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
50
from bzrlib.i18n import gettext
53
from bzrlib.symbol_versioning import (
65
from bzrlib import symbol_versioning
68
# Cross platform wall-clock time functionality with decent resolution.
69
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
70
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
71
# synchronized with ``time.time()``, this is only meant to be used to find
72
# delta times by subtracting from another call to this function.
73
timer_func = time.time
74
if sys.platform == 'win32':
75
timer_func = time.clock
77
# On win32, O_BINARY is used to indicate the file should
78
# be opened in binary mode, rather than text mode.
79
# On other platforms, O_BINARY doesn't exist, because
80
# they always open in binary mode, so it is okay to
81
# OR with 0 on those platforms.
82
# O_NOINHERIT and O_TEXT exists only on win32 too.
83
O_BINARY = getattr(os, 'O_BINARY', 0)
84
O_TEXT = getattr(os, 'O_TEXT', 0)
85
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
88
def get_unicode_argv():
90
user_encoding = get_user_encoding()
91
return [a.decode(user_encoding) for a in sys.argv[1:]]
92
except UnicodeDecodeError:
93
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
94
"application locale.").format(a, user_encoding))
97
def make_readonly(filename):
98
"""Make a filename read-only."""
99
mod = os.lstat(filename).st_mode
100
if not stat.S_ISLNK(mod):
102
os.chmod(filename, mod)
105
def make_writable(filename):
106
mod = os.lstat(filename).st_mode
107
if not stat.S_ISLNK(mod):
109
os.chmod(filename, mod)
112
def minimum_path_selection(paths):
113
"""Return the smallset subset of paths which are outside paths.
115
:param paths: A container (and hence not None) of paths.
116
:return: A set of paths sufficient to include everything in paths via
117
is_inside, drawn from the paths parameter.
123
return path.split('/')
124
sorted_paths = sorted(list(paths), key=sort_key)
126
search_paths = [sorted_paths[0]]
127
for path in sorted_paths[1:]:
128
if not is_inside(search_paths[-1], path):
129
# This path is unique, add it
130
search_paths.append(path)
132
return set(search_paths)
139
"""Return a quoted filename filename
141
This previously used backslash quoting, but that works poorly on
143
# TODO: I'm not really sure this is the best format either.x
145
if _QUOTE_RE is None:
146
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
148
if _QUOTE_RE.search(f):
154
_directory_kind = 'directory'
157
"""Return the current umask"""
158
# Assume that people aren't messing with the umask while running
159
# XXX: This is not thread safe, but there is no way to get the
160
# umask without setting it
168
_directory_kind: "/",
170
'tree-reference': '+',
174
def kind_marker(kind):
176
return _kind_marker_map[kind]
178
# Slightly faster than using .get(, '') when the common case is that
183
lexists = getattr(os.path, 'lexists', None)
187
stat = getattr(os, 'lstat', os.stat)
191
if e.errno == errno.ENOENT:
194
raise errors.BzrError(gettext("lstat/stat of ({0!r}): {1!r}").format(f, e))
197
def fancy_rename(old, new, rename_func, unlink_func):
198
"""A fancy rename, when you don't have atomic rename.
200
:param old: The old path, to rename from
201
:param new: The new path, to rename to
202
:param rename_func: The potentially non-atomic rename function
203
:param unlink_func: A way to delete the target file if the full rename
206
# sftp rename doesn't allow overwriting, so play tricks:
207
base = os.path.basename(new)
208
dirname = os.path.dirname(new)
209
# callers use different encodings for the paths so the following MUST
210
# respect that. We rely on python upcasting to unicode if new is unicode
211
# and keeping a str if not.
212
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
213
os.getpid(), rand_chars(10))
214
tmp_name = pathjoin(dirname, tmp_name)
216
# Rename the file out of the way, but keep track if it didn't exist
217
# We don't want to grab just any exception
218
# something like EACCES should prevent us from continuing
219
# The downside is that the rename_func has to throw an exception
220
# with an errno = ENOENT, or NoSuchFile
223
rename_func(new, tmp_name)
224
except (errors.NoSuchFile,), e:
227
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
228
# function raises an IOError with errno is None when a rename fails.
229
# This then gets caught here.
230
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
233
if (getattr(e, 'errno', None) is None
234
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
243
# This may throw an exception, in which case success will
245
rename_func(old, new)
247
except (IOError, OSError), e:
248
# source and target may be aliases of each other (e.g. on a
249
# case-insensitive filesystem), so we may have accidentally renamed
250
# source by when we tried to rename target
251
failure_exc = sys.exc_info()
252
if (file_existed and e.errno in (None, errno.ENOENT)
253
and old.lower() == new.lower()):
254
# source and target are the same file on a case-insensitive
255
# filesystem, so we don't generate an exception
259
# If the file used to exist, rename it back into place
260
# otherwise just delete it from the tmp location
262
unlink_func(tmp_name)
264
rename_func(tmp_name, new)
265
if failure_exc is not None:
267
raise failure_exc[0], failure_exc[1], failure_exc[2]
272
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
273
# choke on a Unicode string containing a relative path if
274
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
276
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
277
def _posix_abspath(path):
278
# jam 20060426 rather than encoding to fsencoding
279
# copy posixpath.abspath, but use os.getcwdu instead
280
if not posixpath.isabs(path):
281
path = posixpath.join(getcwd(), path)
282
return posixpath.normpath(path)
285
def _posix_realpath(path):
286
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
289
def _win32_fixdrive(path):
290
"""Force drive letters to be consistent.
292
win32 is inconsistent whether it returns lower or upper case
293
and even if it was consistent the user might type the other
294
so we force it to uppercase
295
running python.exe under cmd.exe return capital C:\\
296
running win32 python inside a cygwin shell returns lowercase c:\\
298
drive, path = ntpath.splitdrive(path)
299
return drive.upper() + path
302
def _win32_abspath(path):
303
# Real ntpath.abspath doesn't have a problem with a unicode cwd
304
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
307
def _win98_abspath(path):
308
"""Return the absolute version of a path.
309
Windows 98 safe implementation (python reimplementation
310
of Win32 API function GetFullPathNameW)
315
# \\HOST\path => //HOST/path
316
# //HOST/path => //HOST/path
317
# path => C:/cwd/path
320
# check for absolute path
321
drive = ntpath.splitdrive(path)[0]
322
if drive == '' and path[:2] not in('//','\\\\'):
324
# we cannot simply os.path.join cwd and path
325
# because os.path.join('C:','/path') produce '/path'
326
# and this is incorrect
327
if path[:1] in ('/','\\'):
328
cwd = ntpath.splitdrive(cwd)[0]
330
path = cwd + '\\' + path
331
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
334
def _win32_realpath(path):
335
# Real ntpath.realpath doesn't have a problem with a unicode cwd
336
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
339
def _win32_pathjoin(*args):
340
return ntpath.join(*args).replace('\\', '/')
343
def _win32_normpath(path):
344
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
348
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
351
def _win32_mkdtemp(*args, **kwargs):
352
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
355
def _win32_rename(old, new):
356
"""We expect to be able to atomically replace 'new' with old.
