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
# Cross platform wall-clock time functionality with decent resolution.
81
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
82
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
83
# synchronized with ``time.time()``, this is only meant to be used to find
84
# delta times by subtracting from another call to this function.
85
timer_func = time.time
86
if sys.platform == 'win32':
87
timer_func = time.clock
89
73
# On win32, O_BINARY is used to indicate the file should
90
74
# be opened in binary mode, rather than text mode.
97
81
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
84
class UnsupportedTimezoneFormat(errors.BzrError):
86
_fmt = ('Unsupported timezone format "%(timezone)s", '
87
'options are "utc", "original", "local".')
89
def __init__(self, timezone):
90
self.timezone = timezone
100
93
def get_unicode_argv():
102
97
user_encoding = get_user_encoding()
103
98
return [a.decode(user_encoding) for a in sys.argv[1:]]
104
99
except UnicodeDecodeError:
105
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
100
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
101
"application locale.").format(a, user_encoding))
109
104
def make_readonly(filename):
110
105
"""Make a filename read-only."""
111
106
mod = os.lstat(filename).st_mode
112
107
if not stat.S_ISLNK(mod):
114
os.chmod(filename, mod)
109
chmod_if_possible(filename, mod)
117
112
def make_writable(filename):
118
113
mod = os.lstat(filename).st_mode
119
114
if not stat.S_ISLNK(mod):
121
os.chmod(filename, mod)
116
chmod_if_possible(filename, mod)
119
def chmod_if_possible(filename, mode):
120
# Set file mode if that can be safely done.
121
# Sometimes even on unix the filesystem won't allow it - see
122
# https://bugs.launchpad.net/bzr/+bug/606537
124
# It is probably faster to just do the chmod, rather than
125
# doing a stat, and then trying to compare
126
os.chmod(filename, mode)
127
except (IOError, OSError) as e:
128
# Permission/access denied seems to commonly happen on smbfs; there's
129
# probably no point warning about it.
130
# <https://bugs.launchpad.net/bzr/+bug/606537>
131
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
132
trace.mutter("ignore error on chmod of %r: %r" % (
124
138
def minimum_path_selection(paths):
233
252
file_existed = False
235
254
rename_func(new, tmp_name)
236
except (errors.NoSuchFile,), e:
255
except (errors.NoSuchFile,):
239
258
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
240
259
# function raises an IOError with errno is None when a rename fails.
241
260
# This then gets caught here.
242
261
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
263
except Exception as e:
245
264
if (getattr(e, 'errno', None) is None
246
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
265
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
249
268
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)
272
# This may throw an exception, in which case success will
274
rename_func(old, new)
276
except (IOError, OSError) as e:
277
# source and target may be aliases of each other (e.g. on a
278
# case-insensitive filesystem), so we may have accidentally renamed
279
# source by when we tried to rename target
280
if (file_existed and e.errno in (None, errno.ENOENT)
265
281
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
282
# source and target are the same file on a case-insensitive
283
# filesystem, so we don't generate an exception
271
289
# If the file used to exist, rename it back into place
274
292
unlink_func(tmp_name)
276
294
rename_func(tmp_name, new)
277
if failure_exc is not None:
278
raise failure_exc[0], failure_exc[1], failure_exc[2]
281
297
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
282
298
# choke on a Unicode string containing a relative path if
283
299
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
285
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
286
301
def _posix_abspath(path):
287
302
# jam 20060426 rather than encoding to fsencoding
288
303
# copy posixpath.abspath, but use os.getcwdu instead
289
304
if not posixpath.isabs(path):
290
305
path = posixpath.join(getcwd(), path)
291
return posixpath.normpath(path)
306
return _posix_normpath(path)
294
309
def _posix_realpath(path):
295
310
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
313
def _posix_normpath(path):
314
path = posixpath.normpath(path)
315
# Bug 861008: posixpath.normpath() returns a path normalized according to
316
# the POSIX standard, which stipulates (for compatibility reasons) that two
317
# leading slashes must not be simplified to one, and only if there are 3 or
318
# more should they be simplified as one. So we treat the leading 2 slashes
319
# as a special case here by simply removing the first slash, as we consider
320
# that breaking POSIX compatibility for this obscure feature is acceptable.
321
# This is not a paranoid precaution, as we notably get paths like this when
322
# the repo is hosted at the root of the filesystem, i.e. in "/".
323
if path.startswith('//'):
328
def _posix_path_from_environ(key):
329
"""Get unicode path from `key` in environment or None if not present
331
Note that posix systems use arbitrary byte strings for filesystem objects,
332
so a path that raises BadFilenameEncoding here may still be accessible.
