1
# Copyright (C) 2005-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from __future__ import absolute_import
27
from .lazy_import import lazy_import
28
lazy_import(globals(), """
29
from datetime import datetime
35
# We need to import both shutil and rmtree as we export the later on posix
36
# and need the former on windows
38
from shutil import rmtree
41
# We need to import both tempfile and mkdtemp as we export the later on posix
42
# and need the former on windows
44
from tempfile import mkdtemp
52
from breezy.i18n import gettext
73
# On win32, O_BINARY is used to indicate the file should
74
# be opened in binary mode, rather than text mode.
75
# On other platforms, O_BINARY doesn't exist, because
76
# they always open in binary mode, so it is okay to
77
# OR with 0 on those platforms.
78
# O_NOINHERIT and O_TEXT exists only on win32 too.
79
O_BINARY = getattr(os, 'O_BINARY', 0)
80
O_TEXT = getattr(os, 'O_TEXT', 0)
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
93
def get_unicode_argv():
97
user_encoding = get_user_encoding()
98
return [a.decode(user_encoding) for a in sys.argv[1:]]
99
except UnicodeDecodeError:
100
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
101
"application locale.").format(a, user_encoding))
104
def make_readonly(filename):
105
"""Make a filename read-only."""
106
mod = os.lstat(filename).st_mode
107
if not stat.S_ISLNK(mod):
109
chmod_if_possible(filename, mod)
112
def make_writable(filename):
113
mod = os.lstat(filename).st_mode
114
if not stat.S_ISLNK(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" % (
138
def minimum_path_selection(paths):
139
"""Return the smallset subset of paths which are outside paths.
141
:param paths: A container (and hence not None) of paths.
142
:return: A set of paths sufficient to include everything in paths via
143
is_inside, drawn from the paths parameter.
149
if isinstance(path, bytes):
150
return path.split(b'/')
152
return path.split('/')
153
sorted_paths = sorted(list(paths), key=sort_key)
155
search_paths = [sorted_paths[0]]
156
for path in sorted_paths[1:]:
157
if not is_inside(search_paths[-1], path):
158
# This path is unique, add it
159
search_paths.append(path)
161
return set(search_paths)
168
"""Return a quoted filename filename
170
This previously used backslash quoting, but that works poorly on
172
# TODO: I'm not really sure this is the best format either.x
174
if _QUOTE_RE is None:
175
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
177
if _QUOTE_RE.search(f):
183
_directory_kind = 'directory'
187
"""Return the current umask"""
188
# Assume that people aren't messing with the umask while running
189
# XXX: This is not thread safe, but there is no way to get the
190
# umask without setting it
198
_directory_kind: "/",
200
'tree-reference': '+',
204
def kind_marker(kind):
206
return _kind_marker_map[kind]
208
# Slightly faster than using .get(, '') when the common case is that
213
lexists = getattr(os.path, 'lexists', None)
217
stat = getattr(os, 'lstat', os.stat)
221
if e.errno == errno.ENOENT:
224
raise errors.BzrError(
225
gettext("lstat/stat of ({0!r}): {1!r}").format(f, e))
228
def fancy_rename(old, new, rename_func, unlink_func):
229
"""A fancy rename, when you don't have atomic rename.
231
:param old: The old path, to rename from
232
:param new: The new path, to rename to
233
:param rename_func: The potentially non-atomic rename function
234
:param unlink_func: A way to delete the target file if the full rename
237
# sftp rename doesn't allow overwriting, so play tricks:
238
base = os.path.basename(new)
239
dirname = os.path.dirname(new)
240
# callers use different encodings for the paths so the following MUST
241
# respect that. We rely on python upcasting to unicode if new is unicode
242
# and keeping a str if not.
243
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
244
os.getpid(), rand_chars(10))
245
tmp_name = pathjoin(dirname, tmp_name)
247
# Rename the file out of the way, but keep track if it didn't exist
248
# We don't want to grab just any exception
249
# something like EACCES should prevent us from continuing
250
# The downside is that the rename_func has to throw an exception
251
# with an errno = ENOENT, or NoSuchFile
254
rename_func(new, tmp_name)
255
except (errors.NoSuchFile,):
258
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
259
# function raises an IOError with errno is None when a rename fails.
260
# This then gets caught here.
261
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
263
except Exception as e:
264
if (getattr(e, 'errno', None) is None
265
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
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)
281
and old.lower() == new.lower()):
282
# source and target are the same file on a case-insensitive
283
# filesystem, so we don't generate an exception
289
# If the file used to exist, rename it back into place
290
# otherwise just delete it from the tmp location
292
unlink_func(tmp_name)
294
rename_func(tmp_name, new)
297
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
298
# choke on a Unicode string containing a relative path if
299
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
301
def _posix_abspath(path):
302
# jam 20060426 rather than encoding to fsencoding
303
# copy posixpath.abspath, but use os.getcwdu instead
304
if not posixpath.isabs(path):
305
path = posixpath.join(getcwd(), path)
306
return _posix_normpath(path)
309
def _posix_realpath(path):
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))
368
def _win32_fixdrive(path):
369
"""Force drive letters to be consistent.
371
win32 is inconsistent whether it returns lower or upper case
372
and even if it was consistent the user might type the other
373
so we force it to uppercase
374
running python.exe under cmd.exe return capital C:\\
375
running win32 python inside a cygwin shell returns lowercase c:\\
377
drive, path = ntpath.splitdrive(path)
378
return drive.upper() + path
381
def _win32_abspath(path):
382
# Real ntpath.abspath doesn't have a problem with a unicode cwd
383
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
386
def _win32_realpath(path):
387
# Real ntpath.realpath doesn't have a problem with a unicode cwd
388
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
391
def _win32_pathjoin(*args):
392
return ntpath.join(*args).replace('\\', '/')
395
def _win32_normpath(path):
396
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
400
return _win32_fixdrive(_getcwd().replace('\\', '/'))
403
def _win32_mkdtemp(*args, **kwargs):
404
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
407
def _win32_rename(old, new):
408
"""We expect to be able to atomically replace 'new' with old.
