14
14
# along with this program; if not, write to the Free Software
15
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from __future__ import absolute_import
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
25
from bzrlib.lazy_import import lazy_import
27
from .lazy_import import lazy_import
26
28
lazy_import(globals(), """
27
29
from datetime import datetime
29
from ntpath import (abspath as _nt_abspath,
31
normpath as _nt_normpath,
32
realpath as _nt_realpath,
33
splitdrive as _nt_splitdrive,
35
# We need to import both shutil and rmtree as we export the later on posix
36
# and need the former on windows
38
from shutil import rmtree
41
# We need to import both tempfile and mkdtemp as we export the later on posix
42
# and need the former on windows
43
from tempfile import (
44
from tempfile import mkdtemp
52
from breezy.i18n import gettext
56
from bzrlib.symbol_versioning import (
61
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
63
if sys.version_info < (2, 5):
64
import md5 as _mod_md5
66
import sha as _mod_sha
76
from bzrlib import symbol_versioning
79
# Cross platform wall-clock time functionality with decent resolution.
80
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
81
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
82
# synchronized with ``time.time()``, this is only meant to be used to find
83
# delta times by subtracting from another call to this function.
84
timer_func = time.time
85
if sys.platform == 'win32':
86
timer_func = time.clock
88
73
# On win32, O_BINARY is used to indicate the file should
89
74
# be opened in binary mode, rather than text mode.
96
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
99
93
def get_unicode_argv():
101
97
user_encoding = get_user_encoding()
102
98
return [a.decode(user_encoding) for a in sys.argv[1:]]
103
99
except UnicodeDecodeError:
104
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))
108
104
def make_readonly(filename):
109
105
"""Make a filename read-only."""
110
106
mod = os.lstat(filename).st_mode
111
107
if not stat.S_ISLNK(mod):
113
os.chmod(filename, mod)
109
chmod_if_possible(filename, mod)
116
112
def make_writable(filename):
117
113
mod = os.lstat(filename).st_mode
118
114
if not stat.S_ISLNK(mod):
120
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" % (
123
138
def minimum_path_selection(paths):
232
252
file_existed = False
234
254
rename_func(new, tmp_name)
235
except (errors.NoSuchFile,), e:
255
except (errors.NoSuchFile,):
238
258
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
239
259
# function raises an IOError with errno is None when a rename fails.
240
260
# This then gets caught here.
241
261
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
263
except Exception as e:
244
264
if (getattr(e, 'errno', None) is None
245
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
265
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
248
268
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)
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)
264
281
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
282
# source and target are the same file on a case-insensitive
283
# filesystem, so we don't generate an exception
270
289
# If the file used to exist, rename it back into place
273
292
unlink_func(tmp_name)
275
294
rename_func(tmp_name, new)
276
if failure_exc is not None:
277
raise failure_exc[0], failure_exc[1], failure_exc[2]
280
297
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
281
298
# choke on a Unicode string containing a relative path if
282
299
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
284
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
285
301
def _posix_abspath(path):
286
302
# jam 20060426 rather than encoding to fsencoding
287
303
# copy posixpath.abspath, but use os.getcwdu instead
288
304
if not posixpath.isabs(path):
289
305
path = posixpath.join(getcwd(), path)
290
return posixpath.normpath(path)
306
return _posix_normpath(path)
293
309
def _posix_realpath(path):
294
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))
297
368
def _win32_fixdrive(path):
298
369
"""Force drive letters to be consistent.
