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 datetime import timedelta
30
from ntpath import (abspath as _nt_abspath,
32
normpath as _nt_normpath,
33
realpath as _nt_realpath,
34
splitdrive as _nt_splitdrive,
36
# We need to import both shutil and rmtree as we export the later on posix
37
# and need the former on windows
39
from shutil import rmtree
42
# We need to import both tempfile and mkdtemp as we export the later on posix
43
# and need the former on windows
44
from tempfile import (
45
from tempfile import mkdtemp
54
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
72
# Cross platform wall-clock time functionality with decent resolution.
100
92
def get_unicode_argv():
102
96
user_encoding = get_user_encoding()
103
97
return [a.decode(user_encoding) for a in sys.argv[1:]]
104
98
except UnicodeDecodeError:
105
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
99
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
100
"application locale.").format(a, user_encoding))
109
103
def make_readonly(filename):
110
104
"""Make a filename read-only."""
111
105
mod = os.lstat(filename).st_mode
112
106
if not stat.S_ISLNK(mod):
114
os.chmod(filename, mod)
108
chmod_if_possible(filename, mod)
117
111
def make_writable(filename):
118
112
mod = os.lstat(filename).st_mode
119
113
if not stat.S_ISLNK(mod):
121
os.chmod(filename, mod)
115
chmod_if_possible(filename, mod)
118
def chmod_if_possible(filename, mode):
119
# Set file mode if that can be safely done.
120
# Sometimes even on unix the filesystem won't allow it - see
121
# https://bugs.launchpad.net/bzr/+bug/606537
123
# It is probably faster to just do the chmod, rather than
124
# doing a stat, and then trying to compare
125
os.chmod(filename, mode)
126
except (IOError, OSError) as e:
127
# Permission/access denied seems to commonly happen on smbfs; there's
128
# probably no point warning about it.
129
# <https://bugs.launchpad.net/bzr/+bug/606537>
130
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
131
trace.mutter("ignore error on chmod of %r: %r" % (
124
137
def minimum_path_selection(paths):
233
246
file_existed = False
235
248
rename_func(new, tmp_name)
236
except (errors.NoSuchFile,), e:
249
except (errors.NoSuchFile,) as e:
239
252
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
240
253
# function raises an IOError with errno is None when a rename fails.
241
254
# This then gets caught here.
242
255
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
257
except Exception as e:
245
258
if (getattr(e, 'errno', None) is None
246
259
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
249
262
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
266
# This may throw an exception, in which case success will
268
rename_func(old, new)
270
except (IOError, OSError) as e:
271
# source and target may be aliases of each other (e.g. on a
272
# case-insensitive filesystem), so we may have accidentally renamed
273
# source by when we tried to rename target
274
if (file_existed and e.errno in (None, errno.ENOENT)
275
and old.lower() == new.lower()):
276
# source and target are the same file on a case-insensitive
277
# filesystem, so we don't generate an exception
271
283
# If the file used to exist, rename it back into place
274
286
unlink_func(tmp_name)
276
288
rename_func(tmp_name, new)
277
if failure_exc is not None:
278
raise failure_exc[0], failure_exc[1], failure_exc[2]
281
291
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
282
292
# choke on a Unicode string containing a relative path if
283
293
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
285
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
286
295
def _posix_abspath(path):
287
296
# jam 20060426 rather than encoding to fsencoding
288
297
# copy posixpath.abspath, but use os.getcwdu instead
289
298
if not posixpath.isabs(path):
290
299
path = posixpath.join(getcwd(), path)
291
return posixpath.normpath(path)
300
return _posix_normpath(path)
294
303
def _posix_realpath(path):
295
304
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
307
def _posix_normpath(path):
308
path = posixpath.normpath(path)
309
# Bug 861008: posixpath.normpath() returns a path normalized according to
310
# the POSIX standard, which stipulates (for compatibility reasons) that two
311
# leading slashes must not be simplified to one, and only if there are 3 or
312
# more should they be simplified as one. So we treat the leading 2 slashes
313
# as a special case here by simply removing the first slash, as we consider
314
# that breaking POSIX compatibility for this obscure feature is acceptable.
315
# This is not a paranoid precaution, as we notably get paths like this when
316
# the repo is hosted at the root of the filesystem, i.e. in "/".
317
if path.startswith('//'):
322
def _posix_path_from_environ(key):
323
"""Get unicode path from `key` in environment or None if not present
325
Note that posix systems use arbitrary byte strings for filesystem objects,
326
so a path that raises BadFilenameEncoding here may still be accessible.