334
val = os.environ.get(key, None)
335
if PY3 or val is None:
338
return val.decode(_fs_enc)
339
except UnicodeDecodeError:
340
# GZ 2011-12-12:Ideally want to include `key` in the exception message
341
raise errors.BadFilenameEncoding(val, _fs_enc)
344
def _posix_get_home_dir():
345
"""Get the home directory of the current user as a unicode path"""
346
path = posixpath.expanduser("~")
348
return path.decode(_fs_enc)
349
except AttributeError:
351
except UnicodeDecodeError:
352
raise errors.BadFilenameEncoding(path, _fs_enc)
355
def _posix_getuser_unicode():
356
"""Get username from environment or password database as unicode"""
357
name = getpass.getuser()
360
user_encoding = get_user_encoding()
362
return name.decode(user_encoding)
363
except UnicodeDecodeError:
364
raise errors.BzrError("Encoding of username %r is unsupported by %s "
365
"application locale." % (name, user_encoding))
298
368
def _win32_fixdrive(path):
299
369
"""Force drive letters to be consistent.
304
374
running python.exe under cmd.exe return capital C:\\
305
375
running win32 python inside a cygwin shell returns lowercase c:\\
307
drive, path = _nt_splitdrive(path)
377
drive, path = ntpath.splitdrive(path)
308
378
return drive.upper() + path
311
381
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('\\', '/'))
382
# Real ntpath.abspath doesn't have a problem with a unicode cwd
383
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
343
386
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('\\', '/'))
387
# Real ntpath.realpath doesn't have a problem with a unicode cwd
388
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
348
391
def _win32_pathjoin(*args):
349
return _nt_join(*args).replace('\\', '/')
392
return ntpath.join(*args).replace('\\', '/')
352
395
def _win32_normpath(path):
353
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
396
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
356
399
def _win32_getcwd():
357
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
400
return _win32_fixdrive(_getcwd().replace('\\', '/'))
360
403
def _win32_mkdtemp(*args, **kwargs):
361
404
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
407
def _win32_rename(old, new):
373
408
"""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
425
def _mac_getcwd():
401
return unicodedata.normalize('NFC', os.getcwdu())
426
return unicodedata.normalize('NFC', _getcwd())
429
def _rename_wrap_exception(rename_func):
430
"""Adds extra information to any exceptions that come from rename().
432
The exception has an updated message and 'old_filename' and 'new_filename'
436
def _rename_wrapper(old, new):
438
rename_func(old, new)
440
detailed_error = OSError(e.errno, e.strerror +
441
" [occurred when renaming '%s' to '%s']" %
443
detailed_error.old_filename = old
444
detailed_error.new_filename = new
447
return _rename_wrapper
450
if sys.version_info > (3,):
456
# Default rename wraps os.rename()
457
rename = _rename_wrap_exception(os.rename)
404
459
# Default is to just use the python builtins, but these can be rebound on
405
460
# particular platforms.
406
461
abspath = _posix_abspath
407
462
realpath = _posix_realpath
408
463
pathjoin = os.path.join
409
normpath = os.path.normpath
410
rename = _wrapped_rename # overridden below on win32
464
normpath = _posix_normpath
465
path_from_environ = _posix_path_from_environ
466
_get_home_dir = _posix_get_home_dir
467
getuser_unicode = _posix_getuser_unicode
412
469
dirname = os.path.dirname
413
470
basename = os.path.basename
414
471
split = os.path.split
415
472
splitext = os.path.splitext
416
# These were already imported into local scope
473
# These were already lazily imported into local scope
417
474
# mkdtemp = tempfile.mkdtemp
418
475
# rmtree = shutil.rmtree
420
484
MIN_ABS_PATHLENGTH = 1
423
487
if sys.platform == 'win32':
424
if win32utils.winver == 'Windows 98':
425
abspath = _win98_abspath
427
abspath = _win32_abspath
488
abspath = _win32_abspath
428
489
realpath = _win32_realpath
429
490
pathjoin = _win32_pathjoin
430
491
normpath = _win32_normpath
431
492
getcwd = _win32_getcwd
432
493
mkdtemp = _win32_mkdtemp
433
rename = _win32_rename
494
rename = _rename_wrap_exception(_win32_rename)
496
from . import _walkdirs_win32
500
lstat = _walkdirs_win32.lstat
501
fstat = _walkdirs_win32.fstat
502
wrap_stat = _walkdirs_win32.wrap_stat
435
504
MIN_ABS_PATHLENGTH = 3
472
542
On my standard US Windows XP, the preferred encoding is
473
543
cp1252, but the console is cp437
545
:param trace: If True trace the selected encoding via mutter().
