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,
30
from datetime import timedelta
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
43
from tempfile import (
45
from tempfile import mkdtemp
54
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
72
# Cross platform wall-clock time functionality with decent resolution.
99
92
def get_unicode_argv():
101
96
user_encoding = get_user_encoding()
102
97
return [a.decode(user_encoding) for a in sys.argv[1:]]
103
98
except UnicodeDecodeError:
104
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))
108
103
def make_readonly(filename):
109
104
"""Make a filename read-only."""
110
105
mod = os.lstat(filename).st_mode
111
106
if not stat.S_ISLNK(mod):
113
os.chmod(filename, mod)
108
chmod_if_possible(filename, mod)
116
111
def make_writable(filename):
117
112
mod = os.lstat(filename).st_mode
118
113
if not stat.S_ISLNK(mod):
120
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" % (
123
137
def minimum_path_selection(paths):
232
246
file_existed = False
234
248
rename_func(new, tmp_name)
235
except (errors.NoSuchFile,), e:
249
except (errors.NoSuchFile,) as e:
238
252
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
239
253
# function raises an IOError with errno is None when a rename fails.
240
254
# This then gets caught here.
241
255
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
257
except Exception as e:
244
258
if (getattr(e, 'errno', None) is None
245
259
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
248
262
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
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
270
283
# If the file used to exist, rename it back into place
273
286
unlink_func(tmp_name)
275
288
rename_func(tmp_name, new)
276
if failure_exc is not None:
277
raise failure_exc[0], failure_exc[1], failure_exc[2]
280
291
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
281
292
# choke on a Unicode string containing a relative path if
282
293
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
284
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
285
295
def _posix_abspath(path):
286
296
# jam 20060426 rather than encoding to fsencoding
287
297
# copy posixpath.abspath, but use os.getcwdu instead
288
298
if not posixpath.isabs(path):
289
299
path = posixpath.join(getcwd(), path)
290
return posixpath.normpath(path)
300
return _posix_normpath(path)
293
303
def _posix_realpath(path):
294
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))
297
360
def _win32_fixdrive(path):
298
361
"""Force drive letters to be consistent.
303
366
running python.exe under cmd.exe return capital C:\\
304
367
running win32 python inside a cygwin shell returns lowercase c:\\
306
drive, path = _nt_splitdrive(path)
369
drive, path = ntpath.splitdrive(path)
307
370
return drive.upper() + path
310
373
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('\\', '/'))
374
# Real ntpath.abspath doesn't have a problem with a unicode cwd
375
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
342
378
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('\\', '/'))
379
# Real ntpath.realpath doesn't have a problem with a unicode cwd
380
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
347
383
def _win32_pathjoin(*args):
348
return _nt_join(*args).replace('\\', '/')
384
return ntpath.join(*args).replace('\\', '/')
351
387
def _win32_normpath(path):
352
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
388
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
355
391
def _win32_getcwd():
356
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
392
return _win32_fixdrive(_getcwd().replace('\\', '/'))
359
395
def _win32_mkdtemp(*args, **kwargs):
360
396
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
399
def _win32_rename(old, new):
372
400
"""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
417
def _mac_getcwd():
400
return unicodedata.normalize('NFC', os.getcwdu())
418
return unicodedata.normalize('NFC', _getcwd())
421
def _rename_wrap_exception(rename_func):
422
"""Adds extra information to any exceptions that come from rename().
424
The exception has an updated message and 'old_filename' and 'new_filename'
428
def _rename_wrapper(old, new):
430
rename_func(old, new)
432
detailed_error = OSError(e.errno, e.strerror +
433
" [occurred when renaming '%s' to '%s']" %
435
detailed_error.old_filename = old
436
detailed_error.new_filename = new
439
return _rename_wrapper
442
if sys.version_info > (3,):
448
# Default rename wraps os.rename()
449
rename = _rename_wrap_exception(os.rename)
