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
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
27
from datetime import datetime
31
# We need to import both shutil and rmtree as we export the later on posix
32
# and need the former on windows
34
from shutil import rmtree
37
# We need to import both tempfile and mkdtemp as we export the later on posix
38
# and need the former on windows
40
from tempfile import mkdtemp
51
from bzrlib.symbol_versioning import (
63
from bzrlib import symbol_versioning
66
# Cross platform wall-clock time functionality with decent resolution.
67
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
68
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
69
# synchronized with ``time.time()``, this is only meant to be used to find
70
# delta times by subtracting from another call to this function.
71
timer_func = time.time
72
if sys.platform == 'win32':
73
timer_func = time.clock
75
# On win32, O_BINARY is used to indicate the file should
76
# be opened in binary mode, rather than text mode.
77
# On other platforms, O_BINARY doesn't exist, because
78
# they always open in binary mode, so it is okay to
79
# OR with 0 on those platforms.
80
# O_NOINHERIT and O_TEXT exists only on win32 too.
81
O_BINARY = getattr(os, 'O_BINARY', 0)
82
O_TEXT = getattr(os, 'O_TEXT', 0)
83
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
86
def get_unicode_argv():
88
user_encoding = get_user_encoding()
89
return [a.decode(user_encoding) for a in sys.argv[1:]]
90
except UnicodeDecodeError:
91
raise errors.BzrError("Parameter %r encoding is unsupported by %s "
92
"application locale." % (a, user_encoding))
95
def make_readonly(filename):
96
"""Make a filename read-only."""
97
mod = os.lstat(filename).st_mode
98
if not stat.S_ISLNK(mod):
100
os.chmod(filename, mod)
103
def make_writable(filename):
104
mod = os.lstat(filename).st_mode
105
if not stat.S_ISLNK(mod):
107
os.chmod(filename, mod)
110
def minimum_path_selection(paths):
111
"""Return the smallset subset of paths which are outside paths.
113
:param paths: A container (and hence not None) of paths.
114
:return: A set of paths sufficient to include everything in paths via
115
is_inside, drawn from the paths parameter.
121
return path.split('/')
122
sorted_paths = sorted(list(paths), key=sort_key)
124
search_paths = [sorted_paths[0]]
125
for path in sorted_paths[1:]:
126
if not is_inside(search_paths[-1], path):
127
# This path is unique, add it
128
search_paths.append(path)
130
return set(search_paths)
137
"""Return a quoted filename filename
139
This previously used backslash quoting, but that works poorly on
141
# TODO: I'm not really sure this is the best format either.x
143
if _QUOTE_RE is None:
144
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
146
if _QUOTE_RE.search(f):
152
_directory_kind = 'directory'
155
"""Return the current umask"""
156
# Assume that people aren't messing with the umask while running
157
# XXX: This is not thread safe, but there is no way to get the
158
# umask without setting it
166
_directory_kind: "/",
168
'tree-reference': '+',
172
def kind_marker(kind):
174
return _kind_marker_map[kind]
176
# Slightly faster than using .get(, '') when the common case is that
181
lexists = getattr(os.path, 'lexists', None)
185
stat = getattr(os, 'lstat', os.stat)
189
if e.errno == errno.ENOENT:
192
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
195
def fancy_rename(old, new, rename_func, unlink_func):
196
"""A fancy rename, when you don't have atomic rename.
198
:param old: The old path, to rename from
199
:param new: The new path, to rename to
200
:param rename_func: The potentially non-atomic rename function
201
:param unlink_func: A way to delete the target file if the full rename
204
# sftp rename doesn't allow overwriting, so play tricks:
205
base = os.path.basename(new)
206
dirname = os.path.dirname(new)
207
# callers use different encodings for the paths so the following MUST
208
# respect that. We rely on python upcasting to unicode if new is unicode
209
# and keeping a str if not.
210
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
211
os.getpid(), rand_chars(10))
212
tmp_name = pathjoin(dirname, tmp_name)
214
# Rename the file out of the way, but keep track if it didn't exist
215
# We don't want to grab just any exception
216
# something like EACCES should prevent us from continuing
217
# The downside is that the rename_func has to throw an exception
218
# with an errno = ENOENT, or NoSuchFile
221
rename_func(new, tmp_name)
222
except (errors.NoSuchFile,), e:
225
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
226
# function raises an IOError with errno is None when a rename fails.
227
# This then gets caught here.
228
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
231
if (getattr(e, 'errno', None) is None
232
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
241
# This may throw an exception, in which case success will
243
rename_func(old, new)
245
except (IOError, OSError), e:
246
# source and target may be aliases of each other (e.g. on a
247
# case-insensitive filesystem), so we may have accidentally renamed
248
# source by when we tried to rename target
249
failure_exc = sys.exc_info()
250
if (file_existed and e.errno in (None, errno.ENOENT)
251
and old.lower() == new.lower()):
252
# source and target are the same file on a case-insensitive
253
# filesystem, so we don't generate an exception
257
# If the file used to exist, rename it back into place
258
# otherwise just delete it from the tmp location
260
unlink_func(tmp_name)
262
rename_func(tmp_name, new)
263
if failure_exc is not None:
264
raise failure_exc[0], failure_exc[1], failure_exc[2]
267
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
268
# choke on a Unicode string containing a relative path if
269
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
271
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
272
def _posix_abspath(path):
273
# jam 20060426 rather than encoding to fsencoding
274
# copy posixpath.abspath, but use os.getcwdu instead
275
if not posixpath.isabs(path):
276
path = posixpath.join(getcwd(), path)
277
return posixpath.normpath(path)
280
def _posix_realpath(path):
281
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
284
def _win32_fixdrive(path):
285
"""Force drive letters to be consistent.
287
win32 is inconsistent whether it returns lower or upper case
288
and even if it was consistent the user might type the other
289
so we force it to uppercase
290
running python.exe under cmd.exe return capital C:\\
291
running win32 python inside a cygwin shell returns lowercase c:\\
293
drive, path = ntpath.splitdrive(path)
294
return drive.upper() + path
297
def _win32_abspath(path):
298
# Real ntpath.abspath doesn't have a problem with a unicode cwd
299
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
302
def _win98_abspath(path):
303
"""Return the absolute version of a path.
304
Windows 98 safe implementation (python reimplementation
305
of Win32 API function GetFullPathNameW)
310
# \\HOST\path => //HOST/path
311
# //HOST/path => //HOST/path
312
# path => C:/cwd/path
315
# check for absolute path
316
drive = ntpath.splitdrive(path)[0]
317
if drive == '' and path[:2] not in('//','\\\\'):
319
# we cannot simply os.path.join cwd and path
320
# because os.path.join('C:','/path') produce '/path'
321
# and this is incorrect
322
if path[:1] in ('/','\\'):
323
cwd = ntpath.splitdrive(cwd)[0]
325
path = cwd + '\\' + path
326
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
329
def _win32_realpath(path):
330
# Real ntpath.realpath doesn't have a problem with a unicode cwd
331
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
334
def _win32_pathjoin(*args):
335
return ntpath.join(*args).replace('\\', '/')
338
def _win32_normpath(path):
339
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
343
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
346
def _win32_mkdtemp(*args, **kwargs):
347
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
350
def _win32_rename(old, new):
351
"""We expect to be able to atomically replace 'new' with old.