358
On win32, if new exists, it must be moved out of the way first,
362
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
364
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
365
# If we try to rename a non-existant file onto cwd, we get
366
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
367
# if the old path doesn't exist, sometimes we get EACCES
368
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
374
return unicodedata.normalize('NFC', os.getcwdu())
377
# Default is to just use the python builtins, but these can be rebound on
378
# particular platforms.
379
abspath = _posix_abspath
380
realpath = _posix_realpath
381
pathjoin = os.path.join
382
normpath = os.path.normpath
385
dirname = os.path.dirname
386
basename = os.path.basename
387
split = os.path.split
388
splitext = os.path.splitext
389
# These were already lazily imported into local scope
390
# mkdtemp = tempfile.mkdtemp
391
# rmtree = shutil.rmtree
399
MIN_ABS_PATHLENGTH = 1
402
if sys.platform == 'win32':
403
if win32utils.winver == 'Windows 98':
404
abspath = _win98_abspath
406
abspath = _win32_abspath
407
realpath = _win32_realpath
408
pathjoin = _win32_pathjoin
409
normpath = _win32_normpath
410
getcwd = _win32_getcwd
411
mkdtemp = _win32_mkdtemp
412
rename = _win32_rename
414
from bzrlib import _walkdirs_win32
418
lstat = _walkdirs_win32.lstat
419
fstat = _walkdirs_win32.fstat
420
wrap_stat = _walkdirs_win32.wrap_stat
422
MIN_ABS_PATHLENGTH = 3
424
def _win32_delete_readonly(function, path, excinfo):
425
"""Error handler for shutil.rmtree function [for win32]
426
Helps to remove files and dirs marked as read-only.
428
exception = excinfo[1]
429
if function in (os.remove, os.rmdir) \
430
and isinstance(exception, OSError) \
431
and exception.errno == errno.EACCES:
437
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
438
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
439
return shutil.rmtree(path, ignore_errors, onerror)
441
f = win32utils.get_unicode_argv # special function or None
445
elif sys.platform == 'darwin':
449
def get_terminal_encoding(trace=False):
450
"""Find the best encoding for printing to the screen.
452
This attempts to check both sys.stdout and sys.stdin to see
453
what encoding they are in, and if that fails it falls back to
454
osutils.get_user_encoding().
455
The problem is that on Windows, locale.getpreferredencoding()
456
is not the same encoding as that used by the console:
457
http://mail.python.org/pipermail/python-list/2003-May/162357.html
459
On my standard US Windows XP, the preferred encoding is
460
cp1252, but the console is cp437
462
:param trace: If True trace the selected encoding via mutter().
464
from bzrlib.trace import mutter
465
output_encoding = getattr(sys.stdout, 'encoding', None)
466
if not output_encoding:
467
input_encoding = getattr(sys.stdin, 'encoding', None)
468
if not input_encoding:
469
output_encoding = get_user_encoding()
471
mutter('encoding stdout as osutils.get_user_encoding() %r',
474
output_encoding = input_encoding
476
mutter('encoding stdout as sys.stdin encoding %r',
480
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
481
if output_encoding == 'cp0':
482
# invalid encoding (cp0 means 'no codepage' on Windows)
483
output_encoding = get_user_encoding()
485
mutter('cp0 is invalid encoding.'
486
' encoding stdout as osutils.get_user_encoding() %r',
490
codecs.lookup(output_encoding)
492
sys.stderr.write('bzr: warning:'
493
' unknown terminal encoding %s.\n'
494
' Using encoding %s instead.\n'
495
% (output_encoding, get_user_encoding())
497
output_encoding = get_user_encoding()
499
return output_encoding
502
def normalizepath(f):
503
if getattr(os.path, 'realpath', None) is not None:
507
[p,e] = os.path.split(f)
508
if e == "" or e == "." or e == "..":
511
return pathjoin(F(p), e)
515
"""True if f is an accessible directory."""
517
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
523
"""True if f is a regular file."""
525
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
530
"""True if f is a symlink."""
532
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
536
def is_inside(dir, fname):
537
"""True if fname is inside dir.
539
The parameters should typically be passed to osutils.normpath first, so
540
that . and .. and repeated slashes are eliminated, and the separators
541
are canonical for the platform.
543
The empty string as a dir name is taken as top-of-tree and matches
546
# XXX: Most callers of this can actually do something smarter by
547
# looking at the inventory
557
return fname.startswith(dir)
560
def is_inside_any(dir_list, fname):
561
"""True if fname is inside any of given dirs."""
562
for dirname in dir_list:
563
if is_inside(dirname, fname):
568
def is_inside_or_parent_of_any(dir_list, fname):
569
"""True if fname is a child or a parent of any of the given files."""
570
for dirname in dir_list:
571
if is_inside(dirname, fname) or is_inside(fname, dirname):
576
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
577
report_activity=None, direction='read'):
578
"""Copy contents of one file to another.
580
The read_length can either be -1 to read to end-of-file (EOF) or
581
it can specify the maximum number of bytes to read.
583
The buff_size represents the maximum size for each read operation
584
performed on from_file.
586
:param report_activity: Call this as bytes are read, see
587
Transport._report_activity
588
:param direction: Will be passed to report_activity
590
:return: The number of bytes copied.
594
# read specified number of bytes
596
while read_length > 0:
597
num_bytes_to_read = min(read_length, buff_size)
599
block = from_file.read(num_bytes_to_read)
603
if report_activity is not None:
604
report_activity(len(block), direction)
607
actual_bytes_read = len(block)
608
read_length -= actual_bytes_read
609
length += actual_bytes_read
613
block = from_file.read(buff_size)
617
if report_activity is not None:
618
report_activity(len(block), direction)
624
def pump_string_file(bytes, file_handle, segment_size=None):
625
"""Write bytes to file_handle in many smaller writes.
627
:param bytes: The string to write.
628
:param file_handle: The file to write to.
630
# Write data in chunks rather than all at once, because very large
631
# writes fail on some platforms (e.g. Windows with SMB mounted
634
segment_size = 5242880 # 5MB
635
segments = range(len(bytes) / segment_size + 1)
636
write = file_handle.write
637
for segment_index in segments:
638
segment = buffer(bytes, segment_index * segment_size, segment_size)
642
def file_iterator(input_file, readsize=32768):
644
b = input_file.read(readsize)
651
"""Calculate the hexdigest of an open file.
653
The file cursor should be already at the start.
665
def size_sha_file(f):
666
"""Calculate the size and hexdigest of an open file.
668
The file cursor should be already at the start and
669
the caller is responsible for closing the file afterwards.
680
return size, s.hexdigest()
683
def sha_file_by_name(fname):
684
"""Calculate the SHA1 of a file by reading the full text"""
686
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
689
b = os.read(f, 1<<16)
697
def sha_strings(strings, _factory=sha):
698
"""Return the sha-1 of concatenation of strings"""
700
map(s.update, strings)
704
def sha_string(f, _factory=sha):
705
return _factory(f).hexdigest()
708
def fingerprint_file(f):
710
return {'size': len(b),
711
'sha1': sha(b).hexdigest()}
714
def compare_files(a, b):
715
"""Returns true if equal in contents"""
726
def local_time_offset(t=None):
727
"""Return offset of local zone from GMT, either at present or at time t."""
730
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
731
return offset.days * 86400 + offset.seconds
733
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
734
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
737
def format_date(t, offset=0, timezone='original', date_fmt=None,
739
"""Return a formatted date string.
741
:param t: Seconds since the epoch.
742
:param offset: Timezone offset in seconds east of utc.
743
:param timezone: How to display the time: 'utc', 'original' for the
744
timezone specified by offset, or 'local' for the process's current
746
:param date_fmt: strftime format.
747
:param show_offset: Whether to append the timezone.
749
(date_fmt, tt, offset_str) = \
750
_format_date(t, offset, timezone, date_fmt, show_offset)
751
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
752
date_str = time.strftime(date_fmt, tt)
753
return date_str + offset_str
756
# Cache of formatted offset strings
760
def format_date_with_offset_in_original_timezone(t, offset=0,
761
_cache=_offset_cache):
762
"""Return a formatted date string in the original timezone.
764
This routine may be faster then format_date.
766
:param t: Seconds since the epoch.
767
:param offset: Timezone offset in seconds east of utc.