410
On win32, if new exists, it must be moved out of the way first,
414
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
416
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
417
# If we try to rename a non-existant file onto cwd, we get
418
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
419
# if the old path doesn't exist, sometimes we get EACCES
420
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
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)
459
# Default is to just use the python builtins, but these can be rebound on
460
# particular platforms.
461
abspath = _posix_abspath
462
realpath = _posix_realpath
463
pathjoin = os.path.join
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
469
dirname = os.path.dirname
470
basename = os.path.basename
471
split = os.path.split
472
splitext = os.path.splitext
473
# These were already lazily imported into local scope
474
# mkdtemp = tempfile.mkdtemp
475
# rmtree = shutil.rmtree
484
MIN_ABS_PATHLENGTH = 1
487
if sys.platform == 'win32':
488
abspath = _win32_abspath
489
realpath = _win32_realpath
490
pathjoin = _win32_pathjoin
491
normpath = _win32_normpath
492
getcwd = _win32_getcwd
493
mkdtemp = _win32_mkdtemp
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
504
MIN_ABS_PATHLENGTH = 3
506
def _win32_delete_readonly(function, path, excinfo):
507
"""Error handler for shutil.rmtree function [for win32]
508
Helps to remove files and dirs marked as read-only.
510
exception = excinfo[1]
511
if function in (os.remove, os.rmdir) \
512
and isinstance(exception, OSError) \
513
and exception.errno == errno.EACCES:
519
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
520
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
521
return shutil.rmtree(path, ignore_errors, onerror)
523
get_unicode_argv = getattr(win32utils, 'get_unicode_argv', get_unicode_argv)
524
path_from_environ = win32utils.get_environ_unicode
525
_get_home_dir = win32utils.get_home_location
526
getuser_unicode = win32utils.get_user_name
528
elif sys.platform == 'darwin':
532
def get_terminal_encoding(trace=False):
533
"""Find the best encoding for printing to the screen.
535
This attempts to check both sys.stdout and sys.stdin to see
536
what encoding they are in, and if that fails it falls back to
537
osutils.get_user_encoding().
538
The problem is that on Windows, locale.getpreferredencoding()
539
is not the same encoding as that used by the console:
540
http://mail.python.org/pipermail/python-list/2003-May/162357.html
542
On my standard US Windows XP, the preferred encoding is
543
cp1252, but the console is cp437
545
:param trace: If True trace the selected encoding via mutter().
547
from .trace import mutter
548
output_encoding = getattr(sys.stdout, 'encoding', None)
549
if not output_encoding:
550
input_encoding = getattr(sys.stdin, 'encoding', None)
551
if not input_encoding:
552
output_encoding = get_user_encoding()
554
mutter('encoding stdout as osutils.get_user_encoding() %r',
557
output_encoding = input_encoding
559
mutter('encoding stdout as sys.stdin encoding %r',
563
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
564
if output_encoding == 'cp0':
565
# invalid encoding (cp0 means 'no codepage' on Windows)
566
output_encoding = get_user_encoding()
568
mutter('cp0 is invalid encoding.'
569
' encoding stdout as osutils.get_user_encoding() %r',
573
codecs.lookup(output_encoding)
575
sys.stderr.write('brz: warning:'
576
' unknown terminal encoding %s.\n'
577
' Using encoding %s instead.\n'
578
% (output_encoding, get_user_encoding())
580
output_encoding = get_user_encoding()
582
return output_encoding
585
def normalizepath(f):
586
if getattr(os.path, 'realpath', None) is not None:
590
[p, e] = os.path.split(f)
591
if e == "" or e == "." or e == "..":
594
return pathjoin(F(p), e)
598
"""True if f is an accessible directory."""
600
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
606
"""True if f is a regular file."""
608
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
614
"""True if f is a symlink."""
616
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
621
def is_inside(dir, fname):
622
"""True if fname is inside dir.
624
The parameters should typically be passed to osutils.normpath first, so
625
that . and .. and repeated slashes are eliminated, and the separators
626
are canonical for the platform.
628
The empty string as a dir name is taken as top-of-tree and matches
631
# XXX: Most callers of this can actually do something smarter by
632
# looking at the inventory
639
if isinstance(dir, bytes):
640
if not dir.endswith(b'/'):
643
if not dir.endswith('/'):
646
return fname.startswith(dir)
649
def is_inside_any(dir_list, fname):
650
"""True if fname is inside any of given dirs."""
651
for dirname in dir_list:
652
if is_inside(dirname, fname):
657
def is_inside_or_parent_of_any(dir_list, fname):
658
"""True if fname is a child or a parent of any of the given files."""
659
for dirname in dir_list:
660
if is_inside(dirname, fname) or is_inside(fname, dirname):
665
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
666
report_activity=None, direction='read'):
667
"""Copy contents of one file to another.
669
The read_length can either be -1 to read to end-of-file (EOF) or
670
it can specify the maximum number of bytes to read.
672
The buff_size represents the maximum size for each read operation
673
performed on from_file.
675
:param report_activity: Call this as bytes are read, see
676
Transport._report_activity
677
:param direction: Will be passed to report_activity
679
:return: The number of bytes copied.
683
# read specified number of bytes
685
while read_length > 0:
686
num_bytes_to_read = min(read_length, buff_size)
688
block = from_file.read(num_bytes_to_read)
692
if report_activity is not None:
693
report_activity(len(block), direction)
696
actual_bytes_read = len(block)
697
read_length -= actual_bytes_read
698
length += actual_bytes_read
702
block = from_file.read(buff_size)
706
if report_activity is not None:
707
report_activity(len(block), direction)
713
def pump_string_file(bytes, file_handle, segment_size=None):
714
"""Write bytes to file_handle in many smaller writes.
716
:param bytes: The string to write.
717
:param file_handle: The file to write to.
719
# Write data in chunks rather than all at once, because very large
720
# writes fail on some platforms (e.g. Windows with SMB mounted
723
segment_size = 5242880 # 5MB
724
offsets = range(0, len(bytes), segment_size)
725
view = memoryview(bytes)
726
write = file_handle.write
727
for offset in offsets:
728
write(view[offset:offset + segment_size])
731
def file_iterator(input_file, readsize=32768):
733
b = input_file.read(readsize)
739
# GZ 2017-09-16: Makes sense in general for hexdigest() result to be text, but
740
# used as bytes through most interfaces so encode with this wrapper.
742
def _hexdigest(hashobj):
743
return hashobj.hexdigest().encode()
745
def _hexdigest(hashobj):
746
return hashobj.hexdigest()
750
"""Calculate the hexdigest of an open file.
752
The file cursor should be already at the start.
764
def size_sha_file(f):
765
"""Calculate the size and hexdigest of an open file.
767
The file cursor should be already at the start and
768
the caller is responsible for closing the file afterwards.
