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
68
# On win32, O_BINARY is used to indicate the file should
69
# be opened in binary mode, rather than text mode.
70
# On other platforms, O_BINARY doesn't exist, because
71
# they always open in binary mode, so it is okay to
72
# OR with 0 on those platforms.
73
# O_NOINHERIT and O_TEXT exists only on win32 too.
74
O_BINARY = getattr(os, 'O_BINARY', 0)
75
O_TEXT = getattr(os, 'O_TEXT', 0)
76
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
79
class UnsupportedTimezoneFormat(errors.BzrError):
81
_fmt = ('Unsupported timezone format "%(timezone)s", '
82
'options are "utc", "original", "local".')
84
def __init__(self, timezone):
85
self.timezone = timezone
88
def get_unicode_argv():
92
def make_readonly(filename):
93
"""Make a filename read-only."""
94
mod = os.lstat(filename).st_mode
95
if not stat.S_ISLNK(mod):
97
chmod_if_possible(filename, mod)
100
def make_writable(filename):
101
mod = os.lstat(filename).st_mode
102
if not stat.S_ISLNK(mod):
104
chmod_if_possible(filename, mod)
107
def chmod_if_possible(filename, mode):
108
# Set file mode if that can be safely done.
109
# Sometimes even on unix the filesystem won't allow it - see
110
# https://bugs.launchpad.net/bzr/+bug/606537
112
# It is probably faster to just do the chmod, rather than
113
# doing a stat, and then trying to compare
114
os.chmod(filename, mode)
115
except (IOError, OSError) as e:
116
# Permission/access denied seems to commonly happen on smbfs; there's
117
# probably no point warning about it.
118
# <https://bugs.launchpad.net/bzr/+bug/606537>
119
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
120
trace.mutter("ignore error on chmod of %r: %r" % (
126
def minimum_path_selection(paths):
127
"""Return the smallset subset of paths which are outside paths.
129
:param paths: A container (and hence not None) of paths.
130
:return: A set of paths sufficient to include everything in paths via
131
is_inside, drawn from the paths parameter.
137
if isinstance(path, bytes):
138
return path.split(b'/')
140
return path.split('/')
141
sorted_paths = sorted(list(paths), key=sort_key)
143
search_paths = [sorted_paths[0]]
144
for path in sorted_paths[1:]:
145
if not is_inside(search_paths[-1], path):
146
# This path is unique, add it
147
search_paths.append(path)
149
return set(search_paths)
156
"""Return a quoted filename filename
158
This previously used backslash quoting, but that works poorly on
160
# TODO: I'm not really sure this is the best format either.x
162
if _QUOTE_RE is None:
163
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
165
if _QUOTE_RE.search(f):
171
_directory_kind = 'directory'
175
"""Return the current umask"""
176
# Assume that people aren't messing with the umask while running
177
# XXX: This is not thread safe, but there is no way to get the
178
# umask without setting it
186
_directory_kind: "/",
188
'tree-reference': '+',
192
def kind_marker(kind):
194
return _kind_marker_map[kind]
196
# Slightly faster than using .get(, '') when the common case is that
201
lexists = getattr(os.path, 'lexists', None)
205
stat = getattr(os, 'lstat', os.stat)
209
if e.errno == errno.ENOENT:
212
raise errors.BzrError(
213
gettext("lstat/stat of ({0!r}): {1!r}").format(f, e))
216
def fancy_rename(old, new, rename_func, unlink_func):
217
"""A fancy rename, when you don't have atomic rename.
219
:param old: The old path, to rename from
220
:param new: The new path, to rename to
221
:param rename_func: The potentially non-atomic rename function
222
:param unlink_func: A way to delete the target file if the full rename
225
# sftp rename doesn't allow overwriting, so play tricks:
226
base = os.path.basename(new)
227
dirname = os.path.dirname(new)
228
# callers use different encodings for the paths so the following MUST
229
# respect that. We rely on python upcasting to unicode if new is unicode
230
# and keeping a str if not.
231
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
232
os.getpid(), rand_chars(10))
233
tmp_name = pathjoin(dirname, tmp_name)
235
# Rename the file out of the way, but keep track if it didn't exist
236
# We don't want to grab just any exception
237
# something like EACCES should prevent us from continuing
238
# The downside is that the rename_func has to throw an exception
239
# with an errno = ENOENT, or NoSuchFile
242
rename_func(new, tmp_name)
243
except (errors.NoSuchFile,):
246
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
247
# function raises an IOError with errno is None when a rename fails.
248
# This then gets caught here.
249
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
251
except Exception as e:
252
if (getattr(e, 'errno', None) is None
253
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
260
# This may throw an exception, in which case success will
262
rename_func(old, new)
264
except (IOError, OSError) as e:
265
# source and target may be aliases of each other (e.g. on a
266
# case-insensitive filesystem), so we may have accidentally renamed
267
# source by when we tried to rename target
268
if (file_existed and e.errno in (None, errno.ENOENT)
269
and old.lower() == new.lower()):
270
# source and target are the same file on a case-insensitive
271
# filesystem, so we don't generate an exception
277
# If the file used to exist, rename it back into place
278
# otherwise just delete it from the tmp location
280
unlink_func(tmp_name)
282
rename_func(tmp_name, new)
285
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
286
# choke on a Unicode string containing a relative path if
287
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
289
def _posix_abspath(path):
290
# jam 20060426 rather than encoding to fsencoding
291
# copy posixpath.abspath, but use os.getcwdu instead
292
if not posixpath.isabs(path):
293
path = posixpath.join(getcwd(), path)
294
return _posix_normpath(path)
297
def _posix_realpath(path):
298
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
301
def _posix_normpath(path):
302
path = posixpath.normpath(path)
303
# Bug 861008: posixpath.normpath() returns a path normalized according to
304
# the POSIX standard, which stipulates (for compatibility reasons) that two
305
# leading slashes must not be simplified to one, and only if there are 3 or
306
# more should they be simplified as one. So we treat the leading 2 slashes
307
# as a special case here by simply removing the first slash, as we consider
308
# that breaking POSIX compatibility for this obscure feature is acceptable.
309
# This is not a paranoid precaution, as we notably get paths like this when
310
# the repo is hosted at the root of the filesystem, i.e. in "/".
311
if path.startswith('//'):
316
def _posix_path_from_environ(key):
317
"""Get unicode path from `key` in environment or None if not present
319
Note that posix systems use arbitrary byte strings for filesystem objects,
320
so a path that raises BadFilenameEncoding here may still be accessible.
322
return os.environ.get(key, None)
325
def _posix_get_home_dir():
326
"""Get the home directory of the current user as a unicode path"""
327
path = posixpath.expanduser("~")
329
return path.decode(_fs_enc)
330
except AttributeError:
332
except UnicodeDecodeError:
333
raise errors.BadFilenameEncoding(path, _fs_enc)
336
def _posix_getuser_unicode():
337
"""Get username from environment or password database as unicode"""
338
return getpass.getuser()
341
def _win32_fixdrive(path):
342
"""Force drive letters to be consistent.
344
win32 is inconsistent whether it returns lower or upper case
345
and even if it was consistent the user might type the other
346
so we force it to uppercase
347
running python.exe under cmd.exe return capital C:\\
348
running win32 python inside a cygwin shell returns lowercase c:\\
350
drive, path = ntpath.splitdrive(path)
351
return drive.upper() + path
354
def _win32_abspath(path):
355
# Real ntpath.abspath doesn't have a problem with a unicode cwd
356
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
359
def _win32_realpath(path):
360
# Real ntpath.realpath doesn't have a problem with a unicode cwd
361
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
364
def _win32_pathjoin(*args):
365
return ntpath.join(*args).replace('\\', '/')
368
def _win32_normpath(path):
369
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
373
return _win32_fixdrive(_getcwd().replace('\\', '/'))
376
def _win32_mkdtemp(*args, **kwargs):
377
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
380
def _win32_rename(old, new):
381
"""We expect to be able to atomically replace 'new' with old.
383
On win32, if new exists, it must be moved out of the way first,
387
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
389
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
390
# If we try to rename a non-existant file onto cwd, we get
391
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
392
# if the old path doesn't exist, sometimes we get EACCES
393
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
399
return unicodedata.normalize('NFC', _getcwd())
402
def _rename_wrap_exception(rename_func):
403
"""Adds extra information to any exceptions that come from rename().