328
val = os.environ.get(key, None)
329
if PY3 or val is None:
332
return val.decode(_fs_enc)
333
except UnicodeDecodeError:
334
# GZ 2011-12-12:Ideally want to include `key` in the exception message
335
raise errors.BadFilenameEncoding(val, _fs_enc)
338
def _posix_get_home_dir():
339
"""Get the home directory of the current user as a unicode path"""
340
path = posixpath.expanduser("~")
342
return path.decode(_fs_enc)
343
except AttributeError:
345
except UnicodeDecodeError:
346
raise errors.BadFilenameEncoding(path, _fs_enc)
349
def _posix_getuser_unicode():
350
"""Get username from environment or password database as unicode"""
351
name = getpass.getuser()
352
user_encoding = get_user_encoding()
354
return name.decode(user_encoding)
355
except UnicodeDecodeError:
356
raise errors.BzrError("Encoding of username %r is unsupported by %s "
357
"application locale." % (name, user_encoding))
298
360
def _win32_fixdrive(path):
299
361
"""Force drive letters to be consistent.
327
389
# /path => C:/path
328
390
path = unicode(path)
329
391
# check for absolute path
330
drive = _nt_splitdrive(path)[0]
392
drive = ntpath.splitdrive(path)[0]
331
393
if drive == '' and path[:2] not in('//','\\\\'):
333
395
# we cannot simply os.path.join cwd and path
334
396
# because os.path.join('C:','/path') produce '/path'
335
397
# and this is incorrect
336
398
if path[:1] in ('/','\\'):
337
cwd = _nt_splitdrive(cwd)[0]
399
cwd = ntpath.splitdrive(cwd)[0]
339
401
path = cwd + '\\' + path
340
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
402
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
343
405
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('\\', '/'))
406
# Real ntpath.realpath doesn't have a problem with a unicode cwd
407
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
348
410
def _win32_pathjoin(*args):
349
return _nt_join(*args).replace('\\', '/')
411
return ntpath.join(*args).replace('\\', '/')
352
414
def _win32_normpath(path):
353
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
415
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
356
418
def _win32_getcwd():
357
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
419
return _win32_fixdrive(_getcwd().replace('\\', '/'))
360
422
def _win32_mkdtemp(*args, **kwargs):
361
423
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
426
def _win32_rename(old, new):
373
427
"""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
444
def _mac_getcwd():
401
return unicodedata.normalize('NFC', os.getcwdu())
445
return unicodedata.normalize('NFC', _getcwd())
448
def _rename_wrap_exception(rename_func):
449
"""Adds extra information to any exceptions that come from rename().
451
The exception has an updated message and 'old_filename' and 'new_filename'
455
def _rename_wrapper(old, new):
457
rename_func(old, new)
459
detailed_error = OSError(e.errno, e.strerror +
460
" [occurred when renaming '%s' to '%s']" %
462
detailed_error.old_filename = old
463
detailed_error.new_filename = new
466
return _rename_wrapper
469
if sys.version_info > (3,):
475
# Default rename wraps os.rename()
476
rename = _rename_wrap_exception(os.rename)
404
478
# Default is to just use the python builtins, but these can be rebound on
405
479
# particular platforms.
406
480
abspath = _posix_abspath
407
481
realpath = _posix_realpath
408
482
pathjoin = os.path.join
409
normpath = os.path.normpath
410
rename = _wrapped_rename # overridden below on win32
483
normpath = _posix_normpath
484
path_from_environ = _posix_path_from_environ
485
_get_home_dir = _posix_get_home_dir
486
getuser_unicode = _posix_getuser_unicode
412
488
dirname = os.path.dirname
413
489
basename = os.path.basename
414
490
split = os.path.split
415
491
splitext = os.path.splitext
416
# These were already imported into local scope
492
# These were already lazily imported into local scope
417
493
# mkdtemp = tempfile.mkdtemp
418
494
# rmtree = shutil.rmtree
420
502
MIN_ABS_PATHLENGTH = 1
472
565
On my standard US Windows XP, the preferred encoding is
473
566
cp1252, but the console is cp437
568
:param trace: If True trace the selected encoding via mutter().
