)
1882
2007
key_b = path_prefix_key(path_b)
1883
return cmp(key_a, key_b)
2008
return (key_a > key_b) - (key_a < key_b)
1886
2011
_cached_user_encoding = None
1889
def get_user_encoding(use_cache=True):
2014
def get_user_encoding():
1890
2015
"""Find out what the preferred user encoding is.
1892
2017
This is generally the encoding that is used for command line parameters
1893
2018
and file contents. This may be different from the terminal encoding
1894
2019
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
2021
:return: A string defining the preferred user encoding
1902
2023
global _cached_user_encoding
1903
if _cached_user_encoding is not None and use_cache:
2024
if _cached_user_encoding is not None:
1904
2025
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'
2027
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2028
# Use the existing locale settings and call nl_langinfo directly
2029
# rather than going through getpreferredencoding. This avoids
2030
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2031
# possibility of the setlocale call throwing an error.
2032
user_encoding = locale.nl_langinfo(locale.CODESET)
2034
# GZ 2011-12-19: On windows could call GetACP directly instead.
2035
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:'
2038
user_encoding = codecs.lookup(user_encoding).name
2040
if user_encoding not in ("", "cp0"):
2041
sys.stderr.write('brz: warning:'
1950
2042
' unknown encoding %s.'
1951
2043
' Continuing with ascii encoding.\n'
1952
2044
% user_encoding
1954
user_encoding = 'ascii'
1957
_cached_user_encoding = user_encoding
2046
user_encoding = 'ascii'
2048
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2049
if user_encoding == 'ascii':
2050
if sys.platform == 'darwin':
2051
# OSX is special-cased in Python to have a UTF-8 filesystem
2052
# encoding and previously had LANG set here if not present.
2053
user_encoding = 'utf-8'
2054
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2055
# for some other posix platforms as well.
2057
_cached_user_encoding = user_encoding
1959
2058
return user_encoding
2061
def get_diff_header_encoding():
2062
return get_terminal_encoding()
1962
2065
def get_host_name():
1963
2066
"""Return the current unicode host name.
1987
2099
empty string rather than raise an error), and repeats the recv if
1988
2100
interrupted by a signal.
1992
bytes = sock.recv(max_read_size)
1993
except socket.error, e:
2104
data = sock.recv(max_read_size)
2105
except socket.error as e:
1994
2106
eno = e.args[0]
1995
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2107
if eno in _end_of_stream_errors:
1996
2108
# The connection was closed by the other side. Callers expect
1997
2109
# an empty string to signal end-of-stream.
1999
2111
elif eno == errno.EINTR:
2000
2112
# Retry the interrupted recv.
2004
2116
if report_activity is not None:
2005
report_activity(len(bytes), 'read')
2117
report_activity(len(data), 'read')
2009
2121
def recv_all(socket, count):
2028
2140
def send_all(sock, bytes, report_activity=None):
2029
2141
"""Send all bytes on a socket.
2031
2143
Breaks large blocks in smaller chunks to avoid buffering limitations on
2032
2144
some platforms, and catches EINTR which may be thrown if the send is
2033
2145
interrupted by a signal.
2035
2147
This is preferred to socket.sendall(), because it avoids portability bugs
2036
2148
and provides activity reporting.