771
tt = time.gmtime(t + offset)
772
date_fmt = _default_format_by_weekday_num[tt[6]]
773
date_str = time.strftime(date_fmt, tt)
774
offset_str = _cache.get(offset, None)
775
if offset_str is None:
776
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
777
_cache[offset] = offset_str
778
return date_str + offset_str
781
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
783
"""Return an unicode date string formatted according to the current locale.
785
:param t: Seconds since the epoch.
786
:param offset: Timezone offset in seconds east of utc.
787
:param timezone: How to display the time: 'utc', 'original' for the
788
timezone specified by offset, or 'local' for the process's current
790
:param date_fmt: strftime format.
791
:param show_offset: Whether to append the timezone.
793
(date_fmt, tt, offset_str) = \
794
_format_date(t, offset, timezone, date_fmt, show_offset)
795
date_str = time.strftime(date_fmt, tt)
796
if not isinstance(date_str, unicode):
797
date_str = date_str.decode(get_user_encoding(), 'replace')
798
return date_str + offset_str
801
def _format_date(t, offset, timezone, date_fmt, show_offset):
802
if timezone == 'utc':
805
elif timezone == 'original':
808
tt = time.gmtime(t + offset)
809
elif timezone == 'local':
810
tt = time.localtime(t)
811
offset = local_time_offset(t)
813
raise errors.UnsupportedTimezoneFormat(timezone)
815
date_fmt = "%a %Y-%m-%d %H:%M:%S"
817
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
820
return (date_fmt, tt, offset_str)
823
def compact_date(when):
824
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
827
def format_delta(delta):
828
"""Get a nice looking string for a time delta.
830
:param delta: The time difference in seconds, can be positive or negative.
831
positive indicates time in the past, negative indicates time in the
832
future. (usually time.time() - stored_time)
833
:return: String formatted to show approximate resolution
839
direction = 'in the future'
843
if seconds < 90: # print seconds up to 90 seconds
845
return '%d second %s' % (seconds, direction,)
847
return '%d seconds %s' % (seconds, direction)
849
minutes = int(seconds / 60)
850
seconds -= 60 * minutes
855
if minutes < 90: # print minutes, seconds up to 90 minutes
857
return '%d minute, %d second%s %s' % (
858
minutes, seconds, plural_seconds, direction)
860
return '%d minutes, %d second%s %s' % (
861
minutes, seconds, plural_seconds, direction)
863
hours = int(minutes / 60)
864
minutes -= 60 * hours
871
return '%d hour, %d minute%s %s' % (hours, minutes,
872
plural_minutes, direction)
873
return '%d hours, %d minute%s %s' % (hours, minutes,
874
plural_minutes, direction)
877
"""Return size of given open file."""
878
return os.fstat(f.fileno())[stat.ST_SIZE]
881
# Define rand_bytes based on platform.
883
# Python 2.4 and later have os.urandom,
884
# but it doesn't work on some arches
886
rand_bytes = os.urandom
887
except (NotImplementedError, AttributeError):
888
# If python doesn't have os.urandom, or it doesn't work,
889
# then try to first pull random data from /dev/urandom
891
rand_bytes = file('/dev/urandom', 'rb').read
892
# Otherwise, use this hack as a last resort
893
except (IOError, OSError):
894
# not well seeded, but better than nothing
899
s += chr(random.randint(0, 255))
904
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
906
"""Return a random string of num alphanumeric characters
908
The result only contains lowercase chars because it may be used on
909
case-insensitive filesystems.
912
for raw_byte in rand_bytes(num):
913
s += ALNUM[ord(raw_byte) % 36]
917
## TODO: We could later have path objects that remember their list
918
## decomposition (might be too tricksy though.)
921
"""Turn string into list of parts."""
922
# split on either delimiter because people might use either on
924
ps = re.split(r'[\\/]', p)
929
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
930
elif (f == '.') or (f == ''):
939
if (f == '..') or (f is None) or (f == ''):
940
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
944
def parent_directories(filename):
945
"""Return the list of parent directories, deepest first.
947
For example, parent_directories("a/b/c") -> ["a/b", "a"].
950
parts = splitpath(dirname(filename))
952
parents.append(joinpath(parts))
957
_extension_load_failures = []
960
def failed_to_load_extension(exception):
961
"""Handle failing to load a binary extension.
963
This should be called from the ImportError block guarding the attempt to
964
import the native extension. If this function returns, the pure-Python
965
implementation should be loaded instead::
968
>>> import bzrlib._fictional_extension_pyx
969
>>> except ImportError, e:
970
>>> bzrlib.osutils.failed_to_load_extension(e)
971
>>> import bzrlib._fictional_extension_py
973
# NB: This docstring is just an example, not a doctest, because doctest
974
# currently can't cope with the use of lazy imports in this namespace --
977
# This currently doesn't report the failure at the time it occurs, because
978
# they tend to happen very early in startup when we can't check config
979
# files etc, and also we want to report all failures but not spam the user
981
exception_str = str(exception)
982
if exception_str not in _extension_load_failures:
983
trace.mutter("failed to load compiled extension: %s" % exception_str)
984
_extension_load_failures.append(exception_str)
987
def report_extension_load_failures():
988
if not _extension_load_failures:
990
if config.GlobalStack().get('ignore_missing_extensions'):
992
# the warnings framework should by default show this only once
993
from bzrlib.trace import warning
995
"bzr: warning: some compiled extensions could not be loaded; "
996
"see <https://answers.launchpad.net/bzr/+faq/703>")
997
# we no longer show the specific missing extensions here, because it makes
998
# the message too long and scary - see
999
# https://bugs.launchpad.net/bzr/+bug/430529
1003
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1004
except ImportError, e:
1005
failed_to_load_extension(e)
1006
from bzrlib._chunks_to_lines_py import chunks_to_lines
1010
"""Split s into lines, but without removing the newline characters."""
1011
# Trivially convert a fulltext into a 'chunked' representation, and let
1012
# chunks_to_lines do the heavy lifting.
1013
if isinstance(s, str):
1014
# chunks_to_lines only supports 8-bit strings
1015
return chunks_to_lines([s])
1017
return _split_lines(s)
1020
def _split_lines(s):
1021
"""Split s into lines, but without removing the newline characters.
1023
This supports Unicode or plain string objects.
1025
lines = s.split('\n')
1026
result = [line + '\n' for line in lines[:-1]]
1028
result.append(lines[-1])
1032
def hardlinks_good():
1033
return sys.platform not in ('win32', 'cygwin', 'darwin')
1036
def link_or_copy(src, dest):
1037
"""Hardlink a file, or copy it if it can't be hardlinked."""
1038
if not hardlinks_good():
1039
shutil.copyfile(src, dest)
1043
except (OSError, IOError), e:
1044
if e.errno != errno.EXDEV:
1046
shutil.copyfile(src, dest)
1049
def delete_any(path):
1050
"""Delete a file, symlink or directory.
1052
Will delete even if readonly.
1055
_delete_file_or_dir(path)
1056
except (OSError, IOError), e:
1057
if e.errno in (errno.EPERM, errno.EACCES):
1058
# make writable and try again
1061
except (OSError, IOError):
1063
_delete_file_or_dir(path)
1068
def _delete_file_or_dir(path):
1069
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1070
# Forgiveness than Permission (EAFP) because:
1071
# - root can damage a solaris file system by using unlink,
1072
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1073
# EACCES, OSX: EPERM) when invoked on a directory.
1074
if isdir(path): # Takes care of symlinks
1081
if getattr(os, 'symlink', None) is not None:
1087
def has_hardlinks():
1088
if getattr(os, 'link', None) is not None:
1094
def host_os_dereferences_symlinks():
1095
return (has_symlinks()
1096
and sys.platform not in ('cygwin', 'win32'))
1099
def readlink(abspath):
1100
"""Return a string representing the path to which the symbolic link points.
