1
# Copyright (C) 2005-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from __future__ import absolute_import
27
from bzrlib.lazy_import import lazy_import
28
lazy_import(globals(), """
29
from datetime import datetime
35
# We need to import both shutil and rmtree as we export the later on posix
36
# and need the former on windows
38
from shutil import rmtree
41
# We need to import both tempfile and mkdtemp as we export the later on posix
42
# and need the former on windows
44
from tempfile import mkdtemp
54
from bzrlib.i18n import gettext
57
from bzrlib.symbol_versioning import (
62
warn as warn_deprecated,
72
from bzrlib import symbol_versioning, _fs_enc
75
# Cross platform wall-clock time functionality with decent resolution.
76
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
77
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
78
# synchronized with ``time.time()``, this is only meant to be used to find
79
# delta times by subtracting from another call to this function.
80
timer_func = time.time
81
if sys.platform == 'win32':
82
timer_func = time.clock
84
# On win32, O_BINARY is used to indicate the file should
85
# be opened in binary mode, rather than text mode.
86
# On other platforms, O_BINARY doesn't exist, because
87
# they always open in binary mode, so it is okay to
88
# OR with 0 on those platforms.
89
# O_NOINHERIT and O_TEXT exists only on win32 too.
90
O_BINARY = getattr(os, 'O_BINARY', 0)
91
O_TEXT = getattr(os, 'O_TEXT', 0)
92
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
95
def get_unicode_argv():
97
user_encoding = get_user_encoding()
98
return [a.decode(user_encoding) for a in sys.argv[1:]]
99
except UnicodeDecodeError:
100
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
101
"application locale.").format(a, user_encoding))
104
def make_readonly(filename):
105
"""Make a filename read-only."""
106
mod = os.lstat(filename).st_mode
107
if not stat.S_ISLNK(mod):
109
chmod_if_possible(filename, mod)
112
def make_writable(filename):
113
mod = os.lstat(filename).st_mode
114
if not stat.S_ISLNK(mod):
116
chmod_if_possible(filename, mod)
119
def chmod_if_possible(filename, mode):
120
# Set file mode if that can be safely done.
121
# Sometimes even on unix the filesystem won't allow it - see
122
# https://bugs.launchpad.net/bzr/+bug/606537
124
# It is probably faster to just do the chmod, rather than
125
# doing a stat, and then trying to compare
126
os.chmod(filename, mode)
127
except (IOError, OSError),e:
128
# Permission/access denied seems to commonly happen on smbfs; there's
129
# probably no point warning about it.
130
# <https://bugs.launchpad.net/bzr/+bug/606537>
131
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
132
trace.mutter("ignore error on chmod of %r: %r" % (
138
def minimum_path_selection(paths):
139
"""Return the smallset subset of paths which are outside paths.
141
:param paths: A container (and hence not None) of paths.
142
:return: A set of paths sufficient to include everything in paths via
143
is_inside, drawn from the paths parameter.
149
return path.split('/')
150
sorted_paths = sorted(list(paths), key=sort_key)
152
search_paths = [sorted_paths[0]]
153
for path in sorted_paths[1:]:
154
if not is_inside(search_paths[-1], path):
155
# This path is unique, add it
156
search_paths.append(path)
158
return set(search_paths)
165
"""Return a quoted filename filename
167
This previously used backslash quoting, but that works poorly on
169
# TODO: I'm not really sure this is the best format either.x
171
if _QUOTE_RE is None:
172
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
174
if _QUOTE_RE.search(f):
180
_directory_kind = 'directory'
183
"""Return the current umask"""
184
# Assume that people aren't messing with the umask while running
185
# XXX: This is not thread safe, but there is no way to get the
186
# umask without setting it
194
_directory_kind: "/",
196
'tree-reference': '+',
200
def kind_marker(kind):
202
return _kind_marker_map[kind]
204
# Slightly faster than using .get(, '') when the common case is that
209
lexists = getattr(os.path, 'lexists', None)
213
stat = getattr(os, 'lstat', os.stat)
217
if e.errno == errno.ENOENT:
220
raise errors.BzrError(gettext("lstat/stat of ({0!r}): {1!r}").format(f, e))
223
def fancy_rename(old, new, rename_func, unlink_func):
224
"""A fancy rename, when you don't have atomic rename.
226
:param old: The old path, to rename from
227
:param new: The new path, to rename to
228
:param rename_func: The potentially non-atomic rename function
229
:param unlink_func: A way to delete the target file if the full rename
232
# sftp rename doesn't allow overwriting, so play tricks:
233
base = os.path.basename(new)
234
dirname = os.path.dirname(new)
235
# callers use different encodings for the paths so the following MUST
236
# respect that. We rely on python upcasting to unicode if new is unicode
237
# and keeping a str if not.
238
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
239
os.getpid(), rand_chars(10))
240
tmp_name = pathjoin(dirname, tmp_name)
242
# Rename the file out of the way, but keep track if it didn't exist
243
# We don't want to grab just any exception
244
# something like EACCES should prevent us from continuing
245
# The downside is that the rename_func has to throw an exception
246
# with an errno = ENOENT, or NoSuchFile
249
rename_func(new, tmp_name)
250
except (errors.NoSuchFile,), e:
253
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
254
# function raises an IOError with errno is None when a rename fails.
255
# This then gets caught here.
256
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
259
if (getattr(e, 'errno', None) is None
260
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
269
# This may throw an exception, in which case success will
271
rename_func(old, new)
273
except (IOError, OSError), e:
274
# source and target may be aliases of each other (e.g. on a
275
# case-insensitive filesystem), so we may have accidentally renamed
276
# source by when we tried to rename target
277
failure_exc = sys.exc_info()
278
if (file_existed and e.errno in (None, errno.ENOENT)
279
and old.lower() == new.lower()):
280
# source and target are the same file on a case-insensitive
281
# filesystem, so we don't generate an exception
285
# If the file used to exist, rename it back into place
286
# otherwise just delete it from the tmp location
288
unlink_func(tmp_name)
290
rename_func(tmp_name, new)
291
if failure_exc is not None:
293
raise failure_exc[0], failure_exc[1], failure_exc[2]
298
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
299
# choke on a Unicode string containing a relative path if
300
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
302
def _posix_abspath(path):
303
# jam 20060426 rather than encoding to fsencoding
304
# copy posixpath.abspath, but use os.getcwdu instead
305
if not posixpath.isabs(path):
306
path = posixpath.join(getcwd(), path)
307
return _posix_normpath(path)
310
def _posix_realpath(path):
311
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
314
def _posix_normpath(path):
315
path = posixpath.normpath(path)
316
# Bug 861008: posixpath.normpath() returns a path normalized according to
317
# the POSIX standard, which stipulates (for compatibility reasons) that two
318
# leading slashes must not be simplified to one, and only if there are 3 or
319
# more should they be simplified as one. So we treat the leading 2 slashes
320
# as a special case here by simply removing the first slash, as we consider
321
# that breaking POSIX compatibility for this obscure feature is acceptable.
322
# This is not a paranoid precaution, as we notably get paths like this when
323
# the repo is hosted at the root of the filesystem, i.e. in "/".
324
if path.startswith('//'):
329
def _posix_path_from_environ(key):
330
"""Get unicode path from `key` in environment or None if not present
332
Note that posix systems use arbitrary byte strings for filesystem objects,
333
so a path that raises BadFilenameEncoding here may still be accessible.
335
val = os.environ.get(key, None)
339
return val.decode(_fs_enc)
340
except UnicodeDecodeError:
341
# GZ 2011-12-12:Ideally want to include `key` in the exception message
342
raise errors.BadFilenameEncoding(val, _fs_enc)
345
def _posix_get_home_dir():
346
"""Get the home directory of the current user as a unicode path"""
347
path = posixpath.expanduser("~")
349
return path.decode(_fs_enc)
350
except UnicodeDecodeError:
351
raise errors.BadFilenameEncoding(path, _fs_enc)
354
def _posix_getuser_unicode():
355
"""Get username from environment or password database as unicode"""
356
name = getpass.getuser()
357
user_encoding = get_user_encoding()
359
return name.decode(user_encoding)
360
except UnicodeDecodeError:
361
raise errors.BzrError("Encoding of username %r is unsupported by %s "
362
"application locale." % (name, user_encoding))
365
def _win32_fixdrive(path):
366
"""Force drive letters to be consistent.
