1
# Copyright (C) 2005-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from __future__ import absolute_import
27
from .lazy_import import lazy_import
28
lazy_import(globals(), """
29
from datetime import datetime
30
from datetime import timedelta
36
# We need to import both shutil and rmtree as we export the later on posix
37
# and need the former on windows
39
from shutil import rmtree
42
# We need to import both tempfile and mkdtemp as we export the later on posix
43
# and need the former on windows
45
from tempfile import mkdtemp
54
from breezy.i18n import gettext
72
# Cross platform wall-clock time functionality with decent resolution.
73
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
74
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
75
# synchronized with ``time.time()``, this is only meant to be used to find
76
# delta times by subtracting from another call to this function.
77
timer_func = time.time
78
if sys.platform == 'win32':
79
timer_func = time.clock
81
# On win32, O_BINARY is used to indicate the file should
82
# be opened in binary mode, rather than text mode.
83
# On other platforms, O_BINARY doesn't exist, because
84
# they always open in binary mode, so it is okay to
85
# OR with 0 on those platforms.
86
# O_NOINHERIT and O_TEXT exists only on win32 too.
87
O_BINARY = getattr(os, 'O_BINARY', 0)
88
O_TEXT = getattr(os, 'O_TEXT', 0)
89
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
92
def get_unicode_argv():
96
user_encoding = get_user_encoding()
97
return [a.decode(user_encoding) for a in sys.argv[1:]]
98
except UnicodeDecodeError:
99
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
100
"application locale.").format(a, user_encoding))
103
def make_readonly(filename):
104
"""Make a filename read-only."""
105
mod = os.lstat(filename).st_mode
106
if not stat.S_ISLNK(mod):
108
chmod_if_possible(filename, mod)
111
def make_writable(filename):
112
mod = os.lstat(filename).st_mode
113
if not stat.S_ISLNK(mod):
115
chmod_if_possible(filename, mod)
118
def chmod_if_possible(filename, mode):
119
# Set file mode if that can be safely done.
120
# Sometimes even on unix the filesystem won't allow it - see
121
# https://bugs.launchpad.net/bzr/+bug/606537
123
# It is probably faster to just do the chmod, rather than
124
# doing a stat, and then trying to compare
125
os.chmod(filename, mode)
126
except (IOError, OSError) as e:
127
# Permission/access denied seems to commonly happen on smbfs; there's
128
# probably no point warning about it.
129
# <https://bugs.launchpad.net/bzr/+bug/606537>
130
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
131
trace.mutter("ignore error on chmod of %r: %r" % (
137
def minimum_path_selection(paths):
138
"""Return the smallset subset of paths which are outside paths.
140
:param paths: A container (and hence not None) of paths.
141
:return: A set of paths sufficient to include everything in paths via
142
is_inside, drawn from the paths parameter.
148
return path.split('/')
149
sorted_paths = sorted(list(paths), key=sort_key)
151
search_paths = [sorted_paths[0]]
152
for path in sorted_paths[1:]:
153
if not is_inside(search_paths[-1], path):
154
# This path is unique, add it
155
search_paths.append(path)
157
return set(search_paths)
164
"""Return a quoted filename filename
166
This previously used backslash quoting, but that works poorly on
168
# TODO: I'm not really sure this is the best format either.x
170
if _QUOTE_RE is None:
171
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
173
if _QUOTE_RE.search(f):
179
_directory_kind = 'directory'
182
"""Return the current umask"""
183
# Assume that people aren't messing with the umask while running
184
# XXX: This is not thread safe, but there is no way to get the
185
# umask without setting it
193
_directory_kind: "/",
195
'tree-reference': '+',
199
def kind_marker(kind):
201
return _kind_marker_map[kind]
203
# Slightly faster than using .get(, '') when the common case is that
208
lexists = getattr(os.path, 'lexists', None)
212
stat = getattr(os, 'lstat', os.stat)
216
if e.errno == errno.ENOENT:
219
raise errors.BzrError(gettext("lstat/stat of ({0!r}): {1!r}").format(f, e))
222
def fancy_rename(old, new, rename_func, unlink_func):
223
"""A fancy rename, when you don't have atomic rename.