779
return size, _hexdigest(s)
782
def sha_file_by_name(fname):
783
"""Calculate the SHA1 of a file by reading the full text"""
785
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
788
b = os.read(f, 1 << 16)
796
def sha_strings(strings, _factory=sha):
797
"""Return the sha-1 of concatenation of strings"""
799
for string in strings:
804
def sha_string(f, _factory=sha):
805
# GZ 2017-09-16: Dodgy if factory is ever not sha, probably shouldn't be.
806
return _hexdigest(_factory(f))
809
def fingerprint_file(f):
811
return {'size': len(b),
812
'sha1': _hexdigest(sha(b))}
815
def compare_files(a, b):
816
"""Returns true if equal in contents"""
827
def local_time_offset(t=None):
828
"""Return offset of local zone from GMT, either at present or at time t."""
831
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
832
return offset.days * 86400 + offset.seconds
835
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
836
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
839
def format_date(t, offset=0, timezone='original', date_fmt=None,
841
"""Return a formatted date string.
843
:param t: Seconds since the epoch.
844
:param offset: Timezone offset in seconds east of utc.
845
:param timezone: How to display the time: 'utc', 'original' for the
846
timezone specified by offset, or 'local' for the process's current
848
:param date_fmt: strftime format.
849
:param show_offset: Whether to append the timezone.
851
(date_fmt, tt, offset_str) = \
852
_format_date(t, offset, timezone, date_fmt, show_offset)
853
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
854
date_str = time.strftime(date_fmt, tt)
855
return date_str + offset_str
858
# Cache of formatted offset strings
862
def format_date_with_offset_in_original_timezone(t, offset=0,
863
_cache=_offset_cache):
864
"""Return a formatted date string in the original timezone.
866
This routine may be faster then format_date.
868
:param t: Seconds since the epoch.
869
:param offset: Timezone offset in seconds east of utc.
873
tt = time.gmtime(t + offset)
874
date_fmt = _default_format_by_weekday_num[tt[6]]
875
date_str = time.strftime(date_fmt, tt)
876
offset_str = _cache.get(offset, None)
877
if offset_str is None:
878
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
879
_cache[offset] = offset_str
880
return date_str + offset_str
883
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
885
"""Return an unicode date string formatted according to the current locale.
887
:param t: Seconds since the epoch.
888
:param offset: Timezone offset in seconds east of utc.
889
:param timezone: How to display the time: 'utc', 'original' for the
890
timezone specified by offset, or 'local' for the process's current
892
:param date_fmt: strftime format.
893
:param show_offset: Whether to append the timezone.
895
(date_fmt, tt, offset_str) = \
896
_format_date(t, offset, timezone, date_fmt, show_offset)
897
date_str = time.strftime(date_fmt, tt)
898
if not isinstance(date_str, text_type):
899
date_str = date_str.decode(get_user_encoding(), 'replace')
900
return date_str + offset_str
903
def _format_date(t, offset, timezone, date_fmt, show_offset):
904
if timezone == 'utc':
907
elif timezone == 'original':
910
tt = time.gmtime(t + offset)
911
elif timezone == 'local':
912
tt = time.localtime(t)
913
offset = local_time_offset(t)
915
raise UnsupportedTimezoneFormat(timezone)
917
date_fmt = "%a %Y-%m-%d %H:%M:%S"
919
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
922
return (date_fmt, tt, offset_str)
925
def compact_date(when):
926
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
929
def format_delta(delta):
930
"""Get a nice looking string for a time delta.
932
:param delta: The time difference in seconds, can be positive or negative.
933
positive indicates time in the past, negative indicates time in the
934
future. (usually time.time() - stored_time)
935
:return: String formatted to show approximate resolution
941
direction = 'in the future'
945
if seconds < 90: # print seconds up to 90 seconds
947
return '%d second %s' % (seconds, direction,)
949
return '%d seconds %s' % (seconds, direction)
951
minutes = int(seconds / 60)
952
seconds -= 60 * minutes
957
if minutes < 90: # print minutes, seconds up to 90 minutes
959
return '%d minute, %d second%s %s' % (
960
minutes, seconds, plural_seconds, direction)
962
return '%d minutes, %d second%s %s' % (
963
minutes, seconds, plural_seconds, direction)
965
hours = int(minutes / 60)
966
minutes -= 60 * hours
973
return '%d hour, %d minute%s %s' % (hours, minutes,
974
plural_minutes, direction)
975
return '%d hours, %d minute%s %s' % (hours, minutes,
976
plural_minutes, direction)
980
"""Return size of given open file."""
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":
993
except NotImplementedError:
994
# not well seeded, but better than nothing
999
s += chr(random.randint(0, 255))
1004
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
1007
def rand_chars(num):
1008
"""Return a random string of num alphanumeric characters
1010
The result only contains lowercase chars because it may be used on
1011
case-insensitive filesystems.
1014
for raw_byte in rand_bytes(num):
1016
s += ALNUM[ord(raw_byte) % 36]
1018
s += ALNUM[raw_byte % 36]
1022
# TODO: We could later have path objects that remember their list
1023
# decomposition (might be too tricksy though.)
1026
"""Turn string into list of parts."""
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 = ('.', '')
1051
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1052
elif f in current_empty_dir:
1061
if (f == '..') or (f is None) or (f == ''):
1062
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1066
def parent_directories(filename):
1067
"""Return the list of parent directories, deepest first.
1069
For example, parent_directories("a/b/c") -> ["a/b", "a"].
1072
parts = splitpath(dirname(filename))
1074
parents.append(joinpath(parts))
1079
_extension_load_failures = []
1082
def failed_to_load_extension(exception):
1083
"""Handle failing to load a binary extension.
1085
This should be called from the ImportError block guarding the attempt to
1086
import the native extension. If this function returns, the pure-Python
1087
implementation should be loaded instead::
1090
>>> import breezy._fictional_extension_pyx
1091
>>> except ImportError, e:
1092
>>> breezy.osutils.failed_to_load_extension(e)
1093
>>> import breezy._fictional_extension_py
1095
# NB: This docstring is just an example, not a doctest, because doctest
1096
# currently can't cope with the use of lazy imports in this namespace --
1099
# This currently doesn't report the failure at the time it occurs, because
1100
# they tend to happen very early in startup when we can't check config
1101
# files etc, and also we want to report all failures but not spam the user
1103
exception_str = str(exception)
1104
if exception_str not in _extension_load_failures:
1105
trace.mutter("failed to load compiled extension: %s" % exception_str)
1106
_extension_load_failures.append(exception_str)
1109
def report_extension_load_failures():
1110
if not _extension_load_failures:
1112
if config.GlobalConfig().suppress_warning('missing_extensions'):
1114
# the warnings framework should by default show this only once
1115
from .trace import warning
1117
"brz: warning: some compiled extensions could not be loaded; "
1118
"see ``brz help missing-extensions``")
1119
# we no longer show the specific missing extensions here, because it makes
1120
# the message too long and scary - see
1121
# https://bugs.launchpad.net/bzr/+bug/430529
1125
from ._chunks_to_lines_pyx import chunks_to_lines
1126
except ImportError as e:
1127
failed_to_load_extension(e)
1128
from ._chunks_to_lines_py import chunks_to_lines
1132
"""Split s into lines, but without removing the newline characters."""