368
win32 is inconsistent whether it returns lower or upper case
369
and even if it was consistent the user might type the other
370
so we force it to uppercase
371
running python.exe under cmd.exe return capital C:\\
372
running win32 python inside a cygwin shell returns lowercase c:\\
374
drive, path = ntpath.splitdrive(path)
375
return drive.upper() + path
378
def _win32_abspath(path):
379
# Real ntpath.abspath doesn't have a problem with a unicode cwd
380
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
383
def _win98_abspath(path):
384
"""Return the absolute version of a path.
385
Windows 98 safe implementation (python reimplementation
386
of Win32 API function GetFullPathNameW)
391
# \\HOST\path => //HOST/path
392
# //HOST/path => //HOST/path
393
# path => C:/cwd/path
396
# check for absolute path
397
drive = ntpath.splitdrive(path)[0]
398
if drive == '' and path[:2] not in('//','\\\\'):
400
# we cannot simply os.path.join cwd and path
401
# because os.path.join('C:','/path') produce '/path'
402
# and this is incorrect
403
if path[:1] in ('/','\\'):
404
cwd = ntpath.splitdrive(cwd)[0]
406
path = cwd + '\\' + path
407
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
410
def _win32_realpath(path):
411
# Real ntpath.realpath doesn't have a problem with a unicode cwd
412
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
415
def _win32_pathjoin(*args):
416
return ntpath.join(*args).replace('\\', '/')
419
def _win32_normpath(path):
420
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
424
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
427
def _win32_mkdtemp(*args, **kwargs):
428
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
431
def _win32_rename(old, new):
432
"""We expect to be able to atomically replace 'new' with old.
434
On win32, if new exists, it must be moved out of the way first,
438
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
440
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
441
# If we try to rename a non-existant file onto cwd, we get
442
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
443
# if the old path doesn't exist, sometimes we get EACCES
444
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
450
return unicodedata.normalize('NFC', os.getcwdu())
453
def _rename_wrap_exception(rename_func):
454
"""Adds extra information to any exceptions that come from rename().
456
The exception has an updated message and 'old_filename' and 'new_filename'
460
def _rename_wrapper(old, new):
462
rename_func(old, new)
464
detailed_error = OSError(e.errno, e.strerror +
465
" [occurred when renaming '%s' to '%s']" %
467
detailed_error.old_filename = old
468
detailed_error.new_filename = new
471
return _rename_wrapper
473
# Default rename wraps os.rename()
474
rename = _rename_wrap_exception(os.rename)
476
# Default is to just use the python builtins, but these can be rebound on
477
# particular platforms.
478
abspath = _posix_abspath
479
realpath = _posix_realpath
480
pathjoin = os.path.join
481
normpath = _posix_normpath
482
path_from_environ = _posix_path_from_environ
483
_get_home_dir = _posix_get_home_dir
484
getuser_unicode = _posix_getuser_unicode
486
dirname = os.path.dirname
487
basename = os.path.basename
488
split = os.path.split
489
splitext = os.path.splitext
490
# These were already lazily imported into local scope
491
# mkdtemp = tempfile.mkdtemp
492
# rmtree = shutil.rmtree
500
MIN_ABS_PATHLENGTH = 1
503
if sys.platform == 'win32':
504
if win32utils.winver == 'Windows 98':
505
abspath = _win98_abspath
507
abspath = _win32_abspath
508
realpath = _win32_realpath
509
pathjoin = _win32_pathjoin
510
normpath = _win32_normpath
511
getcwd = _win32_getcwd
512
mkdtemp = _win32_mkdtemp
513
rename = _rename_wrap_exception(_win32_rename)
515
from bzrlib import _walkdirs_win32
519
lstat = _walkdirs_win32.lstat
520
fstat = _walkdirs_win32.fstat
521
wrap_stat = _walkdirs_win32.wrap_stat
523
MIN_ABS_PATHLENGTH = 3
525
def _win32_delete_readonly(function, path, excinfo):
526
"""Error handler for shutil.rmtree function [for win32]
527
Helps to remove files and dirs marked as read-only.
529
exception = excinfo[1]
530
if function in (os.remove, os.rmdir) \
531
and isinstance(exception, OSError) \
532
and exception.errno == errno.EACCES:
538
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
539
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
540
return shutil.rmtree(path, ignore_errors, onerror)
542
f = win32utils.get_unicode_argv # special function or None
545
path_from_environ = win32utils.get_environ_unicode
546
_get_home_dir = win32utils.get_home_location
547
getuser_unicode = win32utils.get_user_name
549
elif sys.platform == 'darwin':
553
def get_terminal_encoding(trace=False):
554
"""Find the best encoding for printing to the screen.
556
This attempts to check both sys.stdout and sys.stdin to see
557
what encoding they are in, and if that fails it falls back to
558
osutils.get_user_encoding().
559
The problem is that on Windows, locale.getpreferredencoding()
560
is not the same encoding as that used by the console:
561
http://mail.python.org/pipermail/python-list/2003-May/162357.html
563
On my standard US Windows XP, the preferred encoding is
564
cp1252, but the console is cp437
566
:param trace: If True trace the selected encoding via mutter().
568
from bzrlib.trace import mutter
569
output_encoding = getattr(sys.stdout, 'encoding', None)
570
if not output_encoding:
571
input_encoding = getattr(sys.stdin, 'encoding', None)
572
if not input_encoding:
573
output_encoding = get_user_encoding()
575
mutter('encoding stdout as osutils.get_user_encoding() %r',
578
output_encoding = input_encoding
580
mutter('encoding stdout as sys.stdin encoding %r',
584
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
585
if output_encoding == 'cp0':
586
# invalid encoding (cp0 means 'no codepage' on Windows)
587
output_encoding = get_user_encoding()
589
mutter('cp0 is invalid encoding.'
590
' encoding stdout as osutils.get_user_encoding() %r',
594
codecs.lookup(output_encoding)
596
sys.stderr.write('bzr: warning:'
597
' unknown terminal encoding %s.\n'
598
' Using encoding %s instead.\n'
599
% (output_encoding, get_user_encoding())
601
output_encoding = get_user_encoding()
603
return output_encoding
606
def normalizepath(f):
607
if getattr(os.path, 'realpath', None) is not None:
611
[p,e] = os.path.split(f)
612
if e == "" or e == "." or e == "..":
615
return pathjoin(F(p), e)
619
"""True if f is an accessible directory."""
621
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
627
"""True if f is a regular file."""
629
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
634
"""True if f is a symlink."""
636
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
640
def is_inside(dir, fname):
641
"""True if fname is inside dir.
643
The parameters should typically be passed to osutils.normpath first, so
644
that . and .. and repeated slashes are eliminated, and the separators
645
are canonical for the platform.
647
The empty string as a dir name is taken as top-of-tree and matches
650
# XXX: Most callers of this can actually do something smarter by
651
# looking at the inventory
661
return fname.startswith(dir)
664
def is_inside_any(dir_list, fname):
665
"""True if fname is inside any of given dirs."""
666
for dirname in dir_list:
667
if is_inside(dirname, fname):
672
def is_inside_or_parent_of_any(dir_list, fname):
673
"""True if fname is a child or a parent of any of the given files."""
674
for dirname in dir_list:
675
if is_inside(dirname, fname) or is_inside(fname, dirname):
680
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
681
report_activity=None, direction='read'):
682
"""Copy contents of one file to another.
684
The read_length can either be -1 to read to end-of-file (EOF) or
685
it can specify the maximum number of bytes to read.
687
The buff_size represents the maximum size for each read operation
688
performed on from_file.
690
:param report_activity: Call this as bytes are read, see
691
Transport._report_activity
692
:param direction: Will be passed to report_activity
694
:return: The number of bytes copied.