353
On win32, if new exists, it must be moved out of the way first,
357
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
359
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
360
# If we try to rename a non-existant file onto cwd, we get
361
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
362
# if the old path doesn't exist, sometimes we get EACCES
363
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
369
return unicodedata.normalize('NFC', os.getcwdu())
372
# Default is to just use the python builtins, but these can be rebound on
373
# particular platforms.
374
abspath = _posix_abspath
375
realpath = _posix_realpath
376
pathjoin = os.path.join
377
normpath = os.path.normpath
380
dirname = os.path.dirname
381
basename = os.path.basename
382
split = os.path.split
383
splitext = os.path.splitext
384
# These were already lazily imported into local scope
385
# mkdtemp = tempfile.mkdtemp
386
# rmtree = shutil.rmtree
394
MIN_ABS_PATHLENGTH = 1
397
if sys.platform == 'win32':
398
if win32utils.winver == 'Windows 98':
399
abspath = _win98_abspath
401
abspath = _win32_abspath
402
realpath = _win32_realpath
403
pathjoin = _win32_pathjoin
404
normpath = _win32_normpath
405
getcwd = _win32_getcwd
406
mkdtemp = _win32_mkdtemp
407
rename = _win32_rename
409
from bzrlib import _walkdirs_win32
413
lstat = _walkdirs_win32.lstat
414
fstat = _walkdirs_win32.fstat
415
wrap_stat = _walkdirs_win32.wrap_stat
417
MIN_ABS_PATHLENGTH = 3
419
def _win32_delete_readonly(function, path, excinfo):
420
"""Error handler for shutil.rmtree function [for win32]
421
Helps to remove files and dirs marked as read-only.
423
exception = excinfo[1]
424
if function in (os.remove, os.rmdir) \
425
and isinstance(exception, OSError) \
426
and exception.errno == errno.EACCES:
432
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
433
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
434
return shutil.rmtree(path, ignore_errors, onerror)
436
f = win32utils.get_unicode_argv # special function or None
440
elif sys.platform == 'darwin':
444
def get_terminal_encoding(trace=False):
445
"""Find the best encoding for printing to the screen.
447
This attempts to check both sys.stdout and sys.stdin to see
448
what encoding they are in, and if that fails it falls back to
449
osutils.get_user_encoding().
450
The problem is that on Windows, locale.getpreferredencoding()
451
is not the same encoding as that used by the console:
452
http://mail.python.org/pipermail/python-list/2003-May/162357.html
454
On my standard US Windows XP, the preferred encoding is
455
cp1252, but the console is cp437
457
:param trace: If True trace the selected encoding via mutter().
459
from bzrlib.trace import mutter
460
output_encoding = getattr(sys.stdout, 'encoding', None)
461
if not output_encoding:
462
input_encoding = getattr(sys.stdin, 'encoding', None)
463
if not input_encoding:
464
output_encoding = get_user_encoding()
466
mutter('encoding stdout as osutils.get_user_encoding() %r',
469
output_encoding = input_encoding
471
mutter('encoding stdout as sys.stdin encoding %r',
475
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
476
if output_encoding == 'cp0':
477
# invalid encoding (cp0 means 'no codepage' on Windows)
478
output_encoding = get_user_encoding()
480
mutter('cp0 is invalid encoding.'
481
' encoding stdout as osutils.get_user_encoding() %r',
485
codecs.lookup(output_encoding)
487
sys.stderr.write('bzr: warning:'
488
' unknown terminal encoding %s.\n'
489
' Using encoding %s instead.\n'
490
% (output_encoding, get_user_encoding())
492
output_encoding = get_user_encoding()
494
return output_encoding
497
def normalizepath(f):
498
if getattr(os.path, 'realpath', None) is not None:
502
[p,e] = os.path.split(f)
503
if e == "" or e == "." or e == "..":
506
return pathjoin(F(p), e)
510
"""True if f is an accessible directory."""
512
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
518
"""True if f is a regular file."""
520
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
525
"""True if f is a symlink."""
527
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
531
def is_inside(dir, fname):
532
"""True if fname is inside dir.
534
The parameters should typically be passed to osutils.normpath first, so
535
that . and .. and repeated slashes are eliminated, and the separators
536
are canonical for the platform.
538
The empty string as a dir name is taken as top-of-tree and matches
541
# XXX: Most callers of this can actually do something smarter by
542
# looking at the inventory
552
return fname.startswith(dir)
555
def is_inside_any(dir_list, fname):
556
"""True if fname is inside any of given dirs."""
557
for dirname in dir_list:
558
if is_inside(dirname, fname):
563
def is_inside_or_parent_of_any(dir_list, fname):
564
"""True if fname is a child or a parent of any of the given files."""
565
for dirname in dir_list:
566
if is_inside(dirname, fname) or is_inside(fname, dirname):
571
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
572
report_activity=None, direction='read'):
573
"""Copy contents of one file to another.
575
The read_length can either be -1 to read to end-of-file (EOF) or
576
it can specify the maximum number of bytes to read.
578
The buff_size represents the maximum size for each read operation
579
performed on from_file.
581
:param report_activity: Call this as bytes are read, see
582
Transport._report_activity
583
:param direction: Will be passed to report_activity
585
:return: The number of bytes copied.
589
# read specified number of bytes
591
while read_length > 0:
592
num_bytes_to_read = min(read_length, buff_size)
594
block = from_file.read(num_bytes_to_read)
598
if report_activity is not None:
599
report_activity(len(block), direction)
602
actual_bytes_read = len(block)
603
read_length -= actual_bytes_read
604
length += actual_bytes_read
608
block = from_file.read(buff_size)
612
if report_activity is not None:
613
report_activity(len(block), direction)
619
def pump_string_file(bytes, file_handle, segment_size=None):
620
"""Write bytes to file_handle in many smaller writes.
622
:param bytes: The string to write.
623
:param file_handle: The file to write to.
625
# Write data in chunks rather than all at once, because very large
626
# writes fail on some platforms (e.g. Windows with SMB mounted
629
segment_size = 5242880 # 5MB
630
segments = range(len(bytes) / segment_size + 1)
631
write = file_handle.write
632
for segment_index in segments:
633
segment = buffer(bytes, segment_index * segment_size, segment_size)
637
def file_iterator(input_file, readsize=32768):
639
b = input_file.read(readsize)
646
"""Calculate the hexdigest of an open file.
648
The file cursor should be already at the start.
660
def size_sha_file(f):
661
"""Calculate the size and hexdigest of an open file.
663
The file cursor should be already at the start and
664
the caller is responsible for closing the file afterwards.
675
return size, s.hexdigest()
678
def sha_file_by_name(fname):
679
"""Calculate the SHA1 of a file by reading the full text"""
681
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
684
b = os.read(f, 1<<16)
692
def sha_strings(strings, _factory=sha):
693
"""Return the sha-1 of concatenation of strings"""
695
map(s.update, strings)
699
def sha_string(f, _factory=sha):
700
return _factory(f).hexdigest()
703
def fingerprint_file(f):
705
return {'size': len(b),
706
'sha1': sha(b).hexdigest()}
709
def compare_files(a, b):
710
"""Returns true if equal in contents"""
721
def local_time_offset(t=None):
722
"""Return offset of local zone from GMT, either at present or at time t."""
725
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
726
return offset.days * 86400 + offset.seconds
728
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
729
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
732
def format_date(t, offset=0, timezone='original', date_fmt=None,
734
"""Return a formatted date string.
736
:param t: Seconds since the epoch.
737
:param offset: Timezone offset in seconds east of utc.
738
:param timezone: How to display the time: 'utc', 'original' for the
739
timezone specified by offset, or 'local' for the process's current
741
:param date_fmt: strftime format.
742
:param show_offset: Whether to append the timezone.