225
:param old: The old path, to rename from
226
:param new: The new path, to rename to
227
:param rename_func: The potentially non-atomic rename function
228
:param unlink_func: A way to delete the target file if the full rename
231
# sftp rename doesn't allow overwriting, so play tricks:
232
base = os.path.basename(new)
233
dirname = os.path.dirname(new)
234
# callers use different encodings for the paths so the following MUST
235
# respect that. We rely on python upcasting to unicode if new is unicode
236
# and keeping a str if not.
237
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
238
os.getpid(), rand_chars(10))
239
tmp_name = pathjoin(dirname, tmp_name)
241
# Rename the file out of the way, but keep track if it didn't exist
242
# We don't want to grab just any exception
243
# something like EACCES should prevent us from continuing
244
# The downside is that the rename_func has to throw an exception
245
# with an errno = ENOENT, or NoSuchFile
248
rename_func(new, tmp_name)
249
except (errors.NoSuchFile,) as e:
252
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
253
# function raises an IOError with errno is None when a rename fails.
254
# This then gets caught here.
255
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
257
except Exception as e:
258
if (getattr(e, 'errno', None) is None
259
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
266
# This may throw an exception, in which case success will
268
rename_func(old, new)
270
except (IOError, OSError) as e:
271
# source and target may be aliases of each other (e.g. on a
272
# case-insensitive filesystem), so we may have accidentally renamed
273
# source by when we tried to rename target
274
if (file_existed and e.errno in (None, errno.ENOENT)
275
and old.lower() == new.lower()):
276
# source and target are the same file on a case-insensitive
277
# filesystem, so we don't generate an exception
283
# If the file used to exist, rename it back into place
284
# otherwise just delete it from the tmp location
286
unlink_func(tmp_name)
288
rename_func(tmp_name, new)
291
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
292
# choke on a Unicode string containing a relative path if
293
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
295
def _posix_abspath(path):
296
# jam 20060426 rather than encoding to fsencoding
297
# copy posixpath.abspath, but use os.getcwdu instead
298
if not posixpath.isabs(path):
299
path = posixpath.join(getcwd(), path)
300
return _posix_normpath(path)
303
def _posix_realpath(path):
304
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
307
def _posix_normpath(path):
308
path = posixpath.normpath(path)
309
# Bug 861008: posixpath.normpath() returns a path normalized according to
310
# the POSIX standard, which stipulates (for compatibility reasons) that two
311
# leading slashes must not be simplified to one, and only if there are 3 or
312
# more should they be simplified as one. So we treat the leading 2 slashes
313
# as a special case here by simply removing the first slash, as we consider
314
# that breaking POSIX compatibility for this obscure feature is acceptable.
315
# This is not a paranoid precaution, as we notably get paths like this when
316
# the repo is hosted at the root of the filesystem, i.e. in "/".
317
if path.startswith('//'):
322
def _posix_path_from_environ(key):
323
"""Get unicode path from `key` in environment or None if not present
325
Note that posix systems use arbitrary byte strings for filesystem objects,
326
so a path that raises BadFilenameEncoding here may still be accessible.
328
val = os.environ.get(key, None)
329
if PY3 or val is None:
332
return val.decode(_fs_enc)
333
except UnicodeDecodeError:
334
# GZ 2011-12-12:Ideally want to include `key` in the exception message
335
raise errors.BadFilenameEncoding(val, _fs_enc)
338
def _posix_get_home_dir():
339
"""Get the home directory of the current user as a unicode path"""
340
path = posixpath.expanduser("~")
342
return path.decode(_fs_enc)
343
except AttributeError:
345
except UnicodeDecodeError:
346
raise errors.BadFilenameEncoding(path, _fs_enc)
349
def _posix_getuser_unicode():
350
"""Get username from environment or password database as unicode"""
351
name = getpass.getuser()
354
user_encoding = get_user_encoding()
356
return name.decode(user_encoding)
357
except UnicodeDecodeError:
358
raise errors.BzrError("Encoding of username %r is unsupported by %s "
359
"application locale." % (name, user_encoding))
362
def _win32_fixdrive(path):
363
"""Force drive letters to be consistent.