698
# read specified number of bytes
700
while read_length > 0:
701
num_bytes_to_read = min(read_length, buff_size)
703
block = from_file.read(num_bytes_to_read)
707
if report_activity is not None:
708
report_activity(len(block), direction)
711
actual_bytes_read = len(block)
712
read_length -= actual_bytes_read
713
length += actual_bytes_read
717
block = from_file.read(buff_size)
721
if report_activity is not None:
722
report_activity(len(block), direction)
728
def pump_string_file(bytes, file_handle, segment_size=None):
729
"""Write bytes to file_handle in many smaller writes.
731
:param bytes: The string to write.
732
:param file_handle: The file to write to.
734
# Write data in chunks rather than all at once, because very large
735
# writes fail on some platforms (e.g. Windows with SMB mounted
738
segment_size = 5242880 # 5MB
739
segments = range(len(bytes) / segment_size + 1)
740
write = file_handle.write
741
for segment_index in segments:
742
segment = buffer(bytes, segment_index * segment_size, segment_size)
746
def file_iterator(input_file, readsize=32768):
748
b = input_file.read(readsize)
755
"""Calculate the hexdigest of an open file.
757
The file cursor should be already at the start.
769
def size_sha_file(f):
770
"""Calculate the size and hexdigest of an open file.
772
The file cursor should be already at the start and
773
the caller is responsible for closing the file afterwards.
784
return size, s.hexdigest()
787
def sha_file_by_name(fname):
788
"""Calculate the SHA1 of a file by reading the full text"""
790
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
793
b = os.read(f, 1<<16)
801
def sha_strings(strings, _factory=sha):
802
"""Return the sha-1 of concatenation of strings"""
804
map(s.update, strings)
808
def sha_string(f, _factory=sha):
809
return _factory(f).hexdigest()
812
def fingerprint_file(f):
814
return {'size': len(b),
815
'sha1': sha(b).hexdigest()}
818
def compare_files(a, b):
819
"""Returns true if equal in contents"""
830
def local_time_offset(t=None):
831
"""Return offset of local zone from GMT, either at present or at time t."""
834
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
835
return offset.days * 86400 + offset.seconds
837
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
838
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
841
def format_date(t, offset=0, timezone='original', date_fmt=None,
843
"""Return a formatted date string.
845
:param t: Seconds since the epoch.
846
:param offset: Timezone offset in seconds east of utc.
847
:param timezone: How to display the time: 'utc', 'original' for the
848
timezone specified by offset, or 'local' for the process's current
850
:param date_fmt: strftime format.
851
:param show_offset: Whether to append the timezone.
853
(date_fmt, tt, offset_str) = \
854
_format_date(t, offset, timezone, date_fmt, show_offset)
855
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
856
date_str = time.strftime(date_fmt, tt)
857
return date_str + offset_str
860
# Cache of formatted offset strings
864
def format_date_with_offset_in_original_timezone(t, offset=0,
865
_cache=_offset_cache):
866
"""Return a formatted date string in the original timezone.
868
This routine may be faster then format_date.
870
:param t: Seconds since the epoch.
871
:param offset: Timezone offset in seconds east of utc.
875
tt = time.gmtime(t + offset)
876
date_fmt = _default_format_by_weekday_num[tt[6]]
877
date_str = time.strftime(date_fmt, tt)
878
offset_str = _cache.get(offset, None)
879
if offset_str is None:
880
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
881
_cache[offset] = offset_str
882
return date_str + offset_str
885
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
887
"""Return an unicode date string formatted according to the current locale.
889
:param t: Seconds since the epoch.
890
:param offset: Timezone offset in seconds east of utc.
891
:param timezone: How to display the time: 'utc', 'original' for the
892
timezone specified by offset, or 'local' for the process's current
894
:param date_fmt: strftime format.
895
:param show_offset: Whether to append the timezone.
897
(date_fmt, tt, offset_str) = \
898
_format_date(t, offset, timezone, date_fmt, show_offset)
899
date_str = time.strftime(date_fmt, tt)
900
if not isinstance(date_str, unicode):
901
date_str = date_str.decode(get_user_encoding(), 'replace')
902
return date_str + offset_str
905
def _format_date(t, offset, timezone, date_fmt, show_offset):
906
if timezone == 'utc':
909
elif timezone == 'original':
912
tt = time.gmtime(t + offset)
913
elif timezone == 'local':
914
tt = time.localtime(t)
915
offset = local_time_offset(t)
917
raise errors.UnsupportedTimezoneFormat(timezone)
919
date_fmt = "%a %Y-%m-%d %H:%M:%S"
921
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
924
return (date_fmt, tt, offset_str)
927
def compact_date(when):
928
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
931
def format_delta(delta):
932
"""Get a nice looking string for a time delta.
934
:param delta: The time difference in seconds, can be positive or negative.
935
positive indicates time in the past, negative indicates time in the
936
future. (usually time.time() - stored_time)
937
:return: String formatted to show approximate resolution
943
direction = 'in the future'
947
if seconds < 90: # print seconds up to 90 seconds
949
return '%d second %s' % (seconds, direction,)
951
return '%d seconds %s' % (seconds, direction)
953
minutes = int(seconds / 60)
954
seconds -= 60 * minutes
959
if minutes < 90: # print minutes, seconds up to 90 minutes
961
return '%d minute, %d second%s %s' % (
962
minutes, seconds, plural_seconds, direction)
964
return '%d minutes, %d second%s %s' % (
965
minutes, seconds, plural_seconds, direction)
967
hours = int(minutes / 60)
968
minutes -= 60 * hours
975
return '%d hour, %d minute%s %s' % (hours, minutes,
976
plural_minutes, direction)
977
return '%d hours, %d minute%s %s' % (hours, minutes,
978
plural_minutes, direction)
981
"""Return size of given open file."""
982
return os.fstat(f.fileno())[stat.ST_SIZE]
985
# Alias os.urandom to support platforms (which?) without /dev/urandom and
986
# override if it doesn't work. Avoid checking on windows where there is
987
# significant initialisation cost that can be avoided for some bzr calls.
989
rand_bytes = os.urandom
991
if rand_bytes.__module__ != "nt":
994
except NotImplementedError:
995
# not well seeded, but better than nothing
1000
s += chr(random.randint(0, 255))
1005
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
1006
def rand_chars(num):
1007
"""Return a random string of num alphanumeric characters
1009
The result only contains lowercase chars because it may be used on
1010
case-insensitive filesystems.
1013
for raw_byte in rand_bytes(num):
1014
s += ALNUM[ord(raw_byte) % 36]
1018
## TODO: We could later have path objects that remember their list
1019
## decomposition (might be too tricksy though.)
1022
"""Turn string into list of parts."""
1023
# split on either delimiter because people might use either on
1025
ps = re.split(r'[\\/]', p)
1030
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1031
elif (f == '.') or (f == ''):
1040
if (f == '..') or (f is None) or (f == ''):
1041
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1045
def parent_directories(filename):
1046
"""Return the list of parent directories, deepest first.
1048
For example, parent_directories("a/b/c") -> ["a/b", "a"].
1051
parts = splitpath(dirname(filename))
1053
parents.append(joinpath(parts))
1058
_extension_load_failures = []
1061
def failed_to_load_extension(exception):
1062
"""Handle failing to load a binary extension.
1064
This should be called from the ImportError block guarding the attempt to
1065
import the native extension. If this function returns, the pure-Python
1066
implementation should be loaded instead::
1069
>>> import bzrlib._fictional_extension_pyx
1070
>>> except ImportError, e:
1071
>>> bzrlib.osutils.failed_to_load_extension(e)
1072
>>> import bzrlib._fictional_extension_py
1074
# NB: This docstring is just an example, not a doctest, because doctest
1075
# currently can't cope with the use of lazy imports in this namespace --
1078
# This currently doesn't report the failure at the time it occurs, because
1079
# they tend to happen very early in startup when we can't check config
1080
# files etc, and also we want to report all failures but not spam the user
1082
exception_str = str(exception)
1083
if exception_str not in _extension_load_failures:
1084
trace.mutter("failed to load compiled extension: %s" % exception_str)
1085
_extension_load_failures.append(exception_str)
1088
def report_extension_load_failures():
1089
if not _extension_load_failures:
1091
if config.GlobalStack().get('ignore_missing_extensions'):
1093
# the warnings framework should by default show this only once
1094
from bzrlib.trace import warning
1096
"bzr: warning: some compiled extensions could not be loaded; "
1097
"see <https://answers.launchpad.net/bzr/+faq/703>")
1098
# we no longer show the specific missing extensions here, because it makes
1099
# the message too long and scary - see
1100
# https://bugs.launchpad.net/bzr/+bug/430529
1104
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1105
except ImportError, e:
1106
failed_to_load_extension(e)
1107
from bzrlib._chunks_to_lines_py import chunks_to_lines
1111
"""Split s into lines, but without removing the newline characters."""