303
374
running python.exe under cmd.exe return capital C:\\
304
375
running win32 python inside a cygwin shell returns lowercase c:\\
306
drive, path = _nt_splitdrive(path)
377
drive, path = ntpath.splitdrive(path)
307
378
return drive.upper() + path
310
381
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('\\', '/'))
382
# Real ntpath.abspath doesn't have a problem with a unicode cwd
383
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
342
386
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('\\', '/'))
387
# Real ntpath.realpath doesn't have a problem with a unicode cwd
388
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
347
391
def _win32_pathjoin(*args):
348
return _nt_join(*args).replace('\\', '/')
392
return ntpath.join(*args).replace('\\', '/')
351
395
def _win32_normpath(path):
352
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
396
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
355
399
def _win32_getcwd():
356
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
400
return _win32_fixdrive(_getcwd().replace('\\', '/'))
359
403
def _win32_mkdtemp(*args, **kwargs):
360
404
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
407
def _win32_rename(old, new):
372
408
"""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
425
def _mac_getcwd():
400
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)
403
459
# Default is to just use the python builtins, but these can be rebound on
404
460
# particular platforms.
405
461
abspath = _posix_abspath
406
462
realpath = _posix_realpath
407
463
pathjoin = os.path.join
408
normpath = os.path.normpath
409
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
411
469
dirname = os.path.dirname
412
470
basename = os.path.basename
413
471
split = os.path.split
414
472
splitext = os.path.splitext
415
# These were already imported into local scope
473
# These were already lazily imported into local scope
416
474
# mkdtemp = tempfile.mkdtemp
417
475
# rmtree = shutil.rmtree
419
484
MIN_ABS_PATHLENGTH = 1
422
487
if sys.platform == 'win32':
423
if win32utils.winver == 'Windows 98':
424
abspath = _win98_abspath
426
abspath = _win32_abspath
488
abspath = _win32_abspath
427
489
realpath = _win32_realpath
428
490
pathjoin = _win32_pathjoin
429
491
normpath = _win32_normpath
430
492
getcwd = _win32_getcwd
431
493
mkdtemp = _win32_mkdtemp
432
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
434
504
MIN_ABS_PATHLENGTH = 3
471
542
On my standard US Windows XP, the preferred encoding is
472
543
cp1252, but the console is cp437
545
:param trace: If True trace the selected encoding via mutter().
474
from bzrlib.trace import mutter
547
from .trace import mutter
475
548
output_encoding = getattr(sys.stdout, 'encoding', None)
476
549
if not output_encoding:
477
550
input_encoding = getattr(sys.stdin, 'encoding', None)
478
551
if not input_encoding:
479
552
output_encoding = get_user_encoding()
480
mutter('encoding stdout as osutils.get_user_encoding() %r',
554
mutter('encoding stdout as osutils.get_user_encoding() %r',
483
557
output_encoding = input_encoding
484
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
559
mutter('encoding stdout as sys.stdin encoding %r',
486
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
563
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
487
564
if output_encoding == 'cp0':
488
565
# invalid encoding (cp0 means 'no codepage' on Windows)
489
566
output_encoding = get_user_encoding()
490
mutter('cp0 is invalid encoding.'
491
' encoding stdout as osutils.get_user_encoding() %r',
568
mutter('cp0 is invalid encoding.'
569
' encoding stdout as osutils.get_user_encoding() %r',
495
573
codecs.lookup(output_encoding)
496
574
except LookupError:
497
sys.stderr.write('bzr: warning:'
575
sys.stderr.write('brz: warning:'
498
576
' unknown terminal encoding %s.\n'
499
577
' Using encoding %s instead.\n'
500
578
% (output_encoding, get_user_encoding())
502
580
output_encoding = get_user_encoding()
504
582
return output_encoding
878
975
return '%d hours, %d minute%s %s' % (hours, minutes,
879
976
plural_minutes, direction)
882
980
"""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
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":
896
rand_bytes = file('/dev/urandom', 'rb').read
897
# Otherwise, use this hack as a last resort
898
except (IOError, OSError):
993
except NotImplementedError:
899
994
# not well seeded, but better than nothing
900
995
def rand_bytes(n):
917
1014
for raw_byte in rand_bytes(num):
918
s += ALNUM[ord(raw_byte) % 36]
1016
s += ALNUM[ord(raw_byte) % 36]
1018
s += ALNUM[raw_byte % 36]
922
## TODO: We could later have path objects that remember their list
923
## 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.)
925
1025
def splitpath(p):
926
1026
"""Turn string into list of parts."""