365
win32 is inconsistent whether it returns lower or upper case
366
and even if it was consistent the user might type the other
367
so we force it to uppercase
368
running python.exe under cmd.exe return capital C:\\
369
running win32 python inside a cygwin shell returns lowercase c:\\
371
drive, path = ntpath.splitdrive(path)
372
return drive.upper() + path
375
def _win32_abspath(path):
376
# Real ntpath.abspath doesn't have a problem with a unicode cwd
377
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
380
def _win32_realpath(path):
381
# Real ntpath.realpath doesn't have a problem with a unicode cwd
382
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
385
def _win32_pathjoin(*args):
386
return ntpath.join(*args).replace('\\', '/')
389
def _win32_normpath(path):
390
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
394
return _win32_fixdrive(_getcwd().replace('\\', '/'))
397
def _win32_mkdtemp(*args, **kwargs):
398
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
401
def _win32_rename(old, new):
402
"""We expect to be able to atomically replace 'new' with old.
404
On win32, if new exists, it must be moved out of the way first,
408
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
410
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
411
# If we try to rename a non-existant file onto cwd, we get
412
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
413
# if the old path doesn't exist, sometimes we get EACCES
414
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
420
return unicodedata.normalize('NFC', _getcwd())
423
def _rename_wrap_exception(rename_func):
424
"""Adds extra information to any exceptions that come from rename().
426
The exception has an updated message and 'old_filename' and 'new_filename'
430
def _rename_wrapper(old, new):
432
rename_func(old, new)
434
detailed_error = OSError(e.errno, e.strerror +
435
" [occurred when renaming '%s' to '%s']" %
437
detailed_error.old_filename = old
438
detailed_error.new_filename = new
441
return _rename_wrapper
444
if sys.version_info > (3,):
450
# Default rename wraps os.rename()
451
rename = _rename_wrap_exception(os.rename)
453
# Default is to just use the python builtins, but these can be rebound on
454
# particular platforms.
455
abspath = _posix_abspath
456
realpath = _posix_realpath
457
pathjoin = os.path.join
458
normpath = _posix_normpath
459
path_from_environ = _posix_path_from_environ
460
_get_home_dir = _posix_get_home_dir
461
getuser_unicode = _posix_getuser_unicode
463
dirname = os.path.dirname
464
basename = os.path.basename
465
split = os.path.split
466
splitext = os.path.splitext
467
# These were already lazily imported into local scope
468
# mkdtemp = tempfile.mkdtemp
469
# rmtree = shutil.rmtree
477
MIN_ABS_PATHLENGTH = 1
480
if sys.platform == 'win32':
481
abspath = _win32_abspath
482
realpath = _win32_realpath
483
pathjoin = _win32_pathjoin
484
normpath = _win32_normpath
485
getcwd = _win32_getcwd
486
mkdtemp = _win32_mkdtemp
487
rename = _rename_wrap_exception(_win32_rename)
489
from . import _walkdirs_win32
493
lstat = _walkdirs_win32.lstat
494
fstat = _walkdirs_win32.fstat
495
wrap_stat = _walkdirs_win32.wrap_stat
497
MIN_ABS_PATHLENGTH = 3
499
def _win32_delete_readonly(function, path, excinfo):
500
"""Error handler for shutil.rmtree function [for win32]
501
Helps to remove files and dirs marked as read-only.
503
exception = excinfo[1]
504
if function in (os.remove, os.rmdir) \
505
and isinstance(exception, OSError) \
506
and exception.errno == errno.EACCES:
512
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
513
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
514
return shutil.rmtree(path, ignore_errors, onerror)
516
f = win32utils.get_unicode_argv # special function or None
519
path_from_environ = win32utils.get_environ_unicode
520
_get_home_dir = win32utils.get_home_location
521
getuser_unicode = win32utils.get_user_name
523
elif sys.platform == 'darwin':
527
def get_terminal_encoding(trace=False):
528
"""Find the best encoding for printing to the screen.
530
This attempts to check both sys.stdout and sys.stdin to see
531
what encoding they are in, and if that fails it falls back to
532
osutils.get_user_encoding().
533
The problem is that on Windows, locale.getpreferredencoding()
534
is not the same encoding as that used by the console:
535
http://mail.python.org/pipermail/python-list/2003-May/162357.html
537
On my standard US Windows XP, the preferred encoding is
538
cp1252, but the console is cp437
540
:param trace: If True trace the selected encoding via mutter().