1112
# Trivially convert a fulltext into a 'chunked' representation, and let
1113
# chunks_to_lines do the heavy lifting.
1114
if isinstance(s, str):
1115
# chunks_to_lines only supports 8-bit strings
1116
return chunks_to_lines([s])
1118
return _split_lines(s)
1121
def _split_lines(s):
1122
"""Split s into lines, but without removing the newline characters.
1124
This supports Unicode or plain string objects.
1126
lines = s.split('\n')
1127
result = [line + '\n' for line in lines[:-1]]
1129
result.append(lines[-1])
1133
def hardlinks_good():
1134
return sys.platform not in ('win32', 'cygwin', 'darwin')
1137
def link_or_copy(src, dest):
1138
"""Hardlink a file, or copy it if it can't be hardlinked."""
1139
if not hardlinks_good():
1140
shutil.copyfile(src, dest)
1144
except (OSError, IOError), e:
1145
if e.errno != errno.EXDEV:
1147
shutil.copyfile(src, dest)
1150
def delete_any(path):
1151
"""Delete a file, symlink or directory.
1153
Will delete even if readonly.
1156
_delete_file_or_dir(path)
1157
except (OSError, IOError), e:
1158
if e.errno in (errno.EPERM, errno.EACCES):
1159
# make writable and try again
1162
except (OSError, IOError):
1164
_delete_file_or_dir(path)
1169
def _delete_file_or_dir(path):
1170
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1171
# Forgiveness than Permission (EAFP) because:
1172
# - root can damage a solaris file system by using unlink,
1173
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1174
# EACCES, OSX: EPERM) when invoked on a directory.
1175
if isdir(path): # Takes care of symlinks
1182
if getattr(os, 'symlink', None) is not None:
1188
def has_hardlinks():
1189
if getattr(os, 'link', None) is not None:
1195
def host_os_dereferences_symlinks():
1196
return (has_symlinks()
1197
and sys.platform not in ('cygwin', 'win32'))
1200
def readlink(abspath):
1201
"""Return a string representing the path to which the symbolic link points.
1203
:param abspath: The link absolute unicode path.
1205
This his guaranteed to return the symbolic link in unicode in all python
1208
link = abspath.encode(_fs_enc)
1209
target = os.readlink(link)
1210
target = target.decode(_fs_enc)
1214
def contains_whitespace(s):
1215
"""True if there are any whitespace characters in s."""
1216
# string.whitespace can include '\xa0' in certain locales, because it is
1217
# considered "non-breaking-space" as part of ISO-8859-1. But it
1218
# 1) Isn't a breaking whitespace
1219
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1221
# 3) '\xa0' isn't unicode safe since it is >128.
1223
# This should *not* be a unicode set of characters in case the source
1224
# string is not a Unicode string. We can auto-up-cast the characters since
1225
# they are ascii, but we don't want to auto-up-cast the string in case it
1227
for ch in ' \t\n\r\v\f':
1234
def contains_linebreaks(s):
1235
"""True if there is any vertical whitespace in s."""
1243
def relpath(base, path):
1244
"""Return path relative to base, or raise PathNotChild exception.
1246
The path may be either an absolute path or a path relative to the
1247
current working directory.
1249
os.path.commonprefix (python2.4) has a bad bug that it works just
1250
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1251
avoids that problem.
1253
NOTE: `base` should not have a trailing slash otherwise you'll get
1254
PathNotChild exceptions regardless of `path`.
1257
if len(base) < MIN_ABS_PATHLENGTH:
1258
# must have space for e.g. a drive letter
1259
raise ValueError(gettext('%r is too short to calculate a relative path')
1267
if len(head) <= len(base) and head != base:
1268
raise errors.PathNotChild(rp, base)
1271
head, tail = split(head)
1276
return pathjoin(*reversed(s))
1281
def _cicp_canonical_relpath(base, path):
1282
"""Return the canonical path relative to base.
1284
Like relpath, but on case-insensitive-case-preserving file-systems, this
1285
will return the relpath as stored on the file-system rather than in the
1286
case specified in the input string, for all existing portions of the path.
1288
This will cause O(N) behaviour if called for every path in a tree; if you
1289
have a number of paths to convert, you should use canonical_relpaths().
1291
# TODO: it should be possible to optimize this for Windows by using the
1292
# win32 API FindFiles function to look for the specified name - but using
1293
# os.listdir() still gives us the correct, platform agnostic semantics in
1296
rel = relpath(base, path)
1297
# '.' will have been turned into ''
1301
abs_base = abspath(base)
1303
_listdir = os.listdir
1305
# use an explicit iterator so we can easily consume the rest on early exit.
1306
bit_iter = iter(rel.split('/'))
1307
for bit in bit_iter:
1310
next_entries = _listdir(current)
1311
except OSError: # enoent, eperm, etc
1312
# We can't find this in the filesystem, so just append the
1314
current = pathjoin(current, bit, *list(bit_iter))
1316
for look in next_entries:
1317
if lbit == look.lower():
1318
current = pathjoin(current, look)
1321
# got to the end, nothing matched, so we just return the
1322
# non-existing bits as they were specified (the filename may be
1323
# the target of a move, for example).
1324
current = pathjoin(current, bit, *list(bit_iter))
1326
return current[len(abs_base):].lstrip('/')
1328
# XXX - TODO - we need better detection/integration of case-insensitive
1329
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1330
# filesystems), for example, so could probably benefit from the same basic
1331
# support there. For now though, only Windows and OSX get that support, and
1332
# they get it for *all* file-systems!
1333
if sys.platform in ('win32', 'darwin'):
1334
canonical_relpath = _cicp_canonical_relpath
1336
canonical_relpath = relpath
1338
def canonical_relpaths(base, paths):
1339
"""Create an iterable to canonicalize a sequence of relative paths.
1341
The intent is for this implementation to use a cache, vastly speeding
1342
up multiple transformations in the same directory.
1344
# but for now, we haven't optimized...
1345
return [canonical_relpath(base, p) for p in paths]
1348
def decode_filename(filename):
1349
"""Decode the filename using the filesystem encoding
1351
If it is unicode, it is returned.
1352
Otherwise it is decoded from the the filesystem's encoding. If decoding
1353
fails, a errors.BadFilenameEncoding exception is raised.
1355
if type(filename) is unicode:
1358
return filename.decode(_fs_enc)
1359
except UnicodeDecodeError:
1360
raise errors.BadFilenameEncoding(filename, _fs_enc)
1363
def safe_unicode(unicode_or_utf8_string):
1364
"""Coerce unicode_or_utf8_string into unicode.
1366
If it is unicode, it is returned.
1367
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1368
wrapped in a BzrBadParameterNotUnicode exception.
1370
if isinstance(unicode_or_utf8_string, unicode):
1371
return unicode_or_utf8_string
1373
return unicode_or_utf8_string.decode('utf8')
1374
except UnicodeDecodeError:
1375
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1378
def safe_utf8(unicode_or_utf8_string):
1379
"""Coerce unicode_or_utf8_string to a utf8 string.
1381
If it is a str, it is returned.
1382
If it is Unicode, it is encoded into a utf-8 string.