475
from bzrlib.trace import mutter
570
from .trace import mutter
476
571
output_encoding = getattr(sys.stdout, 'encoding', None)
477
572
if not output_encoding:
478
573
input_encoding = getattr(sys.stdin, 'encoding', None)
479
574
if not input_encoding:
480
575
output_encoding = get_user_encoding()
481
mutter('encoding stdout as osutils.get_user_encoding() %r',
577
mutter('encoding stdout as osutils.get_user_encoding() %r',
484
580
output_encoding = input_encoding
485
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
582
mutter('encoding stdout as sys.stdin encoding %r',
487
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
586
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
488
587
if output_encoding == 'cp0':
489
588
# invalid encoding (cp0 means 'no codepage' on Windows)
490
589
output_encoding = get_user_encoding()
491
mutter('cp0 is invalid encoding.'
591
mutter('cp0 is invalid encoding.'
492
592
' encoding stdout as osutils.get_user_encoding() %r',
496
596
codecs.lookup(output_encoding)
497
597
except LookupError:
498
sys.stderr.write('bzr: warning:'
598
sys.stderr.write('brz: warning:'
499
599
' unknown terminal encoding %s.\n'
500
600
' Using encoding %s instead.\n'
501
601
% (output_encoding, get_user_encoding())
799
910
(date_fmt, tt, offset_str) = \
800
911
_format_date(t, offset, timezone, date_fmt, show_offset)
801
912
date_str = time.strftime(date_fmt, tt)
802
if not isinstance(date_str, unicode):
913
if not isinstance(date_str, text_type):
803
914
date_str = date_str.decode(get_user_encoding(), 'replace')
804
915
return date_str + offset_str
807
918
def _format_date(t, offset, timezone, date_fmt, show_offset):
808
919
if timezone == 'utc':
811
922
elif timezone == 'original':
812
923
if offset is None:
814
tt = time.gmtime(t + offset)
925
tt = gmtime(t + offset)
815
926
elif timezone == 'local':
816
927
tt = time.localtime(t)
817
928
offset = local_time_offset(t)
883
994
"""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
995
return os.fstat(f.fileno())[stat.ST_SIZE]
998
# Alias os.urandom to support platforms (which?) without /dev/urandom and
999
# override if it doesn't work. Avoid checking on windows where there is
1000
# significant initialisation cost that can be avoided for some bzr calls.
1002
rand_bytes = os.urandom
1004
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):
1007
except NotImplementedError:
900
1008
# not well seeded, but better than nothing
901
1009
def rand_bytes(n):
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)
1285
1407
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):
1410
def safe_revision_id(unicode_or_utf8_string):
1294
1411
"""Revision ids should now be utf8, but at one point they were unicode.
1296
1413
: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
1415
:return: None or a utf8 revision id.
1301
1417
if (unicode_or_utf8_string is None
1302
1418
or unicode_or_utf8_string.__class__ == str):
1303
1419
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):
1420
raise TypeError('Unicode revision ids are no longer supported. '
1421
'Revision id generators should be creating utf8 revision '
1425
def safe_file_id(unicode_or_utf8_string):
1315
1426
"""File ids should now be utf8, but at one point they were unicode.
1317
1428
This is the same as safe_utf8, except it uses the cached encode functions
1425
1539
None is returned if the width can't established precisely.
1428
- if BZR_COLUMNS is set, returns its value
1542
- if BRZ_COLUMNS is set, returns its value
1429
1543
- if there is no controlling terminal, returns None
1544
- query the OS, if the queried size has changed since the last query,
1430
1546
- if COLUMNS is set, returns its value,
1547
- if the OS has a value (even though it's never changed), return its value.
1432
1549
From there, we need to query the OS to get the size of the controlling
1552
On Unices we query the OS by:
1436
1553
- get termios.TIOCGWINSZ
1437
1554
- if an error occurs or a negative value is obtained, returns None
1556
On Windows we query the OS by:
1441
1557
- win32utils.get_console_size() decides,
1442
1558
- returns None on error (provided default value)
1560
# Note to implementors: if changing the rules for determining the width,
1561
# make sure you've considered the behaviour in these cases:
1562
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1563
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1565
# - (add more interesting cases here, if you find any)
1566
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1567
# but we don't want to register a signal handler because it is impossible
1568
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1569
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1570
# 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
1573
# If BRZ_COLUMNS is set, take it, user is always right
1574
# Except if they specified 0 in which case, impose no limit here
1447
return int(os.environ['BZR_COLUMNS'])
1576
width = int(os.environ['BRZ_COLUMNS'])
1448
1577
except (KeyError, ValueError):
1579
if width is not None:
1451
1585
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.