475
from bzrlib.trace import mutter
547
from .trace import mutter
476
548
output_encoding = getattr(sys.stdout, 'encoding', None)
477
549
if not output_encoding:
478
550
input_encoding = getattr(sys.stdin, 'encoding', None)
479
551
if not input_encoding:
480
552
output_encoding = get_user_encoding()
481
mutter('encoding stdout as osutils.get_user_encoding() %r',
554
mutter('encoding stdout as osutils.get_user_encoding() %r',
484
557
output_encoding = input_encoding
485
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
559
mutter('encoding stdout as sys.stdin encoding %r',
487
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
563
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
488
564
if output_encoding == 'cp0':
489
565
# invalid encoding (cp0 means 'no codepage' on Windows)
490
566
output_encoding = get_user_encoding()
491
mutter('cp0 is invalid encoding.'
492
' encoding stdout as osutils.get_user_encoding() %r',
568
mutter('cp0 is invalid encoding.'
569
' encoding stdout as osutils.get_user_encoding() %r',
496
573
codecs.lookup(output_encoding)
497
574
except LookupError:
498
sys.stderr.write('bzr: warning:'
575
sys.stderr.write('brz: warning:'
499
576
' unknown terminal encoding %s.\n'
500
577
' Using encoding %s instead.\n'
501
578
% (output_encoding, get_user_encoding())
503
580
output_encoding = get_user_encoding()
505
582
return output_encoding
879
975
return '%d hours, %d minute%s %s' % (hours, minutes,
880
976
plural_minutes, direction)
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):
918
1014
for raw_byte in rand_bytes(num):
919
s += ALNUM[ord(raw_byte) % 36]
1016
s += ALNUM[ord(raw_byte) % 36]
1018
s += ALNUM[raw_byte % 36]
923
## TODO: We could later have path objects that remember their list
924
## decomposition (might be too tricksy though.)
1022
# TODO: We could later have path objects that remember their list
1023
# decomposition (might be too tricksy though.)
926
1025
def splitpath(p):
927
1026
"""Turn string into list of parts."""
928
# split on either delimiter because people might use either on
930
ps = re.split(r'[\\/]', p)
1027
if os.path.sep == '\\':
1028
# split on either delimiter because people might use either on
1030
if isinstance(p, bytes):
1031
ps = re.split(b'[\\\\/]', p)
1033
ps = re.split(r'[\\/]', p)
1035
if isinstance(p, bytes):
935
raise errors.BzrError("sorry, %r not allowed in path" % f)
936
elif (f == '.') or (f == ''):
1042
if f in ('..', b'..'):
1043
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1044
elif f in ('.', '', b'.', b''):
971
1079
implementation should be loaded instead::
974
>>> import bzrlib._fictional_extension_pyx
1082
>>> import breezy._fictional_extension_pyx
975
1083
>>> except ImportError, e:
976
>>> bzrlib.osutils.failed_to_load_extension(e)
977
>>> import bzrlib._fictional_extension_py
1084
>>> breezy.osutils.failed_to_load_extension(e)