542
from .trace import mutter
543
output_encoding = getattr(sys.stdout, 'encoding', None)
544
if not output_encoding:
545
input_encoding = getattr(sys.stdin, 'encoding', None)
546
if not input_encoding:
547
output_encoding = get_user_encoding()
549
mutter('encoding stdout as osutils.get_user_encoding() %r',
552
output_encoding = input_encoding
554
mutter('encoding stdout as sys.stdin encoding %r',
558
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
559
if output_encoding == 'cp0':
560
# invalid encoding (cp0 means 'no codepage' on Windows)
561
output_encoding = get_user_encoding()
563
mutter('cp0 is invalid encoding.'
564
' encoding stdout as osutils.get_user_encoding() %r',
568
codecs.lookup(output_encoding)
570
sys.stderr.write('brz: warning:'
571
' unknown terminal encoding %s.\n'
572
' Using encoding %s instead.\n'
573
% (output_encoding, get_user_encoding())
575
output_encoding = get_user_encoding()
577
return output_encoding
580
def normalizepath(f):
581
if getattr(os.path, 'realpath', None) is not None:
585
[p,e] = os.path.split(f)
586
if e == "" or e == "." or e == "..":
589
return pathjoin(F(p), e)
593
"""True if f is an accessible directory."""
595
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
601
"""True if f is a regular file."""
603
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
608
"""True if f is a symlink."""
610
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
614
def is_inside(dir, fname):
615
"""True if fname is inside dir.
617
The parameters should typically be passed to osutils.normpath first, so
618
that . and .. and repeated slashes are eliminated, and the separators
619
are canonical for the platform.
621
The empty string as a dir name is taken as top-of-tree and matches
624
# XXX: Most callers of this can actually do something smarter by
625
# looking at the inventory
635
return fname.startswith(dir)
638
def is_inside_any(dir_list, fname):
639
"""True if fname is inside any of given dirs."""
640
for dirname in dir_list:
641
if is_inside(dirname, fname):
646
def is_inside_or_parent_of_any(dir_list, fname):
647
"""True if fname is a child or a parent of any of the given files."""
648
for dirname in dir_list:
649
if is_inside(dirname, fname) or is_inside(fname, dirname):
654
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
655
report_activity=None, direction='read'):
656
"""Copy contents of one file to another.
658
The read_length can either be -1 to read to end-of-file (EOF) or
659
it can specify the maximum number of bytes to read.
661
The buff_size represents the maximum size for each read operation
662
performed on from_file.
664
:param report_activity: Call this as bytes are read, see
665
Transport._report_activity
666
:param direction: Will be passed to report_activity
668
:return: The number of bytes copied.
672
# read specified number of bytes
674
while read_length > 0:
675
num_bytes_to_read = min(read_length, buff_size)
677
block = from_file.read(num_bytes_to_read)
681
if report_activity is not None:
682
report_activity(len(block), direction)
685
actual_bytes_read = len(block)
686
read_length -= actual_bytes_read
687
length += actual_bytes_read
691
block = from_file.read(buff_size)
695
if report_activity is not None:
696
report_activity(len(block), direction)
702
def pump_string_file(bytes, file_handle, segment_size=None):
703
"""Write bytes to file_handle in many smaller writes.
705
:param bytes: The string to write.
706
:param file_handle: The file to write to.
708
# Write data in chunks rather than all at once, because very large
709
# writes fail on some platforms (e.g. Windows with SMB mounted
712
segment_size = 5242880 # 5MB
713
offsets = range(0, len(bytes), segment_size)
714
view = memoryview(bytes)
715
write = file_handle.write
716
for offset in offsets:
717
write(view[offset:offset+segment_size])
720
def file_iterator(input_file, readsize=32768):
722
b = input_file.read(readsize)
729
"""Calculate the hexdigest of an open file.
731
The file cursor should be already at the start.
743
def size_sha_file(f):
744
"""Calculate the size and hexdigest of an open file.
746
The file cursor should be already at the start and
747
the caller is responsible for closing the file afterwards.