744
(date_fmt, tt, offset_str) = \
745
_format_date(t, offset, timezone, date_fmt, show_offset)
746
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
747
date_str = time.strftime(date_fmt, tt)
748
return date_str + offset_str
751
# Cache of formatted offset strings
755
def format_date_with_offset_in_original_timezone(t, offset=0,
756
_cache=_offset_cache):
757
"""Return a formatted date string in the original timezone.
759
This routine may be faster then format_date.
761
:param t: Seconds since the epoch.
762
:param offset: Timezone offset in seconds east of utc.
766
tt = time.gmtime(t + offset)
767
date_fmt = _default_format_by_weekday_num[tt[6]]
768
date_str = time.strftime(date_fmt, tt)
769
offset_str = _cache.get(offset, None)
770
if offset_str is None:
771
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
772
_cache[offset] = offset_str
773
return date_str + offset_str
776
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
778
"""Return an unicode date string formatted according to the current locale.
780
:param t: Seconds since the epoch.
781
:param offset: Timezone offset in seconds east of utc.
782
:param timezone: How to display the time: 'utc', 'original' for the
783
timezone specified by offset, or 'local' for the process's current
785
:param date_fmt: strftime format.
786
:param show_offset: Whether to append the timezone.
788
(date_fmt, tt, offset_str) = \
789
_format_date(t, offset, timezone, date_fmt, show_offset)
790
date_str = time.strftime(date_fmt, tt)
791
if not isinstance(date_str, unicode):
792
date_str = date_str.decode(get_user_encoding(), 'replace')
793
return date_str + offset_str
796
def _format_date(t, offset, timezone, date_fmt, show_offset):
797
if timezone == 'utc':
800
elif timezone == 'original':
803
tt = time.gmtime(t + offset)
804
elif timezone == 'local':
805
tt = time.localtime(t)
806
offset = local_time_offset(t)
808
raise errors.UnsupportedTimezoneFormat(timezone)
810
date_fmt = "%a %Y-%m-%d %H:%M:%S"
812
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
815
return (date_fmt, tt, offset_str)
818
def compact_date(when):
819
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
822
def format_delta(delta):
823
"""Get a nice looking string for a time delta.
825
:param delta: The time difference in seconds, can be positive or negative.
826
positive indicates time in the past, negative indicates time in the
827
future. (usually time.time() - stored_time)
828
:return: String formatted to show approximate resolution
834
direction = 'in the future'
838
if seconds < 90: # print seconds up to 90 seconds
840
return '%d second %s' % (seconds, direction,)
842
return '%d seconds %s' % (seconds, direction)
844
minutes = int(seconds / 60)
845
seconds -= 60 * minutes
850
if minutes < 90: # print minutes, seconds up to 90 minutes
852
return '%d minute, %d second%s %s' % (
853
minutes, seconds, plural_seconds, direction)
855
return '%d minutes, %d second%s %s' % (
856
minutes, seconds, plural_seconds, direction)
858
hours = int(minutes / 60)
859
minutes -= 60 * hours
866
return '%d hour, %d minute%s %s' % (hours, minutes,
867
plural_minutes, direction)
868
return '%d hours, %d minute%s %s' % (hours, minutes,
869
plural_minutes, direction)
872
"""Return size of given open file."""
873
return os.fstat(f.fileno())[stat.ST_SIZE]
876
# Define rand_bytes based on platform.
878
# Python 2.4 and later have os.urandom,
879
# but it doesn't work on some arches
881
rand_bytes = os.urandom
882
except (NotImplementedError, AttributeError):
883
# If python doesn't have os.urandom, or it doesn't work,
884
# then try to first pull random data from /dev/urandom
886
rand_bytes = file('/dev/urandom', 'rb').read
887
# Otherwise, use this hack as a last resort
888
except (IOError, OSError):
889
# not well seeded, but better than nothing
894
s += chr(random.randint(0, 255))
899
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
901
"""Return a random string of num alphanumeric characters
903
The result only contains lowercase chars because it may be used on
904
case-insensitive filesystems.
907
for raw_byte in rand_bytes(num):
908
s += ALNUM[ord(raw_byte) % 36]
912
## TODO: We could later have path objects that remember their list
913
## decomposition (might be too tricksy though.)
916
"""Turn string into list of parts."""
917
# split on either delimiter because people might use either on
919
ps = re.split(r'[\\/]', p)
924
raise errors.BzrError("sorry, %r not allowed in path" % f)
925
elif (f == '.') or (f == ''):
934
if (f == '..') or (f is None) or (f == ''):
935
raise errors.BzrError("sorry, %r not allowed in path" % f)
939
def parent_directories(filename):
940
"""Return the list of parent directories, deepest first.
942
For example, parent_directories("a/b/c") -> ["a/b", "a"].
945
parts = splitpath(dirname(filename))
947
parents.append(joinpath(parts))
952
_extension_load_failures = []
955
def failed_to_load_extension(exception):
956
"""Handle failing to load a binary extension.
958
This should be called from the ImportError block guarding the attempt to
959
import the native extension. If this function returns, the pure-Python
960
implementation should be loaded instead::
963
>>> import bzrlib._fictional_extension_pyx
964
>>> except ImportError, e:
965
>>> bzrlib.osutils.failed_to_load_extension(e)
966
>>> import bzrlib._fictional_extension_py
968
# NB: This docstring is just an example, not a doctest, because doctest
969
# currently can't cope with the use of lazy imports in this namespace --
972
# This currently doesn't report the failure at the time it occurs, because
973
# they tend to happen very early in startup when we can't check config
974
# files etc, and also we want to report all failures but not spam the user
976
exception_str = str(exception)
977
if exception_str not in _extension_load_failures:
978
trace.mutter("failed to load compiled extension: %s" % exception_str)
979
_extension_load_failures.append(exception_str)
982
def report_extension_load_failures():
983
if not _extension_load_failures:
985
from bzrlib.config import GlobalConfig
986
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
988
# the warnings framework should by default show this only once
989
from bzrlib.trace import warning
991
"bzr: warning: some compiled extensions could not be loaded; "
992
"see <https://answers.launchpad.net/bzr/+faq/703>")
993
# we no longer show the specific missing extensions here, because it makes
994
# the message too long and scary - see
995
# https://bugs.launchpad.net/bzr/+bug/430529
999
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1000
except ImportError, e:
1001
failed_to_load_extension(e)
1002
from bzrlib._chunks_to_lines_py import chunks_to_lines
1006
"""Split s into lines, but without removing the newline characters."""
1007
# Trivially convert a fulltext into a 'chunked' representation, and let
1008
# chunks_to_lines do the heavy lifting.
1009
if isinstance(s, str):
1010
# chunks_to_lines only supports 8-bit strings
1011
return chunks_to_lines([s])
1013
return _split_lines(s)
1016
def _split_lines(s):
1017
"""Split s into lines, but without removing the newline characters.
1019
This supports Unicode or plain string objects.
1021
lines = s.split('\n')
1022
result = [line + '\n' for line in lines[:-1]]
1024
result.append(lines[-1])
1028
def hardlinks_good():
1029
return sys.platform not in ('win32', 'cygwin', 'darwin')
1032
def link_or_copy(src, dest):
1033
"""Hardlink a file, or copy it if it can't be hardlinked."""
1034
if not hardlinks_good():
1035
shutil.copyfile(src, dest)
1039
except (OSError, IOError), e:
1040
if e.errno != errno.EXDEV:
1042
shutil.copyfile(src, dest)
1045
def delete_any(path):
1046
"""Delete a file, symlink or directory.
1048
Will delete even if readonly.