1085
>>> import breezy._fictional_extension_py
979
1087
# NB: This docstring is just an example, not a doctest, because doctest
980
1088
# currently can't cope with the use of lazy imports in this namespace --
983
1091
# This currently doesn't report the failure at the time it occurs, because
984
1092
# they tend to happen very early in startup when we can't check config
985
1093
# files etc, and also we want to report all failures but not spam the user
986
1094
# with 10 warnings.
987
from bzrlib import trace
988
1095
exception_str = str(exception)
989
1096
if exception_str not in _extension_load_failures:
990
1097
trace.mutter("failed to load compiled extension: %s" % exception_str)
994
1101
def report_extension_load_failures():
995
1102
if not _extension_load_failures:
997
from bzrlib.config import GlobalConfig
998
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1104
if config.GlobalConfig().suppress_warning('missing_extensions'):
1000
1106
# the warnings framework should by default show this only once
1001
from bzrlib.trace import warning
1107
from .trace import warning
1003
"bzr: warning: some compiled extensions could not be loaded; "
1004
"see <https://answers.launchpad.net/bzr/+faq/703>")
1109
"brz: warning: some compiled extensions could not be loaded; "
1110
"see ``brz help missing-extensions``")
1005
1111
# we no longer show the specific missing extensions here, because it makes
1006
1112
# the message too long and scary - see
1007
1113
# https://bugs.launchpad.net/bzr/+bug/430529
1011
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1012
except ImportError, e:
1117
from ._chunks_to_lines_pyx import chunks_to_lines
1118
except ImportError as e:
1013
1119
failed_to_load_extension(e)
1014
from bzrlib._chunks_to_lines_py import chunks_to_lines
1120
from ._chunks_to_lines_py import chunks_to_lines
1017
1123
def split_lines(s):
1018
1124
"""Split s into lines, but without removing the newline characters."""
1019
1125
# Trivially convert a fulltext into a 'chunked' representation, and let
1020
1126
# chunks_to_lines do the heavy lifting.
1021
if isinstance(s, str):
1127
if isinstance(s, bytes):
1022
1128
# chunks_to_lines only supports 8-bit strings
1023
1129
return chunks_to_lines([s])
1050
1157
os.link(src, dest)
1051
except (OSError, IOError), e:
1158
except (OSError, IOError) as e:
1052
1159
if e.errno != errno.EXDEV:
1054
1161
shutil.copyfile(src, dest)
1057
1164
def delete_any(path):
1058
"""Delete a file, symlink or directory.
1165
"""Delete a file, symlink or directory.
1060
1167
Will delete even if readonly.
1063
_delete_file_or_dir(path)
1064
except (OSError, IOError), e:
1170
_delete_file_or_dir(path)
1171
except (OSError, IOError) as e:
1065
1172
if e.errno in (errno.EPERM, errno.EACCES):
1066
1173
# make writable and try again
1285
1410
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):
1413
def safe_revision_id(unicode_or_utf8_string):
1294
1414
"""Revision ids should now be utf8, but at one point they were unicode.
1296
1416
: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
1418
:return: None or a utf8 revision id.
1301
1420
if (unicode_or_utf8_string is None
1302
or unicode_or_utf8_string.__class__ == str):
1421
or unicode_or_utf8_string.__class__ == bytes):
1303
1422
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):
1423
raise TypeError('Unicode revision ids are no longer supported. '
1424
'Revision id generators should be creating utf8 revision '
1428
def safe_file_id(unicode_or_utf8_string):
1315
1429
"""File ids should now be utf8, but at one point they were unicode.
1317
1431
This is the same as safe_utf8, except it uses the cached encode functions
1425
1547
None is returned if the width can't established precisely.
1428
- if BZR_COLUMNS is set, returns its value
1550
- if BRZ_COLUMNS is set, returns its value
1429
1551
- if there is no controlling terminal, returns None
1552
- query the OS, if the queried size has changed since the last query,
1430
1554
- if COLUMNS is set, returns its value,
1555
- if the OS has a value (even though it's never changed), return its value.
1432
1557
From there, we need to query the OS to get the size of the controlling
1560
On Unices we query the OS by:
1436
1561
- get termios.TIOCGWINSZ
1437
1562
- if an error occurs or a negative value is obtained, returns None
1564
On Windows we query the OS by:
1441
1565
- win32utils.get_console_size() decides,
1442
1566
- returns None on error (provided default value)
1568
# Note to implementors: if changing the rules for determining the width,
1569
# make sure you've considered the behaviour in these cases:
1570
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1571
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1573
# - (add more interesting cases here, if you find any)
1574
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1575
# but we don't want to register a signal handler because it is impossible
1576
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1577
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1578
# 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
1581
# If BRZ_COLUMNS is set, take it, user is always right
1582
# Except if they specified 0 in which case, impose no limit here
1447
return int(os.environ['BZR_COLUMNS'])
1584
width = int(os.environ['BRZ_COLUMNS'])
1448
1585
except (KeyError, ValueError):
1587
if width is not None:
1451
1593
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.