405
The exception has an updated message and 'old_filename' and 'new_filename'
409
def _rename_wrapper(old, new):
411
rename_func(old, new)
413
detailed_error = OSError(e.errno, e.strerror +
414
" [occurred when renaming '%s' to '%s']" %
416
detailed_error.old_filename = old
417
detailed_error.new_filename = new
420
return _rename_wrapper
426
# Default rename wraps os.rename()
427
rename = _rename_wrap_exception(os.rename)
429
# Default is to just use the python builtins, but these can be rebound on
430
# particular platforms.
431
abspath = _posix_abspath
432
realpath = _posix_realpath
433
pathjoin = os.path.join
434
normpath = _posix_normpath
435
path_from_environ = _posix_path_from_environ
436
_get_home_dir = _posix_get_home_dir
437
getuser_unicode = _posix_getuser_unicode
439
dirname = os.path.dirname
440
basename = os.path.basename
441
split = os.path.split
442
splitext = os.path.splitext
443
# These were already lazily imported into local scope
444
# mkdtemp = tempfile.mkdtemp
445
# rmtree = shutil.rmtree
454
MIN_ABS_PATHLENGTH = 1
457
if sys.platform == 'win32':
458
abspath = _win32_abspath
459
realpath = _win32_realpath
460
pathjoin = _win32_pathjoin
461
normpath = _win32_normpath
462
getcwd = _win32_getcwd
463
mkdtemp = _win32_mkdtemp
464
rename = _rename_wrap_exception(_win32_rename)
466
from . import _walkdirs_win32
470
lstat = _walkdirs_win32.lstat
471
fstat = _walkdirs_win32.fstat
472
wrap_stat = _walkdirs_win32.wrap_stat
474
MIN_ABS_PATHLENGTH = 3
476
def _win32_delete_readonly(function, path, excinfo):
477
"""Error handler for shutil.rmtree function [for win32]
478
Helps to remove files and dirs marked as read-only.
480
exception = excinfo[1]
481
if function in (os.remove, os.rmdir) \
482
and isinstance(exception, OSError) \
483
and exception.errno == errno.EACCES:
489
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
490
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
491
return shutil.rmtree(path, ignore_errors, onerror)
493
get_unicode_argv = getattr(win32utils, 'get_unicode_argv', get_unicode_argv)
494
path_from_environ = win32utils.get_environ_unicode
495
_get_home_dir = win32utils.get_home_location
496
getuser_unicode = win32utils.get_user_name
498
elif sys.platform == 'darwin':
502
def get_terminal_encoding(trace=False):
503
"""Find the best encoding for printing to the screen.
505
This attempts to check both sys.stdout and sys.stdin to see
506
what encoding they are in, and if that fails it falls back to
507
osutils.get_user_encoding().
508
The problem is that on Windows, locale.getpreferredencoding()
509
is not the same encoding as that used by the console:
510
http://mail.python.org/pipermail/python-list/2003-May/162357.html
512
On my standard US Windows XP, the preferred encoding is
513
cp1252, but the console is cp437
515
:param trace: If True trace the selected encoding via mutter().
517
from .trace import mutter
518
output_encoding = getattr(sys.stdout, 'encoding', None)
519
if not output_encoding:
520
input_encoding = getattr(sys.stdin, 'encoding', None)
521
if not input_encoding:
522
output_encoding = get_user_encoding()
524
mutter('encoding stdout as osutils.get_user_encoding() %r',
527
output_encoding = input_encoding
529
mutter('encoding stdout as sys.stdin encoding %r',
533
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
534
if output_encoding == 'cp0':
535
# invalid encoding (cp0 means 'no codepage' on Windows)
536
output_encoding = get_user_encoding()
538
mutter('cp0 is invalid encoding.'
539
' encoding stdout as osutils.get_user_encoding() %r',
543
codecs.lookup(output_encoding)
545
sys.stderr.write('brz: warning:'
546
' unknown terminal encoding %s.\n'
547
' Using encoding %s instead.\n'
548
% (output_encoding, get_user_encoding())
550
output_encoding = get_user_encoding()
552
return output_encoding
555
def normalizepath(f):
556
if getattr(os.path, 'realpath', None) is not None:
560
[p, e] = os.path.split(f)
561
if e == "" or e == "." or e == "..":
564
return pathjoin(F(p), e)
568
"""True if f is an accessible directory."""
570
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
576
"""True if f is a regular file."""
578
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
584
"""True if f is a symlink."""
586
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
591
def is_inside(dir, fname):
592
"""True if fname is inside dir.
594
The parameters should typically be passed to osutils.normpath first, so
595
that . and .. and repeated slashes are eliminated, and the separators
596
are canonical for the platform.
598
The empty string as a dir name is taken as top-of-tree and matches
601
# XXX: Most callers of this can actually do something smarter by
602
# looking at the inventory
609
if isinstance(dir, bytes):
610
if not dir.endswith(b'/'):
613
if not dir.endswith('/'):
616
return fname.startswith(dir)
619
def is_inside_any(dir_list, fname):
620
"""True if fname is inside any of given dirs."""
621
for dirname in dir_list:
622
if is_inside(dirname, fname):
627
def is_inside_or_parent_of_any(dir_list, fname):
628
"""True if fname is a child or a parent of any of the given files."""
629
for dirname in dir_list:
630
if is_inside(dirname, fname) or is_inside(fname, dirname):
635
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
636
report_activity=None, direction='read'):
637
"""Copy contents of one file to another.
639
The read_length can either be -1 to read to end-of-file (EOF) or
640
it can specify the maximum number of bytes to read.
642
The buff_size represents the maximum size for each read operation
643
performed on from_file.
645
:param report_activity: Call this as bytes are read, see
646
Transport._report_activity
647
:param direction: Will be passed to report_activity
649
:return: The number of bytes copied.
653
# read specified number of bytes
655
while read_length > 0:
656
num_bytes_to_read = min(read_length, buff_size)
658
block = from_file.read(num_bytes_to_read)
662
if report_activity is not None:
663
report_activity(len(block), direction)
666
actual_bytes_read = len(block)
667
read_length -= actual_bytes_read
668
length += actual_bytes_read
672
block = from_file.read(buff_size)
676
if report_activity is not None:
677
report_activity(len(block), direction)
683
def pump_string_file(bytes, file_handle, segment_size=None):
684
"""Write bytes to file_handle in many smaller writes.
686
:param bytes: The string to write.
687
:param file_handle: The file to write to.
689
# Write data in chunks rather than all at once, because very large
690
# writes fail on some platforms (e.g. Windows with SMB mounted
693
segment_size = 5242880 # 5MB
694
offsets = range(0, len(bytes), segment_size)
695
view = memoryview(bytes)
696
write = file_handle.write
697
for offset in offsets:
698
write(view[offset:offset + segment_size])
701
def file_iterator(input_file, readsize=32768):
703
b = input_file.read(readsize)
709
# GZ 2017-09-16: Makes sense in general for hexdigest() result to be text, but
710
# used as bytes through most interfaces so encode with this wrapper.
711
def _hexdigest(hashobj):
712
return hashobj.hexdigest().encode()
716
"""Calculate the hexdigest of an open file.
718
The file cursor should be already at the start.
730
def size_sha_file(f):
731
"""Calculate the size and hexdigest of an open file.
733
The file cursor should be already at the start and
734
the caller is responsible for closing the file afterwards.
745
return size, _hexdigest(s)
748
def sha_file_by_name(fname):
749
"""Calculate the SHA1 of a file by reading the full text"""
751
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
754
b = os.read(f, 1 << 16)
762
def sha_strings(strings, _factory=sha):
763
"""Return the sha-1 of concatenation of strings"""
765
for string in strings:
770
def sha_string(f, _factory=sha):
771
# GZ 2017-09-16: Dodgy if factory is ever not sha, probably shouldn't be.
772
return _hexdigest(_factory(f))
775
def fingerprint_file(f):
777
return {'size': len(b),
778
'sha1': _hexdigest(sha(b))}
781
def compare_files(a, b):
782
"""Returns true if equal in contents"""
793
def local_time_offset(t=None):
794
"""Return offset of local zone from GMT, either at present or at time t."""