2038
2150
:param report_activity: Call this as bytes are read, see
2039
2151
Transport._report_activity
2042
2154
byte_count = len(bytes)
2155
view = memoryview(bytes)
2043
2156
while sent_total < byte_count:
2045
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2046
except socket.error, e:
2158
sent = sock.send(view[sent_total:sent_total+MAX_SOCKET_CHUNK])
2159
except (socket.error, IOError) as e:
2160
if e.args[0] in _end_of_stream_errors:
2161
raise errors.ConnectionReset(
2162
"Error trying to write to socket", e)
2047
2163
if e.args[0] != errno.EINTR:
2167
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2050
2169
sent_total += sent
2051
report_activity(sent, 'write')
2170
if report_activity is not None:
2171
report_activity(sent, 'write')
2174
def connect_socket(address):
2175
# Slight variation of the socket.create_connection() function (provided by
2176
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2177
# provide it for previous python versions. Also, we don't use the timeout
2178
# parameter (provided by the python implementation) so we don't implement
2180
err = socket.error('getaddrinfo returns an empty list')
2181
host, port = address
2182
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2183
af, socktype, proto, canonname, sa = res
2186
sock = socket.socket(af, socktype, proto)
2190
except socket.error as e:
2192
# 'err' is now the most recent error
2193
if sock is not None:
2054
2198
def dereference_path(path):
2082
2226
If and when pkg_resources becomes a standard library, this routine
2083
2227
can delegate to it.
2085
# Check package name is within bzrlib
2086
if package == "bzrlib":
2229
# Check package name is within breezy
2230
if package == "breezy":
2087
2231
resource_relpath = resource_name
2088
elif package.startswith("bzrlib."):
2089
package = package[len("bzrlib."):].replace('.', os.sep)
2232
elif package.startswith("breezy."):
2233
package = package[len("breezy."):].replace('.', os.sep)
2090
2234
resource_relpath = pathjoin(package, resource_name)
2092
raise errors.BzrError('resource package %s not in bzrlib' % package)
2236
raise errors.BzrError('resource package %s not in breezy' % package)
2094
2238
# Map the resource to a file and read its contents
2095
base = dirname(bzrlib.__file__)
2239
base = dirname(breezy.__file__)
2096
2240
if getattr(sys, 'frozen', None): # bzr.exe
2097
2241
base = abspath(pathjoin(base, '..', '..'))
2098
filename = pathjoin(base, resource_relpath)
2099
return open(filename, 'rU').read()
2242
with open(pathjoin(base, resource_relpath), "rt") as f:
2102
2245
def file_kind_from_stat_mode_thunk(mode):
2103
2246
global file_kind_from_stat_mode
2104
2247
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2106
from bzrlib._readdir_pyx import UTF8DirReader
2249
from ._readdir_pyx import UTF8DirReader
2107
2250
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2108
except ImportError, e:
2251
except ImportError as e:
2109
2252
# This is one time where we won't warn that an extension failed to
2110
2253
# load. The extension is never available on Windows anyway.
2111
from bzrlib._readdir_py import (
2254
from ._readdir_py import (
2112
2255
_kind_from_mode as file_kind_from_stat_mode
2114
2257
return file_kind_from_stat_mode(mode)
2115
2258
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)
2269
def file_kind(f, _lstat=os.lstat):
2270
stat_value = file_stat(f, _lstat)
2271
return file_kind_from_stat_mode(stat_value.st_mode)
2127
2273
def until_no_eintr(f, *a, **kw):
2128
2274
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2130
2276
WARNING: you must be certain that it is safe to retry the call repeatedly
2131
2277
if EINTR does occur. This is typically only true for low-level operations
2132
2278
like os.read. If in any doubt, don't use this.
2134
2280
Keep in mind that this is not a complete solution to EINTR. There is
2135
2281
probably code in the Python standard library and other dependencies that
2136
2282
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
2283
that signal). So this function can reduce the impact for IO that breezy
2138
2284
directly controls, but it is not a complete solution.