1594
if isatty is None or not isatty():
1595
# 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))
1599
width, height = os_size = _terminal_size(None, None)
1600
global _first_terminal_size, _terminal_size_state
1601
if _terminal_size_state == 'no_data':
1602
_first_terminal_size = os_size
1603
_terminal_size_state = 'unchanged'
1604
elif (_terminal_size_state == 'unchanged' and
1605
_first_terminal_size != os_size):
1606
_terminal_size_state = 'changed'
1608
# If the OS claims to know how wide the terminal is, and this value has
1609
# ever changed, use that.
1610
if _terminal_size_state == 'changed':
1611
if width is not None and width > 0:
1614
# If COLUMNS is set, use it.
1460
1616
return int(os.environ['COLUMNS'])
1461
1617
except (KeyError, ValueError):
1464
width, height = _terminal_size(None, None)
1466
# Consider invalid values as meaning no width
1620
# Finally, use an unchanged size from the OS, if we have one.
1621
if _terminal_size_state == 'unchanged':
1622
if width is not None and width > 0:
1625
# The width could not be determined.
1472
1629
def _win32_terminal_size(width, height):
1473
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1630
width, height = win32utils.get_console_size(
1631
defaultx=width, defaulty=height)
1474
1632
return width, height
1477
1635
def _ioctl_terminal_size(width, height):
1479
import struct, fcntl, termios
1480
1640
s = struct.pack('HHHH', 0, 0, 0, 0)
1481
1641
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1482
1642
height, width = struct.unpack('HHHH', x)[0:2]
1499
1660
_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
1663
def supports_executable():
1528
1664
return sys.platform != "win32"
1713
1850
global _selected_dir_reader
1714
1851
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
1852
if sys.platform == "win32":
1723
from bzrlib._walkdirs_win32 import Win32ReadDir
1854
from ._walkdirs_win32 import Win32ReadDir
1724
1855
_selected_dir_reader = Win32ReadDir()
1725
1856
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
1858
elif _fs_enc in ('utf-8', 'ascii'):
1730
from bzrlib._readdir_pyx import UTF8DirReader
1860
from ._readdir_pyx import UTF8DirReader
1731
1861
_selected_dir_reader = UTF8DirReader()
1732
except ImportError, e:
1862
except ImportError as e:
1733
1863
failed_to_load_extension(e)
1780
1910
See DirReader.read_dir for details.
1782
1912
_utf8_encode = self._utf8_encode
1914
def _fs_decode(s): return s.decode(_fs_enc)
1916
def _fs_encode(s): return s.encode(_fs_enc)
1783
1917
_lstat = os.lstat
1784
1918
_listdir = os.listdir
1785
1919
_kind_from_mode = file_kind_from_stat_mode
1788
relprefix = prefix + '/'
1922
relprefix = prefix + b'/'
1791
top_slash = top + u'/'
1925
top_slash = top + '/'
1794
1928
append = dirblock.append
1795
for name in sorted(_listdir(top)):
1929
for name_native in _listdir(top.encode('utf-8')):
1797
name_utf8 = _utf8_encode(name)[0]
1931
name = _fs_decode(name_native)
1798
1932
except UnicodeDecodeError:
1799
1933
raise errors.BadFilenameEncoding(
1800
_utf8_encode(relprefix)[0] + name, _fs_enc)
1934
relprefix + name_native, _fs_enc)
1935
name_utf8 = _utf8_encode(name)[0]
1801
1936
abspath = top_slash + name
1802
1937
statvalue = _lstat(abspath)
1803
1938
kind = _kind_from_mode(statvalue.st_mode)
1804
1939
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1940
return sorted(dirblock)
1808
1943
def copy_tree(from_path, to_path, handlers={}):
1875
2013
This can be used to sort paths in the same way that walkdirs does.
1877
return (dirname(path) , path)
2015
return (dirname(path), path)
1880
2018
def compare_paths_prefix_order(path_a, path_b):
1881
2019
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1882
2020
key_a = path_prefix_key(path_a)
1883
2021
key_b = path_prefix_key(path_b)
1884
return cmp(key_a, key_b)
2022
return (key_a > key_b) - (key_a < key_b)
1887
2025
_cached_user_encoding = None
1890
def get_user_encoding(use_cache=True):
2028
def get_user_encoding():
1891
2029
"""Find out what the preferred user encoding is.