1384
if isinstance(unicode_or_utf8_string, str):
1385
# TODO: jam 20070209 This is overkill, and probably has an impact on
1386
# performance if we are dealing with lots of apis that want a
1389
# Make sure it is a valid utf-8 string
1390
unicode_or_utf8_string.decode('utf-8')
1391
except UnicodeDecodeError:
1392
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1393
return unicode_or_utf8_string
1394
return unicode_or_utf8_string.encode('utf-8')
1397
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1398
' Revision id generators should be creating utf8'
1402
def safe_revision_id(unicode_or_utf8_string, warn=True):
1403
"""Revision ids should now be utf8, but at one point they were unicode.
1405
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1407
:param warn: Functions that are sanitizing user data can set warn=False
1408
:return: None or a utf8 revision id.
1410
if (unicode_or_utf8_string is None
1411
or unicode_or_utf8_string.__class__ == str):
1412
return unicode_or_utf8_string
1414
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1416
return cache_utf8.encode(unicode_or_utf8_string)
1419
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1420
' generators should be creating utf8 file ids.')
1423
def safe_file_id(unicode_or_utf8_string, warn=True):
1424
"""File ids should now be utf8, but at one point they were unicode.
1426
This is the same as safe_utf8, except it uses the cached encode functions
1427
to save a little bit of performance.
1429
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1431
:param warn: Functions that are sanitizing user data can set warn=False
1432
:return: None or a utf8 file id.
1434
if (unicode_or_utf8_string is None
1435
or unicode_or_utf8_string.__class__ == str):
1436
return unicode_or_utf8_string
1438
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1440
return cache_utf8.encode(unicode_or_utf8_string)
1443
_platform_normalizes_filenames = False
1444
if sys.platform == 'darwin':
1445
_platform_normalizes_filenames = True
1448
def normalizes_filenames():
1449
"""Return True if this platform normalizes unicode filenames.
1453
return _platform_normalizes_filenames
1456
def _accessible_normalized_filename(path):
1457
"""Get the unicode normalized path, and if you can access the file.
1459
On platforms where the system normalizes filenames (Mac OSX),
1460
you can access a file by any path which will normalize correctly.
1461
On platforms where the system does not normalize filenames
1462
(everything else), you have to access a file by its exact path.
1464
Internally, bzr only supports NFC normalization, since that is
1465
the standard for XML documents.
1467
So return the normalized path, and a flag indicating if the file
1468
can be accessed by that path.
1471
return unicodedata.normalize('NFC', unicode(path)), True
1474
def _inaccessible_normalized_filename(path):
1475
__doc__ = _accessible_normalized_filename.__doc__
1477
normalized = unicodedata.normalize('NFC', unicode(path))
1478
return normalized, normalized == path
1481
if _platform_normalizes_filenames:
1482
normalized_filename = _accessible_normalized_filename
1484
normalized_filename = _inaccessible_normalized_filename
1487
def set_signal_handler(signum, handler, restart_syscall=True):
1488
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1489
on platforms that support that.
1491
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1492
automatically restart (by calling `signal.siginterrupt(signum,
1493
False)`). May be ignored if the feature is not available on this
1494
platform or Python version.
1498
siginterrupt = signal.siginterrupt
1500
# This python implementation doesn't provide signal support, hence no
1503
except AttributeError:
1504
# siginterrupt doesn't exist on this platform, or for this version
1506
siginterrupt = lambda signum, flag: None
1508
def sig_handler(*args):
1509
# Python resets the siginterrupt flag when a signal is
1510
# received. <http://bugs.python.org/issue8354>
1511
# As a workaround for some cases, set it back the way we want it.
1512
siginterrupt(signum, False)
1513
# Now run the handler function passed to set_signal_handler.
1516
sig_handler = handler
1517
old_handler = signal.signal(signum, sig_handler)
1519
siginterrupt(signum, False)
1523
default_terminal_width = 80
1524
"""The default terminal width for ttys.
1526
This is defined so that higher levels can share a common fallback value when
1527
terminal_width() returns None.
1530
# Keep some state so that terminal_width can detect if _terminal_size has
1531
# returned a different size since the process started. See docstring and
1532
# comments of terminal_width for details.
1533
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1534
_terminal_size_state = 'no_data'
1535
_first_terminal_size = None
1537
def terminal_width():
1538
"""Return terminal width.
1540
None is returned if the width can't established precisely.
1543
- if BZR_COLUMNS is set, returns its value
1544
- if there is no controlling terminal, returns None
1545
- query the OS, if the queried size has changed since the last query,
1547
- if COLUMNS is set, returns its value,
1548
- if the OS has a value (even though it's never changed), return its value.
1550
From there, we need to query the OS to get the size of the controlling
1553
On Unices we query the OS by:
1554
- get termios.TIOCGWINSZ
1555
- if an error occurs or a negative value is obtained, returns None
1557
On Windows we query the OS by:
1558
- win32utils.get_console_size() decides,
1559
- returns None on error (provided default value)
1561
# Note to implementors: if changing the rules for determining the width,
1562
# make sure you've considered the behaviour in these cases:
1563
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1564
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1566
# - (add more interesting cases here, if you find any)
1567
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1568
# but we don't want to register a signal handler because it is impossible
1569
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1570
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1571
# time so we can notice if the reported size has changed, which should have
1574
# If BZR_COLUMNS is set, take it, user is always right
1575
# Except if they specified 0 in which case, impose no limit here
1577
width = int(os.environ['BZR_COLUMNS'])
1578
except (KeyError, ValueError):
1580
if width is not None:
1586
isatty = getattr(sys.stdout, 'isatty', None)
1587
if isatty is None or not isatty():
1588
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1592
width, height = os_size = _terminal_size(None, None)
1593
global _first_terminal_size, _terminal_size_state
1594
if _terminal_size_state == 'no_data':
1595
_first_terminal_size = os_size
1596
_terminal_size_state = 'unchanged'
1597
elif (_terminal_size_state == 'unchanged' and
1598
_first_terminal_size != os_size):
1599
_terminal_size_state = 'changed'
1601
# If the OS claims to know how wide the terminal is, and this value has
1602
# ever changed, use that.
1603
if _terminal_size_state == 'changed':
1604
if width is not None and width > 0:
1607
# If COLUMNS is set, use it.
1609
return int(os.environ['COLUMNS'])
1610
except (KeyError, ValueError):
1613
# Finally, use an unchanged size from the OS, if we have one.
1614
if _terminal_size_state == 'unchanged':
1615
if width is not None and width > 0:
1618
# The width could not be determined.
1622
def _win32_terminal_size(width, height):
1623
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1624
return width, height
1627
def _ioctl_terminal_size(width, height):
1629
import struct, fcntl, termios
1630
s = struct.pack('HHHH', 0, 0, 0, 0)
1631
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1632
height, width = struct.unpack('HHHH', x)[0:2]
1633
except (IOError, AttributeError):
1635
return width, height
1637
_terminal_size = None
1638
"""Returns the terminal size as (width, height).
1640
:param width: Default value for width.
1641
:param height: Default value for height.
1643
This is defined specifically for each OS and query the size of the controlling
1644
terminal. If any error occurs, the provided default values should be returned.
1646
if sys.platform == 'win32':
1647
_terminal_size = _win32_terminal_size
1649
_terminal_size = _ioctl_terminal_size
1652
def supports_executable():
1653
return sys.platform != "win32"
1656
def supports_posix_readonly():
1657
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1659
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1660
directory controls creation/deletion, etc.
1662
And under win32, readonly means that the directory itself cannot be
1663
deleted. The contents of a readonly directory can be changed, unlike POSIX
1664
where files in readonly directories cannot be added, deleted or renamed.
1666
return sys.platform != "win32"
1669
def set_or_unset_env(env_variable, value):
1670
"""Modify the environment, setting or removing the env_variable.
1672
:param env_variable: The environment variable in question
1673
:param value: The value to set the environment to. If None, then
1674
the variable will be removed.
1675
:return: The original value of the environment variable.
1677
orig_val = os.environ.get(env_variable)
1679
if orig_val is not None:
1680
del os.environ[env_variable]
1682
if isinstance(value, unicode):
1683
value = value.encode(get_user_encoding())
1684
os.environ[env_variable] = value
1688
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1691
def check_legal_path(path):
1692
"""Check whether the supplied path is legal.