1051
_delete_file_or_dir(path)
1052
except (OSError, IOError), e:
1053
if e.errno in (errno.EPERM, errno.EACCES):
1054
# make writable and try again
1057
except (OSError, IOError):
1059
_delete_file_or_dir(path)
1064
def _delete_file_or_dir(path):
1065
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1066
# Forgiveness than Permission (EAFP) because:
1067
# - root can damage a solaris file system by using unlink,
1068
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1069
# EACCES, OSX: EPERM) when invoked on a directory.
1070
if isdir(path): # Takes care of symlinks
1077
if getattr(os, 'symlink', None) is not None:
1083
def has_hardlinks():
1084
if getattr(os, 'link', None) is not None:
1090
def host_os_dereferences_symlinks():
1091
return (has_symlinks()
1092
and sys.platform not in ('cygwin', 'win32'))
1095
def readlink(abspath):
1096
"""Return a string representing the path to which the symbolic link points.
1098
:param abspath: The link absolute unicode path.
1100
This his guaranteed to return the symbolic link in unicode in all python
1103
link = abspath.encode(_fs_enc)
1104
target = os.readlink(link)
1105
target = target.decode(_fs_enc)
1109
def contains_whitespace(s):
1110
"""True if there are any whitespace characters in s."""
1111
# string.whitespace can include '\xa0' in certain locales, because it is
1112
# considered "non-breaking-space" as part of ISO-8859-1. But it
1113
# 1) Isn't a breaking whitespace
1114
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1116
# 3) '\xa0' isn't unicode safe since it is >128.
1118
# This should *not* be a unicode set of characters in case the source
1119
# string is not a Unicode string. We can auto-up-cast the characters since
1120
# they are ascii, but we don't want to auto-up-cast the string in case it
1122
for ch in ' \t\n\r\v\f':
1129
def contains_linebreaks(s):
1130
"""True if there is any vertical whitespace in s."""
1138
def relpath(base, path):
1139
"""Return path relative to base, or raise PathNotChild exception.
1141
The path may be either an absolute path or a path relative to the
1142
current working directory.
1144
os.path.commonprefix (python2.4) has a bad bug that it works just
1145
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1146
avoids that problem.
1148
NOTE: `base` should not have a trailing slash otherwise you'll get
1149
PathNotChild exceptions regardless of `path`.
1152
if len(base) < MIN_ABS_PATHLENGTH:
1153
# must have space for e.g. a drive letter
1154
raise ValueError('%r is too short to calculate a relative path'
1162
if len(head) <= len(base) and head != base:
1163
raise errors.PathNotChild(rp, base)
1166
head, tail = split(head)
1171
return pathjoin(*reversed(s))
1176
def _cicp_canonical_relpath(base, path):
1177
"""Return the canonical path relative to base.
1179
Like relpath, but on case-insensitive-case-preserving file-systems, this
1180
will return the relpath as stored on the file-system rather than in the
1181
case specified in the input string, for all existing portions of the path.
1183
This will cause O(N) behaviour if called for every path in a tree; if you
1184
have a number of paths to convert, you should use canonical_relpaths().
1186
# TODO: it should be possible to optimize this for Windows by using the
1187
# win32 API FindFiles function to look for the specified name - but using
1188
# os.listdir() still gives us the correct, platform agnostic semantics in
1191
rel = relpath(base, path)
1192
# '.' will have been turned into ''
1196
abs_base = abspath(base)
1198
_listdir = os.listdir
1200
# use an explicit iterator so we can easily consume the rest on early exit.
1201
bit_iter = iter(rel.split('/'))
1202
for bit in bit_iter:
1205
next_entries = _listdir(current)
1206
except OSError: # enoent, eperm, etc
1207
# We can't find this in the filesystem, so just append the
1209
current = pathjoin(current, bit, *list(bit_iter))
1211
for look in next_entries:
1212
if lbit == look.lower():
1213
current = pathjoin(current, look)
1216
# got to the end, nothing matched, so we just return the
1217
# non-existing bits as they were specified (the filename may be
1218
# the target of a move, for example).
1219
current = pathjoin(current, bit, *list(bit_iter))
1221
return current[len(abs_base):].lstrip('/')
1223
# XXX - TODO - we need better detection/integration of case-insensitive
1224
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1225
# filesystems), for example, so could probably benefit from the same basic
1226
# support there. For now though, only Windows and OSX get that support, and
1227
# they get it for *all* file-systems!
1228
if sys.platform in ('win32', 'darwin'):
1229
canonical_relpath = _cicp_canonical_relpath
1231
canonical_relpath = relpath
1233
def canonical_relpaths(base, paths):
1234
"""Create an iterable to canonicalize a sequence of relative paths.
1236
The intent is for this implementation to use a cache, vastly speeding
1237
up multiple transformations in the same directory.
1239
# but for now, we haven't optimized...
1240
return [canonical_relpath(base, p) for p in paths]
1243
def decode_filename(filename):
1244
"""Decode the filename using the filesystem encoding
1246
If it is unicode, it is returned.
1247
Otherwise it is decoded from the the filesystem's encoding. If decoding
1248
fails, a errors.BadFilenameEncoding exception is raised.
1250
if type(filename) is unicode:
1253
return filename.decode(_fs_enc)
1254
except UnicodeDecodeError:
1255
raise errors.BadFilenameEncoding(filename, _fs_enc)
1258
def safe_unicode(unicode_or_utf8_string):
1259
"""Coerce unicode_or_utf8_string into unicode.
1261
If it is unicode, it is returned.
1262
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1263
wrapped in a BzrBadParameterNotUnicode exception.
1265
if isinstance(unicode_or_utf8_string, unicode):
1266
return unicode_or_utf8_string
1268
return unicode_or_utf8_string.decode('utf8')
1269
except UnicodeDecodeError:
1270
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1273
def safe_utf8(unicode_or_utf8_string):
1274
"""Coerce unicode_or_utf8_string to a utf8 string.
1276
If it is a str, it is returned.
1277
If it is Unicode, it is encoded into a utf-8 string.
1279
if isinstance(unicode_or_utf8_string, str):
1280
# TODO: jam 20070209 This is overkill, and probably has an impact on
1281
# performance if we are dealing with lots of apis that want a
1284
# Make sure it is a valid utf-8 string
1285
unicode_or_utf8_string.decode('utf-8')
1286
except UnicodeDecodeError:
1287
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1288
return unicode_or_utf8_string
1289
return unicode_or_utf8_string.encode('utf-8')
1292
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1293
' Revision id generators should be creating utf8'
1297
def safe_revision_id(unicode_or_utf8_string, warn=True):
1298
"""Revision ids should now be utf8, but at one point they were unicode.
1300
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1302
:param warn: Functions that are sanitizing user data can set warn=False
1303
:return: None or a utf8 revision id.
1305
if (unicode_or_utf8_string is None
1306
or unicode_or_utf8_string.__class__ == str):
1307
return unicode_or_utf8_string
1309
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1311
return cache_utf8.encode(unicode_or_utf8_string)
1314
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1315
' generators should be creating utf8 file ids.')
1318
def safe_file_id(unicode_or_utf8_string, warn=True):
1319
"""File ids should now be utf8, but at one point they were unicode.
1321
This is the same as safe_utf8, except it uses the cached encode functions
1322
to save a little bit of performance.
1324
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1326
:param warn: Functions that are sanitizing user data can set warn=False
1327
:return: None or a utf8 file id.