927
# split on either delimiter because people might use either on
929
ps = re.split(r'[\\/]', p)
1027
use_bytes = isinstance(p, bytes)
1028
if os.path.sep == '\\':
1029
# split on either delimiter because people might use either on
1032
ps = re.split(b'[\\\\/]', p)
1034
ps = re.split(r'[\\/]', p)
1043
current_empty_dir = (b'.', b'')
1046
current_empty_dir = ('.', '')
934
raise errors.BzrError("sorry, %r not allowed in path" % f)
935
elif (f == '.') or (f == ''):
1051
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1052
elif f in current_empty_dir:
993
1109
def report_extension_load_failures():
994
1110
if not _extension_load_failures:
996
from bzrlib.config import GlobalConfig
997
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1112
if config.GlobalConfig().suppress_warning('missing_extensions'):
999
1114
# the warnings framework should by default show this only once
1000
from bzrlib.trace import warning
1115
from .trace import warning
1002
"bzr: warning: some compiled extensions could not be loaded; "
1003
"see <https://answers.launchpad.net/bzr/+faq/703>")
1117
"brz: warning: some compiled extensions could not be loaded; "
1118
"see ``brz help missing-extensions``")
1004
1119
# we no longer show the specific missing extensions here, because it makes
1005
1120
# the message too long and scary - see
1006
1121
# https://bugs.launchpad.net/bzr/+bug/430529
1010
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1011
except ImportError, e:
1125
from ._chunks_to_lines_pyx import chunks_to_lines
1126
except ImportError as e:
1012
1127
failed_to_load_extension(e)
1013
from bzrlib._chunks_to_lines_py import chunks_to_lines
1128
from ._chunks_to_lines_py import chunks_to_lines
1016
1131
def split_lines(s):
1017
1132
"""Split s into lines, but without removing the newline characters."""
1018
1133
# Trivially convert a fulltext into a 'chunked' representation, and let
1019
1134
# chunks_to_lines do the heavy lifting.
1020
if isinstance(s, str):
1135
if isinstance(s, bytes):
1021
1136
# chunks_to_lines only supports 8-bit strings
1022
1137
return chunks_to_lines([s])
1284
1418
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):
1293
"""Revision ids should now be utf8, but at one point they were unicode.
1295
: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
:return: None or a utf8 revision id.
1300
if (unicode_or_utf8_string is None
1301
or unicode_or_utf8_string.__class__ == str):
1302
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):
1314
"""File ids should now be utf8, but at one point they were unicode.
1316
This is the same as safe_utf8, except it uses the cached encode functions
1317
to save a little bit of performance.
1319
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1321
:param warn: Functions that are sanitizing user data can set warn=False
1322
:return: None or a utf8 file id.
1324
if (unicode_or_utf8_string is None
1325
or unicode_or_utf8_string.__class__ == str):
1326
return unicode_or_utf8_string
1328
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1330
return cache_utf8.encode(unicode_or_utf8_string)
1333
1421
_platform_normalizes_filenames = False
1334
1422
if sys.platform == 'darwin':
1335
1423
_platform_normalizes_filenames = True
1424
1523
None is returned if the width can't established precisely.
1427
- if BZR_COLUMNS is set, returns its value
1526
- if BRZ_COLUMNS is set, returns its value
1428
1527
- if there is no controlling terminal, returns None
1528
- query the OS, if the queried size has changed since the last query,
1429
1530
- if COLUMNS is set, returns its value,
1531
- if the OS has a value (even though it's never changed), return its value.