1102
:param abspath: The link absolute unicode path.
1104
This his guaranteed to return the symbolic link in unicode in all python
1107
link = abspath.encode(_fs_enc)
1108
target = os.readlink(link)
1109
target = target.decode(_fs_enc)
1113
def contains_whitespace(s):
1114
"""True if there are any whitespace characters in s."""
1115
# string.whitespace can include '\xa0' in certain locales, because it is
1116
# considered "non-breaking-space" as part of ISO-8859-1. But it
1117
# 1) Isn't a breaking whitespace
1118
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1120
# 3) '\xa0' isn't unicode safe since it is >128.
1122
# This should *not* be a unicode set of characters in case the source
1123
# string is not a Unicode string. We can auto-up-cast the characters since
1124
# they are ascii, but we don't want to auto-up-cast the string in case it
1126
for ch in ' \t\n\r\v\f':
1133
def contains_linebreaks(s):
1134
"""True if there is any vertical whitespace in s."""
1142
def relpath(base, path):
1143
"""Return path relative to base, or raise PathNotChild exception.
1145
The path may be either an absolute path or a path relative to the
1146
current working directory.
1148
os.path.commonprefix (python2.4) has a bad bug that it works just
1149
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1150
avoids that problem.
1152
NOTE: `base` should not have a trailing slash otherwise you'll get
1153
PathNotChild exceptions regardless of `path`.
1156
if len(base) < MIN_ABS_PATHLENGTH:
1157
# must have space for e.g. a drive letter
1158
raise ValueError(gettext('%r is too short to calculate a relative path')
1166
if len(head) <= len(base) and head != base:
1167
raise errors.PathNotChild(rp, base)
1170
head, tail = split(head)
1175
return pathjoin(*reversed(s))
1180
def _cicp_canonical_relpath(base, path):
1181
"""Return the canonical path relative to base.
1183
Like relpath, but on case-insensitive-case-preserving file-systems, this
1184
will return the relpath as stored on the file-system rather than in the
1185
case specified in the input string, for all existing portions of the path.
1187
This will cause O(N) behaviour if called for every path in a tree; if you
1188
have a number of paths to convert, you should use canonical_relpaths().
1190
# TODO: it should be possible to optimize this for Windows by using the
1191
# win32 API FindFiles function to look for the specified name - but using
1192
# os.listdir() still gives us the correct, platform agnostic semantics in
1195
rel = relpath(base, path)
1196
# '.' will have been turned into ''
1200
abs_base = abspath(base)
1202
_listdir = os.listdir
1204
# use an explicit iterator so we can easily consume the rest on early exit.
1205
bit_iter = iter(rel.split('/'))
1206
for bit in bit_iter:
1209
next_entries = _listdir(current)
1210
except OSError: # enoent, eperm, etc
1211
# We can't find this in the filesystem, so just append the
1213
current = pathjoin(current, bit, *list(bit_iter))
1215
for look in next_entries:
1216
if lbit == look.lower():
1217
current = pathjoin(current, look)
1220
# got to the end, nothing matched, so we just return the
1221
# non-existing bits as they were specified (the filename may be
1222
# the target of a move, for example).
1223
current = pathjoin(current, bit, *list(bit_iter))
1225
return current[len(abs_base):].lstrip('/')
1227
# XXX - TODO - we need better detection/integration of case-insensitive
1228
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1229
# filesystems), for example, so could probably benefit from the same basic
1230
# support there. For now though, only Windows and OSX get that support, and
1231
# they get it for *all* file-systems!
1232
if sys.platform in ('win32', 'darwin'):
1233
canonical_relpath = _cicp_canonical_relpath
1235
canonical_relpath = relpath
1237
def canonical_relpaths(base, paths):
1238
"""Create an iterable to canonicalize a sequence of relative paths.
1240
The intent is for this implementation to use a cache, vastly speeding
1241
up multiple transformations in the same directory.
1243
# but for now, we haven't optimized...
1244
return [canonical_relpath(base, p) for p in paths]
1247
def decode_filename(filename):
1248
"""Decode the filename using the filesystem encoding
1250
If it is unicode, it is returned.
1251
Otherwise it is decoded from the the filesystem's encoding. If decoding
1252
fails, a errors.BadFilenameEncoding exception is raised.
1254
if type(filename) is unicode:
1257
return filename.decode(_fs_enc)
1258
except UnicodeDecodeError:
1259
raise errors.BadFilenameEncoding(filename, _fs_enc)
1262
def safe_unicode(unicode_or_utf8_string):
1263
"""Coerce unicode_or_utf8_string into unicode.
1265
If it is unicode, it is returned.
1266
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1267
wrapped in a BzrBadParameterNotUnicode exception.
1269
if isinstance(unicode_or_utf8_string, unicode):
1270
return unicode_or_utf8_string
1272
return unicode_or_utf8_string.decode('utf8')
1273
except UnicodeDecodeError:
1274
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1277
def safe_utf8(unicode_or_utf8_string):
1278
"""Coerce unicode_or_utf8_string to a utf8 string.
1280
If it is a str, it is returned.
1281
If it is Unicode, it is encoded into a utf-8 string.
1283
if isinstance(unicode_or_utf8_string, str):
1284
# TODO: jam 20070209 This is overkill, and probably has an impact on
1285
# performance if we are dealing with lots of apis that want a
1288
# Make sure it is a valid utf-8 string
1289
unicode_or_utf8_string.decode('utf-8')
1290
except UnicodeDecodeError:
1291
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1292
return unicode_or_utf8_string
1293
return unicode_or_utf8_string.encode('utf-8')
1296
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1297
' Revision id generators should be creating utf8'
1301
def safe_revision_id(unicode_or_utf8_string, warn=True):
1302
"""Revision ids should now be utf8, but at one point they were unicode.
1304
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1306
:param warn: Functions that are sanitizing user data can set warn=False
1307
:return: None or a utf8 revision id.
1309
if (unicode_or_utf8_string is None
1310
or unicode_or_utf8_string.__class__ == str):
1311
return unicode_or_utf8_string
1313
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1315
return cache_utf8.encode(unicode_or_utf8_string)
1318
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1319
' generators should be creating utf8 file ids.')
1322
def safe_file_id(unicode_or_utf8_string, warn=True):
1323
"""File ids should now be utf8, but at one point they were unicode.
1325
This is the same as safe_utf8, except it uses the cached encode functions
1326
to save a little bit of performance.
1328
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1330
:param warn: Functions that are sanitizing user data can set warn=False
1331
:return: None or a utf8 file id.
1333
if (unicode_or_utf8_string is None
1334
or unicode_or_utf8_string.__class__ == str):
1335
return unicode_or_utf8_string
1337
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1339
return cache_utf8.encode(unicode_or_utf8_string)
1342
_platform_normalizes_filenames = False
1343
if sys.platform == 'darwin':
1344
_platform_normalizes_filenames = True
1347
def normalizes_filenames():
1348
"""Return True if this platform normalizes unicode filenames.
1352
return _platform_normalizes_filenames
1355
def _accessible_normalized_filename(path):
1356
"""Get the unicode normalized path, and if you can access the file.
1358
On platforms where the system normalizes filenames (Mac OSX),
1359
you can access a file by any path which will normalize correctly.
1360
On platforms where the system does not normalize filenames
1361
(everything else), you have to access a file by its exact path.
1363
Internally, bzr only supports NFC normalization, since that is
1364
the standard for XML documents.
1366
So return the normalized path, and a flag indicating if the file
1367
can be accessed by that path.