1329
if (unicode_or_utf8_string is None
1330
or unicode_or_utf8_string.__class__ == str):
1331
return unicode_or_utf8_string
1333
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1335
return cache_utf8.encode(unicode_or_utf8_string)
1338
_platform_normalizes_filenames = False
1339
if sys.platform == 'darwin':
1340
_platform_normalizes_filenames = True
1343
def normalizes_filenames():
1344
"""Return True if this platform normalizes unicode filenames.
1348
return _platform_normalizes_filenames
1351
def _accessible_normalized_filename(path):
1352
"""Get the unicode normalized path, and if you can access the file.
1354
On platforms where the system normalizes filenames (Mac OSX),
1355
you can access a file by any path which will normalize correctly.
1356
On platforms where the system does not normalize filenames
1357
(everything else), you have to access a file by its exact path.
1359
Internally, bzr only supports NFC normalization, since that is
1360
the standard for XML documents.
1362
So return the normalized path, and a flag indicating if the file
1363
can be accessed by that path.
1366
return unicodedata.normalize('NFC', unicode(path)), True
1369
def _inaccessible_normalized_filename(path):
1370
__doc__ = _accessible_normalized_filename.__doc__
1372
normalized = unicodedata.normalize('NFC', unicode(path))
1373
return normalized, normalized == path
1376
if _platform_normalizes_filenames:
1377
normalized_filename = _accessible_normalized_filename
1379
normalized_filename = _inaccessible_normalized_filename
1382
def set_signal_handler(signum, handler, restart_syscall=True):
1383
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1384
on platforms that support that.
1386
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1387
automatically restart (by calling `signal.siginterrupt(signum,
1388
False)`). May be ignored if the feature is not available on this
1389
platform or Python version.
1393
siginterrupt = signal.siginterrupt
1395
# This python implementation doesn't provide signal support, hence no
1398
except AttributeError:
1399
# siginterrupt doesn't exist on this platform, or for this version
1401
siginterrupt = lambda signum, flag: None
1403
def sig_handler(*args):
1404
# Python resets the siginterrupt flag when a signal is
1405
# received. <http://bugs.python.org/issue8354>
1406
# As a workaround for some cases, set it back the way we want it.
1407
siginterrupt(signum, False)
1408
# Now run the handler function passed to set_signal_handler.
1411
sig_handler = handler
1412
old_handler = signal.signal(signum, sig_handler)
1414
siginterrupt(signum, False)
1418
default_terminal_width = 80
1419
"""The default terminal width for ttys.
1421
This is defined so that higher levels can share a common fallback value when
1422
terminal_width() returns None.
1425
# Keep some state so that terminal_width can detect if _terminal_size has
1426
# returned a different size since the process started. See docstring and
1427
# comments of terminal_width for details.
1428
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1429
_terminal_size_state = 'no_data'
1430
_first_terminal_size = None
1432
def terminal_width():
1433
"""Return terminal width.
1435
None is returned if the width can't established precisely.
1438
- if BZR_COLUMNS is set, returns its value
1439
- if there is no controlling terminal, returns None
1440
- query the OS, if the queried size has changed since the last query,
1442
- if COLUMNS is set, returns its value,
1443
- if the OS has a value (even though it's never changed), return its value.
1445
From there, we need to query the OS to get the size of the controlling
1448
On Unices we query the OS by:
1449
- get termios.TIOCGWINSZ
1450
- if an error occurs or a negative value is obtained, returns None
1452
On Windows we query the OS by:
1453
- win32utils.get_console_size() decides,
1454
- returns None on error (provided default value)
1456
# Note to implementors: if changing the rules for determining the width,
1457
# make sure you've considered the behaviour in these cases:
1458
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1459
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1461
# - (add more interesting cases here, if you find any)
1462
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1463
# but we don't want to register a signal handler because it is impossible
1464
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1465
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1466
# time so we can notice if the reported size has changed, which should have
1469
# If BZR_COLUMNS is set, take it, user is always right
1470
# Except if they specified 0 in which case, impose no limit here
1472
width = int(os.environ['BZR_COLUMNS'])
1473
except (KeyError, ValueError):
1475
if width is not None:
1481
isatty = getattr(sys.stdout, 'isatty', None)
1482
if isatty is None or not isatty():
1483
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1487
width, height = os_size = _terminal_size(None, None)
1488
global _first_terminal_size, _terminal_size_state
1489
if _terminal_size_state == 'no_data':
1490
_first_terminal_size = os_size
1491
_terminal_size_state = 'unchanged'
1492
elif (_terminal_size_state == 'unchanged' and
1493
_first_terminal_size != os_size):
1494
_terminal_size_state = 'changed'
1496
# If the OS claims to know how wide the terminal is, and this value has
1497
# ever changed, use that.
1498
if _terminal_size_state == 'changed':
1499
if width is not None and width > 0:
1502
# If COLUMNS is set, use it.
1504
return int(os.environ['COLUMNS'])
1505
except (KeyError, ValueError):
1508
# Finally, use an unchanged size from the OS, if we have one.
1509
if _terminal_size_state == 'unchanged':
1510
if width is not None and width > 0:
1513
# The width could not be determined.
1517
def _win32_terminal_size(width, height):
1518
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1519
return width, height
1522
def _ioctl_terminal_size(width, height):
1524
import struct, fcntl, termios
1525
s = struct.pack('HHHH', 0, 0, 0, 0)
1526
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1527
height, width = struct.unpack('HHHH', x)[0:2]
1528
except (IOError, AttributeError):
1530
return width, height
1532
_terminal_size = None
1533
"""Returns the terminal size as (width, height).
1535
:param width: Default value for width.
1536
:param height: Default value for height.
1538
This is defined specifically for each OS and query the size of the controlling
1539
terminal. If any error occurs, the provided default values should be returned.
1541
if sys.platform == 'win32':
1542
_terminal_size = _win32_terminal_size
1544
_terminal_size = _ioctl_terminal_size
1547
def supports_executable():
1548
return sys.platform != "win32"
1551
def supports_posix_readonly():
1552
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1554
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1555
directory controls creation/deletion, etc.
1557
And under win32, readonly means that the directory itself cannot be
1558
deleted. The contents of a readonly directory can be changed, unlike POSIX
1559
where files in readonly directories cannot be added, deleted or renamed.
1561
return sys.platform != "win32"
1564
def set_or_unset_env(env_variable, value):
1565
"""Modify the environment, setting or removing the env_variable.
1567
:param env_variable: The environment variable in question
1568
:param value: The value to set the environment to. If None, then
1569
the variable will be removed.
1570
:return: The original value of the environment variable.
1572
orig_val = os.environ.get(env_variable)
1574
if orig_val is not None:
1575
del os.environ[env_variable]
1577
if isinstance(value, unicode):
1578
value = value.encode(get_user_encoding())
1579
os.environ[env_variable] = value
1583
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1586
def check_legal_path(path):
1587
"""Check whether the supplied path is legal.
1588
This is only required on Windows, so we don't test on other platforms
1591
if sys.platform != "win32":
1593
if _validWin32PathRE.match(path) is None:
1594
raise errors.IllegalPath(path)
1597
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1599
def _is_error_enotdir(e):
1600
"""Check if this exception represents ENOTDIR.
1602
Unfortunately, python is very inconsistent about the exception
1603
here. The cases are:
1604
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1605
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1606
which is the windows error code.
1607
3) Windows, Python2.5 uses errno == EINVAL and
1608
winerror == ERROR_DIRECTORY
1610
:param e: An Exception object (expected to be OSError with an errno
1611
attribute, but we should be able to cope with anything)
1612
:return: True if this represents an ENOTDIR error. False otherwise.