403
451
# Default is to just use the python builtins, but these can be rebound on
404
452
# particular platforms.
405
453
abspath = _posix_abspath
406
454
realpath = _posix_realpath
407
455
pathjoin = os.path.join
408
normpath = os.path.normpath
409
rename = _wrapped_rename # overridden below on win32
456
normpath = _posix_normpath
457
path_from_environ = _posix_path_from_environ
458
_get_home_dir = _posix_get_home_dir
459
getuser_unicode = _posix_getuser_unicode
411
461
dirname = os.path.dirname
412
462
basename = os.path.basename
413
463
split = os.path.split
414
464
splitext = os.path.splitext
415
# These were already imported into local scope
465
# These were already lazily imported into local scope
416
466
# mkdtemp = tempfile.mkdtemp
417
467
# rmtree = shutil.rmtree
419
475
MIN_ABS_PATHLENGTH = 1
422
478
if sys.platform == 'win32':
423
if win32utils.winver == 'Windows 98':
424
abspath = _win98_abspath
426
abspath = _win32_abspath
479
abspath = _win32_abspath
427
480
realpath = _win32_realpath
428
481
pathjoin = _win32_pathjoin
429
482
normpath = _win32_normpath
430
483
getcwd = _win32_getcwd
431
484
mkdtemp = _win32_mkdtemp
432
rename = _win32_rename
485
rename = _rename_wrap_exception(_win32_rename)
487
from . import _walkdirs_win32
491
lstat = _walkdirs_win32.lstat
492
fstat = _walkdirs_win32.fstat
493
wrap_stat = _walkdirs_win32.wrap_stat
434
495
MIN_ABS_PATHLENGTH = 3
471
535
On my standard US Windows XP, the preferred encoding is
472
536
cp1252, but the console is cp437
538
:param trace: If True trace the selected encoding via mutter().
474
from bzrlib.trace import mutter
540
from .trace import mutter
475
541
output_encoding = getattr(sys.stdout, 'encoding', None)
476
542
if not output_encoding:
477
543
input_encoding = getattr(sys.stdin, 'encoding', None)
478
544
if not input_encoding:
479
545
output_encoding = get_user_encoding()
480
mutter('encoding stdout as osutils.get_user_encoding() %r',
547
mutter('encoding stdout as osutils.get_user_encoding() %r',
483
550
output_encoding = input_encoding
484
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
552
mutter('encoding stdout as sys.stdin encoding %r',
486
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
556
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
487
557
if output_encoding == 'cp0':
488
558
# invalid encoding (cp0 means 'no codepage' on Windows)
489
559
output_encoding = get_user_encoding()
490
mutter('cp0 is invalid encoding.'
561
mutter('cp0 is invalid encoding.'
491
562
' encoding stdout as osutils.get_user_encoding() %r',
495
566
codecs.lookup(output_encoding)
496
567
except LookupError:
497
sys.stderr.write('bzr: warning:'
568
sys.stderr.write('brz: warning:'
498
569
' unknown terminal encoding %s.\n'
499
570
' Using encoding %s instead.\n'
500
571
% (output_encoding, get_user_encoding())
798
880
(date_fmt, tt, offset_str) = \
799
881
_format_date(t, offset, timezone, date_fmt, show_offset)
800
882
date_str = time.strftime(date_fmt, tt)
801
if not isinstance(date_str, unicode):
883
if not isinstance(date_str, text_type):
802
884
date_str = date_str.decode(get_user_encoding(), 'replace')
803
885
return date_str + offset_str
806
888
def _format_date(t, offset, timezone, date_fmt, show_offset):
807
889
if timezone == 'utc':
810
892
elif timezone == 'original':
811
893
if offset is None:
813
tt = time.gmtime(t + offset)
895
tt = gmtime(t + offset)
814
896
elif timezone == 'local':
815
897
tt = time.localtime(t)
816
898
offset = local_time_offset(t)
882
964
"""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
965
return os.fstat(f.fileno())[stat.ST_SIZE]
968
# Alias os.urandom to support platforms (which?) without /dev/urandom and
969
# override if it doesn't work. Avoid checking on windows where there is
970
# significant initialisation cost that can be avoided for some bzr calls.
972
rand_bytes = os.urandom
974
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):
977
except NotImplementedError:
899
978
# not well seeded, but better than nothing
900
979
def rand_bytes(n):
970
1049
implementation should be loaded instead::
973
>>> import bzrlib._fictional_extension_pyx
1052
>>> import breezy._fictional_extension_pyx
974
1053
>>> except ImportError, e:
975
>>> bzrlib.osutils.failed_to_load_extension(e)
976
>>> import bzrlib._fictional_extension_py
1054
>>> breezy.osutils.failed_to_load_extension(e)