1431
1533
From there, we need to query the OS to get the size of the controlling
1536
On Unices we query the OS by:
1435
1537
- get termios.TIOCGWINSZ
1436
1538
- if an error occurs or a negative value is obtained, returns None
1540
On Windows we query the OS by:
1440
1541
- win32utils.get_console_size() decides,
1441
1542
- returns None on error (provided default value)
1544
# Note to implementors: if changing the rules for determining the width,
1545
# make sure you've considered the behaviour in these cases:
1546
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1547
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1549
# - (add more interesting cases here, if you find any)
1550
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1551
# but we don't want to register a signal handler because it is impossible
1552
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1553
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1554
# 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
1557
# If BRZ_COLUMNS is set, take it, user is always right
1558
# Except if they specified 0 in which case, impose no limit here
1446
return int(os.environ['BZR_COLUMNS'])
1560
width = int(os.environ['BRZ_COLUMNS'])
1447
1561
except (KeyError, ValueError):
1563
if width is not None:
1450
1569
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.
1570
if isatty is None or not isatty():
1571
# 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))
1575
width, height = os_size = _terminal_size(None, None)
1576
global _first_terminal_size, _terminal_size_state
1577
if _terminal_size_state == 'no_data':
1578
_first_terminal_size = os_size
1579
_terminal_size_state = 'unchanged'
1580
elif (_terminal_size_state == 'unchanged' and
1581
_first_terminal_size != os_size):
1582
_terminal_size_state = 'changed'
1584
# If the OS claims to know how wide the terminal is, and this value has
1585
# ever changed, use that.
1586
if _terminal_size_state == 'changed':
1587
if width is not None and width > 0:
1590
# If COLUMNS is set, use it.
1459
1592
return int(os.environ['COLUMNS'])
1460
1593
except (KeyError, ValueError):
1463
width, height = _terminal_size(None, None)
1465
# Consider invalid values as meaning no width
1596
# Finally, use an unchanged size from the OS, if we have one.
1597
if _terminal_size_state == 'unchanged':
1598
if width is not None and width > 0:
1601
# The width could not be determined.
1471
1605
def _win32_terminal_size(width, height):
1472
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1606
width, height = win32utils.get_console_size(
1607
defaultx=width, defaulty=height)
1473
1608
return width, height
1476
1611
def _ioctl_terminal_size(width, height):
1478
import struct, fcntl, termios
1479
1616
s = struct.pack('HHHH', 0, 0, 0, 0)
1480
1617
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1481
1618
height, width = struct.unpack('HHHH', x)[0:2]
1498
1636
_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
def supports_executable():
1527
return sys.platform != "win32"
1639
def supports_executable(path):
1640
"""Return if filesystem at path supports executable bit.
1642
:param path: Path for which to check the file system
1643
:return: boolean indicating whether executable bit can be stored/relied upon
1645
if sys.platform == 'win32':
1648
fs_type = get_fs_type(path)
1649
except errors.DependencyNotPresent as e:
1650
trace.mutter('Unable to get fs type for %r: %s', path, e)
1652
if fs_type in ('vfat', 'ntfs'):
1653
# filesystems known to not support executable bit
1658
def supports_symlinks(path):
1659
"""Return if the filesystem at path supports the creation of symbolic links.
1662
if not has_symlinks():
1665
fs_type = get_fs_type(path)
1666
except errors.DependencyNotPresent as e:
1667
trace.mutter('Unable to get fs type for %r: %s', path, e)
1669
if fs_type in ('vfat', 'ntfs'):
1670
# filesystems known to not support symlinks
1530
1675
def supports_posix_readonly():
1779
1918
See DirReader.read_dir for details.
1781
1920
_utf8_encode = self._utf8_encode
1922
def _fs_decode(s): return s.decode(_fs_enc)
1924
def _fs_encode(s): return s.encode(_fs_enc)
1782
1925
_lstat = os.lstat
1783
1926
_listdir = os.listdir
1784
1927
_kind_from_mode = file_kind_from_stat_mode
1787
relprefix = prefix + '/'
1930
relprefix = prefix + b'/'
1790
top_slash = top + u'/'
1933
top_slash = top + '/'
1793
1936
append = dirblock.append
1794
for name in sorted(_listdir(top)):
1937
for name_native in _listdir(top.encode('utf-8')):
1796
name_utf8 = _utf8_encode(name)[0]
1939
name = _fs_decode(name_native)
1797
1940
except UnicodeDecodeError:
1798
1941
raise errors.BadFilenameEncoding(
1799
_utf8_encode(relprefix)[0] + name, _fs_enc)
1942
relprefix + name_native, _fs_enc)
1943
name_utf8 = _utf8_encode(name)[0]
1800
1944
abspath = top_slash + name
1801
1945
statvalue = _lstat(abspath)
1802
1946
kind = _kind_from_mode(statvalue.st_mode)
1803
1947
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1948
return sorted(dirblock)
1807
1951
def copy_tree(from_path, to_path, handlers={}):
1874
2021
This can be used to sort paths in the same way that walkdirs does.