1614
en = getattr(e, 'errno', None)
1615
if (en == errno.ENOTDIR
1616
or (sys.platform == 'win32'
1617
and (en == _WIN32_ERROR_DIRECTORY
1618
or (en == errno.EINVAL
1619
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1625
def walkdirs(top, prefix=""):
1626
"""Yield data about all the directories in a tree.
1628
This yields all the data about the contents of a directory at a time.
1629
After each directory has been yielded, if the caller has mutated the list
1630
to exclude some directories, they are then not descended into.
1632
The data yielded is of the form:
1633
((directory-relpath, directory-path-from-top),
1634
[(relpath, basename, kind, lstat, path-from-top), ...]),
1635
- directory-relpath is the relative path of the directory being returned
1636
with respect to top. prefix is prepended to this.
1637
- directory-path-from-root is the path including top for this directory.
1638
It is suitable for use with os functions.
1639
- relpath is the relative path within the subtree being walked.
1640
- basename is the basename of the path
1641
- kind is the kind of the file now. If unknown then the file is not
1642
present within the tree - but it may be recorded as versioned. See
1644
- lstat is the stat data *if* the file was statted.
1645
- planned, not implemented:
1646
path_from_tree_root is the path from the root of the tree.
1648
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1649
allows one to walk a subtree but get paths that are relative to a tree
1651
:return: an iterator over the dirs.
1653
#TODO there is a bit of a smell where the results of the directory-
1654
# summary in this, and the path from the root, may not agree
1655
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1656
# potentially confusing output. We should make this more robust - but
1657
# not at a speed cost. RBC 20060731
1659
_directory = _directory_kind
1660
_listdir = os.listdir
1661
_kind_from_mode = file_kind_from_stat_mode
1662
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1664
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1665
relroot, _, _, _, top = pending.pop()
1667
relprefix = relroot + u'/'
1670
top_slash = top + u'/'
1673
append = dirblock.append
1675
names = sorted(map(decode_filename, _listdir(top)))
1677
if not _is_error_enotdir(e):
1681
abspath = top_slash + name
1682
statvalue = _lstat(abspath)
1683
kind = _kind_from_mode(statvalue.st_mode)
1684
append((relprefix + name, name, kind, statvalue, abspath))
1685
yield (relroot, top), dirblock
1687
# push the user specified dirs from dirblock
1688
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1691
class DirReader(object):
1692
"""An interface for reading directories."""
1694
def top_prefix_to_starting_dir(self, top, prefix=""):
1695
"""Converts top and prefix to a starting dir entry
1697
:param top: A utf8 path
1698
:param prefix: An optional utf8 path to prefix output relative paths
1700
:return: A tuple starting with prefix, and ending with the native
1703
raise NotImplementedError(self.top_prefix_to_starting_dir)
1705
def read_dir(self, prefix, top):
1706
"""Read a specific dir.
1708
:param prefix: A utf8 prefix to be preprended to the path basenames.
1709
:param top: A natively encoded path to read.
1710
:return: A list of the directories contents. Each item contains:
1711
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1713
raise NotImplementedError(self.read_dir)
1716
_selected_dir_reader = None
1719
def _walkdirs_utf8(top, prefix=""):
1720
"""Yield data about all the directories in a tree.
1722
This yields the same information as walkdirs() only each entry is yielded
1723
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1724
are returned as exact byte-strings.
1726
:return: yields a tuple of (dir_info, [file_info])
1727
dir_info is (utf8_relpath, path-from-top)
1728
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1729
if top is an absolute path, path-from-top is also an absolute path.
1730
path-from-top might be unicode or utf8, but it is the correct path to
1731
pass to os functions to affect the file in question. (such as os.lstat)
1733
global _selected_dir_reader
1734
if _selected_dir_reader is None:
1735
fs_encoding = _fs_enc.upper()
1736
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1737
# Win98 doesn't have unicode apis like FindFirstFileW
1738
# TODO: We possibly could support Win98 by falling back to the
1739
# original FindFirstFile, and using TCHAR instead of WCHAR,
1740
# but that gets a bit tricky, and requires custom compiling
1743
from bzrlib._walkdirs_win32 import Win32ReadDir
1744
_selected_dir_reader = Win32ReadDir()
1747
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1748
# ANSI_X3.4-1968 is a form of ASCII
1750
from bzrlib._readdir_pyx import UTF8DirReader
1751
_selected_dir_reader = UTF8DirReader()
1752
except ImportError, e:
1753
failed_to_load_extension(e)
1756
if _selected_dir_reader is None:
1757
# Fallback to the python version
1758
_selected_dir_reader = UnicodeDirReader()
1760
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1761
# But we don't actually uses 1-3 in pending, so set them to None
1762
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1763
read_dir = _selected_dir_reader.read_dir
1764
_directory = _directory_kind
1766
relroot, _, _, _, top = pending[-1].pop()
1769
dirblock = sorted(read_dir(relroot, top))
1770
yield (relroot, top), dirblock
1771
# push the user specified dirs from dirblock
1772
next = [d for d in reversed(dirblock) if d[2] == _directory]
1774
pending.append(next)
1777
class UnicodeDirReader(DirReader):
1778
"""A dir reader for non-utf8 file systems, which transcodes."""
1780
__slots__ = ['_utf8_encode']
1783
self._utf8_encode = codecs.getencoder('utf8')
1785
def top_prefix_to_starting_dir(self, top, prefix=""):
1786
"""See DirReader.top_prefix_to_starting_dir."""
1787
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1789
def read_dir(self, prefix, top):
1790
"""Read a single directory from a non-utf8 file system.
1792
top, and the abspath element in the output are unicode, all other paths
1793
are utf8. Local disk IO is done via unicode calls to listdir etc.
1795
This is currently the fallback code path when the filesystem encoding is
1796
not UTF-8. It may be better to implement an alternative so that we can
1797
safely handle paths that are not properly decodable in the current
1800
See DirReader.read_dir for details.
1802
_utf8_encode = self._utf8_encode
1804
_listdir = os.listdir
1805
_kind_from_mode = file_kind_from_stat_mode
1808
relprefix = prefix + '/'
1811
top_slash = top + u'/'
1814
append = dirblock.append
1815
for name in sorted(_listdir(top)):
1817
name_utf8 = _utf8_encode(name)[0]
1818
except UnicodeDecodeError:
1819
raise errors.BadFilenameEncoding(
1820
_utf8_encode(relprefix)[0] + name, _fs_enc)
1821
abspath = top_slash + name
1822
statvalue = _lstat(abspath)
1823
kind = _kind_from_mode(statvalue.st_mode)
1824
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1828
def copy_tree(from_path, to_path, handlers={}):
1829
"""Copy all of the entries in from_path into to_path.
1831
:param from_path: The base directory to copy.
1832
:param to_path: The target directory. If it does not exist, it will
1834
:param handlers: A dictionary of functions, which takes a source and
1835
destinations for files, directories, etc.
1836
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1837
'file', 'directory', and 'symlink' should always exist.
1838
If they are missing, they will be replaced with 'os.mkdir()',
1839
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1841
# Now, just copy the existing cached tree to the new location
1842
# We use a cheap trick here.
1843
# Absolute paths are prefixed with the first parameter
1844
# relative paths are prefixed with the second.
1845
# So we can get both the source and target returned
1846
# without any extra work.
1848
def copy_dir(source, dest):
1851
def copy_link(source, dest):
1852
"""Copy the contents of a symlink"""
1853
link_to = os.readlink(source)
1854
os.symlink(link_to, dest)
1856
real_handlers = {'file':shutil.copy2,
1857
'symlink':copy_link,
1858
'directory':copy_dir,
1860
real_handlers.update(handlers)
1862
if not os.path.exists(to_path):
1863
real_handlers['directory'](from_path, to_path)
1865
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1866
for relpath, name, kind, st, abspath in entries:
1867
real_handlers[kind](abspath, relpath)
1870
def copy_ownership_from_path(dst, src=None):
1871
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1873
If src is None, the containing directory is used as source. If chown
1874
fails, the error is ignored and a warning is printed.