797
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
798
return offset.days * 86400 + offset.seconds
801
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
802
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
805
def format_date(t, offset=0, timezone='original', date_fmt=None,
807
"""Return a formatted date string.
809
:param t: Seconds since the epoch.
810
:param offset: Timezone offset in seconds east of utc.
811
:param timezone: How to display the time: 'utc', 'original' for the
812
timezone specified by offset, or 'local' for the process's current
814
:param date_fmt: strftime format.
815
:param show_offset: Whether to append the timezone.
817
(date_fmt, tt, offset_str) = \
818
_format_date(t, offset, timezone, date_fmt, show_offset)
819
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
820
date_str = time.strftime(date_fmt, tt)
821
return date_str + offset_str
824
# Cache of formatted offset strings
828
def format_date_with_offset_in_original_timezone(t, offset=0,
829
_cache=_offset_cache):
830
"""Return a formatted date string in the original timezone.
832
This routine may be faster then format_date.
834
:param t: Seconds since the epoch.
835
:param offset: Timezone offset in seconds east of utc.
839
tt = time.gmtime(t + offset)
840
date_fmt = _default_format_by_weekday_num[tt[6]]
841
date_str = time.strftime(date_fmt, tt)
842
offset_str = _cache.get(offset, None)
843
if offset_str is None:
844
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
845
_cache[offset] = offset_str
846
return date_str + offset_str
849
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
851
"""Return an unicode date string formatted according to the current locale.
853
:param t: Seconds since the epoch.
854
:param offset: Timezone offset in seconds east of utc.
855
:param timezone: How to display the time: 'utc', 'original' for the
856
timezone specified by offset, or 'local' for the process's current
858
:param date_fmt: strftime format.
859
:param show_offset: Whether to append the timezone.
861
(date_fmt, tt, offset_str) = \
862
_format_date(t, offset, timezone, date_fmt, show_offset)
863
date_str = time.strftime(date_fmt, tt)
864
if not isinstance(date_str, str):
865
date_str = date_str.decode(get_user_encoding(), 'replace')
866
return date_str + offset_str
869
def _format_date(t, offset, timezone, date_fmt, show_offset):
870
if timezone == 'utc':
873
elif timezone == 'original':
876
tt = time.gmtime(t + offset)
877
elif timezone == 'local':
878
tt = time.localtime(t)
879
offset = local_time_offset(t)
881
raise UnsupportedTimezoneFormat(timezone)
883
date_fmt = "%a %Y-%m-%d %H:%M:%S"
885
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
888
return (date_fmt, tt, offset_str)
891
def compact_date(when):
892
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
895
def format_delta(delta):
896
"""Get a nice looking string for a time delta.
898
:param delta: The time difference in seconds, can be positive or negative.
899
positive indicates time in the past, negative indicates time in the
900
future. (usually time.time() - stored_time)
901
:return: String formatted to show approximate resolution
907
direction = 'in the future'
911
if seconds < 90: # print seconds up to 90 seconds
913
return '%d second %s' % (seconds, direction,)
915
return '%d seconds %s' % (seconds, direction)
917
minutes = int(seconds / 60)
918
seconds -= 60 * minutes
923
if minutes < 90: # print minutes, seconds up to 90 minutes
925
return '%d minute, %d second%s %s' % (
926
minutes, seconds, plural_seconds, direction)
928
return '%d minutes, %d second%s %s' % (
929
minutes, seconds, plural_seconds, direction)
931
hours = int(minutes / 60)
932
minutes -= 60 * hours
939
return '%d hour, %d minute%s %s' % (hours, minutes,
940
plural_minutes, direction)
941
return '%d hours, %d minute%s %s' % (hours, minutes,
942
plural_minutes, direction)
946
"""Return size of given open file."""
947
return os.fstat(f.fileno())[stat.ST_SIZE]
950
# Alias os.urandom to support platforms (which?) without /dev/urandom and
951
# override if it doesn't work. Avoid checking on windows where there is
952
# significant initialisation cost that can be avoided for some bzr calls.
954
rand_bytes = os.urandom
956
if rand_bytes.__module__ != "nt":
959
except NotImplementedError:
960
# not well seeded, but better than nothing
965
s += chr(random.randint(0, 255))
970
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
974
"""Return a random string of num alphanumeric characters
976
The result only contains lowercase chars because it may be used on
977
case-insensitive filesystems.
980
for raw_byte in rand_bytes(num):
981
s += ALNUM[raw_byte % 36]
985
# TODO: We could later have path objects that remember their list
986
# decomposition (might be too tricksy though.)
989
"""Turn string into list of parts."""
990
use_bytes = isinstance(p, bytes)
991
if os.path.sep == '\\':
992
# split on either delimiter because people might use either on
995
ps = re.split(b'[\\\\/]', p)
997
ps = re.split(r'[\\/]', p)
1006
current_empty_dir = (b'.', b'')
1009
current_empty_dir = ('.', '')
1014
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1015
elif f in current_empty_dir:
1024
if (f == '..') or (f is None) or (f == ''):
1025
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1029
def parent_directories(filename):
1030
"""Return the list of parent directories, deepest first.
1032
For example, parent_directories("a/b/c") -> ["a/b", "a"].
1035
parts = splitpath(dirname(filename))
1037
parents.append(joinpath(parts))
1042
_extension_load_failures = []
1045
def failed_to_load_extension(exception):
1046
"""Handle failing to load a binary extension.
1048
This should be called from the ImportError block guarding the attempt to
1049
import the native extension. If this function returns, the pure-Python
1050
implementation should be loaded instead::
1053
>>> import breezy._fictional_extension_pyx
1054
>>> except ImportError, e:
1055
>>> breezy.osutils.failed_to_load_extension(e)
1056
>>> import breezy._fictional_extension_py
1058
# NB: This docstring is just an example, not a doctest, because doctest
1059
# currently can't cope with the use of lazy imports in this namespace --
1062
# This currently doesn't report the failure at the time it occurs, because
1063
# they tend to happen very early in startup when we can't check config
1064
# files etc, and also we want to report all failures but not spam the user
1066
exception_str = str(exception)
1067
if exception_str not in _extension_load_failures:
1068
trace.mutter("failed to load compiled extension: %s" % exception_str)
1069
_extension_load_failures.append(exception_str)
1072
def report_extension_load_failures():
1073
if not _extension_load_failures:
1075
if config.GlobalConfig().suppress_warning('missing_extensions'):
1077
# the warnings framework should by default show this only once
1078
from .trace import warning
1080
"brz: warning: some compiled extensions could not be loaded; "
1081
"see ``brz help missing-extensions``")
1082
# we no longer show the specific missing extensions here, because it makes
1083
# the message too long and scary - see
1084
# https://bugs.launchpad.net/bzr/+bug/430529
1088
from ._chunks_to_lines_pyx import chunks_to_lines
1089
except ImportError as e:
1090
failed_to_load_extension(e)
1091
from ._chunks_to_lines_py import chunks_to_lines
1095
"""Split s into lines, but without removing the newline characters."""
1096
# Trivially convert a fulltext into a 'chunked' representation, and let
1097
# chunks_to_lines do the heavy lifting.
1098
if isinstance(s, bytes):
1099
# chunks_to_lines only supports 8-bit strings
1100
return chunks_to_lines([s])
1102
return _split_lines(s)
1105
def _split_lines(s):
1106
"""Split s into lines, but without removing the newline characters.
1108
This supports Unicode or plain string objects.
1110
nl = b'\n' if isinstance(s, bytes) else u'\n'
1112
result = [line + nl for line in lines[:-1]]
1114
result.append(lines[-1])
1118
def hardlinks_good():
1119
return sys.platform not in ('win32', 'cygwin', 'darwin')
1122
def link_or_copy(src, dest):
1123
"""Hardlink a file, or copy it if it can't be hardlinked."""
1124
if not hardlinks_good():
1125
shutil.copyfile(src, dest)
1129
except (OSError, IOError) as e:
1130
if e.errno != errno.EXDEV:
1132
shutil.copyfile(src, dest)
1135
def delete_any(path):
1136
"""Delete a file, symlink or directory.