1586
if isatty is None or not isatty():
1587
# 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))
1591
width, height = os_size = _terminal_size(None, None)
1592
global _first_terminal_size, _terminal_size_state
1593
if _terminal_size_state == 'no_data':
1594
_first_terminal_size = os_size
1595
_terminal_size_state = 'unchanged'
1596
elif (_terminal_size_state == 'unchanged' and
1597
_first_terminal_size != os_size):
1598
_terminal_size_state = 'changed'
1600
# If the OS claims to know how wide the terminal is, and this value has
1601
# ever changed, use that.
1602
if _terminal_size_state == 'changed':
1603
if width is not None and width > 0:
1606
# If COLUMNS is set, use it.
1460
1608
return int(os.environ['COLUMNS'])
1461
1609
except (KeyError, ValueError):
1464
width, height = _terminal_size(None, None)
1466
# Consider invalid values as meaning no width
1612
# Finally, use an unchanged size from the OS, if we have one.
1613
if _terminal_size_state == 'unchanged':
1614
if width is not None and width > 0:
1617
# The width could not be determined.
1472
1621
def _win32_terminal_size(width, height):
1499
1648
_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
1651
def supports_executable():
1528
1652
return sys.platform != "win32"
1887
2012
_cached_user_encoding = None
1890
def get_user_encoding(use_cache=True):
2015
def get_user_encoding():
1891
2016
"""Find out what the preferred user encoding is.
1893
2018
This is generally the encoding that is used for command line parameters
1894
2019
and file contents. This may be different from the terminal encoding
1895
2020
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
2022
:return: A string defining the preferred user encoding
1903
2024
global _cached_user_encoding
1904
if _cached_user_encoding is not None and use_cache:
2025
if _cached_user_encoding is not None:
1905
2026
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'
2028
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2029
# Use the existing locale settings and call nl_langinfo directly
2030
# rather than going through getpreferredencoding. This avoids
2031
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2032
# possibility of the setlocale call throwing an error.
2033
user_encoding = locale.nl_langinfo(locale.CODESET)
2035
# GZ 2011-12-19: On windows could call GetACP directly instead.
2036
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:'
2039
user_encoding = codecs.lookup(user_encoding).name
2041
if user_encoding not in ("", "cp0"):
2042
sys.stderr.write('brz: warning:'
1951
2043
' unknown encoding %s.'
1952
2044
' Continuing with ascii encoding.\n'
1953
2045
% user_encoding
1955
user_encoding = 'ascii'
1958
_cached_user_encoding = user_encoding
2047
user_encoding = 'ascii'
2049
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2050
if user_encoding == 'ascii':
2051
if sys.platform == 'darwin':
2052
# OSX is special-cased in Python to have a UTF-8 filesystem
2053
# encoding and previously had LANG set here if not present.
2054
user_encoding = 'utf-8'
2055
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2056
# for some other posix platforms as well.
2058
_cached_user_encoding = user_encoding
1960
2059
return user_encoding
2062
def get_diff_header_encoding():
2063
return get_terminal_encoding()
1963
2066
def get_host_name():
1964
2067
"""Return the current unicode host name.
2029
2139
def send_all(sock, bytes, report_activity=None):
2030
2140
"""Send all bytes on a socket.
2032
2142
Breaks large blocks in smaller chunks to avoid buffering limitations on
2033
2143
some platforms, and catches EINTR which may be thrown if the send is
2034
2144
interrupted by a signal.
2036
2146
This is preferred to socket.sendall(), because it avoids portability bugs
2037
2147
and provides activity reporting.