758
return size, s.hexdigest()
761
def sha_file_by_name(fname):
762
"""Calculate the SHA1 of a file by reading the full text"""
764
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
767
b = os.read(f, 1<<16)
775
def sha_strings(strings, _factory=sha):
776
"""Return the sha-1 of concatenation of strings"""
778
for string in strings:
783
def sha_string(f, _factory=sha):
784
return _factory(f).hexdigest()
787
def fingerprint_file(f):
789
return {'size': len(b),
790
'sha1': sha(b).hexdigest()}
793
def compare_files(a, b):
794
"""Returns true if equal in contents"""
805
def gmtime(seconds=None):
806
"""Convert seconds since the Epoch to a time tuple expressing UTC (a.k.a.
807
GMT). When 'seconds' is not passed in, convert the current time instead.
808
Handy replacement for time.gmtime() buggy on Windows and 32-bit platforms.
811
seconds = time.time()
812
return (datetime(1970, 1, 1) + timedelta(seconds=seconds)).timetuple()
815
def local_time_offset(t=None):
816
"""Return offset of local zone from GMT, either at present or at time t."""
819
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
820
return offset.days * 86400 + offset.seconds
822
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
823
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
826
def format_date(t, offset=0, timezone='original', date_fmt=None,
828
"""Return a formatted date string.
830
:param t: Seconds since the epoch.
831
:param offset: Timezone offset in seconds east of utc.
832
:param timezone: How to display the time: 'utc', 'original' for the
833
timezone specified by offset, or 'local' for the process's current
835
:param date_fmt: strftime format.
836
:param show_offset: Whether to append the timezone.
838
(date_fmt, tt, offset_str) = \
839
_format_date(t, offset, timezone, date_fmt, show_offset)
840
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
841
date_str = time.strftime(date_fmt, tt)
842
return date_str + offset_str
845
# Cache of formatted offset strings
849
def format_date_with_offset_in_original_timezone(t, offset=0,
850
_cache=_offset_cache):
851
"""Return a formatted date string in the original timezone.
853
This routine may be faster then format_date.
855
:param t: Seconds since the epoch.
856
:param offset: Timezone offset in seconds east of utc.
860
tt = gmtime(t + offset)
861
date_fmt = _default_format_by_weekday_num[tt[6]]
862
date_str = time.strftime(date_fmt, tt)
863
offset_str = _cache.get(offset, None)
864
if offset_str is None:
865
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
866
_cache[offset] = offset_str
867
return date_str + offset_str
870
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
872
"""Return an unicode date string formatted according to the current locale.
874
:param t: Seconds since the epoch.
875
:param offset: Timezone offset in seconds east of utc.
876
:param timezone: How to display the time: 'utc', 'original' for the
877
timezone specified by offset, or 'local' for the process's current
879
:param date_fmt: strftime format.
880
:param show_offset: Whether to append the timezone.
882
(date_fmt, tt, offset_str) = \
883
_format_date(t, offset, timezone, date_fmt, show_offset)
884
date_str = time.strftime(date_fmt, tt)
885
if not isinstance(date_str, text_type):
886
date_str = date_str.decode(get_user_encoding(), 'replace')
887
return date_str + offset_str
890
def _format_date(t, offset, timezone, date_fmt, show_offset):
891
if timezone == 'utc':
894
elif timezone == 'original':
897
tt = gmtime(t + offset)
898
elif timezone == 'local':
899
tt = time.localtime(t)
900
offset = local_time_offset(t)
902
raise errors.UnsupportedTimezoneFormat(timezone)
904
date_fmt = "%a %Y-%m-%d %H:%M:%S"
906
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
909
return (date_fmt, tt, offset_str)
912
def compact_date(when):
913
return time.strftime('%Y%m%d%H%M%S', gmtime(when))
916
def format_delta(delta):
917
"""Get a nice looking string for a time delta.
919
:param delta: The time difference in seconds, can be positive or negative.