1138
Will delete even if readonly.
1141
_delete_file_or_dir(path)
1142
except (OSError, IOError) as e:
1143
if e.errno in (errno.EPERM, errno.EACCES):
1144
# make writable and try again
1147
except (OSError, IOError):
1149
_delete_file_or_dir(path)
1154
def _delete_file_or_dir(path):
1155
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1156
# Forgiveness than Permission (EAFP) because:
1157
# - root can damage a solaris file system by using unlink,
1158
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1159
# EACCES, OSX: EPERM) when invoked on a directory.
1160
if isdir(path): # Takes care of symlinks
1167
if getattr(os, 'symlink', None) is not None:
1173
def has_hardlinks():
1174
if getattr(os, 'link', None) is not None:
1180
def host_os_dereferences_symlinks():
1181
return (has_symlinks()
1182
and sys.platform not in ('cygwin', 'win32'))
1185
def readlink(abspath):
1186
"""Return a string representing the path to which the symbolic link points.
1188
:param abspath: The link absolute unicode path.
1190
This his guaranteed to return the symbolic link in unicode in all python
1193
link = abspath.encode(_fs_enc)
1194
target = os.readlink(link)
1195
target = target.decode(_fs_enc)
1199
def contains_whitespace(s):
1200
"""True if there are any whitespace characters in s."""
1201
# string.whitespace can include '\xa0' in certain locales, because it is
1202
# considered "non-breaking-space" as part of ISO-8859-1. But it
1203
# 1) Isn't a breaking whitespace
1204
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1206
# 3) '\xa0' isn't unicode safe since it is >128.
1208
if isinstance(s, str):
1211
ws = (b' ', b'\t', b'\n', b'\r', b'\v', b'\f')
1219
def contains_linebreaks(s):
1220
"""True if there is any vertical whitespace in s."""
1228
def relpath(base, path):
1229
"""Return path relative to base, or raise PathNotChild exception.
1231
The path may be either an absolute path or a path relative to the
1232
current working directory.
1234
os.path.commonprefix (python2.4) has a bad bug that it works just
1235
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1236
avoids that problem.
1238
NOTE: `base` should not have a trailing slash otherwise you'll get
1239
PathNotChild exceptions regardless of `path`.
1242
if len(base) < MIN_ABS_PATHLENGTH:
1243
# must have space for e.g. a drive letter
1244
raise ValueError(gettext('%r is too short to calculate a relative path')
1252
if len(head) <= len(base) and head != base:
1253
raise errors.PathNotChild(rp, base)
1256
head, tail = split(head)
1261
return pathjoin(*reversed(s))
1266
def _cicp_canonical_relpath(base, path):
1267
"""Return the canonical path relative to base.
1269
Like relpath, but on case-insensitive-case-preserving file-systems, this
1270
will return the relpath as stored on the file-system rather than in the
1271
case specified in the input string, for all existing portions of the path.
1273
This will cause O(N) behaviour if called for every path in a tree; if you
1274
have a number of paths to convert, you should use canonical_relpaths().
1276
# TODO: it should be possible to optimize this for Windows by using the
1277
# win32 API FindFiles function to look for the specified name - but using
1278
# os.listdir() still gives us the correct, platform agnostic semantics in
1281
rel = relpath(base, path)
1282
# '.' will have been turned into ''
1286
abs_base = abspath(base)
1288
_listdir = os.listdir
1290
# use an explicit iterator so we can easily consume the rest on early exit.
1291
bit_iter = iter(rel.split('/'))
1292
for bit in bit_iter:
1295
next_entries = _listdir(current)
1296
except OSError: # enoent, eperm, etc
1297
# We can't find this in the filesystem, so just append the
1299
current = pathjoin(current, bit, *list(bit_iter))
1301
for look in next_entries:
1302
if lbit == look.lower():
1303
current = pathjoin(current, look)
1306
# got to the end, nothing matched, so we just return the
1307
# non-existing bits as they were specified (the filename may be
1308
# the target of a move, for example).
1309
current = pathjoin(current, bit, *list(bit_iter))
1311
return current[len(abs_base):].lstrip('/')
1314
# XXX - TODO - we need better detection/integration of case-insensitive
1315
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1316
# filesystems), for example, so could probably benefit from the same basic
1317
# support there. For now though, only Windows and OSX get that support, and
1318
# they get it for *all* file-systems!
1319
if sys.platform in ('win32', 'darwin'):
1320
canonical_relpath = _cicp_canonical_relpath
1322
canonical_relpath = relpath
1325
def canonical_relpaths(base, paths):
1326
"""Create an iterable to canonicalize a sequence of relative paths.
1328
The intent is for this implementation to use a cache, vastly speeding
1329
up multiple transformations in the same directory.
1331
# but for now, we haven't optimized...
1332
return [canonical_relpath(base, p) for p in paths]
1335
def decode_filename(filename):
1336
"""Decode the filename using the filesystem encoding
1338
If it is unicode, it is returned.
1339
Otherwise it is decoded from the the filesystem's encoding. If decoding
1340
fails, a errors.BadFilenameEncoding exception is raised.
1342
if isinstance(filename, str):
1345
return filename.decode(_fs_enc)
1346
except UnicodeDecodeError:
1347
raise errors.BadFilenameEncoding(filename, _fs_enc)
1350
def safe_unicode(unicode_or_utf8_string):
1351
"""Coerce unicode_or_utf8_string into unicode.
1353
If it is unicode, it is returned.
1354
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1355
wrapped in a BzrBadParameterNotUnicode exception.
1357
if isinstance(unicode_or_utf8_string, str):
1358
return unicode_or_utf8_string
1360
return unicode_or_utf8_string.decode('utf8')
1361
except UnicodeDecodeError:
1362
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1365
def safe_utf8(unicode_or_utf8_string):
1366
"""Coerce unicode_or_utf8_string to a utf8 string.
1368
If it is a str, it is returned.
1369
If it is Unicode, it is encoded into a utf-8 string.
1371
if isinstance(unicode_or_utf8_string, bytes):
1372
# TODO: jam 20070209 This is overkill, and probably has an impact on
1373
# performance if we are dealing with lots of apis that want a
1376
# Make sure it is a valid utf-8 string
1377
unicode_or_utf8_string.decode('utf-8')
1378
except UnicodeDecodeError:
1379
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1380
return unicode_or_utf8_string
1381
return unicode_or_utf8_string.encode('utf-8')
1384
def safe_revision_id(unicode_or_utf8_string):
1385
"""Revision ids should now be utf8, but at one point they were unicode.
1387
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1389
:return: None or a utf8 revision id.
1391
if (unicode_or_utf8_string is None
1392
or unicode_or_utf8_string.__class__ == bytes):
1393
return unicode_or_utf8_string
1394
raise TypeError('Unicode revision ids are no longer supported. '
1395
'Revision id generators should be creating utf8 revision '
1399
def safe_file_id(unicode_or_utf8_string):
1400
"""File ids should now be utf8, but at one point they were unicode.
1402
This is the same as safe_utf8, except it uses the cached encode functions
1403
to save a little bit of performance.
1405
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1407
:return: None or a utf8 file id.
1409
if (unicode_or_utf8_string is None
1410
or unicode_or_utf8_string.__class__ == bytes):
1411
return unicode_or_utf8_string
1412
raise TypeError('Unicode file ids are no longer supported. '
1413
'File id generators should be creating utf8 file ids.')
1416
_platform_normalizes_filenames = False
1417
if sys.platform == 'darwin':
1418
_platform_normalizes_filenames = True
1421
def normalizes_filenames():
1422
"""Return True if this platform normalizes unicode filenames.
1426
return _platform_normalizes_filenames
1429
def _accessible_normalized_filename(path):
1430
"""Get the unicode normalized path, and if you can access the file.
1432
On platforms where the system normalizes filenames (Mac OSX),
1433
you can access a file by any path which will normalize correctly.
1434
On platforms where the system does not normalize filenames
1435
(everything else), you have to access a file by its exact path.
1437
Internally, bzr only supports NFC normalization, since that is
1438
the standard for XML documents.
1440
So return the normalized path, and a flag indicating if the file
1441
can be accessed by that path.