1876
chown = getattr(os, 'chown', None)
1881
src = os.path.dirname(dst)
1887
chown(dst, s.st_uid, s.st_gid)
1890
'Unable to copy ownership from "%s" to "%s". '
1891
'You may want to set it manually.', src, dst)
1892
trace.log_exception_quietly()
1895
def path_prefix_key(path):
1896
"""Generate a prefix-order path key for path.
1898
This can be used to sort paths in the same way that walkdirs does.
1900
return (dirname(path) , path)
1903
def compare_paths_prefix_order(path_a, path_b):
1904
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1905
key_a = path_prefix_key(path_a)
1906
key_b = path_prefix_key(path_b)
1907
return cmp(key_a, key_b)
1910
_cached_user_encoding = None
1913
def get_user_encoding(use_cache=True):
1914
"""Find out what the preferred user encoding is.
1916
This is generally the encoding that is used for command line parameters
1917
and file contents. This may be different from the terminal encoding
1918
or the filesystem encoding.
1920
:param use_cache: Enable cache for detected encoding.
1921
(This parameter is turned on by default,
1922
and required only for selftesting)
1924
:return: A string defining the preferred user encoding
1926
global _cached_user_encoding
1927
if _cached_user_encoding is not None and use_cache:
1928
return _cached_user_encoding
1930
if sys.platform == 'darwin':
1931
# python locale.getpreferredencoding() always return
1932
# 'mac-roman' on darwin. That's a lie.
1933
sys.platform = 'posix'
1935
if os.environ.get('LANG', None) is None:
1936
# If LANG is not set, we end up with 'ascii', which is bad
1937
# ('mac-roman' is more than ascii), so we set a default which
1938
# will give us UTF-8 (which appears to work in all cases on
1939
# OSX). Users are still free to override LANG of course, as
1940
# long as it give us something meaningful. This work-around
1941
# *may* not be needed with python 3k and/or OSX 10.5, but will
1942
# work with them too -- vila 20080908
1943
os.environ['LANG'] = 'en_US.UTF-8'
1946
sys.platform = 'darwin'
1951
user_encoding = locale.getpreferredencoding()
1952
except locale.Error, e:
1953
sys.stderr.write('bzr: warning: %s\n'
1954
' Could not determine what text encoding to use.\n'
1955
' This error usually means your Python interpreter\n'
1956
' doesn\'t support the locale set by $LANG (%s)\n'
1957
" Continuing with ascii encoding.\n"
1958
% (e, os.environ.get('LANG')))
1959
user_encoding = 'ascii'
1961
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1962
# treat that as ASCII, and not support printing unicode characters to the
1965
# For python scripts run under vim, we get '', so also treat that as ASCII
1966
if user_encoding in (None, 'cp0', ''):
1967
user_encoding = 'ascii'
1971
codecs.lookup(user_encoding)
1973
sys.stderr.write('bzr: warning:'
1974
' unknown encoding %s.'
1975
' Continuing with ascii encoding.\n'
1978
user_encoding = 'ascii'
1981
_cached_user_encoding = user_encoding
1983
return user_encoding
1986
def get_diff_header_encoding():
1987
return get_terminal_encoding()
1990
def get_host_name():
1991
"""Return the current unicode host name.
1993
This is meant to be used in place of socket.gethostname() because that
1994
behaves inconsistently on different platforms.
1996
if sys.platform == "win32":
1998
return win32utils.get_host_name()
2001
return socket.gethostname().decode(get_user_encoding())
2004
# We must not read/write any more than 64k at a time from/to a socket so we
2005
# don't risk "no buffer space available" errors on some platforms. Windows in
2006
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2008
MAX_SOCKET_CHUNK = 64 * 1024
2010
_end_of_stream_errors = [errno.ECONNRESET]
2011
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2012
_eno = getattr(errno, _eno, None)
2013
if _eno is not None:
2014
_end_of_stream_errors.append(_eno)
2018
def read_bytes_from_socket(sock, report_activity=None,
2019
max_read_size=MAX_SOCKET_CHUNK):
2020
"""Read up to max_read_size of bytes from sock and notify of progress.
2022
Translates "Connection reset by peer" into file-like EOF (return an
2023
empty string rather than raise an error), and repeats the recv if
2024
interrupted by a signal.
2028
bytes = sock.recv(max_read_size)
2029
except socket.error, e:
2031
if eno in _end_of_stream_errors:
2032
# The connection was closed by the other side. Callers expect
2033
# an empty string to signal end-of-stream.
2035
elif eno == errno.EINTR:
2036
# Retry the interrupted recv.
2040
if report_activity is not None:
2041
report_activity(len(bytes), 'read')
2045
def recv_all(socket, count):
2046
"""Receive an exact number of bytes.
2048
Regular Socket.recv() may return less than the requested number of bytes,
2049
depending on what's in the OS buffer. MSG_WAITALL is not available
2050
on all platforms, but this should work everywhere. This will return
2051
less than the requested amount if the remote end closes.
2053
This isn't optimized and is intended mostly for use in testing.
2056
while len(b) < count:
2057
new = read_bytes_from_socket(socket, None, count - len(b))
2064
def send_all(sock, bytes, report_activity=None):
2065
"""Send all bytes on a socket.
2067
Breaks large blocks in smaller chunks to avoid buffering limitations on
2068
some platforms, and catches EINTR which may be thrown if the send is
2069
interrupted by a signal.
2071
This is preferred to socket.sendall(), because it avoids portability bugs
2072
and provides activity reporting.
2074
:param report_activity: Call this as bytes are read, see
2075
Transport._report_activity
2078
byte_count = len(bytes)
2079
while sent_total < byte_count:
2081
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2082
except socket.error, e:
2083
if e.args[0] != errno.EINTR:
2087
report_activity(sent, 'write')
2090
def connect_socket(address):
2091
# Slight variation of the socket.create_connection() function (provided by
2092
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2093
# provide it for previous python versions. Also, we don't use the timeout
2094
# parameter (provided by the python implementation) so we don't implement
2096
err = socket.error('getaddrinfo returns an empty list')
2097
host, port = address
2098
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2099
af, socktype, proto, canonname, sa = res
2102
sock = socket.socket(af, socktype, proto)
2106
except socket.error, err:
2107
# 'err' is now the most recent error
2108
if sock is not None:
2113
def dereference_path(path):
2114
"""Determine the real path to a file.
2116
All parent elements are dereferenced. But the file itself is not
2118
:param path: The original path. May be absolute or relative.
2119
:return: the real path *to* the file
2121
parent, base = os.path.split(path)
2122
# The pathjoin for '.' is a workaround for Python bug #1213894.
2123
# (initial path components aren't dereferenced)
2124
return pathjoin(realpath(pathjoin('.', parent)), base)
2127
def supports_mapi():
2128
"""Return True if we can use MAPI to launch a mail client."""
2129
return sys.platform == "win32"
2132
def resource_string(package, resource_name):
2133
"""Load a resource from a package and return it as a string.
2135
Note: Only packages that start with bzrlib are currently supported.