1055
>>> import breezy._fictional_extension_py
978
1057
# NB: This docstring is just an example, not a doctest, because doctest
979
1058
# currently can't cope with the use of lazy imports in this namespace --
982
1061
# This currently doesn't report the failure at the time it occurs, because
983
1062
# they tend to happen very early in startup when we can't check config
984
1063
# files etc, and also we want to report all failures but not spam the user
985
1064
# with 10 warnings.
986
from bzrlib import trace
987
1065
exception_str = str(exception)
988
1066
if exception_str not in _extension_load_failures:
989
1067
trace.mutter("failed to load compiled extension: %s" % exception_str)
1049
1126
os.link(src, dest)
1050
except (OSError, IOError), e:
1127
except (OSError, IOError) as e:
1051
1128
if e.errno != errno.EXDEV:
1053
1130
shutil.copyfile(src, dest)
1056
1133
def delete_any(path):
1057
"""Delete a file, symlink or directory.
1134
"""Delete a file, symlink or directory.
1059
1136
Will delete even if readonly.
1062
1139
_delete_file_or_dir(path)
1063
except (OSError, IOError), e:
1140
except (OSError, IOError) as e:
1064
1141
if e.errno in (errno.EPERM, errno.EACCES):
1065
1142
# make writable and try again
1250
1327
# but for now, we haven't optimized...
1251
1328
return [canonical_relpath(base, p) for p in paths]
1331
def decode_filename(filename):
1332
"""Decode the filename using the filesystem encoding
1334
If it is unicode, it is returned.
1335
Otherwise it is decoded from the the filesystem's encoding. If decoding
1336
fails, a errors.BadFilenameEncoding exception is raised.
1338
if isinstance(filename, text_type):
1341
return filename.decode(_fs_enc)
1342
except UnicodeDecodeError:
1343
raise errors.BadFilenameEncoding(filename, _fs_enc)
1253
1346
def safe_unicode(unicode_or_utf8_string):
1254
1347
"""Coerce unicode_or_utf8_string into unicode.
1284
1377
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):
1380
def safe_revision_id(unicode_or_utf8_string):
1293
1381
"""Revision ids should now be utf8, but at one point they were unicode.
1295
1383
: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
1385
:return: None or a utf8 revision id.
1300
1387
if (unicode_or_utf8_string is None
1301
1388
or unicode_or_utf8_string.__class__ == str):
1302
1389
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):
1390
raise TypeError('Unicode revision ids are no longer supported. '
1391
'Revision id generators should be creating utf8 revision '
1395
def safe_file_id(unicode_or_utf8_string):
1314
1396
"""File ids should now be utf8, but at one point they were unicode.
1316
1398
This is the same as safe_utf8, except it uses the cached encode functions
1424
1509
None is returned if the width can't established precisely.
1427
- if BZR_COLUMNS is set, returns its value
1512
- if BRZ_COLUMNS is set, returns its value
1428
1513
- if there is no controlling terminal, returns None
1514
- query the OS, if the queried size has changed since the last query,
1429
1516
- if COLUMNS is set, returns its value,
1517
- if the OS has a value (even though it's never changed), return its value.
1431
1519
From there, we need to query the OS to get the size of the controlling
1522
On Unices we query the OS by:
1435
1523
- get termios.TIOCGWINSZ
1436
1524
- if an error occurs or a negative value is obtained, returns None
1526
On Windows we query the OS by:
1440
1527
- win32utils.get_console_size() decides,
1441
1528
- returns None on error (provided default value)
1530
# Note to implementors: if changing the rules for determining the width,
1531
# make sure you've considered the behaviour in these cases:
1532
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1533
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1535
# - (add more interesting cases here, if you find any)
1536
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1537
# but we don't want to register a signal handler because it is impossible
1538
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1539
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1540
# 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
1543
# If BRZ_COLUMNS is set, take it, user is always right
1544
# Except if they specified 0 in which case, impose no limit here
1446
return int(os.environ['BZR_COLUMNS'])
1546
width = int(os.environ['BRZ_COLUMNS'])
1447
1547
except (KeyError, ValueError):
1549
if width is not None:
1450
1555
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.