1133
# Trivially convert a fulltext into a 'chunked' representation, and let
1134
# chunks_to_lines do the heavy lifting.
1135
if isinstance(s, bytes):
1136
# chunks_to_lines only supports 8-bit strings
1137
return chunks_to_lines([s])
1139
return _split_lines(s)
1142
def _split_lines(s):
1143
"""Split s into lines, but without removing the newline characters.
1145
This supports Unicode or plain string objects.
1147
nl = b'\n' if isinstance(s, bytes) else u'\n'
1149
result = [line + nl for line in lines[:-1]]
1151
result.append(lines[-1])
1155
def hardlinks_good():
1156
return sys.platform not in ('win32', 'cygwin', 'darwin')
1159
def link_or_copy(src, dest):
1160
"""Hardlink a file, or copy it if it can't be hardlinked."""
1161
if not hardlinks_good():
1162
shutil.copyfile(src, dest)
1166
except (OSError, IOError) as e:
1167
if e.errno != errno.EXDEV:
1169
shutil.copyfile(src, dest)
1172
def delete_any(path):
1173
"""Delete a file, symlink or directory.
1175
Will delete even if readonly.
1178
_delete_file_or_dir(path)
1179
except (OSError, IOError) as e:
1180
if e.errno in (errno.EPERM, errno.EACCES):
1181
# make writable and try again
1184
except (OSError, IOError):
1186
_delete_file_or_dir(path)
1191
def _delete_file_or_dir(path):
1192
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1193
# Forgiveness than Permission (EAFP) because:
1194
# - root can damage a solaris file system by using unlink,
1195
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1196
# EACCES, OSX: EPERM) when invoked on a directory.
1197
if isdir(path): # Takes care of symlinks
1204
if getattr(os, 'symlink', None) is not None:
1210
def has_hardlinks():
1211
if getattr(os, 'link', None) is not None:
1217
def host_os_dereferences_symlinks():
1218
return (has_symlinks()
1219
and sys.platform not in ('cygwin', 'win32'))
1222
def readlink(abspath):
1223
"""Return a string representing the path to which the symbolic link points.
1225
:param abspath: The link absolute unicode path.
1227
This his guaranteed to return the symbolic link in unicode in all python
1230
link = abspath.encode(_fs_enc)
1231
target = os.readlink(link)
1232
target = target.decode(_fs_enc)
1236
def contains_whitespace(s):
1237
"""True if there are any whitespace characters in s."""
1238
# string.whitespace can include '\xa0' in certain locales, because it is
1239
# considered "non-breaking-space" as part of ISO-8859-1. But it
1240
# 1) Isn't a breaking whitespace
1241
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1243
# 3) '\xa0' isn't unicode safe since it is >128.
1245
if isinstance(s, str):
1248
ws = (b' ', b'\t', b'\n', b'\r', b'\v', b'\f')
1256
def contains_linebreaks(s):
1257
"""True if there is any vertical whitespace in s."""
1265
def relpath(base, path):
1266
"""Return path relative to base, or raise PathNotChild exception.
1268
The path may be either an absolute path or a path relative to the
1269
current working directory.
1271
os.path.commonprefix (python2.4) has a bad bug that it works just
1272
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1273
avoids that problem.
1275
NOTE: `base` should not have a trailing slash otherwise you'll get
1276
PathNotChild exceptions regardless of `path`.
1279
if len(base) < MIN_ABS_PATHLENGTH:
1280
# must have space for e.g. a drive letter
1281
raise ValueError(gettext('%r is too short to calculate a relative path')
1289
if len(head) <= len(base) and head != base:
1290
raise errors.PathNotChild(rp, base)
1293
head, tail = split(head)
1298
return pathjoin(*reversed(s))
1303
def _cicp_canonical_relpath(base, path):
1304
"""Return the canonical path relative to base.
1306
Like relpath, but on case-insensitive-case-preserving file-systems, this
1307
will return the relpath as stored on the file-system rather than in the
1308
case specified in the input string, for all existing portions of the path.
1310
This will cause O(N) behaviour if called for every path in a tree; if you
1311
have a number of paths to convert, you should use canonical_relpaths().
1313
# TODO: it should be possible to optimize this for Windows by using the
1314
# win32 API FindFiles function to look for the specified name - but using
1315
# os.listdir() still gives us the correct, platform agnostic semantics in
1318
rel = relpath(base, path)
1319
# '.' will have been turned into ''
1323
abs_base = abspath(base)
1325
_listdir = os.listdir
1327
# use an explicit iterator so we can easily consume the rest on early exit.
1328
bit_iter = iter(rel.split('/'))
1329
for bit in bit_iter:
1332
next_entries = _listdir(current)
1333
except OSError: # enoent, eperm, etc
1334
# We can't find this in the filesystem, so just append the
1336
current = pathjoin(current, bit, *list(bit_iter))
1338
for look in next_entries:
1339
if lbit == look.lower():
1340
current = pathjoin(current, look)
1343
# got to the end, nothing matched, so we just return the
1344
# non-existing bits as they were specified (the filename may be
1345
# the target of a move, for example).
1346
current = pathjoin(current, bit, *list(bit_iter))
1348
return current[len(abs_base):].lstrip('/')
1351
# XXX - TODO - we need better detection/integration of case-insensitive
1352
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1353
# filesystems), for example, so could probably benefit from the same basic
1354
# support there. For now though, only Windows and OSX get that support, and
1355
# they get it for *all* file-systems!
1356
if sys.platform in ('win32', 'darwin'):
1357
canonical_relpath = _cicp_canonical_relpath
1359
canonical_relpath = relpath
1362
def canonical_relpaths(base, paths):
1363
"""Create an iterable to canonicalize a sequence of relative paths.