1693
This is only required on Windows, so we don't test on other platforms
1696
if sys.platform != "win32":
1698
if _validWin32PathRE.match(path) is None:
1699
raise errors.IllegalPath(path)
1702
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1704
def _is_error_enotdir(e):
1705
"""Check if this exception represents ENOTDIR.
1707
Unfortunately, python is very inconsistent about the exception
1708
here. The cases are:
1709
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1710
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1711
which is the windows error code.
1712
3) Windows, Python2.5 uses errno == EINVAL and
1713
winerror == ERROR_DIRECTORY
1715
:param e: An Exception object (expected to be OSError with an errno
1716
attribute, but we should be able to cope with anything)
1717
:return: True if this represents an ENOTDIR error. False otherwise.
1719
en = getattr(e, 'errno', None)
1720
if (en == errno.ENOTDIR
1721
or (sys.platform == 'win32'
1722
and (en == _WIN32_ERROR_DIRECTORY
1723
or (en == errno.EINVAL
1724
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1730
def walkdirs(top, prefix=""):
1731
"""Yield data about all the directories in a tree.
1733
This yields all the data about the contents of a directory at a time.
1734
After each directory has been yielded, if the caller has mutated the list
1735
to exclude some directories, they are then not descended into.
1737
The data yielded is of the form:
1738
((directory-relpath, directory-path-from-top),
1739
[(relpath, basename, kind, lstat, path-from-top), ...]),
1740
- directory-relpath is the relative path of the directory being returned
1741
with respect to top. prefix is prepended to this.
1742
- directory-path-from-root is the path including top for this directory.
1743
It is suitable for use with os functions.
1744
- relpath is the relative path within the subtree being walked.
1745
- basename is the basename of the path
1746
- kind is the kind of the file now. If unknown then the file is not
1747
present within the tree - but it may be recorded as versioned. See
1749
- lstat is the stat data *if* the file was statted.
1750
- planned, not implemented:
1751
path_from_tree_root is the path from the root of the tree.
1753
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1754
allows one to walk a subtree but get paths that are relative to a tree
1756
:return: an iterator over the dirs.
1758
#TODO there is a bit of a smell where the results of the directory-
1759
# summary in this, and the path from the root, may not agree
1760
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1761
# potentially confusing output. We should make this more robust - but
1762
# not at a speed cost. RBC 20060731
1764
_directory = _directory_kind
1765
_listdir = os.listdir
1766
_kind_from_mode = file_kind_from_stat_mode
1767
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1769
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1770
relroot, _, _, _, top = pending.pop()
1772
relprefix = relroot + u'/'
1775
top_slash = top + u'/'
1778
append = dirblock.append
1780
names = sorted(map(decode_filename, _listdir(top)))
1782
if not _is_error_enotdir(e):
1786
abspath = top_slash + name
1787
statvalue = _lstat(abspath)
1788
kind = _kind_from_mode(statvalue.st_mode)
1789
append((relprefix + name, name, kind, statvalue, abspath))
1790
yield (relroot, top), dirblock
1792
# push the user specified dirs from dirblock
1793
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1796
class DirReader(object):
1797
"""An interface for reading directories."""
1799
def top_prefix_to_starting_dir(self, top, prefix=""):
1800
"""Converts top and prefix to a starting dir entry
1802
:param top: A utf8 path
1803
:param prefix: An optional utf8 path to prefix output relative paths
1805
:return: A tuple starting with prefix, and ending with the native
1808
raise NotImplementedError(self.top_prefix_to_starting_dir)
1810
def read_dir(self, prefix, top):
1811
"""Read a specific dir.
1813
:param prefix: A utf8 prefix to be preprended to the path basenames.
1814
:param top: A natively encoded path to read.
1815
:return: A list of the directories contents. Each item contains:
1816
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1818
raise NotImplementedError(self.read_dir)
1821
_selected_dir_reader = None
1824
def _walkdirs_utf8(top, prefix=""):
1825
"""Yield data about all the directories in a tree.
1827
This yields the same information as walkdirs() only each entry is yielded
1828
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1829
are returned as exact byte-strings.
1831
:return: yields a tuple of (dir_info, [file_info])
1832
dir_info is (utf8_relpath, path-from-top)
1833
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1834
if top is an absolute path, path-from-top is also an absolute path.
1835
path-from-top might be unicode or utf8, but it is the correct path to
1836
pass to os functions to affect the file in question. (such as os.lstat)
1838
global _selected_dir_reader
1839
if _selected_dir_reader is None:
1840
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1841
# Win98 doesn't have unicode apis like FindFirstFileW
1842
# TODO: We possibly could support Win98 by falling back to the
1843
# original FindFirstFile, and using TCHAR instead of WCHAR,
1844
# but that gets a bit tricky, and requires custom compiling
1847
from bzrlib._walkdirs_win32 import Win32ReadDir
1848
_selected_dir_reader = Win32ReadDir()
1851
elif _fs_enc in ('utf-8', 'ascii'):
1853
from bzrlib._readdir_pyx import UTF8DirReader
1854
_selected_dir_reader = UTF8DirReader()
1855
except ImportError, e:
1856
failed_to_load_extension(e)
1859
if _selected_dir_reader is None:
1860
# Fallback to the python version
1861
_selected_dir_reader = UnicodeDirReader()
1863
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1864
# But we don't actually uses 1-3 in pending, so set them to None
1865
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1866
read_dir = _selected_dir_reader.read_dir
1867
_directory = _directory_kind
1869
relroot, _, _, _, top = pending[-1].pop()
1872
dirblock = sorted(read_dir(relroot, top))
1873
yield (relroot, top), dirblock
1874
# push the user specified dirs from dirblock
1875
next = [d for d in reversed(dirblock) if d[2] == _directory]
1877
pending.append(next)
1880
class UnicodeDirReader(DirReader):
1881
"""A dir reader for non-utf8 file systems, which transcodes."""
1883
__slots__ = ['_utf8_encode']
1886
self._utf8_encode = codecs.getencoder('utf8')
1888
def top_prefix_to_starting_dir(self, top, prefix=""):
1889
"""See DirReader.top_prefix_to_starting_dir."""
1890
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1892
def read_dir(self, prefix, top):
1893
"""Read a single directory from a non-utf8 file system.
1895
top, and the abspath element in the output are unicode, all other paths
1896
are utf8. Local disk IO is done via unicode calls to listdir etc.
1898
This is currently the fallback code path when the filesystem encoding is
1899
not UTF-8. It may be better to implement an alternative so that we can
1900
safely handle paths that are not properly decodable in the current
1903
See DirReader.read_dir for details.
1905
_utf8_encode = self._utf8_encode
1907
_listdir = os.listdir
1908
_kind_from_mode = file_kind_from_stat_mode
1911
relprefix = prefix + '/'
1914
top_slash = top + u'/'
1917
append = dirblock.append
1918
for name in sorted(_listdir(top)):
1920
name_utf8 = _utf8_encode(name)[0]
1921
except UnicodeDecodeError:
1922
raise errors.BadFilenameEncoding(
1923
_utf8_encode(relprefix)[0] + name, _fs_enc)
1924
abspath = top_slash + name
1925
statvalue = _lstat(abspath)
1926
kind = _kind_from_mode(statvalue.st_mode)
1927
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1931
def copy_tree(from_path, to_path, handlers={}):
1932
"""Copy all of the entries in from_path into to_path.
1934
:param from_path: The base directory to copy.
1935
:param to_path: The target directory. If it does not exist, it will
1937
:param handlers: A dictionary of functions, which takes a source and
1938
destinations for files, directories, etc.
1939
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1940
'file', 'directory', and 'symlink' should always exist.