1893
2031
This is generally the encoding that is used for command line parameters
1894
2032
and file contents. This may be different from the terminal encoding
1895
2033
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
2035
:return: A string defining the preferred user encoding
1903
2037
global _cached_user_encoding
1904
if _cached_user_encoding is not None and use_cache:
2038
if _cached_user_encoding is not None:
1905
2039
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'
2041
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2042
# Use the existing locale settings and call nl_langinfo directly
2043
# rather than going through getpreferredencoding. This avoids
2044
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2045
# possibility of the setlocale call throwing an error.
2046
user_encoding = locale.nl_langinfo(locale.CODESET)
2048
# GZ 2011-12-19: On windows could call GetACP directly instead.
2049
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:'
2052
user_encoding = codecs.lookup(user_encoding).name
2054
if user_encoding not in ("", "cp0"):
2055
sys.stderr.write('brz: warning:'
1951
2056
' unknown encoding %s.'
1952
2057
' Continuing with ascii encoding.\n'
1953
2058
% user_encoding
1955
user_encoding = 'ascii'
1958
_cached_user_encoding = user_encoding
2060
user_encoding = 'ascii'
2062
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2063
if user_encoding == 'ascii':
2064
if sys.platform == 'darwin':
2065
# OSX is special-cased in Python to have a UTF-8 filesystem
2066
# encoding and previously had LANG set here if not present.
2067
user_encoding = 'utf-8'
2068
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2069
# for some other posix platforms as well.
2071
_cached_user_encoding = user_encoding
1960
2072
return user_encoding
2075
def get_diff_header_encoding():
2076
return get_terminal_encoding()
1963
2079
def get_host_name():
1964
2080
"""Return the current unicode host name.
1980
2097
# data at once.
1981
2098
MAX_SOCKET_CHUNK = 64 * 1024
2100
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2101
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2102
_eno = getattr(errno, _eno, None)
2103
if _eno is not None:
2104
_end_of_stream_errors.append(_eno)
1983
2108
def read_bytes_from_socket(sock, report_activity=None,
1984
max_read_size=MAX_SOCKET_CHUNK):
2109
max_read_size=MAX_SOCKET_CHUNK):
1985
2110
"""Read up to max_read_size of bytes from sock and notify of progress.
1987
2112
Translates "Connection reset by peer" into file-like EOF (return an
1988
2113
empty string rather than raise an error), and repeats the recv if
1989
2114
interrupted by a signal.
1993
bytes = sock.recv(max_read_size)
1994
except socket.error, e:
2118
data = sock.recv(max_read_size)
2119
except socket.error as e:
1995
2120
eno = e.args[0]
1996
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2121
if eno in _end_of_stream_errors:
1997
2122
# The connection was closed by the other side. Callers expect
1998
2123
# an empty string to signal end-of-stream.
2000
2125
elif eno == errno.EINTR:
2001
2126
# Retry the interrupted recv.
2005
2130
if report_activity is not None:
2006
report_activity(len(bytes), 'read')
2131
report_activity(len(data), 'read')
2010
2135
def recv_all(socket, count):
2018
2143
This isn't optimized and is intended mostly for use in testing.
2021
2146
while len(b) < count:
2022
2147
new = read_bytes_from_socket(socket, None, count - len(b))
2029
2154
def send_all(sock, bytes, report_activity=None):
2030
2155
"""Send all bytes on a socket.
2032
2157
Breaks large blocks in smaller chunks to avoid buffering limitations on
2033
2158
some platforms, and catches EINTR which may be thrown if the send is
2034
2159
interrupted by a signal.
2036
2161
This is preferred to socket.sendall(), because it avoids portability bugs
2037
2162
and provides activity reporting.
2039
2164
:param report_activity: Call this as bytes are read, see
2040
2165
Transport._report_activity
2043
2168
byte_count = len(bytes)
2169
view = memoryview(bytes)
2044
2170
while sent_total < byte_count:
2046
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2047
except socket.error, e:
2172
sent = sock.send(view[sent_total:sent_total + MAX_SOCKET_CHUNK])
2173
except (socket.error, IOError) as e:
2174
if e.args[0] in _end_of_stream_errors:
2175
raise errors.ConnectionReset(
2176
"Error trying to write to socket", e)
2048
2177
if e.args[0] != errno.EINTR:
2181
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2051
2183
sent_total += sent
2052
report_activity(sent, 'write')
2184
if report_activity is not None:
2185
report_activity(sent, 'write')
2188
def connect_socket(address):
2189
# Slight variation of the socket.create_connection() function (provided by
2190
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2191
# provide it for previous python versions. Also, we don't use the timeout
2192
# parameter (provided by the python implementation) so we don't implement
2194
err = socket.error('getaddrinfo returns an empty list')
2195
host, port = address
2196
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2197
af, socktype, proto, canonname, sa = res
2200
sock = socket.socket(af, socktype, proto)
2204
except socket.error as e:
2206
# 'err' is now the most recent error
2207
if sock is not None:
2055
2212
def dereference_path(path):
2083
2240
If and when pkg_resources becomes a standard library, this routine
2084
2241
can delegate to it.