2039
2149
:param report_activity: Call this as bytes are read, see
2040
2150
Transport._report_activity
2043
2153
byte_count = len(bytes)
2154
view = memoryview(bytes)
2044
2155
while sent_total < byte_count:
2046
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2047
except socket.error, e:
2157
sent = sock.send(view[sent_total:sent_total+MAX_SOCKET_CHUNK])
2158
except (socket.error, IOError) as e:
2159
if e.args[0] in _end_of_stream_errors:
2160
raise errors.ConnectionReset(
2161
"Error trying to write to socket", e)
2048
2162
if e.args[0] != errno.EINTR:
2166
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2051
2168
sent_total += sent
2052
report_activity(sent, 'write')
2169
if report_activity is not None:
2170
report_activity(sent, 'write')
2173
def connect_socket(address):
2174
# Slight variation of the socket.create_connection() function (provided by
2175
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2176
# provide it for previous python versions. Also, we don't use the timeout
2177
# parameter (provided by the python implementation) so we don't implement
2179
err = socket.error('getaddrinfo returns an empty list')
2180
host, port = address
2181
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2182
af, socktype, proto, canonname, sa = res
2185
sock = socket.socket(af, socktype, proto)
2189
except socket.error as err:
2190
# 'err' is now the most recent error
2191
if sock is not None:
2055
2196
def dereference_path(path):
2083
2224
If and when pkg_resources becomes a standard library, this routine
2084
2225
can delegate to it.
2086
# Check package name is within bzrlib
2087
if package == "bzrlib":
2227
# Check package name is within breezy
2228
if package == "breezy":
2088
2229
resource_relpath = resource_name
2089
elif package.startswith("bzrlib."):
2090
package = package[len("bzrlib."):].replace('.', os.sep)
2230
elif package.startswith("breezy."):
2231
package = package[len("breezy."):].replace('.', os.sep)
2091
2232
resource_relpath = pathjoin(package, resource_name)
2093
raise errors.BzrError('resource package %s not in bzrlib' % package)
2234
raise errors.BzrError('resource package %s not in breezy' % package)
2095
2236
# Map the resource to a file and read its contents
2096
base = dirname(bzrlib.__file__)
2237
base = dirname(breezy.__file__)
2097
2238
if getattr(sys, 'frozen', None): # bzr.exe
2098
2239
base = abspath(pathjoin(base, '..', '..'))
2099
filename = pathjoin(base, resource_relpath)
2100
return open(filename, 'rU').read()
2240
f = file(pathjoin(base, resource_relpath), "rU")
2103
2246
def file_kind_from_stat_mode_thunk(mode):
2104
2247
global file_kind_from_stat_mode
2105
2248
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2107
from bzrlib._readdir_pyx import UTF8DirReader
2250
from ._readdir_pyx import UTF8DirReader
2108
2251
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2109
except ImportError, e:
2252
except ImportError as e:
2110
2253
# This is one time where we won't warn that an extension failed to
2111
2254
# load. The extension is never available on Windows anyway.
2112
from bzrlib._readdir_py import (
2255
from ._readdir_py import (
2113
2256
_kind_from_mode as file_kind_from_stat_mode
2115
2258
return file_kind_from_stat_mode(mode)
2116
2259
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2119
def file_kind(f, _lstat=os.lstat):
2261
def file_stat(f, _lstat=os.lstat):
2121
return file_kind_from_stat_mode(_lstat(f).st_mode)
2265
except OSError as e:
2123
2266
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2124
2267
raise errors.NoSuchFile(f)
2270
def file_kind(f, _lstat=os.lstat):
2271
stat_value = file_stat(f, _lstat)
2272
return file_kind_from_stat_mode(stat_value.st_mode)
2128
2274
def until_no_eintr(f, *a, **kw):
2129
2275
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2131
2277
WARNING: you must be certain that it is safe to retry the call repeatedly
2132
2278
if EINTR does occur. This is typically only true for low-level operations
2133
2279
like os.read. If in any doubt, don't use this.
2135
2281
Keep in mind that this is not a complete solution to EINTR. There is
2136
2282
probably code in the Python standard library and other dependencies that
2137
2283
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
2284
that signal). So this function can reduce the impact for IO that breezy
2139
2285
directly controls, but it is not a complete solution.