1876
return (dirname(path) , path)
2023
return (dirname(path), path)
1879
2026
def compare_paths_prefix_order(path_a, path_b):
1880
2027
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1881
2028
key_a = path_prefix_key(path_a)
1882
2029
key_b = path_prefix_key(path_b)
1883
return cmp(key_a, key_b)
2030
return (key_a > key_b) - (key_a < key_b)
1886
2033
_cached_user_encoding = None
1889
def get_user_encoding(use_cache=True):
2036
def get_user_encoding():
1890
2037
"""Find out what the preferred user encoding is.
1892
2039
This is generally the encoding that is used for command line parameters
1893
2040
and file contents. This may be different from the terminal encoding
1894
2041
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
2043
:return: A string defining the preferred user encoding
1902
2045
global _cached_user_encoding
1903
if _cached_user_encoding is not None and use_cache:
2046
if _cached_user_encoding is not None:
1904
2047
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'
2049
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2050
# Use the existing locale settings and call nl_langinfo directly
2051
# rather than going through getpreferredencoding. This avoids
2052
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2053
# possibility of the setlocale call throwing an error.
2054
user_encoding = locale.nl_langinfo(locale.CODESET)
2056
# GZ 2011-12-19: On windows could call GetACP directly instead.
2057
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:'
2060
user_encoding = codecs.lookup(user_encoding).name
2062
if user_encoding not in ("", "cp0"):
2063
sys.stderr.write('brz: warning:'
1950
2064
' unknown encoding %s.'
1951
2065
' Continuing with ascii encoding.\n'
1952
2066
% user_encoding
1954
user_encoding = 'ascii'
1957
_cached_user_encoding = user_encoding
2068
user_encoding = 'ascii'
2070
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2071
if user_encoding == 'ascii':
2072
if sys.platform == 'darwin':
2073
# OSX is special-cased in Python to have a UTF-8 filesystem
2074
# encoding and previously had LANG set here if not present.
2075
user_encoding = 'utf-8'
2076
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2077
# for some other posix platforms as well.
2079
_cached_user_encoding = user_encoding
1959
2080
return user_encoding
2083
def get_diff_header_encoding():
2084
return get_terminal_encoding()
1962
2087
def get_host_name():
1963
2088
"""Return the current unicode host name.
1979
2105
# data at once.
1980
2106
MAX_SOCKET_CHUNK = 64 * 1024
2108
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2109
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2110
_eno = getattr(errno, _eno, None)
2111
if _eno is not None:
2112
_end_of_stream_errors.append(_eno)
1982
2116
def read_bytes_from_socket(sock, report_activity=None,
1983
max_read_size=MAX_SOCKET_CHUNK):
2117
max_read_size=MAX_SOCKET_CHUNK):
1984
2118
"""Read up to max_read_size of bytes from sock and notify of progress.
1986
2120
Translates "Connection reset by peer" into file-like EOF (return an
1987
2121
empty string rather than raise an error), and repeats the recv if
1988
2122
interrupted by a signal.
1992
bytes = sock.recv(max_read_size)
1993
except socket.error, e:
2126
data = sock.recv(max_read_size)
2127
except socket.error as e:
1994
2128
eno = e.args[0]
1995
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2129
if eno in _end_of_stream_errors:
1996
2130
# The connection was closed by the other side. Callers expect
1997
2131
# an empty string to signal end-of-stream.
1999
2133
elif eno == errno.EINTR:
2000
2134
# Retry the interrupted recv.
2004
2138
if report_activity is not None:
2005
report_activity(len(bytes), 'read')
2139
report_activity(len(data), 'read')
2009
2143
def recv_all(socket, count):
2017
2151
This isn't optimized and is intended mostly for use in testing.