1556
if isatty is None or not isatty():
1557
# 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))
1561
width, height = os_size = _terminal_size(None, None)
1562
global _first_terminal_size, _terminal_size_state
1563
if _terminal_size_state == 'no_data':
1564
_first_terminal_size = os_size
1565
_terminal_size_state = 'unchanged'
1566
elif (_terminal_size_state == 'unchanged' and
1567
_first_terminal_size != os_size):
1568
_terminal_size_state = 'changed'
1570
# If the OS claims to know how wide the terminal is, and this value has
1571
# ever changed, use that.
1572
if _terminal_size_state == 'changed':
1573
if width is not None and width > 0:
1576
# If COLUMNS is set, use it.
1459
1578
return int(os.environ['COLUMNS'])
1460
1579
except (KeyError, ValueError):
1463
width, height = _terminal_size(None, None)
1465
# Consider invalid values as meaning no width
1582
# Finally, use an unchanged size from the OS, if we have one.
1583
if _terminal_size_state == 'unchanged':
1584
if width is not None and width > 0:
1587
# The width could not be determined.
1471
1591
def _win32_terminal_size(width, height):
1498
1618
_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
1621
def supports_executable():
1527
1622
return sys.platform != "win32"
1712
1807
global _selected_dir_reader
1713
1808
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
1809
if sys.platform == "win32":
1722
from bzrlib._walkdirs_win32 import Win32ReadDir
1811
from ._walkdirs_win32 import Win32ReadDir
1723
1812
_selected_dir_reader = Win32ReadDir()
1724
1813
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
1815
elif _fs_enc in ('utf-8', 'ascii'):
1729
from bzrlib._readdir_pyx import UTF8DirReader
1817
from ._readdir_pyx import UTF8DirReader
1730
1818
_selected_dir_reader = UTF8DirReader()
1731
except ImportError, e:
1819
except ImportError as e:
1732
1820
failed_to_load_extension(e)
1886
1977
_cached_user_encoding = None
1889
def get_user_encoding(use_cache=True):
1980
def get_user_encoding():
1890
1981
"""Find out what the preferred user encoding is.
1892
1983
This is generally the encoding that is used for command line parameters
1893
1984
and file contents. This may be different from the terminal encoding
1894
1985
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
1987
:return: A string defining the preferred user encoding
1902
1989
global _cached_user_encoding
1903
if _cached_user_encoding is not None and use_cache:
1990
if _cached_user_encoding is not None:
1904
1991
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'
1993
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
1994
# Use the existing locale settings and call nl_langinfo directly
1995
# rather than going through getpreferredencoding. This avoids
1996
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
1997
# possibility of the setlocale call throwing an error.
1998
user_encoding = locale.nl_langinfo(locale.CODESET)
2000
# GZ 2011-12-19: On windows could call GetACP directly instead.
2001
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:'
2004
user_encoding = codecs.lookup(user_encoding).name
2006
if user_encoding not in ("", "cp0"):
2007
sys.stderr.write('brz: warning:'
1950
2008
' unknown encoding %s.'
1951
2009
' Continuing with ascii encoding.\n'
1952
2010
% user_encoding
1954
user_encoding = 'ascii'
1957
_cached_user_encoding = user_encoding
2012
user_encoding = 'ascii'
2014
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2015
if user_encoding == 'ascii':
2016
if sys.platform == 'darwin':
2017
# OSX is special-cased in Python to have a UTF-8 filesystem
2018
# encoding and previously had LANG set here if not present.
2019
user_encoding = 'utf-8'
2020
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2021
# for some other posix platforms as well.
2023
_cached_user_encoding = user_encoding
1959
2024
return user_encoding
2027
def get_diff_header_encoding():
2028
return get_terminal_encoding()
1962
2031
def get_host_name():
1963
2032
"""Return the current unicode host name.
2028
2104
def send_all(sock, bytes, report_activity=None):
2029
2105
"""Send all bytes on a socket.
2031
2107
Breaks large blocks in smaller chunks to avoid buffering limitations on
2032
2108
some platforms, and catches EINTR which may be thrown if the send is
2033
2109
interrupted by a signal.
2035
2111
This is preferred to socket.sendall(), because it avoids portability bugs
2036
2112
and provides activity reporting.