2141
2287
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2144
2290
return f(*a, **kw)
2145
except (IOError, OSError), e:
2291
except (IOError, OSError) as e:
2146
2292
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
2297
if sys.platform == "win32":
2177
2300
return msvcrt.getch()
2182
2305
fd = sys.stdin.fileno()
2183
2306
settings = termios.tcgetattr(fd)
2188
2311
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2192
if sys.platform == 'linux2':
2314
if sys.platform.startswith('linux'):
2193
2315
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
2317
return os.sysconf('SC_NPROCESSORS_ONLN')
2318
except (ValueError, OSError, AttributeError):
2200
2320
elif sys.platform == 'darwin':
2201
2321
def _local_concurrency():
2202
2322
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2203
2323
stdout=subprocess.PIPE).communicate()[0]
2204
elif sys.platform[0:7] == 'freebsd':
2324
elif "bsd" in sys.platform:
2205
2325
def _local_concurrency():
2206
2326
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2207
2327
stdout=subprocess.PIPE).communicate()[0]
2232
2352
if _cached_local_concurrency is not None and use_cache:
2233
2353
return _cached_local_concurrency
2235
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2355
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2236
2356
if concurrency is None:
2238
concurrency = _local_concurrency()
2239
except (OSError, IOError):
2358
import multiprocessing
2359
concurrency = multiprocessing.cpu_count()
2360
except (ImportError, NotImplementedError):
2361
# multiprocessing is only available on Python >= 2.6
2362
# and multiprocessing.cpu_count() isn't implemented on all
2365
concurrency = _local_concurrency()
2366
except (OSError, IOError):
2242
2369
concurrency = int(concurrency)
2243
2370
except (TypeError, ValueError):
2305
2432
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." % \
2435
def available_backup_name(base, exists):
2436
"""Find a non-existing backup file name.
2438
This will *not* create anything, this only return a 'free' entry. This
2439
should be used for checking names in a directory below a locked
2440
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2441
Leap) and generally discouraged.
2443
:param base: The base name.
2445
:param exists: A callable returning True if the path parameter exists.
2448
name = "%s.~%d~" % (base, counter)
2451
name = "%s.~%d~" % (base, counter)
2455
def set_fd_cloexec(fd):
2456
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2457
support for this is not available.
2461
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2462
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2463
except (ImportError, AttributeError):
2464
# Either the fcntl module or specific constants are not present
2468
def find_executable_on_path(name):
2469
"""Finds an executable on the PATH.
2471
On Windows, this will try to append each extension in the PATHEXT
2472
environment variable to the name, if it cannot be found with the name
2475
:param name: The base name of the executable.
2476
:return: The path to the executable found or None.
2478
if sys.platform == 'win32':
2479
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2480
exts = [ext.lower() for ext in exts]
2481
base, ext = os.path.splitext(name)
2483
if ext.lower() not in exts:
2489
path = os.environ.get('PATH')
2490
if path is not None:
2491
path = path.split(os.pathsep)
2494
f = os.path.join(d, name) + ext
2495
if os.access(f, os.X_OK):
2497
if sys.platform == 'win32':
2498
app_path = win32utils.get_app_path(name)
2499
if app_path != name:
2504
def _posix_is_local_pid_dead(pid):
2505
"""True if pid doesn't correspond to live process on this machine"""
2507
# Special meaning of unix kill: just check if it's there.
2509
except OSError as e:
2510
if e.errno == errno.ESRCH:
2511
# On this machine, and really not found: as sure as we can be
2514
elif e.errno == errno.EPERM:
2515
# exists, though not ours
2518
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2519
# Don't really know.
2522
# Exists and our process: not dead.
2525
if sys.platform == "win32":
2526
is_local_pid_dead = win32utils.is_local_pid_dead
2528
is_local_pid_dead = _posix_is_local_pid_dead
2530
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2531
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2532
if getattr(errno, name, None) is not None]
2535
def fdatasync(fileno):
2536
"""Flush file contents to disk if possible.
2538
:param fileno: Integer OS file handle.
2539
:raises TransportNotPossible: If flushing to disk is not possible.
2541
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2545
except IOError as e:
2546
# See bug #1075108, on some platforms fdatasync exists, but can
2547
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2548
# and reduce the chance of corruption-on-powerloss situations. It
2549
# is not a mandatory call, so it is ok to suppress failures.
2550
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2551
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2555
def ensure_empty_directory_exists(path, exception_class):
2556
"""Make sure a local directory exists and is empty.
2558
If it does not exist, it is created. If it exists and is not empty, an
2559
instance of exception_class is raised.
2563
except OSError as e:
2564
if e.errno != errno.EEXIST:
2566
if os.listdir(path) != []:
2567
raise exception_class(path)
2570
def is_environment_error(evalue):
2571
"""True if exception instance is due to a process environment issue
2573
This includes OSError and IOError, but also other errors that come from
2574
the operating system or core libraries but are not subclasses of those.
2576
if isinstance(evalue, (EnvironmentError, select.error)):
2578
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):