2020
2154
while len(b) < count:
2021
2155
new = read_bytes_from_socket(socket, None, count - len(b))
2028
2162
def send_all(sock, bytes, report_activity=None):
2029
2163
"""Send all bytes on a socket.
2031
2165
Breaks large blocks in smaller chunks to avoid buffering limitations on
2032
2166
some platforms, and catches EINTR which may be thrown if the send is
2033
2167
interrupted by a signal.
2035
2169
This is preferred to socket.sendall(), because it avoids portability bugs
2036
2170
and provides activity reporting.
2038
2172
:param report_activity: Call this as bytes are read, see
2039
2173
Transport._report_activity
2042
2176
byte_count = len(bytes)
2177
view = memoryview(bytes)
2043
2178
while sent_total < byte_count:
2045
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2046
except socket.error, e:
2180
sent = sock.send(view[sent_total:sent_total + MAX_SOCKET_CHUNK])
2181
except (socket.error, IOError) as e:
2182
if e.args[0] in _end_of_stream_errors:
2183
raise errors.ConnectionReset(
2184
"Error trying to write to socket", e)
2047
2185
if e.args[0] != errno.EINTR:
2189
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2050
2191
sent_total += sent
2051
report_activity(sent, 'write')
2192
if report_activity is not None:
2193
report_activity(sent, 'write')
2196
def connect_socket(address):
2197
# Slight variation of the socket.create_connection() function (provided by
2198
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2199
# provide it for previous python versions. Also, we don't use the timeout
2200
# parameter (provided by the python implementation) so we don't implement
2202
err = socket.error('getaddrinfo returns an empty list')
2203
host, port = address
2204
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2205
af, socktype, proto, canonname, sa = res
2208
sock = socket.socket(af, socktype, proto)
2212
except socket.error as e:
2214
# 'err' is now the most recent error
2215
if sock is not None:
2054
2220
def dereference_path(path):
2082
2248
If and when pkg_resources becomes a standard library, this routine
2083
2249
can delegate to it.
2085
# Check package name is within bzrlib
2086
if package == "bzrlib":
2251
# Check package name is within breezy
2252
if package == "breezy":
2087
2253
resource_relpath = resource_name
2088
elif package.startswith("bzrlib."):
2089
package = package[len("bzrlib."):].replace('.', os.sep)
2254
elif package.startswith("breezy."):
2255
package = package[len("breezy."):].replace('.', os.sep)
2090
2256
resource_relpath = pathjoin(package, resource_name)
2092
raise errors.BzrError('resource package %s not in bzrlib' % package)
2258
raise errors.BzrError('resource package %s not in breezy' % package)
2094
2260
# Map the resource to a file and read its contents
2095
base = dirname(bzrlib.__file__)
2261
base = dirname(breezy.__file__)
2096
2262
if getattr(sys, 'frozen', None): # bzr.exe
2097
2263
base = abspath(pathjoin(base, '..', '..'))
2098
filename = pathjoin(base, resource_relpath)
2099
return open(filename, 'rU').read()
2264
with open(pathjoin(base, resource_relpath), "rt") as f:
2102
2268
def file_kind_from_stat_mode_thunk(mode):
2103
2269
global file_kind_from_stat_mode
2104
2270
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2106
from bzrlib._readdir_pyx import UTF8DirReader
2272
from ._readdir_pyx import UTF8DirReader
2107
2273
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2108
except ImportError, e:
2109
2275
# This is one time where we won't warn that an extension failed to
2110
2276
# load. The extension is never available on Windows anyway.
2111
from bzrlib._readdir_py import (
2277
from ._readdir_py import (
2112
2278
_kind_from_mode as file_kind_from_stat_mode
2114
2280
return file_kind_from_stat_mode(mode)
2115
2283
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2118
def file_kind(f, _lstat=os.lstat):
2286
def file_stat(f, _lstat=os.lstat):
2120
return file_kind_from_stat_mode(_lstat(f).st_mode)
2290
except OSError as e:
2122
2291
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2123
2292
raise errors.NoSuchFile(f)
2296
def file_kind(f, _lstat=os.lstat):
2297
stat_value = file_stat(f, _lstat)
2298
return file_kind_from_stat_mode(stat_value.st_mode)
2127
2301
def until_no_eintr(f, *a, **kw):
2128
2302
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2130
2304
WARNING: you must be certain that it is safe to retry the call repeatedly
2131
2305
if EINTR does occur. This is typically only true for low-level operations
2132
2306
like os.read. If in any doubt, don't use this.
2134
2308
Keep in mind that this is not a complete solution to EINTR. There is
2135
2309
probably code in the Python standard library and other dependencies that
2136
2310
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
2311
that signal). So this function can reduce the impact for IO that breezy
2138
2312
directly controls, but it is not a complete solution.
2140
2314
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2143
2317
return f(*a, **kw)
2144
except (IOError, OSError), e:
2318
except (IOError, OSError) as e:
2145
2319
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
2324
if sys.platform == "win32":
2176
2327
return msvcrt.getch()
2181
2332
fd = sys.stdin.fileno()
2182
2333
settings = termios.tcgetattr(fd)
2187
2338
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2191
if sys.platform == 'linux2':
2341
if sys.platform.startswith('linux'):
2192
2342
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
2344
return os.sysconf('SC_NPROCESSORS_ONLN')
2345
except (ValueError, OSError, AttributeError):
2199
2347
elif sys.platform == 'darwin':
2200
2348
def _local_concurrency():
2201
2349
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2202
2350
stdout=subprocess.PIPE).communicate()[0]
2203
elif sys.platform[0:7] == 'freebsd':
2351
elif "bsd" in sys.platform:
2204
2352
def _local_concurrency():
2205
2353
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2206
2354
stdout=subprocess.PIPE).communicate()[0]
2207
2355
elif sys.platform == 'sunos5':
2208
2356
def _local_concurrency():
2209
return subprocess.Popen(['psrinfo', '-p',],
2357
return subprocess.Popen(['psrinfo', '-p', ],
2210
2358
stdout=subprocess.PIPE).communicate()[0]
2211
2359
elif sys.platform == "win32":
2212
2360
def _local_concurrency():
2302
2457
return os.fdopen(os.open(filename, flags), mode, bufsize)
2304
2459
open_file = open
2462
def available_backup_name(base, exists):
2463
"""Find a non-existing backup file name.
2465
This will *not* create anything, this only return a 'free' entry. This
2466
should be used for checking names in a directory below a locked
2467
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2468
Leap) and generally discouraged.
2470
:param base: The base name.
2472
:param exists: A callable returning True if the path parameter exists.
2475
name = "%s.~%d~" % (base, counter)
2478
name = "%s.~%d~" % (base, counter)
2482
def set_fd_cloexec(fd):
2483
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2484
support for this is not available.
2488
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2489
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2490
except (ImportError, AttributeError):
2491
# Either the fcntl module or specific constants are not present
2495
def find_executable_on_path(name):
2496
"""Finds an executable on the PATH.
2498
On Windows, this will try to append each extension in the PATHEXT
2499
environment variable to the name, if it cannot be found with the name
2502
:param name: The base name of the executable.
2503
:return: The path to the executable found or None.