1370
return unicodedata.normalize('NFC', unicode(path)), True
1373
def _inaccessible_normalized_filename(path):
1374
__doc__ = _accessible_normalized_filename.__doc__
1376
normalized = unicodedata.normalize('NFC', unicode(path))
1377
return normalized, normalized == path
1380
if _platform_normalizes_filenames:
1381
normalized_filename = _accessible_normalized_filename
1383
normalized_filename = _inaccessible_normalized_filename
1386
def set_signal_handler(signum, handler, restart_syscall=True):
1387
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1388
on platforms that support that.
1390
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1391
automatically restart (by calling `signal.siginterrupt(signum,
1392
False)`). May be ignored if the feature is not available on this
1393
platform or Python version.
1397
siginterrupt = signal.siginterrupt
1399
# This python implementation doesn't provide signal support, hence no
1402
except AttributeError:
1403
# siginterrupt doesn't exist on this platform, or for this version
1405
siginterrupt = lambda signum, flag: None
1407
def sig_handler(*args):
1408
# Python resets the siginterrupt flag when a signal is
1409
# received. <http://bugs.python.org/issue8354>
1410
# As a workaround for some cases, set it back the way we want it.
1411
siginterrupt(signum, False)
1412
# Now run the handler function passed to set_signal_handler.
1415
sig_handler = handler
1416
old_handler = signal.signal(signum, sig_handler)
1418
siginterrupt(signum, False)
1422
default_terminal_width = 80
1423
"""The default terminal width for ttys.
1425
This is defined so that higher levels can share a common fallback value when
1426
terminal_width() returns None.
1429
# Keep some state so that terminal_width can detect if _terminal_size has
1430
# returned a different size since the process started. See docstring and
1431
# comments of terminal_width for details.
1432
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1433
_terminal_size_state = 'no_data'
1434
_first_terminal_size = None
1436
def terminal_width():
1437
"""Return terminal width.
1439
None is returned if the width can't established precisely.
1442
- if BZR_COLUMNS is set, returns its value
1443
- if there is no controlling terminal, returns None
1444
- query the OS, if the queried size has changed since the last query,
1446
- if COLUMNS is set, returns its value,
1447
- if the OS has a value (even though it's never changed), return its value.
1449
From there, we need to query the OS to get the size of the controlling
1452
On Unices we query the OS by:
1453
- get termios.TIOCGWINSZ
1454
- if an error occurs or a negative value is obtained, returns None
1456
On Windows we query the OS by:
1457
- win32utils.get_console_size() decides,
1458
- returns None on error (provided default value)
1460
# Note to implementors: if changing the rules for determining the width,
1461
# make sure you've considered the behaviour in these cases:
1462
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1463
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1465
# - (add more interesting cases here, if you find any)
1466
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1467
# but we don't want to register a signal handler because it is impossible
1468
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1469
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1470
# time so we can notice if the reported size has changed, which should have
1473
# If BZR_COLUMNS is set, take it, user is always right
1474
# Except if they specified 0 in which case, impose no limit here
1476
width = int(os.environ['BZR_COLUMNS'])
1477
except (KeyError, ValueError):
1479
if width is not None:
1485
isatty = getattr(sys.stdout, 'isatty', None)
1486
if isatty is None or not isatty():
1487
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1491
width, height = os_size = _terminal_size(None, None)
1492
global _first_terminal_size, _terminal_size_state
1493
if _terminal_size_state == 'no_data':
1494
_first_terminal_size = os_size
1495
_terminal_size_state = 'unchanged'
1496
elif (_terminal_size_state == 'unchanged' and
1497
_first_terminal_size != os_size):
1498
_terminal_size_state = 'changed'
1500
# If the OS claims to know how wide the terminal is, and this value has
1501
# ever changed, use that.
1502
if _terminal_size_state == 'changed':
1503
if width is not None and width > 0:
1506
# If COLUMNS is set, use it.
1508
return int(os.environ['COLUMNS'])
1509
except (KeyError, ValueError):
1512
# Finally, use an unchanged size from the OS, if we have one.
1513
if _terminal_size_state == 'unchanged':
1514
if width is not None and width > 0:
1517
# The width could not be determined.
1521
def _win32_terminal_size(width, height):
1522
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1523
return width, height
1526
def _ioctl_terminal_size(width, height):
1528
import struct, fcntl, termios
1529
s = struct.pack('HHHH', 0, 0, 0, 0)
1530
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1531
height, width = struct.unpack('HHHH', x)[0:2]
1532
except (IOError, AttributeError):
1534
return width, height
1536
_terminal_size = None
1537
"""Returns the terminal size as (width, height).
1539
:param width: Default value for width.
1540
:param height: Default value for height.
1542
This is defined specifically for each OS and query the size of the controlling
1543
terminal. If any error occurs, the provided default values should be returned.
1545
if sys.platform == 'win32':
1546
_terminal_size = _win32_terminal_size
1548
_terminal_size = _ioctl_terminal_size
1551
def supports_executable():
1552
return sys.platform != "win32"
1555
def supports_posix_readonly():
1556
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1558
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1559
directory controls creation/deletion, etc.
1561
And under win32, readonly means that the directory itself cannot be
1562
deleted. The contents of a readonly directory can be changed, unlike POSIX
1563
where files in readonly directories cannot be added, deleted or renamed.
1565
return sys.platform != "win32"
1568
def set_or_unset_env(env_variable, value):
1569
"""Modify the environment, setting or removing the env_variable.
1571
:param env_variable: The environment variable in question
1572
:param value: The value to set the environment to. If None, then
1573
the variable will be removed.
1574
:return: The original value of the environment variable.
1576
orig_val = os.environ.get(env_variable)
1578
if orig_val is not None:
1579
del os.environ[env_variable]
1581
if isinstance(value, unicode):
1582
value = value.encode(get_user_encoding())
1583
os.environ[env_variable] = value
1587
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1590
def check_legal_path(path):
1591
"""Check whether the supplied path is legal.
1592
This is only required on Windows, so we don't test on other platforms
1595
if sys.platform != "win32":
1597
if _validWin32PathRE.match(path) is None:
1598
raise errors.IllegalPath(path)
1601
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1603
def _is_error_enotdir(e):
1604
"""Check if this exception represents ENOTDIR.
1606
Unfortunately, python is very inconsistent about the exception
1607
here. The cases are:
1608
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1609
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1610
which is the windows error code.
1611
3) Windows, Python2.5 uses errno == EINVAL and
1612
winerror == ERROR_DIRECTORY
1614
:param e: An Exception object (expected to be OSError with an errno
1615
attribute, but we should be able to cope with anything)
1616
:return: True if this represents an ENOTDIR error. False otherwise.
1618
en = getattr(e, 'errno', None)
1619
if (en == errno.ENOTDIR
1620
or (sys.platform == 'win32'
1621
and (en == _WIN32_ERROR_DIRECTORY
1622
or (en == errno.EINVAL
1623
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1629
def walkdirs(top, prefix=""):
1630
"""Yield data about all the directories in a tree.
1632
This yields all the data about the contents of a directory at a time.
1633
After each directory has been yielded, if the caller has mutated the list
1634
to exclude some directories, they are then not descended into.
1636
The data yielded is of the form:
1637
((directory-relpath, directory-path-from-top),
1638
[(relpath, basename, kind, lstat, path-from-top), ...]),
1639
- directory-relpath is the relative path of the directory being returned
1640
with respect to top. prefix is prepended to this.
1641
- directory-path-from-root is the path including top for this directory.
1642
It is suitable for use with os functions.
1643
- relpath is the relative path within the subtree being walked.