1941
If they are missing, they will be replaced with 'os.mkdir()',
1942
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1944
# Now, just copy the existing cached tree to the new location
1945
# We use a cheap trick here.
1946
# Absolute paths are prefixed with the first parameter
1947
# relative paths are prefixed with the second.
1948
# So we can get both the source and target returned
1949
# without any extra work.
1951
def copy_dir(source, dest):
1954
def copy_link(source, dest):
1955
"""Copy the contents of a symlink"""
1956
link_to = os.readlink(source)
1957
os.symlink(link_to, dest)
1959
real_handlers = {'file':shutil.copy2,
1960
'symlink':copy_link,
1961
'directory':copy_dir,
1963
real_handlers.update(handlers)
1965
if not os.path.exists(to_path):
1966
real_handlers['directory'](from_path, to_path)
1968
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1969
for relpath, name, kind, st, abspath in entries:
1970
real_handlers[kind](abspath, relpath)
1973
def copy_ownership_from_path(dst, src=None):
1974
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1976
If src is None, the containing directory is used as source. If chown
1977
fails, the error is ignored and a warning is printed.
1979
chown = getattr(os, 'chown', None)
1984
src = os.path.dirname(dst)
1990
chown(dst, s.st_uid, s.st_gid)
1993
'Unable to copy ownership from "%s" to "%s". '
1994
'You may want to set it manually.', src, dst)
1995
trace.log_exception_quietly()
1998
def path_prefix_key(path):
1999
"""Generate a prefix-order path key for path.
2001
This can be used to sort paths in the same way that walkdirs does.
2003
return (dirname(path) , path)
2006
def compare_paths_prefix_order(path_a, path_b):
2007
"""Compare path_a and path_b to generate the same order walkdirs uses."""
2008
key_a = path_prefix_key(path_a)
2009
key_b = path_prefix_key(path_b)
2010
return cmp(key_a, key_b)
2013
_cached_user_encoding = None
2016
def get_user_encoding(use_cache=DEPRECATED_PARAMETER):
2017
"""Find out what the preferred user encoding is.
2019
This is generally the encoding that is used for command line parameters
2020
and file contents. This may be different from the terminal encoding
2021
or the filesystem encoding.
2023
:return: A string defining the preferred user encoding
2025
global _cached_user_encoding
2026
if deprecated_passed(use_cache):
2027
warn_deprecated("use_cache should only have been used for tests",
2028
DeprecationWarning, stacklevel=2)
2029
if _cached_user_encoding is not None:
2030
return _cached_user_encoding
2032
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2033
# Use the existing locale settings and call nl_langinfo directly
2034
# rather than going through getpreferredencoding. This avoids
2035
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2036
# possibility of the setlocale call throwing an error.
2037
user_encoding = locale.nl_langinfo(locale.CODESET)
2039
# GZ 2011-12-19: On windows could call GetACP directly instead.
2040
user_encoding = locale.getpreferredencoding(False)
2043
user_encoding = codecs.lookup(user_encoding).name
2045
if user_encoding not in ("", "cp0"):
2046
sys.stderr.write('bzr: warning:'
2047
' unknown encoding %s.'
2048
' Continuing with ascii encoding.\n'
2051
user_encoding = 'ascii'
2053
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2054
if user_encoding == 'ascii':
2055
if sys.platform == 'darwin':
2056
# OSX is special-cased in Python to have a UTF-8 filesystem
2057
# encoding and previously had LANG set here if not present.
2058
user_encoding = 'utf-8'
2059
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2060
# for some other posix platforms as well.
2062
_cached_user_encoding = user_encoding
2063
return user_encoding
2066
def get_diff_header_encoding():
2067
return get_terminal_encoding()
2070
def get_host_name():
2071
"""Return the current unicode host name.
2073
This is meant to be used in place of socket.gethostname() because that
2074
behaves inconsistently on different platforms.
2076
if sys.platform == "win32":
2077
return win32utils.get_host_name()
2080
return socket.gethostname().decode(get_user_encoding())
2083
# We must not read/write any more than 64k at a time from/to a socket so we
2084
# don't risk "no buffer space available" errors on some platforms. Windows in
2085
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2087
MAX_SOCKET_CHUNK = 64 * 1024
2089
_end_of_stream_errors = [errno.ECONNRESET]
2090
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2091
_eno = getattr(errno, _eno, None)
2092
if _eno is not None:
2093
_end_of_stream_errors.append(_eno)
2097
def read_bytes_from_socket(sock, report_activity=None,
2098
max_read_size=MAX_SOCKET_CHUNK):
2099
"""Read up to max_read_size of bytes from sock and notify of progress.
2101
Translates "Connection reset by peer" into file-like EOF (return an
2102
empty string rather than raise an error), and repeats the recv if
2103
interrupted by a signal.
2107
bytes = sock.recv(max_read_size)
2108
except socket.error, e:
2110
if eno in _end_of_stream_errors:
2111
# The connection was closed by the other side. Callers expect
2112
# an empty string to signal end-of-stream.
2114
elif eno == errno.EINTR:
2115
# Retry the interrupted recv.
2119
if report_activity is not None:
2120
report_activity(len(bytes), 'read')
2124
def recv_all(socket, count):
2125
"""Receive an exact number of bytes.
2127
Regular Socket.recv() may return less than the requested number of bytes,
2128
depending on what's in the OS buffer. MSG_WAITALL is not available
2129
on all platforms, but this should work everywhere. This will return
2130
less than the requested amount if the remote end closes.
2132
This isn't optimized and is intended mostly for use in testing.
2135
while len(b) < count:
2136
new = read_bytes_from_socket(socket, None, count - len(b))
2143
def send_all(sock, bytes, report_activity=None):
2144
"""Send all bytes on a socket.
2146
Breaks large blocks in smaller chunks to avoid buffering limitations on
2147
some platforms, and catches EINTR which may be thrown if the send is
2148
interrupted by a signal.
2150
This is preferred to socket.sendall(), because it avoids portability bugs
2151
and provides activity reporting.
2153
:param report_activity: Call this as bytes are read, see
2154
Transport._report_activity
2157
byte_count = len(bytes)
2158
while sent_total < byte_count:
2160
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2161
except socket.error, e:
2162
if e.args[0] != errno.EINTR:
2166
report_activity(sent, 'write')
2169
def connect_socket(address):
2170
# Slight variation of the socket.create_connection() function (provided by
2171
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2172
# provide it for previous python versions. Also, we don't use the timeout
2173
# parameter (provided by the python implementation) so we don't implement
2175
err = socket.error('getaddrinfo returns an empty list')
2176
host, port = address
2177
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2178
af, socktype, proto, canonname, sa = res
2181
sock = socket.socket(af, socktype, proto)
2185
except socket.error, err:
2186
# 'err' is now the most recent error
2187
if sock is not None:
2192
def dereference_path(path):
2193
"""Determine the real path to a file.
2195
All parent elements are dereferenced. But the file itself is not
2197
:param path: The original path. May be absolute or relative.
2198
:return: the real path *to* the file
2200
parent, base = os.path.split(path)
2201
# The pathjoin for '.' is a workaround for Python bug #1213894.
2202
# (initial path components aren't dereferenced)
2203
return pathjoin(realpath(pathjoin('.', parent)), base)
2206
def supports_mapi():
2207
"""Return True if we can use MAPI to launch a mail client."""
2208
return sys.platform == "win32"
2211
def resource_string(package, resource_name):
2212
"""Load a resource from a package and return it as a string.
2214
Note: Only packages that start with bzrlib are currently supported.
2216
This is designed to be a lightweight implementation of resource
2217
loading in a way which is API compatible with the same API from
2219
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2220
If and when pkg_resources becomes a standard library, this routine
2223
# Check package name is within bzrlib
2224
if package == "bzrlib":
2225
resource_relpath = resource_name
2226
elif package.startswith("bzrlib."):
2227
package = package[len("bzrlib."):].replace('.', os.sep)
2228
resource_relpath = pathjoin(package, resource_name)
2230
raise errors.BzrError('resource package %s not in bzrlib' % package)
2232
# Map the resource to a file and read its contents
2233
base = dirname(bzrlib.__file__)
2234
if getattr(sys, 'frozen', None): # bzr.exe
2235
base = abspath(pathjoin(base, '..', '..'))