1444
if isinstance(path, bytes):
1445
path = path.decode(sys.getfilesystemencoding())
1446
return unicodedata.normalize('NFC', path), True
1449
def _inaccessible_normalized_filename(path):
1450
__doc__ = _accessible_normalized_filename.__doc__
1452
if isinstance(path, bytes):
1453
path = path.decode(sys.getfilesystemencoding())
1454
normalized = unicodedata.normalize('NFC', path)
1455
return normalized, normalized == path
1458
if _platform_normalizes_filenames:
1459
normalized_filename = _accessible_normalized_filename
1461
normalized_filename = _inaccessible_normalized_filename
1464
def set_signal_handler(signum, handler, restart_syscall=True):
1465
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1466
on platforms that support that.
1468
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1469
automatically restart (by calling `signal.siginterrupt(signum,
1470
False)`). May be ignored if the feature is not available on this
1471
platform or Python version.
1475
siginterrupt = signal.siginterrupt
1477
# This python implementation doesn't provide signal support, hence no
1480
except AttributeError:
1481
# siginterrupt doesn't exist on this platform, or for this version
1483
def siginterrupt(signum, flag): return None
1485
def sig_handler(*args):
1486
# Python resets the siginterrupt flag when a signal is
1487
# received. <http://bugs.python.org/issue8354>
1488
# As a workaround for some cases, set it back the way we want it.
1489
siginterrupt(signum, False)
1490
# Now run the handler function passed to set_signal_handler.
1493
sig_handler = handler
1494
old_handler = signal.signal(signum, sig_handler)
1496
siginterrupt(signum, False)
1500
default_terminal_width = 80
1501
"""The default terminal width for ttys.
1503
This is defined so that higher levels can share a common fallback value when
1504
terminal_width() returns None.
1507
# Keep some state so that terminal_width can detect if _terminal_size has
1508
# returned a different size since the process started. See docstring and
1509
# comments of terminal_width for details.
1510
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1511
_terminal_size_state = 'no_data'
1512
_first_terminal_size = None
1515
def terminal_width():
1516
"""Return terminal width.
1518
None is returned if the width can't established precisely.
1521
- if BRZ_COLUMNS is set, returns its value
1522
- if there is no controlling terminal, returns None
1523
- query the OS, if the queried size has changed since the last query,
1525
- if COLUMNS is set, returns its value,
1526
- if the OS has a value (even though it's never changed), return its value.
1528
From there, we need to query the OS to get the size of the controlling
1531
On Unices we query the OS by:
1532
- get termios.TIOCGWINSZ
1533
- if an error occurs or a negative value is obtained, returns None
1535
On Windows we query the OS by:
1536
- win32utils.get_console_size() decides,
1537
- returns None on error (provided default value)
1539
# Note to implementors: if changing the rules for determining the width,
1540
# make sure you've considered the behaviour in these cases:
1541
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1542
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1544
# - (add more interesting cases here, if you find any)
1545
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1546
# but we don't want to register a signal handler because it is impossible
1547
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1548
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1549
# time so we can notice if the reported size has changed, which should have
1552
# If BRZ_COLUMNS is set, take it, user is always right
1553
# Except if they specified 0 in which case, impose no limit here
1555
width = int(os.environ['BRZ_COLUMNS'])
1556
except (KeyError, ValueError):
1558
if width is not None:
1564
isatty = getattr(sys.stdout, 'isatty', None)
1565
if isatty is None or not isatty():
1566
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
1570
width, height = os_size = _terminal_size(None, None)
1571
global _first_terminal_size, _terminal_size_state
1572
if _terminal_size_state == 'no_data':
1573
_first_terminal_size = os_size
1574
_terminal_size_state = 'unchanged'
1575
elif (_terminal_size_state == 'unchanged' and
1576
_first_terminal_size != os_size):
1577
_terminal_size_state = 'changed'
1579
# If the OS claims to know how wide the terminal is, and this value has
1580
# ever changed, use that.
1581
if _terminal_size_state == 'changed':
1582
if width is not None and width > 0:
1585
# If COLUMNS is set, use it.
1587
return int(os.environ['COLUMNS'])
1588
except (KeyError, ValueError):
1591
# Finally, use an unchanged size from the OS, if we have one.
1592
if _terminal_size_state == 'unchanged':
1593
if width is not None and width > 0:
1596
# The width could not be determined.
1600
def _win32_terminal_size(width, height):
1601
width, height = win32utils.get_console_size(
1602
defaultx=width, defaulty=height)
1603
return width, height
1606
def _ioctl_terminal_size(width, height):
1611
s = struct.pack('HHHH', 0, 0, 0, 0)
1612
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1613
height, width = struct.unpack('HHHH', x)[0:2]
1614
except (IOError, AttributeError):
1616
return width, height
1619
_terminal_size = None
1620
"""Returns the terminal size as (width, height).
1622
:param width: Default value for width.
1623
:param height: Default value for height.
1625
This is defined specifically for each OS and query the size of the controlling
1626
terminal. If any error occurs, the provided default values should be returned.
1628
if sys.platform == 'win32':
1629
_terminal_size = _win32_terminal_size
1631
_terminal_size = _ioctl_terminal_size
1634
def supports_executable(path):
1635
"""Return if filesystem at path supports executable bit.
1637
:param path: Path for which to check the file system
1638
:return: boolean indicating whether executable bit can be stored/relied upon
1640
if sys.platform == 'win32':
1643
fs_type = get_fs_type(path)
1644
except errors.DependencyNotPresent as e:
1645
trace.mutter('Unable to get fs type for %r: %s', path, e)
1647
if fs_type in ('vfat', 'ntfs'):
1648
# filesystems known to not support executable bit
1653
def supports_symlinks(path):
1654
"""Return if the filesystem at path supports the creation of symbolic links.
1657
if not has_symlinks():
1660
fs_type = get_fs_type(path)
1661
except errors.DependencyNotPresent as e:
1662
trace.mutter('Unable to get fs type for %r: %s', path, e)
1664
if fs_type in ('vfat', 'ntfs'):
1665
# filesystems known to not support symlinks
1670
def supports_posix_readonly():
1671
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1673
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1674
directory controls creation/deletion, etc.
1676
And under win32, readonly means that the directory itself cannot be
1677
deleted. The contents of a readonly directory can be changed, unlike POSIX
1678
where files in readonly directories cannot be added, deleted or renamed.
1680
return sys.platform != "win32"
1683
def set_or_unset_env(env_variable, value):
1684
"""Modify the environment, setting or removing the env_variable.
1686
:param env_variable: The environment variable in question
1687
:param value: The value to set the environment to. If None, then
1688
the variable will be removed.
1689
:return: The original value of the environment variable.
1691
orig_val = os.environ.get(env_variable)
1693
if orig_val is not None:
1694
del os.environ[env_variable]
1696
os.environ[env_variable] = value
1700
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1703
def check_legal_path(path):
1704
"""Check whether the supplied path is legal.
1705
This is only required on Windows, so we don't test on other platforms
1708
if sys.platform != "win32":
1710
if _validWin32PathRE.match(path) is None:
1711
raise errors.IllegalPath(path)
1714
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1717
def _is_error_enotdir(e):
1718
"""Check if this exception represents ENOTDIR.
1720
Unfortunately, python is very inconsistent about the exception
1721
here. The cases are:
1722
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1723
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1724
which is the windows error code.
1725
3) Windows, Python2.5 uses errno == EINVAL and
1726
winerror == ERROR_DIRECTORY
1728
:param e: An Exception object (expected to be OSError with an errno
1729
attribute, but we should be able to cope with anything)
1730
:return: True if this represents an ENOTDIR error. False otherwise.
1732
en = getattr(e, 'errno', None)
1733
if (en == errno.ENOTDIR or
1734
(sys.platform == 'win32' and
1735
(en == _WIN32_ERROR_DIRECTORY or
1737
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1743
def walkdirs(top, prefix=""):
1744
"""Yield data about all the directories in a tree.
1746
This yields all the data about the contents of a directory at a time.
1747
After each directory has been yielded, if the caller has mutated the list
1748
to exclude some directories, they are then not descended into.