920
positive indicates time in the past, negative indicates time in the
921
future. (usually time.time() - stored_time)
922
:return: String formatted to show approximate resolution
928
direction = 'in the future'
932
if seconds < 90: # print seconds up to 90 seconds
934
return '%d second %s' % (seconds, direction,)
936
return '%d seconds %s' % (seconds, direction)
938
minutes = int(seconds / 60)
939
seconds -= 60 * minutes
944
if minutes < 90: # print minutes, seconds up to 90 minutes
946
return '%d minute, %d second%s %s' % (
947
minutes, seconds, plural_seconds, direction)
949
return '%d minutes, %d second%s %s' % (
950
minutes, seconds, plural_seconds, direction)
952
hours = int(minutes / 60)
953
minutes -= 60 * hours
960
return '%d hour, %d minute%s %s' % (hours, minutes,
961
plural_minutes, direction)
962
return '%d hours, %d minute%s %s' % (hours, minutes,
963
plural_minutes, direction)
966
"""Return size of given open file."""
967
return os.fstat(f.fileno())[stat.ST_SIZE]
970
# Alias os.urandom to support platforms (which?) without /dev/urandom and
971
# override if it doesn't work. Avoid checking on windows where there is
972
# significant initialisation cost that can be avoided for some bzr calls.
974
rand_bytes = os.urandom
976
if rand_bytes.__module__ != "nt":
979
except NotImplementedError:
980
# not well seeded, but better than nothing
985
s += chr(random.randint(0, 255))
990
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
992
"""Return a random string of num alphanumeric characters
994
The result only contains lowercase chars because it may be used on
995
case-insensitive filesystems.
998
for raw_byte in rand_bytes(num):
1000
s += ALNUM[ord(raw_byte) % 36]
1002
s += ALNUM[raw_byte % 36]
1006
## TODO: We could later have path objects that remember their list
1007
## decomposition (might be too tricksy though.)
1010
"""Turn string into list of parts."""
1011
# split on either delimiter because people might use either on
1013
ps = re.split(r'[\\/]', p)
1018
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1019
elif (f == '.') or (f == ''):
1028
if (f == '..') or (f is None) or (f == ''):
1029
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1033
def parent_directories(filename):
1034
"""Return the list of parent directories, deepest first.
1036
For example, parent_directories("a/b/c") -> ["a/b", "a"].
1039
parts = splitpath(dirname(filename))
1041
parents.append(joinpath(parts))
1046
_extension_load_failures = []
1049
def failed_to_load_extension(exception):
1050
"""Handle failing to load a binary extension.
1052
This should be called from the ImportError block guarding the attempt to
1053
import the native extension. If this function returns, the pure-Python
1054
implementation should be loaded instead::
1057
>>> import breezy._fictional_extension_pyx
1058
>>> except ImportError, e:
1059
>>> breezy.osutils.failed_to_load_extension(e)
1060
>>> import breezy._fictional_extension_py
1062
# NB: This docstring is just an example, not a doctest, because doctest
1063
# currently can't cope with the use of lazy imports in this namespace --
1066
# This currently doesn't report the failure at the time it occurs, because
1067
# they tend to happen very early in startup when we can't check config
1068
# files etc, and also we want to report all failures but not spam the user
1070
exception_str = str(exception)
1071
if exception_str not in _extension_load_failures:
1072
trace.mutter("failed to load compiled extension: %s" % exception_str)
1073
_extension_load_failures.append(exception_str)
1076
def report_extension_load_failures():
1077
if not _extension_load_failures:
1079
if config.GlobalStack().get('ignore_missing_extensions'):
1081
# the warnings framework should by default show this only once
1082
from .trace import warning
1084
"brz: warning: some compiled extensions could not be loaded; "
1085
"see <https://answers.launchpad.net/bzr/+faq/703>")
1086
# we no longer show the specific missing extensions here, because it makes
1087
# the message too long and scary - see
1088
# https://bugs.launchpad.net/bzr/+bug/430529
1092
from ._chunks_to_lines_pyx import chunks_to_lines
1093
except ImportError as e:
1094
failed_to_load_extension(e)
1095
from ._chunks_to_lines_py import chunks_to_lines
1099
"""Split s into lines, but without removing the newline characters."""
1100
# Trivially convert a fulltext into a 'chunked' representation, and let
1101
# chunks_to_lines do the heavy lifting.
1102
if isinstance(s, str):
1103
# chunks_to_lines only supports 8-bit strings
1104
return chunks_to_lines([s])
1106
return _split_lines(s)
1109
def _split_lines(s):
1110
"""Split s into lines, but without removing the newline characters.