2086
# Check package name is within bzrlib
2087
if package == "bzrlib":
2243
# Check package name is within breezy
2244
if package == "breezy":
2088
2245
resource_relpath = resource_name
2089
elif package.startswith("bzrlib."):
2090
package = package[len("bzrlib."):].replace('.', os.sep)
2246
elif package.startswith("breezy."):
2247
package = package[len("breezy."):].replace('.', os.sep)
2091
2248
resource_relpath = pathjoin(package, resource_name)
2093
raise errors.BzrError('resource package %s not in bzrlib' % package)
2250
raise errors.BzrError('resource package %s not in breezy' % package)
2095
2252
# Map the resource to a file and read its contents
2096
base = dirname(bzrlib.__file__)
2253
base = dirname(breezy.__file__)
2097
2254
if getattr(sys, 'frozen', None): # bzr.exe
2098
2255
base = abspath(pathjoin(base, '..', '..'))
2099
filename = pathjoin(base, resource_relpath)
2100
return open(filename, 'rU').read()
2256
with open(pathjoin(base, resource_relpath), "rt") as f:
2103
2260
def file_kind_from_stat_mode_thunk(mode):
2104
2261
global file_kind_from_stat_mode
2105
2262
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2107
from bzrlib._readdir_pyx import UTF8DirReader
2264
from ._readdir_pyx import UTF8DirReader
2108
2265
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2109
except ImportError, e:
2110
2267
# This is one time where we won't warn that an extension failed to
2111
2268
# load. The extension is never available on Windows anyway.
2112
from bzrlib._readdir_py import (
2269
from ._readdir_py import (
2113
2270
_kind_from_mode as file_kind_from_stat_mode
2115
2272
return file_kind_from_stat_mode(mode)
2116
2275
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2119
def file_kind(f, _lstat=os.lstat):
2278
def file_stat(f, _lstat=os.lstat):
2121
return file_kind_from_stat_mode(_lstat(f).st_mode)
2282
except OSError as e:
2123
2283
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2124
2284
raise errors.NoSuchFile(f)
2288
def file_kind(f, _lstat=os.lstat):
2289
stat_value = file_stat(f, _lstat)
2290
return file_kind_from_stat_mode(stat_value.st_mode)
2128
2293
def until_no_eintr(f, *a, **kw):
2129
2294
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2131
2296
WARNING: you must be certain that it is safe to retry the call repeatedly
2132
2297
if EINTR does occur. This is typically only true for low-level operations
2133
2298
like os.read. If in any doubt, don't use this.
2135
2300
Keep in mind that this is not a complete solution to EINTR. There is
2136
2301
probably code in the Python standard library and other dependencies that
2137
2302
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
2303
that signal). So this function can reduce the impact for IO that breezy
2139
2304
directly controls, but it is not a complete solution.
2141
2306
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2144
2309
return f(*a, **kw)
2145
except (IOError, OSError), e:
2310
except (IOError, OSError) as e:
2146
2311
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
2316
if sys.platform == "win32":
2177
2319
return msvcrt.getch()
2182
2324
fd = sys.stdin.fileno()
2183
2325
settings = termios.tcgetattr(fd)
2188
2330
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2192
if sys.platform == 'linux2':
2333
if sys.platform.startswith('linux'):
2193
2334
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
2336
return os.sysconf('SC_NPROCESSORS_ONLN')
2337
except (ValueError, OSError, AttributeError):
2200
2339
elif sys.platform == 'darwin':
2201
2340
def _local_concurrency():
2202
2341
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2203
2342
stdout=subprocess.PIPE).communicate()[0]
2204
elif sys.platform[0:7] == 'freebsd':
2343
elif "bsd" in sys.platform:
2205
2344
def _local_concurrency():
2206
2345
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2207
2346
stdout=subprocess.PIPE).communicate()[0]
2208
2347
elif sys.platform == 'sunos5':
2209
2348
def _local_concurrency():
2210
return subprocess.Popen(['psrinfo', '-p',],
2349
return subprocess.Popen(['psrinfo', '-p', ],
2211
2350
stdout=subprocess.PIPE).communicate()[0]
2212
2351
elif sys.platform == "win32":
2213
2352
def _local_concurrency():
2232
2372
if _cached_local_concurrency is not None and use_cache:
2233
2373
return _cached_local_concurrency
2235
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2375
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2236
2376
if concurrency is None:
2377
import multiprocessing
2238
concurrency = _local_concurrency()
2239
except (OSError, IOError):
2379
concurrency = multiprocessing.cpu_count()
2380
except NotImplementedError:
2381
# multiprocessing.cpu_count() isn't implemented on all platforms
2383
concurrency = _local_concurrency()
2384
except (OSError, IOError):
2242
2387
concurrency = int(concurrency)
2243
2388
except (TypeError, ValueError):
2244
2389
concurrency = 1
2246
_cached_concurrency = concurrency
2391
_cached_local_concurrency = concurrency
2247
2392
return concurrency
2305
2451
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." % \
2454
def available_backup_name(base, exists):
2455
"""Find a non-existing backup file name.