2140
2286
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2143
2289
return f(*a, **kw)
2144
except (IOError, OSError), e:
2290
except (IOError, OSError) as e:
2145
2291
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
2296
if sys.platform == "win32":
2176
2299
return msvcrt.getch()
2181
2304
fd = sys.stdin.fileno()
2182
2305
settings = termios.tcgetattr(fd)
2187
2310
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2191
if sys.platform == 'linux2':
2313
if sys.platform.startswith('linux'):
2192
2314
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
2316
return os.sysconf('SC_NPROCESSORS_ONLN')
2317
except (ValueError, OSError, AttributeError):
2199
2319
elif sys.platform == 'darwin':
2200
2320
def _local_concurrency():
2201
2321
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2202
2322
stdout=subprocess.PIPE).communicate()[0]
2203
elif sys.platform[0:7] == 'freebsd':
2323
elif "bsd" in sys.platform:
2204
2324
def _local_concurrency():
2205
2325
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2206
2326
stdout=subprocess.PIPE).communicate()[0]
2302
2428
return os.fdopen(os.open(filename, flags), mode, bufsize)
2304
2430
open_file = open
2433
def available_backup_name(base, exists):
2434
"""Find a non-existing backup file name.
2436
This will *not* create anything, this only return a 'free' entry. This
2437
should be used for checking names in a directory below a locked
2438
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2439
Leap) and generally discouraged.
2441
:param base: The base name.
2443
:param exists: A callable returning True if the path parameter exists.
2446
name = "%s.~%d~" % (base, counter)
2449
name = "%s.~%d~" % (base, counter)
2453
def set_fd_cloexec(fd):
2454
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2455
support for this is not available.
2459
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2460
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2461
except (ImportError, AttributeError):
2462
# Either the fcntl module or specific constants are not present
2466
def find_executable_on_path(name):
2467
"""Finds an executable on the PATH.
2469
On Windows, this will try to append each extension in the PATHEXT
2470
environment variable to the name, if it cannot be found with the name
2473
:param name: The base name of the executable.
2474
:return: The path to the executable found or None.
2476
if sys.platform == 'win32':
2477
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2478
exts = [ext.lower() for ext in exts]
2479
base, ext = os.path.splitext(name)
2481
if ext.lower() not in exts:
2487
path = os.environ.get('PATH')
2488
if path is not None:
2489
path = path.split(os.pathsep)
2492
f = os.path.join(d, name) + ext
2493
if os.access(f, os.X_OK):
2495
if sys.platform == 'win32':
2496
app_path = win32utils.get_app_path(name)
2497
if app_path != name:
2502
def _posix_is_local_pid_dead(pid):
2503
"""True if pid doesn't correspond to live process on this machine"""
2505
# Special meaning of unix kill: just check if it's there.
2507
except OSError as e:
2508
if e.errno == errno.ESRCH:
2509
# On this machine, and really not found: as sure as we can be
2512
elif e.errno == errno.EPERM:
2513
# exists, though not ours
2516
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2517
# Don't really know.
2520
# Exists and our process: not dead.
2523
if sys.platform == "win32":
2524
is_local_pid_dead = win32utils.is_local_pid_dead
2526
is_local_pid_dead = _posix_is_local_pid_dead
2528
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2529
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2530
if getattr(errno, name, None) is not None]
2533
def fdatasync(fileno):
2534
"""Flush file contents to disk if possible.
2536
:param fileno: Integer OS file handle.
2537
:raises TransportNotPossible: If flushing to disk is not possible.
2539
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2543
except IOError as e:
2544
# See bug #1075108, on some platforms fdatasync exists, but can
2545
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2546
# and reduce the chance of corruption-on-powerloss situations. It
2547
# is not a mandatory call, so it is ok to suppress failures.
2548
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2549
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2553
def ensure_empty_directory_exists(path, exception_class):
2554
"""Make sure a local directory exists and is empty.
2556
If it does not exist, it is created. If it exists and is not empty, an
2557
instance of exception_class is raised.
2561
except OSError as e:
2562
if e.errno != errno.EEXIST:
2564
if os.listdir(path) != []:
2565
raise exception_class(path)
2568
def is_environment_error(evalue):
2569
"""True if exception instance is due to a process environment issue
2571
This includes OSError and IOError, but also other errors that come from
2572
the operating system or core libraries but are not subclasses of those.
2574
if isinstance(evalue, (EnvironmentError, select.error)):
2576
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
added
removed
breezy/osutils.py