2038
2114
:param report_activity: Call this as bytes are read, see
2039
2115
Transport._report_activity
2042
2118
byte_count = len(bytes)
2119
view = memoryview(bytes)
2043
2120
while sent_total < byte_count:
2045
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2046
except socket.error, e:
2122
sent = sock.send(view[sent_total:sent_total+MAX_SOCKET_CHUNK])
2123
except (socket.error, IOError) as e:
2124
if e.args[0] in _end_of_stream_errors:
2125
raise errors.ConnectionReset(
2126
"Error trying to write to socket", e)
2047
2127
if e.args[0] != errno.EINTR:
2131
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2050
2133
sent_total += sent
2051
report_activity(sent, 'write')
2134
if report_activity is not None:
2135
report_activity(sent, 'write')
2138
def connect_socket(address):
2139
# Slight variation of the socket.create_connection() function (provided by
2140
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2141
# provide it for previous python versions. Also, we don't use the timeout
2142
# parameter (provided by the python implementation) so we don't implement
2144
err = socket.error('getaddrinfo returns an empty list')
2145
host, port = address
2146
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2147
af, socktype, proto, canonname, sa = res
2150
sock = socket.socket(af, socktype, proto)
2154
except socket.error as err:
2155
# 'err' is now the most recent error
2156
if sock is not None:
2054
2161
def dereference_path(path):
2082
2189
If and when pkg_resources becomes a standard library, this routine
2083
2190
can delegate to it.
2085
# Check package name is within bzrlib
2086
if package == "bzrlib":
2192
# Check package name is within breezy
2193
if package == "breezy":
2087
2194
resource_relpath = resource_name
2088
elif package.startswith("bzrlib."):
2089
package = package[len("bzrlib."):].replace('.', os.sep)
2195
elif package.startswith("breezy."):
2196
package = package[len("breezy."):].replace('.', os.sep)
2090
2197
resource_relpath = pathjoin(package, resource_name)
2092
raise errors.BzrError('resource package %s not in bzrlib' % package)
2199
raise errors.BzrError('resource package %s not in breezy' % package)
2094
2201
# Map the resource to a file and read its contents
2095
base = dirname(bzrlib.__file__)
2202
base = dirname(breezy.__file__)
2096
2203
if getattr(sys, 'frozen', None): # bzr.exe
2097
2204
base = abspath(pathjoin(base, '..', '..'))
2098
filename = pathjoin(base, resource_relpath)
2099
return open(filename, 'rU').read()
2205
f = file(pathjoin(base, resource_relpath), "rU")
2102
2211
def file_kind_from_stat_mode_thunk(mode):
2103
2212
global file_kind_from_stat_mode
2104
2213
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2106
from bzrlib._readdir_pyx import UTF8DirReader
2215
from ._readdir_pyx import UTF8DirReader
2107
2216
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2108
except ImportError, e:
2217
except ImportError as e:
2109
2218
# This is one time where we won't warn that an extension failed to
2110
2219
# load. The extension is never available on Windows anyway.
2111
from bzrlib._readdir_py import (
2220
from ._readdir_py import (
2112
2221
_kind_from_mode as file_kind_from_stat_mode
2114
2223
return file_kind_from_stat_mode(mode)
2115
2224
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2118
def file_kind(f, _lstat=os.lstat):
2226
def file_stat(f, _lstat=os.lstat):
2120
return file_kind_from_stat_mode(_lstat(f).st_mode)
2230
except OSError as e:
2122
2231
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2123
2232
raise errors.NoSuchFile(f)
2235
def file_kind(f, _lstat=os.lstat):
2236
stat_value = file_stat(f, _lstat)
2237
return file_kind_from_stat_mode(stat_value.st_mode)
2127
2239
def until_no_eintr(f, *a, **kw):
2128
2240
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2130
2242
WARNING: you must be certain that it is safe to retry the call repeatedly
2131
2243
if EINTR does occur. This is typically only true for low-level operations
2132
2244
like os.read. If in any doubt, don't use this.
2134
2246
Keep in mind that this is not a complete solution to EINTR. There is
2135
2247
probably code in the Python standard library and other dependencies that
2136
2248
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
2249
that signal). So this function can reduce the impact for IO that breezy
2138
2250
directly controls, but it is not a complete solution.