1644
- basename is the basename of the path
1645
- kind is the kind of the file now. If unknown then the file is not
1646
present within the tree - but it may be recorded as versioned. See
1648
- lstat is the stat data *if* the file was statted.
1649
- planned, not implemented:
1650
path_from_tree_root is the path from the root of the tree.
1652
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1653
allows one to walk a subtree but get paths that are relative to a tree
1655
:return: an iterator over the dirs.
1657
#TODO there is a bit of a smell where the results of the directory-
1658
# summary in this, and the path from the root, may not agree
1659
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1660
# potentially confusing output. We should make this more robust - but
1661
# not at a speed cost. RBC 20060731
1663
_directory = _directory_kind
1664
_listdir = os.listdir
1665
_kind_from_mode = file_kind_from_stat_mode
1666
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1668
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1669
relroot, _, _, _, top = pending.pop()
1671
relprefix = relroot + u'/'
1674
top_slash = top + u'/'
1677
append = dirblock.append
1679
names = sorted(map(decode_filename, _listdir(top)))
1681
if not _is_error_enotdir(e):
1685
abspath = top_slash + name
1686
statvalue = _lstat(abspath)
1687
kind = _kind_from_mode(statvalue.st_mode)
1688
append((relprefix + name, name, kind, statvalue, abspath))
1689
yield (relroot, top), dirblock
1691
# push the user specified dirs from dirblock
1692
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1695
class DirReader(object):
1696
"""An interface for reading directories."""
1698
def top_prefix_to_starting_dir(self, top, prefix=""):
1699
"""Converts top and prefix to a starting dir entry
1701
:param top: A utf8 path
1702
:param prefix: An optional utf8 path to prefix output relative paths
1704
:return: A tuple starting with prefix, and ending with the native
1707
raise NotImplementedError(self.top_prefix_to_starting_dir)
1709
def read_dir(self, prefix, top):
1710
"""Read a specific dir.
1712
:param prefix: A utf8 prefix to be preprended to the path basenames.
1713
:param top: A natively encoded path to read.
1714
:return: A list of the directories contents. Each item contains:
1715
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1717
raise NotImplementedError(self.read_dir)
1720
_selected_dir_reader = None
1723
def _walkdirs_utf8(top, prefix=""):
1724
"""Yield data about all the directories in a tree.
1726
This yields the same information as walkdirs() only each entry is yielded
1727
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1728
are returned as exact byte-strings.
1730
:return: yields a tuple of (dir_info, [file_info])
1731
dir_info is (utf8_relpath, path-from-top)
1732
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1733
if top is an absolute path, path-from-top is also an absolute path.
1734
path-from-top might be unicode or utf8, but it is the correct path to
1735
pass to os functions to affect the file in question. (such as os.lstat)
1737
global _selected_dir_reader
1738
if _selected_dir_reader is None:
1739
fs_encoding = _fs_enc.upper()
1740
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1741
# Win98 doesn't have unicode apis like FindFirstFileW
1742
# TODO: We possibly could support Win98 by falling back to the
1743
# original FindFirstFile, and using TCHAR instead of WCHAR,
1744
# but that gets a bit tricky, and requires custom compiling
1747
from bzrlib._walkdirs_win32 import Win32ReadDir
1748
_selected_dir_reader = Win32ReadDir()
1751
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1752
# ANSI_X3.4-1968 is a form of ASCII
1754
from bzrlib._readdir_pyx import UTF8DirReader
1755
_selected_dir_reader = UTF8DirReader()
1756
except ImportError, e:
1757
failed_to_load_extension(e)
1760
if _selected_dir_reader is None:
1761
# Fallback to the python version
1762
_selected_dir_reader = UnicodeDirReader()
1764
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1765
# But we don't actually uses 1-3 in pending, so set them to None
1766
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1767
read_dir = _selected_dir_reader.read_dir
1768
_directory = _directory_kind
1770
relroot, _, _, _, top = pending[-1].pop()
1773
dirblock = sorted(read_dir(relroot, top))
1774
yield (relroot, top), dirblock
1775
# push the user specified dirs from dirblock
1776
next = [d for d in reversed(dirblock) if d[2] == _directory]
1778
pending.append(next)
1781
class UnicodeDirReader(DirReader):
1782
"""A dir reader for non-utf8 file systems, which transcodes."""
1784
__slots__ = ['_utf8_encode']
1787
self._utf8_encode = codecs.getencoder('utf8')
1789
def top_prefix_to_starting_dir(self, top, prefix=""):
1790
"""See DirReader.top_prefix_to_starting_dir."""
1791
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1793
def read_dir(self, prefix, top):
1794
"""Read a single directory from a non-utf8 file system.
1796
top, and the abspath element in the output are unicode, all other paths
1797
are utf8. Local disk IO is done via unicode calls to listdir etc.
1799
This is currently the fallback code path when the filesystem encoding is
1800
not UTF-8. It may be better to implement an alternative so that we can
1801
safely handle paths that are not properly decodable in the current
1804
See DirReader.read_dir for details.
1806
_utf8_encode = self._utf8_encode
1808
_listdir = os.listdir
1809
_kind_from_mode = file_kind_from_stat_mode
1812
relprefix = prefix + '/'
1815
top_slash = top + u'/'
1818
append = dirblock.append
1819
for name in sorted(_listdir(top)):
1821
name_utf8 = _utf8_encode(name)[0]
1822
except UnicodeDecodeError:
1823
raise errors.BadFilenameEncoding(
1824
_utf8_encode(relprefix)[0] + name, _fs_enc)
1825
abspath = top_slash + name
1826
statvalue = _lstat(abspath)
1827
kind = _kind_from_mode(statvalue.st_mode)
1828
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1832
def copy_tree(from_path, to_path, handlers={}):
1833
"""Copy all of the entries in from_path into to_path.
1835
:param from_path: The base directory to copy.
1836
:param to_path: The target directory. If it does not exist, it will
1838
:param handlers: A dictionary of functions, which takes a source and
1839
destinations for files, directories, etc.
1840
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1841
'file', 'directory', and 'symlink' should always exist.
1842
If they are missing, they will be replaced with 'os.mkdir()',
1843
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1845
# Now, just copy the existing cached tree to the new location
1846
# We use a cheap trick here.
1847
# Absolute paths are prefixed with the first parameter
1848
# relative paths are prefixed with the second.
1849
# So we can get both the source and target returned
1850
# without any extra work.
1852
def copy_dir(source, dest):
1855
def copy_link(source, dest):
1856
"""Copy the contents of a symlink"""
1857
link_to = os.readlink(source)
1858
os.symlink(link_to, dest)
1860
real_handlers = {'file':shutil.copy2,
1861
'symlink':copy_link,
1862
'directory':copy_dir,
1864
real_handlers.update(handlers)
1866
if not os.path.exists(to_path):
1867
real_handlers['directory'](from_path, to_path)
1869
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1870
for relpath, name, kind, st, abspath in entries:
1871
real_handlers[kind](abspath, relpath)
1874
def copy_ownership_from_path(dst, src=None):
1875
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1877
If src is None, the containing directory is used as source. If chown
1878
fails, the error is ignored and a warning is printed.
1880
chown = getattr(os, 'chown', None)
1885
src = os.path.dirname(dst)
1891
chown(dst, s.st_uid, s.st_gid)
1894
'Unable to copy ownership from "%s" to "%s". '
1895
'You may want to set it manually.', src, dst)
1896
trace.log_exception_quietly()
1899
def path_prefix_key(path):
1900
"""Generate a prefix-order path key for path.
1902
This can be used to sort paths in the same way that walkdirs does.