1365
The intent is for this implementation to use a cache, vastly speeding
1366
up multiple transformations in the same directory.
1368
# but for now, we haven't optimized...
1369
return [canonical_relpath(base, p) for p in paths]
1372
def decode_filename(filename):
1373
"""Decode the filename using the filesystem encoding
1375
If it is unicode, it is returned.
1376
Otherwise it is decoded from the the filesystem's encoding. If decoding
1377
fails, a errors.BadFilenameEncoding exception is raised.
1379
if isinstance(filename, text_type):
1382
return filename.decode(_fs_enc)
1383
except UnicodeDecodeError:
1384
raise errors.BadFilenameEncoding(filename, _fs_enc)
1387
def safe_unicode(unicode_or_utf8_string):
1388
"""Coerce unicode_or_utf8_string into unicode.
1390
If it is unicode, it is returned.
1391
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1392
wrapped in a BzrBadParameterNotUnicode exception.
1394
if isinstance(unicode_or_utf8_string, text_type):
1395
return unicode_or_utf8_string
1397
return unicode_or_utf8_string.decode('utf8')
1398
except UnicodeDecodeError:
1399
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1402
def safe_utf8(unicode_or_utf8_string):
1403
"""Coerce unicode_or_utf8_string to a utf8 string.
1405
If it is a str, it is returned.
1406
If it is Unicode, it is encoded into a utf-8 string.
1408
if isinstance(unicode_or_utf8_string, bytes):
1409
# TODO: jam 20070209 This is overkill, and probably has an impact on
1410
# performance if we are dealing with lots of apis that want a
1413
# Make sure it is a valid utf-8 string
1414
unicode_or_utf8_string.decode('utf-8')
1415
except UnicodeDecodeError:
1416
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1417
return unicode_or_utf8_string
1418
return unicode_or_utf8_string.encode('utf-8')
1421
_platform_normalizes_filenames = False
1422
if sys.platform == 'darwin':
1423
_platform_normalizes_filenames = True
1426
def normalizes_filenames():
1427
"""Return True if this platform normalizes unicode filenames.
1431
return _platform_normalizes_filenames
1434
def _accessible_normalized_filename(path):
1435
"""Get the unicode normalized path, and if you can access the file.
1437
On platforms where the system normalizes filenames (Mac OSX),
1438
you can access a file by any path which will normalize correctly.
1439
On platforms where the system does not normalize filenames
1440
(everything else), you have to access a file by its exact path.
1442
Internally, bzr only supports NFC normalization, since that is
1443
the standard for XML documents.
1445
So return the normalized path, and a flag indicating if the file
1446
can be accessed by that path.
1449
if isinstance(path, bytes):
1450
path = path.decode(sys.getfilesystemencoding())
1451
return unicodedata.normalize('NFC', path), True
1454
def _inaccessible_normalized_filename(path):
1455
__doc__ = _accessible_normalized_filename.__doc__
1457
if isinstance(path, bytes):
1458
path = path.decode(sys.getfilesystemencoding())
1459
normalized = unicodedata.normalize('NFC', path)
1460
return normalized, normalized == path
1463
if _platform_normalizes_filenames:
1464
normalized_filename = _accessible_normalized_filename
1466
normalized_filename = _inaccessible_normalized_filename
1469
def set_signal_handler(signum, handler, restart_syscall=True):
1470
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1471
on platforms that support that.
1473
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1474
automatically restart (by calling `signal.siginterrupt(signum,
1475
False)`). May be ignored if the feature is not available on this
1476
platform or Python version.
1480
siginterrupt = signal.siginterrupt
1482
# This python implementation doesn't provide signal support, hence no
1485
except AttributeError:
1486
# siginterrupt doesn't exist on this platform, or for this version
1488
def siginterrupt(signum, flag): return None
1490
def sig_handler(*args):
1491
# Python resets the siginterrupt flag when a signal is
1492
# received. <http://bugs.python.org/issue8354>
1493
# As a workaround for some cases, set it back the way we want it.
1494
siginterrupt(signum, False)
1495
# Now run the handler function passed to set_signal_handler.
1498
sig_handler = handler
1499
old_handler = signal.signal(signum, sig_handler)
1501
siginterrupt(signum, False)
1505
default_terminal_width = 80
1506
"""The default terminal width for ttys.
1508
This is defined so that higher levels can share a common fallback value when
1509
terminal_width() returns None.
1512
# Keep some state so that terminal_width can detect if _terminal_size has
1513
# returned a different size since the process started. See docstring and
1514
# comments of terminal_width for details.
1515
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1516
_terminal_size_state = 'no_data'
1517
_first_terminal_size = None
1520
def terminal_width():
1521
"""Return terminal width.
1523
None is returned if the width can't established precisely.
1526
- if BRZ_COLUMNS is set, returns its value
1527
- if there is no controlling terminal, returns None
1528
- query the OS, if the queried size has changed since the last query,
1530
- if COLUMNS is set, returns its value,
1531
- if the OS has a value (even though it's never changed), return its value.
1533
From there, we need to query the OS to get the size of the controlling
1536
On Unices we query the OS by:
1537
- get termios.TIOCGWINSZ
1538
- if an error occurs or a negative value is obtained, returns None
1540
On Windows we query the OS by:
1541
- win32utils.get_console_size() decides,
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
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
1560
width = int(os.environ['BRZ_COLUMNS'])
1561
except (KeyError, ValueError):
1563
if width is not None:
1569
isatty = getattr(sys.stdout, 'isatty', None)
1570
if isatty is None or not isatty():
1571
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
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.
1592
return int(os.environ['COLUMNS'])
1593
except (KeyError, ValueError):
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.
1605
def _win32_terminal_size(width, height):
1606
width, height = win32utils.get_console_size(
1607
defaultx=width, defaulty=height)
1608
return width, height
1611
def _ioctl_terminal_size(width, height):
1616
s = struct.pack('HHHH', 0, 0, 0, 0)
1617
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1618
height, width = struct.unpack('HHHH', x)[0:2]
1619
except (IOError, AttributeError):
1621
return width, height
1624
_terminal_size = None
1625
"""Returns the terminal size as (width, height).
1627
:param width: Default value for width.
1628
:param height: Default value for height.
1630
This is defined specifically for each OS and query the size of the controlling
1631
terminal. If any error occurs, the provided default values should be returned.
1633
if sys.platform == 'win32':
1634
_terminal_size = _win32_terminal_size
1636
_terminal_size = _ioctl_terminal_size
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
1675
def supports_posix_readonly():
1676
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1678
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1679
directory controls creation/deletion, etc.