2140
2252
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2143
2255
return f(*a, **kw)
2144
except (IOError, OSError), e:
2256
except (IOError, OSError) as e:
2145
2257
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
2262
if sys.platform == "win32":
2176
2265
return msvcrt.getch()
2181
2270
fd = sys.stdin.fileno()
2182
2271
settings = termios.tcgetattr(fd)
2187
2276
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2191
if sys.platform == 'linux2':
2279
if sys.platform.startswith('linux'):
2192
2280
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
2282
return os.sysconf('SC_NPROCESSORS_ONLN')
2283
except (ValueError, OSError, AttributeError):
2199
2285
elif sys.platform == 'darwin':
2200
2286
def _local_concurrency():
2201
2287
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2202
2288
stdout=subprocess.PIPE).communicate()[0]
2203
elif sys.platform[0:7] == 'freebsd':
2289
elif "bsd" in sys.platform:
2204
2290
def _local_concurrency():
2205
2291
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2206
2292
stdout=subprocess.PIPE).communicate()[0]
2231
2317
if _cached_local_concurrency is not None and use_cache:
2232
2318
return _cached_local_concurrency
2234
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2320
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2235
2321
if concurrency is None:
2237
concurrency = _local_concurrency()
2238
except (OSError, IOError):
2323
import multiprocessing
2324
concurrency = multiprocessing.cpu_count()
2325
except (ImportError, NotImplementedError):
2326
# multiprocessing is only available on Python >= 2.6
2327
# and multiprocessing.cpu_count() isn't implemented on all
2330
concurrency = _local_concurrency()
2331
except (OSError, IOError):
2241
2334
concurrency = int(concurrency)
2242
2335
except (TypeError, ValueError):
2302
2395
return os.fdopen(os.open(filename, flags), mode, bufsize)
2304
2397
open_file = open
2400
def available_backup_name(base, exists):
2401
"""Find a non-existing backup file name.
2403
This will *not* create anything, this only return a 'free' entry. This
2404
should be used for checking names in a directory below a locked
2405
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2406
Leap) and generally discouraged.
2408
:param base: The base name.
2410
:param exists: A callable returning True if the path parameter exists.
2413
name = "%s.~%d~" % (base, counter)
2416
name = "%s.~%d~" % (base, counter)
2420
def set_fd_cloexec(fd):
2421
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2422
support for this is not available.
2426
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2427
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2428
except (ImportError, AttributeError):
2429
# Either the fcntl module or specific constants are not present
2433
def find_executable_on_path(name):
2434
"""Finds an executable on the PATH.
2436
On Windows, this will try to append each extension in the PATHEXT
2437
environment variable to the name, if it cannot be found with the name
2440
:param name: The base name of the executable.
2441
:return: The path to the executable found or None.
2443
if sys.platform == 'win32':
2444
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2445
exts = [ext.lower() for ext in exts]
2446
base, ext = os.path.splitext(name)
2448
if ext.lower() not in exts:
2454
path = os.environ.get('PATH')
2455
if path is not None:
2456
path = path.split(os.pathsep)
2459
f = os.path.join(d, name) + ext
2460
if os.access(f, os.X_OK):
2462
if sys.platform == 'win32':
2463
app_path = win32utils.get_app_path(name)
2464
if app_path != name:
2469
def _posix_is_local_pid_dead(pid):
2470
"""True if pid doesn't correspond to live process on this machine"""
2472
# Special meaning of unix kill: just check if it's there.
2474
except OSError as e:
2475
if e.errno == errno.ESRCH:
2476
# On this machine, and really not found: as sure as we can be
2479
elif e.errno == errno.EPERM:
2480
# exists, though not ours
2483
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2484
# Don't really know.
2487
# Exists and our process: not dead.
2490
if sys.platform == "win32":
2491
is_local_pid_dead = win32utils.is_local_pid_dead
2493
is_local_pid_dead = _posix_is_local_pid_dead
2495
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2496
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2497
if getattr(errno, name, None) is not None]
2500
def fdatasync(fileno):
2501
"""Flush file contents to disk if possible.
2503
:param fileno: Integer OS file handle.
2504
:raises TransportNotPossible: If flushing to disk is not possible.
2506
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2510
except IOError as e:
2511
# See bug #1075108, on some platforms fdatasync exists, but can
2512
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2513
# and reduce the chance of corruption-on-powerloss situations. It
2514
# is not a mandatory call, so it is ok to suppress failures.
2515
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2516
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2520
def ensure_empty_directory_exists(path, exception_class):
2521
"""Make sure a local directory exists and is empty.
2523
If it does not exist, it is created. If it exists and is not empty, an
2524
instance of exception_class is raised.
2528
except OSError as e:
2529
if e.errno != errno.EEXIST:
2531
if os.listdir(path) != []:
2532
raise exception_class(path)
2535
def is_environment_error(evalue):
2536
"""True if exception instance is due to a process environment issue
2538
This includes OSError and IOError, but also other errors that come from
2539
the operating system or core libraries but are not subclasses of those.
2541
if isinstance(evalue, (EnvironmentError, select.error)):
2543
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):