2236
f = file(pathjoin(base, resource_relpath), "rU")
2242
def file_kind_from_stat_mode_thunk(mode):
2243
global file_kind_from_stat_mode
2244
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2246
from bzrlib._readdir_pyx import UTF8DirReader
2247
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2248
except ImportError, e:
2249
# This is one time where we won't warn that an extension failed to
2250
# load. The extension is never available on Windows anyway.
2251
from bzrlib._readdir_py import (
2252
_kind_from_mode as file_kind_from_stat_mode
2254
return file_kind_from_stat_mode(mode)
2255
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2257
def file_stat(f, _lstat=os.lstat):
2262
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2263
raise errors.NoSuchFile(f)
2266
def file_kind(f, _lstat=os.lstat):
2267
stat_value = file_stat(f, _lstat)
2268
return file_kind_from_stat_mode(stat_value.st_mode)
2270
def until_no_eintr(f, *a, **kw):
2271
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2273
WARNING: you must be certain that it is safe to retry the call repeatedly
2274
if EINTR does occur. This is typically only true for low-level operations
2275
like os.read. If in any doubt, don't use this.
2277
Keep in mind that this is not a complete solution to EINTR. There is
2278
probably code in the Python standard library and other dependencies that
2279
may encounter EINTR if a signal arrives (and there is signal handler for
2280
that signal). So this function can reduce the impact for IO that bzrlib
2281
directly controls, but it is not a complete solution.
2283
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2287
except (IOError, OSError), e:
2288
if e.errno == errno.EINTR:
2293
@deprecated_function(deprecated_in((2, 2, 0)))
2294
def re_compile_checked(re_string, flags=0, where=""):
2295
"""Return a compiled re, or raise a sensible error.
2297
This should only be used when compiling user-supplied REs.
2299
:param re_string: Text form of regular expression.
2300
:param flags: eg re.IGNORECASE
2301
:param where: Message explaining to the user the context where
2302
it occurred, eg 'log search filter'.
2304
# from https://bugs.launchpad.net/bzr/+bug/251352
2306
re_obj = re.compile(re_string, flags)
2309
except errors.InvalidPattern, e:
2311
where = ' in ' + where
2312
# despite the name 'error' is a type
2313
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2317
if sys.platform == "win32":
2320
return msvcrt.getch()
2325
fd = sys.stdin.fileno()
2326
settings = termios.tcgetattr(fd)
2329
ch = sys.stdin.read(1)
2331
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2334
if sys.platform.startswith('linux'):
2335
def _local_concurrency():
2337
return os.sysconf('SC_NPROCESSORS_ONLN')
2338
except (ValueError, OSError, AttributeError):
2340
elif sys.platform == 'darwin':
2341
def _local_concurrency():
2342
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2343
stdout=subprocess.PIPE).communicate()[0]
2344
elif "bsd" in sys.platform:
2345
def _local_concurrency():
2346
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2347
stdout=subprocess.PIPE).communicate()[0]
2348
elif sys.platform == 'sunos5':
2349
def _local_concurrency():
2350
return subprocess.Popen(['psrinfo', '-p',],
2351
stdout=subprocess.PIPE).communicate()[0]
2352
elif sys.platform == "win32":
2353
def _local_concurrency():
2354
# This appears to return the number of cores.
2355
return os.environ.get('NUMBER_OF_PROCESSORS')
2357
def _local_concurrency():
2362
_cached_local_concurrency = None
2364
def local_concurrency(use_cache=True):
2365
"""Return how many processes can be run concurrently.
2367
Rely on platform specific implementations and default to 1 (one) if
2368
anything goes wrong.
2370
global _cached_local_concurrency
2372
if _cached_local_concurrency is not None and use_cache:
2373
return _cached_local_concurrency
2375
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2376
if concurrency is None:
2378
import multiprocessing
2379
concurrency = multiprocessing.cpu_count()
2380
except (ImportError, NotImplementedError):
2381
# multiprocessing is only available on Python >= 2.6
2382
# and multiprocessing.cpu_count() isn't implemented on all
2385
concurrency = _local_concurrency()
2386
except (OSError, IOError):
2389
concurrency = int(concurrency)
2390
except (TypeError, ValueError):
2393
_cached_concurrency = concurrency
2397
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2398
"""A stream writer that doesn't decode str arguments."""
2400
def __init__(self, encode, stream, errors='strict'):
2401
codecs.StreamWriter.__init__(self, stream, errors)
2402
self.encode = encode
2404
def write(self, object):
2405
if type(object) is str:
2406
self.stream.write(object)
2408
data, _ = self.encode(object, self.errors)
2409
self.stream.write(data)
2411
if sys.platform == 'win32':
2412
def open_file(filename, mode='r', bufsize=-1):
2413
"""This function is used to override the ``open`` builtin.
2415
But it uses O_NOINHERIT flag so the file handle is not inherited by
2416
child processes. Deleting or renaming a closed file opened with this
2417
function is not blocking child processes.
2419
writing = 'w' in mode
2420
appending = 'a' in mode
2421
updating = '+' in mode
2422
binary = 'b' in mode
2425
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2426
# for flags for each modes.
2436
flags |= os.O_WRONLY
2437
flags |= os.O_CREAT | os.O_TRUNC
2442
flags |= os.O_WRONLY
2443
flags |= os.O_CREAT | os.O_APPEND
2448
flags |= os.O_RDONLY
2450
return os.fdopen(os.open(filename, flags), mode, bufsize)
2455
def available_backup_name(base, exists):
2456
"""Find a non-existing backup file name.
2458
This will *not* create anything, this only return a 'free' entry. This
2459
should be used for checking names in a directory below a locked
2460
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2461
Leap) and generally discouraged.
2463
:param base: The base name.
2465
:param exists: A callable returning True if the path parameter exists.
2468
name = "%s.~%d~" % (base, counter)
2471
name = "%s.~%d~" % (base, counter)
2475
def set_fd_cloexec(fd):
2476
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2477
support for this is not available.
2481
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2482
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2483
except (ImportError, AttributeError):
2484
# Either the fcntl module or specific constants are not present
2488
def find_executable_on_path(name):
2489
"""Finds an executable on the PATH.
2491
On Windows, this will try to append each extension in the PATHEXT
2492
environment variable to the name, if it cannot be found with the name
2495
:param name: The base name of the executable.
2496
:return: The path to the executable found or None.
2498
if sys.platform == 'win32':
2499
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2500
exts = [ext.lower() for ext in exts]
2501
base, ext = os.path.splitext(name)
2503
if ext.lower() not in exts:
2509
path = os.environ.get('PATH')
2510
if path is not None:
2511
path = path.split(os.pathsep)
2514
f = os.path.join(d, name) + ext
2515
if os.access(f, os.X_OK):
2517
if sys.platform == 'win32':
2518
app_path = win32utils.get_app_path(name)
2519
if app_path != name:
2524
def _posix_is_local_pid_dead(pid):
2525
"""True if pid doesn't correspond to live process on this machine"""
2527
# Special meaning of unix kill: just check if it's there.
2530
if e.errno == errno.ESRCH:
2531
# On this machine, and really not found: as sure as we can be
2534
elif e.errno == errno.EPERM:
2535
# exists, though not ours
2538
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2539
# Don't really know.
2542
# Exists and our process: not dead.
2545
if sys.platform == "win32":
2546
is_local_pid_dead = win32utils.is_local_pid_dead
2548
is_local_pid_dead = _posix_is_local_pid_dead
2551
def fdatasync(fileno):
2552
"""Flush file contents to disk if possible.
2554
:param fileno: Integer OS file handle.
2555
:raises TransportNotPossible: If flushing to disk is not possible.
2557
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2562
def ensure_empty_directory_exists(path, exception_class):
2563
"""Make sure a local directory exists and is empty.
2565
If it does not exist, it is created. If it exists and is not empty, an
2566
instance of exception_class is raised.
2571
if e.errno != errno.EEXIST:
2573
if os.listdir(path) != []:
2574
raise exception_class(path)
2577
def is_environment_error(evalue):
2578
"""True if exception instance is due to a process environment issue
2580
This includes OSError and IOError, but also other errors that come from
2581
the operating system or core libraries but are not subclasses of those.
2583
if isinstance(evalue, (EnvironmentError, select.error)):
2585
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
added
removed
bzrlib/osutils.py