1681
And under win32, readonly means that the directory itself cannot be
1682
deleted. The contents of a readonly directory can be changed, unlike POSIX
1683
where files in readonly directories cannot be added, deleted or renamed.
1685
return sys.platform != "win32"
1688
def set_or_unset_env(env_variable, value):
1689
"""Modify the environment, setting or removing the env_variable.
1691
:param env_variable: The environment variable in question
1692
:param value: The value to set the environment to. If None, then
1693
the variable will be removed.
1694
:return: The original value of the environment variable.
1696
orig_val = os.environ.get(env_variable)
1698
if orig_val is not None:
1699
del os.environ[env_variable]
1701
if not PY3 and isinstance(value, text_type):
1702
value = value.encode(get_user_encoding())
1703
os.environ[env_variable] = value
1707
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1710
def check_legal_path(path):
1711
"""Check whether the supplied path is legal.
1712
This is only required on Windows, so we don't test on other platforms
1715
if sys.platform != "win32":
1717
if _validWin32PathRE.match(path) is None:
1718
raise errors.IllegalPath(path)
1721
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1724
def _is_error_enotdir(e):
1725
"""Check if this exception represents ENOTDIR.
1727
Unfortunately, python is very inconsistent about the exception
1728
here. The cases are:
1729
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1730
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1731
which is the windows error code.
1732
3) Windows, Python2.5 uses errno == EINVAL and
1733
winerror == ERROR_DIRECTORY
1735
:param e: An Exception object (expected to be OSError with an errno
1736
attribute, but we should be able to cope with anything)
1737
:return: True if this represents an ENOTDIR error. False otherwise.
1739
en = getattr(e, 'errno', None)
1740
if (en == errno.ENOTDIR or
1741
(sys.platform == 'win32' and
1742
(en == _WIN32_ERROR_DIRECTORY or
1744
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1750
def walkdirs(top, prefix=""):
1751
"""Yield data about all the directories in a tree.
1753
This yields all the data about the contents of a directory at a time.
1754
After each directory has been yielded, if the caller has mutated the list
1755
to exclude some directories, they are then not descended into.
1757
The data yielded is of the form:
1758
((directory-relpath, directory-path-from-top),
1759
[(relpath, basename, kind, lstat, path-from-top), ...]),
1760
- directory-relpath is the relative path of the directory being returned
1761
with respect to top. prefix is prepended to this.
1762
- directory-path-from-root is the path including top for this directory.
1763
It is suitable for use with os functions.
1764
- relpath is the relative path within the subtree being walked.
1765
- basename is the basename of the path
1766
- kind is the kind of the file now. If unknown then the file is not
1767
present within the tree - but it may be recorded as versioned. See
1769
- lstat is the stat data *if* the file was statted.
1770
- planned, not implemented:
1771
path_from_tree_root is the path from the root of the tree.
1773
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1774
allows one to walk a subtree but get paths that are relative to a tree
1776
:return: an iterator over the dirs.
1778
# TODO there is a bit of a smell where the results of the directory-
1779
# summary in this, and the path from the root, may not agree
1780
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1781
# potentially confusing output. We should make this more robust - but
1782
# not at a speed cost. RBC 20060731
1784
_directory = _directory_kind
1785
_listdir = os.listdir
1786
_kind_from_mode = file_kind_from_stat_mode
1787
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1789
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1790
relroot, _, _, _, top = pending.pop()
1792
relprefix = relroot + u'/'
1795
top_slash = top + u'/'
1798
append = dirblock.append
1800
names = sorted(map(decode_filename, _listdir(top)))
1801
except OSError as e:
1802
if not _is_error_enotdir(e):
1806
abspath = top_slash + name
1807
statvalue = _lstat(abspath)
1808
kind = _kind_from_mode(statvalue.st_mode)
1809
append((relprefix + name, name, kind, statvalue, abspath))
1810
yield (relroot, top), dirblock
1812
# push the user specified dirs from dirblock
1813
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1816
class DirReader(object):
1817
"""An interface for reading directories."""
1819
def top_prefix_to_starting_dir(self, top, prefix=""):
1820
"""Converts top and prefix to a starting dir entry
1822
:param top: A utf8 path
1823
:param prefix: An optional utf8 path to prefix output relative paths
1825
:return: A tuple starting with prefix, and ending with the native
1828
raise NotImplementedError(self.top_prefix_to_starting_dir)
1830
def read_dir(self, prefix, top):
1831
"""Read a specific dir.
1833
:param prefix: A utf8 prefix to be preprended to the path basenames.
1834
:param top: A natively encoded path to read.
1835
:return: A list of the directories contents. Each item contains:
1836
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1838
raise NotImplementedError(self.read_dir)
1841
_selected_dir_reader = None
1844
def _walkdirs_utf8(top, prefix=""):
1845
"""Yield data about all the directories in a tree.
1847
This yields the same information as walkdirs() only each entry is yielded
1848
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1849
are returned as exact byte-strings.
1851
:return: yields a tuple of (dir_info, [file_info])
1852
dir_info is (utf8_relpath, path-from-top)
1853
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1854
if top is an absolute path, path-from-top is also an absolute path.
1855
path-from-top might be unicode or utf8, but it is the correct path to
1856
pass to os functions to affect the file in question. (such as os.lstat)
1858
global _selected_dir_reader
1859
if _selected_dir_reader is None:
1860
if sys.platform == "win32":
1862
from ._walkdirs_win32 import Win32ReadDir
1863
_selected_dir_reader = Win32ReadDir()
1866
elif _fs_enc in ('utf-8', 'ascii'):
1868
from ._readdir_pyx import UTF8DirReader
1869
_selected_dir_reader = UTF8DirReader()
1870
except ImportError as e:
1871
failed_to_load_extension(e)
1874
if _selected_dir_reader is None:
1875
# Fallback to the python version
1876
_selected_dir_reader = UnicodeDirReader()
1878
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1879
# But we don't actually uses 1-3 in pending, so set them to None
1880
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1881
read_dir = _selected_dir_reader.read_dir
1882
_directory = _directory_kind
1884
relroot, _, _, _, top = pending[-1].pop()
1887
dirblock = sorted(read_dir(relroot, top))
1888
yield (relroot, top), dirblock
1889
# push the user specified dirs from dirblock
1890
next = [d for d in reversed(dirblock) if d[2] == _directory]
1892
pending.append(next)
1895
class UnicodeDirReader(DirReader):
1896
"""A dir reader for non-utf8 file systems, which transcodes."""