2505
if sys.platform == 'win32':
2506
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2507
exts = [ext.lower() for ext in exts]
2508
base, ext = os.path.splitext(name)
2510
if ext.lower() not in exts:
2516
path = os.environ.get('PATH')
2517
if path is not None:
2518
path = path.split(os.pathsep)
2521
f = os.path.join(d, name) + ext
2522
if os.access(f, os.X_OK):
2524
if sys.platform == 'win32':
2525
app_path = win32utils.get_app_path(name)
2526
if app_path != name:
2531
def _posix_is_local_pid_dead(pid):
2532
"""True if pid doesn't correspond to live process on this machine"""
2534
# Special meaning of unix kill: just check if it's there.
2536
except OSError as e:
2537
if e.errno == errno.ESRCH:
2538
# On this machine, and really not found: as sure as we can be
2541
elif e.errno == errno.EPERM:
2542
# exists, though not ours
2545
trace.mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2546
# Don't really know.
2549
# Exists and our process: not dead.
2553
if sys.platform == "win32":
2554
is_local_pid_dead = win32utils.is_local_pid_dead
2556
is_local_pid_dead = _posix_is_local_pid_dead
2558
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2559
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2560
if getattr(errno, name, None) is not None]
2563
def fdatasync(fileno):
2564
"""Flush file contents to disk if possible.
2566
:param fileno: Integer OS file handle.
2567
:raises TransportNotPossible: If flushing to disk is not possible.
2569
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2573
except IOError as e:
2574
# See bug #1075108, on some platforms fdatasync exists, but can
2575
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2576
# and reduce the chance of corruption-on-powerloss situations. It
2577
# is not a mandatory call, so it is ok to suppress failures.
2578
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2579
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2583
def ensure_empty_directory_exists(path, exception_class):
2584
"""Make sure a local directory exists and is empty.
2586
If it does not exist, it is created. If it exists and is not empty, an
2587
instance of exception_class is raised.
2591
except OSError as e:
2592
if e.errno != errno.EEXIST:
2594
if os.listdir(path) != []:
2595
raise exception_class(path)
2598
def is_environment_error(evalue):
2599
"""True if exception instance is due to a process environment issue
2601
This includes OSError and IOError, but also other errors that come from
2602
the operating system or core libraries but are not subclasses of those.
2604
if isinstance(evalue, (EnvironmentError, select.error)):
2606
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
2611
def read_mtab(path):
2612
"""Read an fstab-style file and extract mountpoint+filesystem information.
2614
:param path: Path to read from
2615
:yield: Tuples with mountpoints (as bytestrings) and filesystem names
2617
with open(path, 'rb') as f:
2619
if line.startswith(b'#'):
2624
yield cols[1], cols[2].decode('ascii', 'replace')
2627
MTAB_PATH = '/etc/mtab'
2629
class FilesystemFinder(object):
2630
"""Find the filesystem for a particular path."""
2632
def __init__(self, mountpoints):
2635
self._mountpoints = sorted(mountpoints, key=key, reverse=True)
2639
"""Create a FilesystemFinder from an mtab-style file.
2641
Note that this will silenty ignore mtab if it doesn't exist or can not
2644
# TODO(jelmer): Use inotify to be notified when /etc/mtab changes and
2645
# we need to re-read it.
2647
return cls(read_mtab(MTAB_PATH))
2648
except EnvironmentError as e:
2649
trace.mutter('Unable to read mtab: %s', e)
2652
def find(self, path):
2653
"""Find the filesystem used by a particular path.
2655
:param path: Path to find (bytestring or text type)
2656
:return: Filesystem name (as text type) or None, if the filesystem is
2659
for mountpoint, filesystem in self._mountpoints:
2660
if is_inside(mountpoint, path):
2665
_FILESYSTEM_FINDER = None
2668
def get_fs_type(path):
2669
"""Return the filesystem type for the partition a path is in.
2671
:param path: Path to search filesystem type for
2672
:return: A FS type, as string. E.g. "ext2"
2674
global _FILESYSTEM_FINDER
2675
if _FILESYSTEM_FINDER is None:
2676
_FILESYSTEM_FINDER = FilesystemFinder.from_mtab()
2678
if not isinstance(path, bytes):
2679
path = path.encode(_fs_enc)
2681
return _FILESYSTEM_FINDER.find(path)
2685
perf_counter = time.perf_counter
2687
perf_counter = time.clock
added
removed
breezy/osutils.py