2137
This is designed to be a lightweight implementation of resource
2138
loading in a way which is API compatible with the same API from
2140
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2141
If and when pkg_resources becomes a standard library, this routine
2144
# Check package name is within bzrlib
2145
if package == "bzrlib":
2146
resource_relpath = resource_name
2147
elif package.startswith("bzrlib."):
2148
package = package[len("bzrlib."):].replace('.', os.sep)
2149
resource_relpath = pathjoin(package, resource_name)
2151
raise errors.BzrError('resource package %s not in bzrlib' % package)
2153
# Map the resource to a file and read its contents
2154
base = dirname(bzrlib.__file__)
2155
if getattr(sys, 'frozen', None): # bzr.exe
2156
base = abspath(pathjoin(base, '..', '..'))
2157
f = file(pathjoin(base, resource_relpath), "rU")
2163
def file_kind_from_stat_mode_thunk(mode):
2164
global file_kind_from_stat_mode
2165
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2167
from bzrlib._readdir_pyx import UTF8DirReader
2168
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2169
except ImportError, e:
2170
# This is one time where we won't warn that an extension failed to
2171
# load. The extension is never available on Windows anyway.
2172
from bzrlib._readdir_py import (
2173
_kind_from_mode as file_kind_from_stat_mode
2175
return file_kind_from_stat_mode(mode)
2176
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2179
def file_kind(f, _lstat=os.lstat):
2181
return file_kind_from_stat_mode(_lstat(f).st_mode)
2183
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2184
raise errors.NoSuchFile(f)
2188
def until_no_eintr(f, *a, **kw):
2189
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2191
WARNING: you must be certain that it is safe to retry the call repeatedly
2192
if EINTR does occur. This is typically only true for low-level operations
2193
like os.read. If in any doubt, don't use this.
2195
Keep in mind that this is not a complete solution to EINTR. There is
2196
probably code in the Python standard library and other dependencies that
2197
may encounter EINTR if a signal arrives (and there is signal handler for
2198
that signal). So this function can reduce the impact for IO that bzrlib
2199
directly controls, but it is not a complete solution.
2201
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2205
except (IOError, OSError), e:
2206
if e.errno == errno.EINTR:
2211
@deprecated_function(deprecated_in((2, 2, 0)))
2212
def re_compile_checked(re_string, flags=0, where=""):
2213
"""Return a compiled re, or raise a sensible error.
2215
This should only be used when compiling user-supplied REs.
2217
:param re_string: Text form of regular expression.
2218
:param flags: eg re.IGNORECASE
2219
:param where: Message explaining to the user the context where
2220
it occurred, eg 'log search filter'.
2222
# from https://bugs.launchpad.net/bzr/+bug/251352
2224
re_obj = re.compile(re_string, flags)
2227
except errors.InvalidPattern, e:
2229
where = ' in ' + where
2230
# despite the name 'error' is a type
2231
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2235
if sys.platform == "win32":
2238
return msvcrt.getch()
2243
fd = sys.stdin.fileno()
2244
settings = termios.tcgetattr(fd)
2247
ch = sys.stdin.read(1)
2249
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2252
if sys.platform == 'linux2':
2253
def _local_concurrency():
2255
return os.sysconf('SC_NPROCESSORS_ONLN')
2256
except (ValueError, OSError, AttributeError):
2258
elif sys.platform == 'darwin':
2259
def _local_concurrency():
2260
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2261
stdout=subprocess.PIPE).communicate()[0]
2262
elif "bsd" in sys.platform:
2263
def _local_concurrency():
2264
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2265
stdout=subprocess.PIPE).communicate()[0]
2266
elif sys.platform == 'sunos5':
2267
def _local_concurrency():
2268
return subprocess.Popen(['psrinfo', '-p',],
2269
stdout=subprocess.PIPE).communicate()[0]
2270
elif sys.platform == "win32":
2271
def _local_concurrency():
2272
# This appears to return the number of cores.
2273
return os.environ.get('NUMBER_OF_PROCESSORS')
2275
def _local_concurrency():
2280
_cached_local_concurrency = None
2282
def local_concurrency(use_cache=True):
2283
"""Return how many processes can be run concurrently.
2285
Rely on platform specific implementations and default to 1 (one) if
2286
anything goes wrong.
2288
global _cached_local_concurrency
2290
if _cached_local_concurrency is not None and use_cache:
2291
return _cached_local_concurrency
2293
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2294
if concurrency is None:
2296
import multiprocessing
2298
# multiprocessing is only available on Python >= 2.6
2300
concurrency = _local_concurrency()
2301
except (OSError, IOError):
2304
concurrency = multiprocessing.cpu_count()
2306
concurrency = int(concurrency)
2307
except (TypeError, ValueError):
2310
_cached_concurrency = concurrency
2314
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2315
"""A stream writer that doesn't decode str arguments."""
2317
def __init__(self, encode, stream, errors='strict'):
2318
codecs.StreamWriter.__init__(self, stream, errors)
2319
self.encode = encode
2321
def write(self, object):
2322
if type(object) is str:
2323
self.stream.write(object)
2325
data, _ = self.encode(object, self.errors)
2326
self.stream.write(data)
2328
if sys.platform == 'win32':
2329
def open_file(filename, mode='r', bufsize=-1):
2330
"""This function is used to override the ``open`` builtin.
2332
But it uses O_NOINHERIT flag so the file handle is not inherited by
2333
child processes. Deleting or renaming a closed file opened with this
2334
function is not blocking child processes.
2336
writing = 'w' in mode
2337
appending = 'a' in mode
2338
updating = '+' in mode
2339
binary = 'b' in mode
2342
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2343
# for flags for each modes.
2353
flags |= os.O_WRONLY
2354
flags |= os.O_CREAT | os.O_TRUNC
2359
flags |= os.O_WRONLY
2360
flags |= os.O_CREAT | os.O_APPEND
2365
flags |= os.O_RDONLY
2367
return os.fdopen(os.open(filename, flags), mode, bufsize)
2372
def getuser_unicode():
2373
"""Return the username as unicode.
2376
user_encoding = get_user_encoding()
2377
username = getpass.getuser().decode(user_encoding)
2378
except UnicodeDecodeError:
2379
raise errors.BzrError("Can't decode username as %s." % \
2384
def available_backup_name(base, exists):
2385
"""Find a non-existing backup file name.
2387
This will *not* create anything, this only return a 'free' entry. This
2388
should be used for checking names in a directory below a locked
2389
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2390
Leap) and generally discouraged.
2392
:param base: The base name.
2394
:param exists: A callable returning True if the path parameter exists.
2397
name = "%s.~%d~" % (base, counter)
2400
name = "%s.~%d~" % (base, counter)
2404
def set_fd_cloexec(fd):
2405
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2406
support for this is not available.
2410
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2411
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2412
except (ImportError, AttributeError):
2413
# Either the fcntl module or specific constants are not present
2417
def find_executable_on_path(name):
2418
"""Finds an executable on the PATH.
2420
On Windows, this will try to append each extension in the PATHEXT
2421
environment variable to the name, if it cannot be found with the name
2424
:param name: The base name of the executable.
2425
:return: The path to the executable found or None.
2427
path = os.environ.get('PATH')
2430
path = path.split(os.pathsep)
2431
if sys.platform == 'win32':
2432
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2433
exts = [ext.lower() for ext in exts]
2434
base, ext = os.path.splitext(name)
2436
if ext.lower() not in exts:
2444
f = os.path.join(d, name) + ext
2445
if os.access(f, os.X_OK):
added
removed
bzrlib/osutils.py