1904
return (dirname(path) , path)
1907
def compare_paths_prefix_order(path_a, path_b):
1908
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1909
key_a = path_prefix_key(path_a)
1910
key_b = path_prefix_key(path_b)
1911
return cmp(key_a, key_b)
1914
_cached_user_encoding = None
1917
def get_user_encoding(use_cache=True):
1918
"""Find out what the preferred user encoding is.
1920
This is generally the encoding that is used for command line parameters
1921
and file contents. This may be different from the terminal encoding
1922
or the filesystem encoding.
1924
:param use_cache: Enable cache for detected encoding.
1925
(This parameter is turned on by default,
1926
and required only for selftesting)
1928
:return: A string defining the preferred user encoding
1930
global _cached_user_encoding
1931
if _cached_user_encoding is not None and use_cache:
1932
return _cached_user_encoding
1934
if sys.platform == 'darwin':
1935
# python locale.getpreferredencoding() always return
1936
# 'mac-roman' on darwin. That's a lie.
1937
sys.platform = 'posix'
1939
if os.environ.get('LANG', None) is None:
1940
# If LANG is not set, we end up with 'ascii', which is bad
1941
# ('mac-roman' is more than ascii), so we set a default which
1942
# will give us UTF-8 (which appears to work in all cases on
1943
# OSX). Users are still free to override LANG of course, as
1944
# long as it give us something meaningful. This work-around
1945
# *may* not be needed with python 3k and/or OSX 10.5, but will
1946
# work with them too -- vila 20080908
1947
os.environ['LANG'] = 'en_US.UTF-8'
1950
sys.platform = 'darwin'
1955
user_encoding = locale.getpreferredencoding()
1956
except locale.Error, e:
1957
sys.stderr.write('bzr: warning: %s\n'
1958
' Could not determine what text encoding to use.\n'
1959
' This error usually means your Python interpreter\n'
1960
' doesn\'t support the locale set by $LANG (%s)\n'
1961
" Continuing with ascii encoding.\n"
1962
% (e, os.environ.get('LANG')))
1963
user_encoding = 'ascii'
1965
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1966
# treat that as ASCII, and not support printing unicode characters to the
1969
# For python scripts run under vim, we get '', so also treat that as ASCII
1970
if user_encoding in (None, 'cp0', ''):
1971
user_encoding = 'ascii'
1975
codecs.lookup(user_encoding)
1977
sys.stderr.write('bzr: warning:'
1978
' unknown encoding %s.'
1979
' Continuing with ascii encoding.\n'
1982
user_encoding = 'ascii'
1985
_cached_user_encoding = user_encoding
1987
return user_encoding
1990
def get_diff_header_encoding():
1991
return get_terminal_encoding()
1994
def get_host_name():
1995
"""Return the current unicode host name.
1997
This is meant to be used in place of socket.gethostname() because that
1998
behaves inconsistently on different platforms.
2000
if sys.platform == "win32":
2002
return win32utils.get_host_name()
2005
return socket.gethostname().decode(get_user_encoding())
2008
# We must not read/write any more than 64k at a time from/to a socket so we
2009
# don't risk "no buffer space available" errors on some platforms. Windows in
2010
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2012
MAX_SOCKET_CHUNK = 64 * 1024
2014
_end_of_stream_errors = [errno.ECONNRESET]
2015
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2016
_eno = getattr(errno, _eno, None)
2017
if _eno is not None:
2018
_end_of_stream_errors.append(_eno)
2022
def read_bytes_from_socket(sock, report_activity=None,
2023
max_read_size=MAX_SOCKET_CHUNK):
2024
"""Read up to max_read_size of bytes from sock and notify of progress.
2026
Translates "Connection reset by peer" into file-like EOF (return an
2027
empty string rather than raise an error), and repeats the recv if
2028
interrupted by a signal.
2032
bytes = sock.recv(max_read_size)
2033
except socket.error, e:
2035
if eno in _end_of_stream_errors:
2036
# The connection was closed by the other side. Callers expect
2037
# an empty string to signal end-of-stream.
2039
elif eno == errno.EINTR:
2040
# Retry the interrupted recv.
2044
if report_activity is not None:
2045
report_activity(len(bytes), 'read')
2049
def recv_all(socket, count):
2050
"""Receive an exact number of bytes.
2052
Regular Socket.recv() may return less than the requested number of bytes,
2053
depending on what's in the OS buffer. MSG_WAITALL is not available
2054
on all platforms, but this should work everywhere. This will return
2055
less than the requested amount if the remote end closes.
2057
This isn't optimized and is intended mostly for use in testing.
2060
while len(b) < count:
2061
new = read_bytes_from_socket(socket, None, count - len(b))
2068
def send_all(sock, bytes, report_activity=None):
2069
"""Send all bytes on a socket.
2071
Breaks large blocks in smaller chunks to avoid buffering limitations on
2072
some platforms, and catches EINTR which may be thrown if the send is
2073
interrupted by a signal.
2075
This is preferred to socket.sendall(), because it avoids portability bugs
2076
and provides activity reporting.
2078
:param report_activity: Call this as bytes are read, see
2079
Transport._report_activity
2082
byte_count = len(bytes)
2083
while sent_total < byte_count:
2085
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2086
except socket.error, e:
2087
if e.args[0] != errno.EINTR:
2091
report_activity(sent, 'write')
2094
def connect_socket(address):
2095
# Slight variation of the socket.create_connection() function (provided by
2096
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2097
# provide it for previous python versions. Also, we don't use the timeout
2098
# parameter (provided by the python implementation) so we don't implement
2100
err = socket.error('getaddrinfo returns an empty list')
2101
host, port = address
2102
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2103
af, socktype, proto, canonname, sa = res
2106
sock = socket.socket(af, socktype, proto)
2110
except socket.error, err:
2111
# 'err' is now the most recent error
2112
if sock is not None:
2117
def dereference_path(path):
2118
"""Determine the real path to a file.
2120
All parent elements are dereferenced. But the file itself is not
2122
:param path: The original path. May be absolute or relative.
2123
:return: the real path *to* the file
2125
parent, base = os.path.split(path)
2126
# The pathjoin for '.' is a workaround for Python bug #1213894.
2127
# (initial path components aren't dereferenced)
2128
return pathjoin(realpath(pathjoin('.', parent)), base)
2131
def supports_mapi():
2132
"""Return True if we can use MAPI to launch a mail client."""
2133
return sys.platform == "win32"
2136
def resource_string(package, resource_name):
2137
"""Load a resource from a package and return it as a string.
2139
Note: Only packages that start with bzrlib are currently supported.
2141
This is designed to be a lightweight implementation of resource
2142
loading in a way which is API compatible with the same API from
2144
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2145
If and when pkg_resources becomes a standard library, this routine
2148
# Check package name is within bzrlib
2149
if package == "bzrlib":
2150
resource_relpath = resource_name
2151
elif package.startswith("bzrlib."):
2152
package = package[len("bzrlib."):].replace('.', os.sep)
2153
resource_relpath = pathjoin(package, resource_name)
2155
raise errors.BzrError('resource package %s not in bzrlib' % package)
2157
# Map the resource to a file and read its contents
2158
base = dirname(bzrlib.__file__)
2159
if getattr(sys, 'frozen', None): # bzr.exe
2160
base = abspath(pathjoin(base, '..', '..'))