1750
The data yielded is of the form:
1751
((directory-relpath, directory-path-from-top),
1752
[(relpath, basename, kind, lstat, path-from-top), ...]),
1753
- directory-relpath is the relative path of the directory being returned
1754
with respect to top. prefix is prepended to this.
1755
- directory-path-from-root is the path including top for this directory.
1756
It is suitable for use with os functions.
1757
- relpath is the relative path within the subtree being walked.
1758
- basename is the basename of the path
1759
- kind is the kind of the file now. If unknown then the file is not
1760
present within the tree - but it may be recorded as versioned. See
1762
- lstat is the stat data *if* the file was statted.
1763
- planned, not implemented:
1764
path_from_tree_root is the path from the root of the tree.
1766
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1767
allows one to walk a subtree but get paths that are relative to a tree
1769
:return: an iterator over the dirs.
1771
# TODO there is a bit of a smell where the results of the directory-
1772
# summary in this, and the path from the root, may not agree
1773
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1774
# potentially confusing output. We should make this more robust - but
1775
# not at a speed cost. RBC 20060731
1777
_directory = _directory_kind
1778
_listdir = os.listdir
1779
_kind_from_mode = file_kind_from_stat_mode
1780
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1782
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1783
relroot, _, _, _, top = pending.pop()
1785
relprefix = relroot + u'/'
1788
top_slash = top + u'/'
1791
append = dirblock.append
1793
names = sorted(map(decode_filename, _listdir(top)))
1794
except OSError as e:
1795
if not _is_error_enotdir(e):
1799
abspath = top_slash + name
1800
statvalue = _lstat(abspath)
1801
kind = _kind_from_mode(statvalue.st_mode)
1802
append((relprefix + name, name, kind, statvalue, abspath))
1803
yield (relroot, top), dirblock
1805
# push the user specified dirs from dirblock
1806
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1809
class DirReader(object):
1810
"""An interface for reading directories."""
1812
def top_prefix_to_starting_dir(self, top, prefix=""):
1813
"""Converts top and prefix to a starting dir entry
1815
:param top: A utf8 path
1816
:param prefix: An optional utf8 path to prefix output relative paths
1818
:return: A tuple starting with prefix, and ending with the native
1821
raise NotImplementedError(self.top_prefix_to_starting_dir)
1823
def read_dir(self, prefix, top):
1824
"""Read a specific dir.
1826
:param prefix: A utf8 prefix to be preprended to the path basenames.
1827
:param top: A natively encoded path to read.
1828
:return: A list of the directories contents. Each item contains:
1829
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1831
raise NotImplementedError(self.read_dir)
1834
_selected_dir_reader = None
1837
def _walkdirs_utf8(top, prefix=""):
1838
"""Yield data about all the directories in a tree.
1840
This yields the same information as walkdirs() only each entry is yielded
1841
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1842
are returned as exact byte-strings.
1844
:return: yields a tuple of (dir_info, [file_info])
1845
dir_info is (utf8_relpath, path-from-top)
1846
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1847
if top is an absolute path, path-from-top is also an absolute path.
1848
path-from-top might be unicode or utf8, but it is the correct path to
1849
pass to os functions to affect the file in question. (such as os.lstat)
1851
global _selected_dir_reader
1852
if _selected_dir_reader is None:
1853
if sys.platform == "win32":
1855
from ._walkdirs_win32 import Win32ReadDir
1856
_selected_dir_reader = Win32ReadDir()
1859
elif _fs_enc in ('utf-8', 'ascii'):
1861
from ._readdir_pyx import UTF8DirReader
1862
_selected_dir_reader = UTF8DirReader()
1863
except ImportError as e:
1864
failed_to_load_extension(e)
1867
if _selected_dir_reader is None:
1868
# Fallback to the python version
1869
_selected_dir_reader = UnicodeDirReader()
1871
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1872
# But we don't actually uses 1-3 in pending, so set them to None
1873
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1874
read_dir = _selected_dir_reader.read_dir
1875
_directory = _directory_kind
1877
relroot, _, _, _, top = pending[-1].pop()
1880
dirblock = sorted(read_dir(relroot, top))
1881
yield (relroot, top), dirblock
1882
# push the user specified dirs from dirblock
1883
next = [d for d in reversed(dirblock) if d[2] == _directory]
1885
pending.append(next)
1888
class UnicodeDirReader(DirReader):
1889
"""A dir reader for non-utf8 file systems, which transcodes."""
1891
__slots__ = ['_utf8_encode']
1894
self._utf8_encode = codecs.getencoder('utf8')
1896
def top_prefix_to_starting_dir(self, top, prefix=""):
1897
"""See DirReader.top_prefix_to_starting_dir."""
1898
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1900
def read_dir(self, prefix, top):
1901
"""Read a single directory from a non-utf8 file system.
1903
top, and the abspath element in the output are unicode, all other paths
1904
are utf8. Local disk IO is done via unicode calls to listdir etc.
1906
This is currently the fallback code path when the filesystem encoding is
1907
not UTF-8. It may be better to implement an alternative so that we can
1908
safely handle paths that are not properly decodable in the current
1911
See DirReader.read_dir for details.
1913
_utf8_encode = self._utf8_encode
1915
def _fs_decode(s): return s.decode(_fs_enc)
1917
def _fs_encode(s): return s.encode(_fs_enc)
1919
_listdir = os.listdir
1920
_kind_from_mode = file_kind_from_stat_mode
1923
relprefix = prefix + b'/'
1926
top_slash = top + '/'
1929
append = dirblock.append
1930
for name_native in _listdir(top.encode('utf-8')):
1932
name = _fs_decode(name_native)
1933
except UnicodeDecodeError:
1934
raise errors.BadFilenameEncoding(
1935
relprefix + name_native, _fs_enc)
1936
name_utf8 = _utf8_encode(name)[0]
1937
abspath = top_slash + name
1938
statvalue = _lstat(abspath)
1939
kind = _kind_from_mode(statvalue.st_mode)
1940
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1941
return sorted(dirblock)
1944
def copy_tree(from_path, to_path, handlers={}):
1945
"""Copy all of the entries in from_path into to_path.
1947
:param from_path: The base directory to copy.
1948
:param to_path: The target directory. If it does not exist, it will
1950
:param handlers: A dictionary of functions, which takes a source and
1951
destinations for files, directories, etc.
1952
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1953
'file', 'directory', and 'symlink' should always exist.
1954
If they are missing, they will be replaced with 'os.mkdir()',
1955
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1957
# Now, just copy the existing cached tree to the new location
1958
# We use a cheap trick here.
1959
# Absolute paths are prefixed with the first parameter
1960
# relative paths are prefixed with the second.
1961
# So we can get both the source and target returned
1962
# without any extra work.
1964
def copy_dir(source, dest):
1967
def copy_link(source, dest):
1968
"""Copy the contents of a symlink"""
1969
link_to = os.readlink(source)
1970
os.symlink(link_to, dest)
1972
real_handlers = {'file': shutil.copy2,
1973
'symlink': copy_link,
1974
'directory': copy_dir,
1976
real_handlers.update(handlers)
1978
if not os.path.exists(to_path):
1979
real_handlers['directory'](from_path, to_path)
1981
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1982
for relpath, name, kind, st, abspath in entries:
1983
real_handlers[kind](abspath, relpath)
1986
def copy_ownership_from_path(dst, src=None):
1987
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1989
If src is None, the containing directory is used as source. If chown
1990
fails, the error is ignored and a warning is printed.
1992
chown = getattr(os, 'chown', None)
1997
src = os.path.dirname(dst)
2003
chown(dst, s.st_uid, s.st_gid)
2006
'Unable to copy ownership from "%s" to "%s". '
2007
'You may want to set it manually.', src, dst)
2008
trace.log_exception_quietly()
2011
def path_prefix_key(path):
2012
"""Generate a prefix-order path key for path.
2014
This can be used to sort paths in the same way that walkdirs does.
2016
return (dirname(path), path)
2019
def compare_paths_prefix_order(path_a, path_b):
2020
"""Compare path_a and path_b to generate the same order walkdirs uses."""