1112
This supports Unicode or plain string objects.
1114
nl = b'\n' if isinstance(s, bytes) else u'\n'
1116
result = [line + nl for line in lines[:-1]]
1118
result.append(lines[-1])
1122
def hardlinks_good():
1123
return sys.platform not in ('win32', 'cygwin', 'darwin')
1126
def link_or_copy(src, dest):
1127
"""Hardlink a file, or copy it if it can't be hardlinked."""
1128
if not hardlinks_good():
1129
shutil.copyfile(src, dest)
1133
except (OSError, IOError) as e:
1134
if e.errno != errno.EXDEV:
1136
shutil.copyfile(src, dest)
1139
def delete_any(path):
1140
"""Delete a file, symlink or directory.
1142
Will delete even if readonly.
1145
_delete_file_or_dir(path)
1146
except (OSError, IOError) as e:
1147
if e.errno in (errno.EPERM, errno.EACCES):
1148
# make writable and try again
1151
except (OSError, IOError):
1153
_delete_file_or_dir(path)
1158
def _delete_file_or_dir(path):
1159
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1160
# Forgiveness than Permission (EAFP) because:
1161
# - root can damage a solaris file system by using unlink,
1162
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1163
# EACCES, OSX: EPERM) when invoked on a directory.
1164
if isdir(path): # Takes care of symlinks
1171
if getattr(os, 'symlink', None) is not None:
1177
def has_hardlinks():
1178
if getattr(os, 'link', None) is not None:
1184
def host_os_dereferences_symlinks():
1185
return (has_symlinks()
1186
and sys.platform not in ('cygwin', 'win32'))
1189
def readlink(abspath):
1190
"""Return a string representing the path to which the symbolic link points.
1192
:param abspath: The link absolute unicode path.
1194
This his guaranteed to return the symbolic link in unicode in all python
1197
link = abspath.encode(_fs_enc)
1198
target = os.readlink(link)
1199
target = target.decode(_fs_enc)
1203
def contains_whitespace(s):
1204
"""True if there are any whitespace characters in s."""
1205
# string.whitespace can include '\xa0' in certain locales, because it is
1206
# considered "non-breaking-space" as part of ISO-8859-1. But it
1207
# 1) Isn't a breaking whitespace
1208
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1210
# 3) '\xa0' isn't unicode safe since it is >128.
1212
# This should *not* be a unicode set of characters in case the source
1213
# string is not a Unicode string. We can auto-up-cast the characters since
1214
# they are ascii, but we don't want to auto-up-cast the string in case it
1216
for ch in ' \t\n\r\v\f':
1223
def contains_linebreaks(s):
1224
"""True if there is any vertical whitespace in s."""
1232
def relpath(base, path):
1233
"""Return path relative to base, or raise PathNotChild exception.
1235
The path may be either an absolute path or a path relative to the
1236
current working directory.
1238
os.path.commonprefix (python2.4) has a bad bug that it works just
1239
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1240
avoids that problem.
1242
NOTE: `base` should not have a trailing slash otherwise you'll get
1243
PathNotChild exceptions regardless of `path`.
1246
if len(base) < MIN_ABS_PATHLENGTH:
1247
# must have space for e.g. a drive letter
1248
raise ValueError(gettext('%r is too short to calculate a relative path')
1256
if len(head) <= len(base) and head != base:
1257
raise errors.PathNotChild(rp, base)
1260
head, tail = split(head)
1265
return pathjoin(*reversed(s))
1270
def _cicp_canonical_relpath(base, path):
1271
"""Return the canonical path relative to base.
1273
Like relpath, but on case-insensitive-case-preserving file-systems, this
1274
will return the relpath as stored on the file-system rather than in the
1275
case specified in the input string, for all existing portions of the path.
1277
This will cause O(N) behaviour if called for every path in a tree; if you
1278
have a number of paths to convert, you should use canonical_relpaths().