1898
__slots__ = ['_utf8_encode']
1901
self._utf8_encode = codecs.getencoder('utf8')
1903
def top_prefix_to_starting_dir(self, top, prefix=""):
1904
"""See DirReader.top_prefix_to_starting_dir."""
1905
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1907
def read_dir(self, prefix, top):
1908
"""Read a single directory from a non-utf8 file system.
1910
top, and the abspath element in the output are unicode, all other paths
1911
are utf8. Local disk IO is done via unicode calls to listdir etc.
1913
This is currently the fallback code path when the filesystem encoding is
1914
not UTF-8. It may be better to implement an alternative so that we can
1915
safely handle paths that are not properly decodable in the current
1918
See DirReader.read_dir for details.
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)
1926
_listdir = os.listdir
1927
_kind_from_mode = file_kind_from_stat_mode
1930
relprefix = prefix + b'/'
1933
top_slash = top + '/'
1936
append = dirblock.append
1937
for name_native in _listdir(top.encode('utf-8')):
1939
name = _fs_decode(name_native)
1940
except UnicodeDecodeError:
1941
raise errors.BadFilenameEncoding(
1942
relprefix + name_native, _fs_enc)
1943
name_utf8 = _utf8_encode(name)[0]
1944
abspath = top_slash + name
1945
statvalue = _lstat(abspath)
1946
kind = _kind_from_mode(statvalue.st_mode)
1947
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1948
return sorted(dirblock)
1951
def copy_tree(from_path, to_path, handlers={}):
1952
"""Copy all of the entries in from_path into to_path.
1954
:param from_path: The base directory to copy.
1955
:param to_path: The target directory. If it does not exist, it will
1957
:param handlers: A dictionary of functions, which takes a source and
1958
destinations for files, directories, etc.
1959
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1960
'file', 'directory', and 'symlink' should always exist.
1961
If they are missing, they will be replaced with 'os.mkdir()',
1962
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1964
# Now, just copy the existing cached tree to the new location
1965
# We use a cheap trick here.
1966
# Absolute paths are prefixed with the first parameter
1967
# relative paths are prefixed with the second.
1968
# So we can get both the source and target returned
1969
# without any extra work.
1971
def copy_dir(source, dest):
1974
def copy_link(source, dest):
1975
"""Copy the contents of a symlink"""
1976
link_to = os.readlink(source)
1977
os.symlink(link_to, dest)
1979
real_handlers = {'file': shutil.copy2,
1980
'symlink': copy_link,
1981
'directory': copy_dir,
1983
real_handlers.update(handlers)
1985
if not os.path.exists(to_path):
1986
real_handlers['directory'](from_path, to_path)
1988
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1989
for relpath, name, kind, st, abspath in entries:
1990
real_handlers[kind](abspath, relpath)
1993
def copy_ownership_from_path(dst, src=None):
1994
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1996
If src is None, the containing directory is used as source. If chown
1997
fails, the error is ignored and a warning is printed.
1999
chown = getattr(os, 'chown', None)
2004
src = os.path.dirname(dst)
2010
chown(dst, s.st_uid, s.st_gid)
2013
'Unable to copy ownership from "%s" to "%s". '
2014
'You may want to set it manually.', src, dst)
2015
trace.log_exception_quietly()
2018
def path_prefix_key(path):
2019
"""Generate a prefix-order path key for path.
2021
This can be used to sort paths in the same way that walkdirs does.
2023
return (dirname(path), path)
2026
def compare_paths_prefix_order(path_a, path_b):
2027
"""Compare path_a and path_b to generate the same order walkdirs uses."""
2028
key_a = path_prefix_key(path_a)
2029
key_b = path_prefix_key(path_b)
2030
return (key_a > key_b) - (key_a < key_b)
2033
_cached_user_encoding = None
2036
def get_user_encoding():
2037
"""Find out what the preferred user encoding is.
2039
This is generally the encoding that is used for command line parameters
2040
and file contents. This may be different from the terminal encoding
2041
or the filesystem encoding.
2043
:return: A string defining the preferred user encoding
2045
global _cached_user_encoding
2046
if _cached_user_encoding is not None:
2047
return _cached_user_encoding
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)
2060
user_encoding = codecs.lookup(user_encoding).name
2062
if user_encoding not in ("", "cp0"):
2063
sys.stderr.write('brz: warning:'
2064
' unknown encoding %s.'
2065
' Continuing with ascii encoding.\n'
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
2080
return user_encoding
2083
def get_diff_header_encoding():
2084
return get_terminal_encoding()
2087
def get_host_name():
2088
"""Return the current unicode host name.
2090
This is meant to be used in place of socket.gethostname() because that
2091
behaves inconsistently on different platforms.
2093
if sys.platform == "win32":
2094
return win32utils.get_host_name()
2098
return socket.gethostname()
2099
return socket.gethostname().decode(get_user_encoding())
2102
# We must not read/write any more than 64k at a time from/to a socket so we
2103
# don't risk "no buffer space available" errors on some platforms. Windows in
2104
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
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)
2116
def read_bytes_from_socket(sock, report_activity=None,
2117
max_read_size=MAX_SOCKET_CHUNK):
2118
"""Read up to max_read_size of bytes from sock and notify of progress.
2120
Translates "Connection reset by peer" into file-like EOF (return an
2121
empty string rather than raise an error), and repeats the recv if
2122
interrupted by a signal.
2126
data = sock.recv(max_read_size)
2127
except socket.error as e:
2129
if eno in _end_of_stream_errors:
2130
# The connection was closed by the other side. Callers expect
2131
# an empty string to signal end-of-stream.
2133
elif eno == errno.EINTR:
2134
# Retry the interrupted recv.
2138
if report_activity is not None:
2139
report_activity(len(data), 'read')
2143
def recv_all(socket, count):
2144
"""Receive an exact number of bytes.
2146
Regular Socket.recv() may return less than the requested number of bytes,
2147
depending on what's in the OS buffer. MSG_WAITALL is not available
2148
on all platforms, but this should work everywhere. This will return
2149
less than the requested amount if the remote end closes.
2151
This isn't optimized and is intended mostly for use in testing.
2154
while len(b) < count:
2155
new = read_bytes_from_socket(socket, None, count - len(b))
2162
def send_all(sock, bytes, report_activity=None):
2163
"""Send all bytes on a socket.