2457
This will *not* create anything, this only return a 'free' entry. This
2458
should be used for checking names in a directory below a locked
2459
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2460
Leap) and generally discouraged.
2462
:param base: The base name.
2464
:param exists: A callable returning True if the path parameter exists.
2467
name = "%s.~%d~" % (base, counter)
2470
name = "%s.~%d~" % (base, counter)
2474
def set_fd_cloexec(fd):
2475
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2476
support for this is not available.
2480
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2481
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2482
except (ImportError, AttributeError):
2483
# Either the fcntl module or specific constants are not present
2487
def find_executable_on_path(name):
2488
"""Finds an executable on the PATH.
2490
On Windows, this will try to append each extension in the PATHEXT
2491
environment variable to the name, if it cannot be found with the name
2494
:param name: The base name of the executable.
2495
:return: The path to the executable found or None.
2497
if sys.platform == 'win32':
2498
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2499
exts = [ext.lower() for ext in exts]
2500
base, ext = os.path.splitext(name)
2502
if ext.lower() not in exts:
2508
path = os.environ.get('PATH')
2509
if path is not None:
2510
path = path.split(os.pathsep)
2513
f = os.path.join(d, name) + ext
2514
if os.access(f, os.X_OK):
2516
if sys.platform == 'win32':
2517
app_path = win32utils.get_app_path(name)
2518
if app_path != name:
2523
def _posix_is_local_pid_dead(pid):
2524
"""True if pid doesn't correspond to live process on this machine"""
2526
# Special meaning of unix kill: just check if it's there.
2528
except OSError as e:
2529
if e.errno == errno.ESRCH:
2530
# On this machine, and really not found: as sure as we can be
2533
elif e.errno == errno.EPERM:
2534
# exists, though not ours
2537
trace.mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2538
# Don't really know.
2541
# Exists and our process: not dead.
2545
if sys.platform == "win32":
2546
is_local_pid_dead = win32utils.is_local_pid_dead
2548
is_local_pid_dead = _posix_is_local_pid_dead
2550
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2551
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2552
if getattr(errno, name, None) is not None]
2555
def fdatasync(fileno):
2556
"""Flush file contents to disk if possible.
2558
:param fileno: Integer OS file handle.
2559
:raises TransportNotPossible: If flushing to disk is not possible.
2561
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2565
except IOError as e:
2566
# See bug #1075108, on some platforms fdatasync exists, but can
2567
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2568
# and reduce the chance of corruption-on-powerloss situations. It
2569
# is not a mandatory call, so it is ok to suppress failures.
2570
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2571
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2575
def ensure_empty_directory_exists(path, exception_class):
2576
"""Make sure a local directory exists and is empty.
2578
If it does not exist, it is created. If it exists and is not empty, an
2579
instance of exception_class is raised.
2583
except OSError as e:
2584
if e.errno != errno.EEXIST:
2586
if os.listdir(path) != []:
2587
raise exception_class(path)
2590
def is_environment_error(evalue):
2591
"""True if exception instance is due to a process environment issue
2593
This includes OSError and IOError, but also other errors that come from
2594
the operating system or core libraries but are not subclasses of those.
2596
if isinstance(evalue, (EnvironmentError, select.error)):
2598
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
2604
perf_counter = time.perf_counter
2606
perf_counter = time.clock
added
removed
breezy/osutils.py