2021
key_a = path_prefix_key(path_a)
2022
key_b = path_prefix_key(path_b)
2023
return (key_a > key_b) - (key_a < key_b)
2026
_cached_user_encoding = None
2029
def get_user_encoding():
2030
"""Find out what the preferred user encoding is.
2032
This is generally the encoding that is used for command line parameters
2033
and file contents. This may be different from the terminal encoding
2034
or the filesystem encoding.
2036
:return: A string defining the preferred user encoding
2038
global _cached_user_encoding
2039
if _cached_user_encoding is not None:
2040
return _cached_user_encoding
2042
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2043
# Use the existing locale settings and call nl_langinfo directly
2044
# rather than going through getpreferredencoding. This avoids
2045
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2046
# possibility of the setlocale call throwing an error.
2047
user_encoding = locale.nl_langinfo(locale.CODESET)
2049
# GZ 2011-12-19: On windows could call GetACP directly instead.
2050
user_encoding = locale.getpreferredencoding(False)
2053
user_encoding = codecs.lookup(user_encoding).name
2055
if user_encoding not in ("", "cp0"):
2056
sys.stderr.write('brz: warning:'
2057
' unknown encoding %s.'
2058
' Continuing with ascii encoding.\n'
2061
user_encoding = 'ascii'
2063
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2064
if user_encoding == 'ascii':
2065
if sys.platform == 'darwin':
2066
# OSX is special-cased in Python to have a UTF-8 filesystem
2067
# encoding and previously had LANG set here if not present.
2068
user_encoding = 'utf-8'
2069
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2070
# for some other posix platforms as well.
2072
_cached_user_encoding = user_encoding
2073
return user_encoding
2076
def get_diff_header_encoding():
2077
return get_terminal_encoding()
2080
def get_host_name():
2081
"""Return the current unicode host name.
2083
This is meant to be used in place of socket.gethostname() because that
2084
behaves inconsistently on different platforms.
2086
if sys.platform == "win32":
2087
return win32utils.get_host_name()
2090
return socket.gethostname()
2093
# We must not read/write any more than 64k at a time from/to a socket so we
2094
# don't risk "no buffer space available" errors on some platforms. Windows in
2095
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2097
MAX_SOCKET_CHUNK = 64 * 1024
2099
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2100
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2101
_eno = getattr(errno, _eno, None)
2102
if _eno is not None:
2103
_end_of_stream_errors.append(_eno)
2107
def read_bytes_from_socket(sock, report_activity=None,
2108
max_read_size=MAX_SOCKET_CHUNK):
2109
"""Read up to max_read_size of bytes from sock and notify of progress.
2111
Translates "Connection reset by peer" into file-like EOF (return an
2112
empty string rather than raise an error), and repeats the recv if
2113
interrupted by a signal.
2117
data = sock.recv(max_read_size)
2118
except socket.error as e:
2120
if eno in _end_of_stream_errors:
2121
# The connection was closed by the other side. Callers expect
2122
# an empty string to signal end-of-stream.
2124
elif eno == errno.EINTR:
2125
# Retry the interrupted recv.
2129
if report_activity is not None:
2130
report_activity(len(data), 'read')
2134
def recv_all(socket, count):
2135
"""Receive an exact number of bytes.
2137
Regular Socket.recv() may return less than the requested number of bytes,
2138
depending on what's in the OS buffer. MSG_WAITALL is not available
2139
on all platforms, but this should work everywhere. This will return
2140
less than the requested amount if the remote end closes.
2142
This isn't optimized and is intended mostly for use in testing.
2145
while len(b) < count:
2146
new = read_bytes_from_socket(socket, None, count - len(b))
2153
def send_all(sock, bytes, report_activity=None):
2154
"""Send all bytes on a socket.
2156
Breaks large blocks in smaller chunks to avoid buffering limitations on
2157
some platforms, and catches EINTR which may be thrown if the send is
2158
interrupted by a signal.
2160
This is preferred to socket.sendall(), because it avoids portability bugs
2161
and provides activity reporting.
2163
:param report_activity: Call this as bytes are read, see
2164
Transport._report_activity
2167
byte_count = len(bytes)
2168
view = memoryview(bytes)
2169
while sent_total < byte_count:
2171
sent = sock.send(view[sent_total:sent_total + MAX_SOCKET_CHUNK])
2172
except (socket.error, IOError) as e:
2173
if e.args[0] in _end_of_stream_errors:
2174
raise errors.ConnectionReset(
2175
"Error trying to write to socket", e)
2176
if e.args[0] != errno.EINTR:
2180
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2183
if report_activity is not None:
2184
report_activity(sent, 'write')
2187
def connect_socket(address):
2188
# Slight variation of the socket.create_connection() function (provided by
2189
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2190
# provide it for previous python versions. Also, we don't use the timeout
2191
# parameter (provided by the python implementation) so we don't implement
2193
err = socket.error('getaddrinfo returns an empty list')
2194
host, port = address
2195
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2196
af, socktype, proto, canonname, sa = res
2199
sock = socket.socket(af, socktype, proto)
2203
except socket.error as e:
2205
# 'err' is now the most recent error
2206
if sock is not None:
2211
def dereference_path(path):
2212
"""Determine the real path to a file.
2214
All parent elements are dereferenced. But the file itself is not
2216
:param path: The original path. May be absolute or relative.
2217
:return: the real path *to* the file
2219
parent, base = os.path.split(path)
2220
# The pathjoin for '.' is a workaround for Python bug #1213894.
2221
# (initial path components aren't dereferenced)
2222
return pathjoin(realpath(pathjoin('.', parent)), base)
2225
def supports_mapi():
2226
"""Return True if we can use MAPI to launch a mail client."""
2227
return sys.platform == "win32"
2230
def resource_string(package, resource_name):
2231
"""Load a resource from a package and return it as a string.
2233
Note: Only packages that start with breezy are currently supported.
2235
This is designed to be a lightweight implementation of resource
2236
loading in a way which is API compatible with the same API from
2238
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2239
If and when pkg_resources becomes a standard library, this routine
2242
# Check package name is within breezy
2243
if package == "breezy":
2244
resource_relpath = resource_name
2245
elif package.startswith("breezy."):
2246
package = package[len("breezy."):].replace('.', os.sep)
2247
resource_relpath = pathjoin(package, resource_name)
2249
raise errors.BzrError('resource package %s not in breezy' % package)
2251
# Map the resource to a file and read its contents
2252
base = dirname(breezy.__file__)
2253
if getattr(sys, 'frozen', None): # bzr.exe
2254
base = abspath(pathjoin(base, '..', '..'))
2255
with open(pathjoin(base, resource_relpath), "rt") as f:
2259
def file_kind_from_stat_mode_thunk(mode):
2260
global file_kind_from_stat_mode
2261
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2263
from ._readdir_pyx import UTF8DirReader
2264
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2266
# This is one time where we won't warn that an extension failed to
2267
# load. The extension is never available on Windows anyway.
2268
from ._readdir_py import (
2269
_kind_from_mode as file_kind_from_stat_mode
2271
return file_kind_from_stat_mode(mode)
2274
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2277
def file_stat(f, _lstat=os.lstat):
2281
except OSError as e:
2282
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2283
raise errors.NoSuchFile(f)
2287
def file_kind(f, _lstat=os.lstat):
2288
stat_value = file_stat(f, _lstat)
2289
return file_kind_from_stat_mode(stat_value.st_mode)
2292
def until_no_eintr(f, *a, **kw):
2293
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2295
WARNING: you must be certain that it is safe to retry the call repeatedly
2296
if EINTR does occur. This is typically only true for low-level operations
2297
like os.read. If in any doubt, don't use this.
2299
Keep in mind that this is not a complete solution to EINTR. There is
2300
probably code in the Python standard library and other dependencies that
2301
may encounter EINTR if a signal arrives (and there is signal handler for
2302
that signal). So this function can reduce the impact for IO that breezy
2303
directly controls, but it is not a complete solution.