2165
Breaks large blocks in smaller chunks to avoid buffering limitations on
2166
some platforms, and catches EINTR which may be thrown if the send is
2167
interrupted by a signal.
2169
This is preferred to socket.sendall(), because it avoids portability bugs
2170
and provides activity reporting.
2172
:param report_activity: Call this as bytes are read, see
2173
Transport._report_activity
2176
byte_count = len(bytes)
2177
view = memoryview(bytes)
2178
while sent_total < byte_count:
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)
2185
if e.args[0] != errno.EINTR:
2189
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
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:
2220
def dereference_path(path):
2221
"""Determine the real path to a file.
2223
All parent elements are dereferenced. But the file itself is not
2225
:param path: The original path. May be absolute or relative.
2226
:return: the real path *to* the file
2228
parent, base = os.path.split(path)
2229
# The pathjoin for '.' is a workaround for Python bug #1213894.
2230
# (initial path components aren't dereferenced)
2231
return pathjoin(realpath(pathjoin('.', parent)), base)
2234
def supports_mapi():
2235
"""Return True if we can use MAPI to launch a mail client."""
2236
return sys.platform == "win32"
2239
def resource_string(package, resource_name):
2240
"""Load a resource from a package and return it as a string.
2242
Note: Only packages that start with breezy are currently supported.
2244
This is designed to be a lightweight implementation of resource
2245
loading in a way which is API compatible with the same API from
2247
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2248
If and when pkg_resources becomes a standard library, this routine
2251
# Check package name is within breezy
2252
if package == "breezy":
2253
resource_relpath = resource_name
2254
elif package.startswith("breezy."):
2255
package = package[len("breezy."):].replace('.', os.sep)
2256
resource_relpath = pathjoin(package, resource_name)
2258
raise errors.BzrError('resource package %s not in breezy' % package)
2260
# Map the resource to a file and read its contents
2261
base = dirname(breezy.__file__)
2262
if getattr(sys, 'frozen', None): # bzr.exe
2263
base = abspath(pathjoin(base, '..', '..'))
2264
with open(pathjoin(base, resource_relpath), "rt") as f:
2268
def file_kind_from_stat_mode_thunk(mode):
2269
global file_kind_from_stat_mode
2270
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2272
from ._readdir_pyx import UTF8DirReader
2273
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2275
# This is one time where we won't warn that an extension failed to
2276
# load. The extension is never available on Windows anyway.
2277
from ._readdir_py import (
2278
_kind_from_mode as file_kind_from_stat_mode
2280
return file_kind_from_stat_mode(mode)
2283
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2286
def file_stat(f, _lstat=os.lstat):
2290
except OSError as e:
2291
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
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)
2301
def until_no_eintr(f, *a, **kw):
2302
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2304
WARNING: you must be certain that it is safe to retry the call repeatedly
2305
if EINTR does occur. This is typically only true for low-level operations
2306
like os.read. If in any doubt, don't use this.
2308
Keep in mind that this is not a complete solution to EINTR. There is
2309
probably code in the Python standard library and other dependencies that
2310
may encounter EINTR if a signal arrives (and there is signal handler for
2311
that signal). So this function can reduce the impact for IO that breezy
2312
directly controls, but it is not a complete solution.
2314
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2318
except (IOError, OSError) as e:
2319
if e.errno == errno.EINTR:
2324
if sys.platform == "win32":
2327
return msvcrt.getch()
2332
fd = sys.stdin.fileno()
2333
settings = termios.tcgetattr(fd)
2336
ch = sys.stdin.read(1)
2338
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2341
if sys.platform.startswith('linux'):
2342
def _local_concurrency():
2344
return os.sysconf('SC_NPROCESSORS_ONLN')
2345
except (ValueError, OSError, AttributeError):
2347
elif sys.platform == 'darwin':
2348
def _local_concurrency():
2349
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2350
stdout=subprocess.PIPE).communicate()[0]
2351
elif "bsd" in sys.platform:
2352
def _local_concurrency():
2353
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2354
stdout=subprocess.PIPE).communicate()[0]
2355
elif sys.platform == 'sunos5':
2356
def _local_concurrency():
2357
return subprocess.Popen(['psrinfo', '-p', ],
2358
stdout=subprocess.PIPE).communicate()[0]
2359
elif sys.platform == "win32":
2360
def _local_concurrency():
2361
# This appears to return the number of cores.
2362
return os.environ.get('NUMBER_OF_PROCESSORS')
2364
def _local_concurrency():
2369
_cached_local_concurrency = None
2372
def local_concurrency(use_cache=True):
2373
"""Return how many processes can be run concurrently.
2375
Rely on platform specific implementations and default to 1 (one) if
2376
anything goes wrong.
2378
global _cached_local_concurrency
2380
if _cached_local_concurrency is not None and use_cache:
2381
return _cached_local_concurrency
2383
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2384
if concurrency is None:
2385
import multiprocessing
2387
concurrency = multiprocessing.cpu_count()
2388
except NotImplementedError:
2389
# multiprocessing.cpu_count() isn't implemented on all platforms
2391
concurrency = _local_concurrency()
2392
except (OSError, IOError):
2395
concurrency = int(concurrency)
2396
except (TypeError, ValueError):
2399
_cached_local_concurrency = concurrency
2403
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2404
"""A stream writer that doesn't decode str arguments."""
2406
def __init__(self, encode, stream, errors='strict'):
2407
codecs.StreamWriter.__init__(self, stream, errors)
2408
self.encode = encode
2410
def write(self, object):
2411
if isinstance(object, str):
2412
self.stream.write(object)
2414
data, _ = self.encode(object, self.errors)
2415
self.stream.write(data)
2418
if sys.platform == 'win32':
2419
def open_file(filename, mode='r', bufsize=-1):
2420
"""This function is used to override the ``open`` builtin.
2422
But it uses O_NOINHERIT flag so the file handle is not inherited by
2423
child processes. Deleting or renaming a closed file opened with this
2424
function is not blocking child processes.
2426
writing = 'w' in mode
2427
appending = 'a' in mode
2428
updating = '+' in mode
2429
binary = 'b' in mode
2432
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2433
# for flags for each modes.
2443
flags |= os.O_WRONLY
2444
flags |= os.O_CREAT | os.O_TRUNC
2449
flags |= os.O_WRONLY
2450
flags |= os.O_CREAT | os.O_APPEND
2455
flags |= os.O_RDONLY
2457
return os.fdopen(os.open(filename, flags), mode, bufsize)
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