2161
f = file(pathjoin(base, resource_relpath), "rU")
2167
def file_kind_from_stat_mode_thunk(mode):
2168
global file_kind_from_stat_mode
2169
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2171
from bzrlib._readdir_pyx import UTF8DirReader
2172
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2173
except ImportError, e:
2174
# This is one time where we won't warn that an extension failed to
2175
# load. The extension is never available on Windows anyway.
2176
from bzrlib._readdir_py import (
2177
_kind_from_mode as file_kind_from_stat_mode
2179
return file_kind_from_stat_mode(mode)
2180
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2182
def file_stat(f, _lstat=os.lstat):
2187
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2188
raise errors.NoSuchFile(f)
2191
def file_kind(f, _lstat=os.lstat):
2192
stat_value = file_stat(f, _lstat)
2193
return file_kind_from_stat_mode(stat_value.st_mode)
2195
def until_no_eintr(f, *a, **kw):
2196
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2198
WARNING: you must be certain that it is safe to retry the call repeatedly
2199
if EINTR does occur. This is typically only true for low-level operations
2200
like os.read. If in any doubt, don't use this.
2202
Keep in mind that this is not a complete solution to EINTR. There is
2203
probably code in the Python standard library and other dependencies that
2204
may encounter EINTR if a signal arrives (and there is signal handler for
2205
that signal). So this function can reduce the impact for IO that bzrlib
2206
directly controls, but it is not a complete solution.
2208
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2212
except (IOError, OSError), e:
2213
if e.errno == errno.EINTR:
2218
@deprecated_function(deprecated_in((2, 2, 0)))
2219
def re_compile_checked(re_string, flags=0, where=""):
2220
"""Return a compiled re, or raise a sensible error.
2222
This should only be used when compiling user-supplied REs.
2224
:param re_string: Text form of regular expression.
2225
:param flags: eg re.IGNORECASE
2226
:param where: Message explaining to the user the context where
2227
it occurred, eg 'log search filter'.
2229
# from https://bugs.launchpad.net/bzr/+bug/251352
2231
re_obj = re.compile(re_string, flags)
2234
except errors.InvalidPattern, e:
2236
where = ' in ' + where
2237
# despite the name 'error' is a type
2238
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2242
if sys.platform == "win32":
2245
return msvcrt.getch()
2250
fd = sys.stdin.fileno()
2251
settings = termios.tcgetattr(fd)
2254
ch = sys.stdin.read(1)
2256
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2259
if sys.platform.startswith('linux'):
2260
def _local_concurrency():
2262
return os.sysconf('SC_NPROCESSORS_ONLN')
2263
except (ValueError, OSError, AttributeError):
2265
elif sys.platform == 'darwin':
2266
def _local_concurrency():
2267
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2268
stdout=subprocess.PIPE).communicate()[0]
2269
elif "bsd" in sys.platform:
2270
def _local_concurrency():
2271
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2272
stdout=subprocess.PIPE).communicate()[0]
2273
elif sys.platform == 'sunos5':
2274
def _local_concurrency():
2275
return subprocess.Popen(['psrinfo', '-p',],
2276
stdout=subprocess.PIPE).communicate()[0]
2277
elif sys.platform == "win32":
2278
def _local_concurrency():
2279
# This appears to return the number of cores.
2280
return os.environ.get('NUMBER_OF_PROCESSORS')
2282
def _local_concurrency():
2287
_cached_local_concurrency = None
2289
def local_concurrency(use_cache=True):
2290
"""Return how many processes can be run concurrently.
2292
Rely on platform specific implementations and default to 1 (one) if
2293
anything goes wrong.
2295
global _cached_local_concurrency
2297
if _cached_local_concurrency is not None and use_cache:
2298
return _cached_local_concurrency
2300
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2301
if concurrency is None:
2303
import multiprocessing
2305
# multiprocessing is only available on Python >= 2.6
2307
concurrency = _local_concurrency()
2308
except (OSError, IOError):
2311
concurrency = multiprocessing.cpu_count()
2313
concurrency = int(concurrency)
2314
except (TypeError, ValueError):
2317
_cached_concurrency = concurrency
2321
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2322
"""A stream writer that doesn't decode str arguments."""
2324
def __init__(self, encode, stream, errors='strict'):
2325
codecs.StreamWriter.__init__(self, stream, errors)
2326
self.encode = encode
2328
def write(self, object):
2329
if type(object) is str:
2330
self.stream.write(object)
2332
data, _ = self.encode(object, self.errors)
2333
self.stream.write(data)
2335
if sys.platform == 'win32':
2336
def open_file(filename, mode='r', bufsize=-1):
2337
"""This function is used to override the ``open`` builtin.
2339
But it uses O_NOINHERIT flag so the file handle is not inherited by
2340
child processes. Deleting or renaming a closed file opened with this
2341
function is not blocking child processes.
2343
writing = 'w' in mode
2344
appending = 'a' in mode
2345
updating = '+' in mode
2346
binary = 'b' in mode
2349
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2350
# for flags for each modes.
2360
flags |= os.O_WRONLY
2361
flags |= os.O_CREAT | os.O_TRUNC
2366
flags |= os.O_WRONLY
2367
flags |= os.O_CREAT | os.O_APPEND
2372
flags |= os.O_RDONLY
2374
return os.fdopen(os.open(filename, flags), mode, bufsize)
2379
def getuser_unicode():
2380
"""Return the username as unicode.
2383
user_encoding = get_user_encoding()
2384
username = getpass.getuser().decode(user_encoding)
2385
except UnicodeDecodeError:
2386
raise errors.BzrError("Can't decode username as %s." % \
2388
except ImportError, e:
2389
if sys.platform != 'win32':
2391
if str(e) != 'No module named pwd':
2393
# https://bugs.launchpad.net/bzr/+bug/660174
2394
# getpass.getuser() is unable to return username on Windows
2395
# if there is no USERNAME environment variable set.
2396
# That could be true if bzr is running as a service,
2397
# e.g. running `bzr serve` as a service on Windows.
2398
# We should not fail with traceback in this case.
2399
username = u'UNKNOWN'
2403
def available_backup_name(base, exists):
2404
"""Find a non-existing backup file name.
2406
This will *not* create anything, this only return a 'free' entry. This
2407
should be used for checking names in a directory below a locked
2408
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2409
Leap) and generally discouraged.
2411
:param base: The base name.
2413
:param exists: A callable returning True if the path parameter exists.
2416
name = "%s.~%d~" % (base, counter)
2419
name = "%s.~%d~" % (base, counter)
2423
def set_fd_cloexec(fd):
2424
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2425
support for this is not available.
2429
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2430
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2431
except (ImportError, AttributeError):
2432
# Either the fcntl module or specific constants are not present
2436
def find_executable_on_path(name):
2437
"""Finds an executable on the PATH.
2439
On Windows, this will try to append each extension in the PATHEXT
2440
environment variable to the name, if it cannot be found with the name
2443
:param name: The base name of the executable.
2444
:return: The path to the executable found or None.
2446
path = os.environ.get('PATH')
2449
path = path.split(os.pathsep)
2450
if sys.platform == 'win32':
2451
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2452
exts = [ext.lower() for ext in exts]
2453
base, ext = os.path.splitext(name)
2455
if ext.lower() not in exts:
2463
f = os.path.join(d, name) + ext
2464
if os.access(f, os.X_OK):
2469
def _posix_is_local_pid_dead(pid):
2470
"""True if pid doesn't correspond to live process on this machine"""
2472
# Special meaning of unix kill: just check if it's there.
2475
if e.errno == errno.ESRCH:
2476
# On this machine, and really not found: as sure as we can be
2479
elif e.errno == errno.EPERM:
2480
# exists, though not ours
2483
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2484
# Don't really know.
2487
# Exists and our process: not dead.
2490
if sys.platform == "win32":
2491
is_local_pid_dead = win32utils.is_local_pid_dead
2493
is_local_pid_dead = _posix_is_local_pid_dead
2496
def fdatasync(fileno):
2497
"""Flush file contents to disk if possible.
2499
:param fileno: Integer OS file handle.
2500
:raises TransportNotPossible: If flushing to disk is not possible.
2502
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
added
removed
bzrlib/osutils.py