2305
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2309
except (IOError, OSError) as e:
2310
if e.errno == errno.EINTR:
2315
if sys.platform == "win32":
2318
return msvcrt.getch()
2323
fd = sys.stdin.fileno()
2324
settings = termios.tcgetattr(fd)
2327
ch = sys.stdin.read(1)
2329
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2332
if sys.platform.startswith('linux'):
2333
def _local_concurrency():
2335
return os.sysconf('SC_NPROCESSORS_ONLN')
2336
except (ValueError, OSError, AttributeError):
2338
elif sys.platform == 'darwin':
2339
def _local_concurrency():
2340
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2341
stdout=subprocess.PIPE).communicate()[0]
2342
elif "bsd" in sys.platform:
2343
def _local_concurrency():
2344
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2345
stdout=subprocess.PIPE).communicate()[0]
2346
elif sys.platform == 'sunos5':
2347
def _local_concurrency():
2348
return subprocess.Popen(['psrinfo', '-p', ],
2349
stdout=subprocess.PIPE).communicate()[0]
2350
elif sys.platform == "win32":
2351
def _local_concurrency():
2352
# This appears to return the number of cores.
2353
return os.environ.get('NUMBER_OF_PROCESSORS')
2355
def _local_concurrency():
2360
_cached_local_concurrency = None
2363
def local_concurrency(use_cache=True):
2364
"""Return how many processes can be run concurrently.
2366
Rely on platform specific implementations and default to 1 (one) if
2367
anything goes wrong.
2369
global _cached_local_concurrency
2371
if _cached_local_concurrency is not None and use_cache:
2372
return _cached_local_concurrency
2374
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2375
if concurrency is None:
2376
import multiprocessing
2378
concurrency = multiprocessing.cpu_count()
2379
except NotImplementedError:
2380
# multiprocessing.cpu_count() isn't implemented on all platforms
2382
concurrency = _local_concurrency()
2383
except (OSError, IOError):
2386
concurrency = int(concurrency)
2387
except (TypeError, ValueError):
2390
_cached_local_concurrency = concurrency
2394
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2395
"""A stream writer that doesn't decode str arguments."""
2397
def __init__(self, encode, stream, errors='strict'):
2398
codecs.StreamWriter.__init__(self, stream, errors)
2399
self.encode = encode
2401
def write(self, object):
2402
if isinstance(object, str):
2403
self.stream.write(object)
2405
data, _ = self.encode(object, self.errors)
2406
self.stream.write(data)
2409
if sys.platform == 'win32':
2410
def open_file(filename, mode='r', bufsize=-1):
2411
"""This function is used to override the ``open`` builtin.
2413
But it uses O_NOINHERIT flag so the file handle is not inherited by
2414
child processes. Deleting or renaming a closed file opened with this
2415
function is not blocking child processes.
2417
writing = 'w' in mode
2418
appending = 'a' in mode
2419
updating = '+' in mode
2420
binary = 'b' in mode
2423
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2424
# for flags for each modes.
2434
flags |= os.O_WRONLY
2435
flags |= os.O_CREAT | os.O_TRUNC
2440
flags |= os.O_WRONLY
2441
flags |= os.O_CREAT | os.O_APPEND
2446
flags |= os.O_RDONLY
2448
return os.fdopen(os.open(filename, flags), mode, bufsize)
2453
def available_backup_name(base, exists):
2454
"""Find a non-existing backup file name.
2456
This will *not* create anything, this only return a 'free' entry. This
2457
should be used for checking names in a directory below a locked
2458
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2459
Leap) and generally discouraged.
2461
:param base: The base name.
2463
:param exists: A callable returning True if the path parameter exists.
2466
name = "%s.~%d~" % (base, counter)
2469
name = "%s.~%d~" % (base, counter)
2473
def set_fd_cloexec(fd):
2474
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2475
support for this is not available.
2479
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2480
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2481
except (ImportError, AttributeError):
2482
# Either the fcntl module or specific constants are not present
2486
def find_executable_on_path(name):
2487
"""Finds an executable on the PATH.
2489
On Windows, this will try to append each extension in the PATHEXT
2490
environment variable to the name, if it cannot be found with the name
2493
:param name: The base name of the executable.
2494
:return: The path to the executable found or None.
2496
if sys.platform == 'win32':
2497
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2498
exts = [ext.lower() for ext in exts]
2499
base, ext = os.path.splitext(name)
2501
if ext.lower() not in exts:
2507
path = os.environ.get('PATH')
2508
if path is not None:
2509
path = path.split(os.pathsep)
2512
f = os.path.join(d, name) + ext
2513
if os.access(f, os.X_OK):
2515
if sys.platform == 'win32':
2516
app_path = win32utils.get_app_path(name)
2517
if app_path != name:
2522
def _posix_is_local_pid_dead(pid):
2523
"""True if pid doesn't correspond to live process on this machine"""
2525
# Special meaning of unix kill: just check if it's there.
2527
except OSError as e:
2528
if e.errno == errno.ESRCH:
2529
# On this machine, and really not found: as sure as we can be
2532
elif e.errno == errno.EPERM:
2533
# exists, though not ours
2536
trace.mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2537
# Don't really know.
2540
# Exists and our process: not dead.
2544
if sys.platform == "win32":
2545
is_local_pid_dead = win32utils.is_local_pid_dead
2547
is_local_pid_dead = _posix_is_local_pid_dead
2549
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2550
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2551
if getattr(errno, name, None) is not None]
2554
def fdatasync(fileno):
2555
"""Flush file contents to disk if possible.
2557
:param fileno: Integer OS file handle.
2558
:raises TransportNotPossible: If flushing to disk is not possible.
2560
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2564
except IOError as e:
2565
# See bug #1075108, on some platforms fdatasync exists, but can
2566
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2567
# and reduce the chance of corruption-on-powerloss situations. It
2568
# is not a mandatory call, so it is ok to suppress failures.
2569
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2570
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2574
def ensure_empty_directory_exists(path, exception_class):
2575
"""Make sure a local directory exists and is empty.
2577
If it does not exist, it is created. If it exists and is not empty, an
2578
instance of exception_class is raised.
2582
except OSError as e:
2583
if e.errno != errno.EEXIST:
2585
if os.listdir(path) != []:
2586
raise exception_class(path)
2589
def is_environment_error(evalue):
2590
"""True if exception instance is due to a process environment issue
2592
This includes OSError and IOError, but also other errors that come from
2593
the operating system or core libraries but are not subclasses of those.
2595
if isinstance(evalue, (EnvironmentError, select.error)):
2597
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
2602
def read_mtab(path):
2603
"""Read an fstab-style file and extract mountpoint+filesystem information.
2605
:param path: Path to read from
2606
:yield: Tuples with mountpoints (as bytestrings) and filesystem names
2608
with open(path, 'rb') as f:
2610
if line.startswith(b'#'):
2615
yield cols[1], cols[2].decode('ascii', 'replace')
2618
MTAB_PATH = '/etc/mtab'
2620
class FilesystemFinder(object):
2621
"""Find the filesystem for a particular path."""
2623
def __init__(self, mountpoints):
2626
self._mountpoints = sorted(mountpoints, key=key, reverse=True)
2630
"""Create a FilesystemFinder from an mtab-style file.
2632
Note that this will silenty ignore mtab if it doesn't exist or can not
2635
# TODO(jelmer): Use inotify to be notified when /etc/mtab changes and
2636
# we need to re-read it.
2638
return cls(read_mtab(MTAB_PATH))
2639
except EnvironmentError as e:
2640
trace.mutter('Unable to read mtab: %s', e)
2643
def find(self, path):
2644
"""Find the filesystem used by a particular path.
2646
:param path: Path to find (bytestring or text type)
2647
:return: Filesystem name (as text type) or None, if the filesystem is
2650
for mountpoint, filesystem in self._mountpoints:
2651
if is_inside(mountpoint, path):
2656
_FILESYSTEM_FINDER = None
2659
def get_fs_type(path):
2660
"""Return the filesystem type for the partition a path is in.
2662
:param path: Path to search filesystem type for
2663
:return: A FS type, as string. E.g. "ext2"
2665
global _FILESYSTEM_FINDER
2666
if _FILESYSTEM_FINDER is None:
2667
_FILESYSTEM_FINDER = FilesystemFinder.from_mtab()
2669
if not isinstance(path, bytes):
2670
path = path.encode(_fs_enc)
2672
return _FILESYSTEM_FINDER.find(path)
2675
perf_counter = time.perf_counter
added
removed
breezy/osutils.py