1280
# TODO: it should be possible to optimize this for Windows by using the
1281
# win32 API FindFiles function to look for the specified name - but using
1282
# os.listdir() still gives us the correct, platform agnostic semantics in
1285
rel = relpath(base, path)
1286
# '.' will have been turned into ''
1290
abs_base = abspath(base)
1292
_listdir = os.listdir
1294
# use an explicit iterator so we can easily consume the rest on early exit.
1295
bit_iter = iter(rel.split('/'))
1296
for bit in bit_iter:
1299
next_entries = _listdir(current)
1300
except OSError: # enoent, eperm, etc
1301
# We can't find this in the filesystem, so just append the
1303
current = pathjoin(current, bit, *list(bit_iter))
1305
for look in next_entries:
1306
if lbit == look.lower():
1307
current = pathjoin(current, look)
1310
# got to the end, nothing matched, so we just return the
1311
# non-existing bits as they were specified (the filename may be
1312
# the target of a move, for example).
1313
current = pathjoin(current, bit, *list(bit_iter))
1315
return current[len(abs_base):].lstrip('/')
1317
# XXX - TODO - we need better detection/integration of case-insensitive
1318
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1319
# filesystems), for example, so could probably benefit from the same basic
1320
# support there. For now though, only Windows and OSX get that support, and
1321
# they get it for *all* file-systems!
1322
if sys.platform in ('win32', 'darwin'):
1323
canonical_relpath = _cicp_canonical_relpath
1325
canonical_relpath = relpath
1327
def canonical_relpaths(base, paths):
1328
"""Create an iterable to canonicalize a sequence of relative paths.
1330
The intent is for this implementation to use a cache, vastly speeding
1331
up multiple transformations in the same directory.
1333
# but for now, we haven't optimized...
1334
return [canonical_relpath(base, p) for p in paths]
1337
def decode_filename(filename):
1338
"""Decode the filename using the filesystem encoding
1340
If it is unicode, it is returned.
1341
Otherwise it is decoded from the the filesystem's encoding. If decoding
1342
fails, a errors.BadFilenameEncoding exception is raised.
1344
if isinstance(filename, text_type):
1347
return filename.decode(_fs_enc)
1348
except UnicodeDecodeError:
1349
raise errors.BadFilenameEncoding(filename, _fs_enc)
1352
def safe_unicode(unicode_or_utf8_string):
1353
"""Coerce unicode_or_utf8_string into unicode.
1355
If it is unicode, it is returned.
1356
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1357
wrapped in a BzrBadParameterNotUnicode exception.
1359
if isinstance(unicode_or_utf8_string, text_type):
1360
return unicode_or_utf8_string
1362
return unicode_or_utf8_string.decode('utf8')
1363
except UnicodeDecodeError:
1364
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1367
def safe_utf8(unicode_or_utf8_string):
1368
"""Coerce unicode_or_utf8_string to a utf8 string.
1370
If it is a str, it is returned.
1371
If it is Unicode, it is encoded into a utf-8 string.
1373
if isinstance(unicode_or_utf8_string, str):
1374
# TODO: jam 20070209 This is overkill, and probably has an impact on
1375
# performance if we are dealing with lots of apis that want a
1378
# Make sure it is a valid utf-8 string
1379
unicode_or_utf8_string.decode('utf-8')
1380
except UnicodeDecodeError:
1381
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1382
return unicode_or_utf8_string
1383
return unicode_or_utf8_string.encode('utf-8')
1386
def safe_revision_id(unicode_or_utf8_string):
1387
"""Revision ids should now be utf8, but at one point they were unicode.
1389
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1391
:return: None or a utf8 revision id.
1393
if (unicode_or_utf8_string is None
1394
or unicode_or_utf8_string.__class__ == str):
1395
return unicode_or_utf8_string
1396
raise TypeError('Unicode revision ids are no longer supported. '
1397
'Revision id generators should be creating utf8 revision '
1401
def safe_file_id(unicode_or_utf8_string):
1402
"""File ids should now be utf8, but at one point they were unicode.
1404
This is the same as safe_utf8, except it uses the cached encode functions
1405
to save a little bit of performance.
1407
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1409
:return: None or a utf8 file id.
1411
if (unicode_or_utf8_string is None
1412
or unicode_or_utf8_string.__class__ == str):
1413
return unicode_or_utf8_string
1414
raise TypeError('Unicode file ids are no longer supported. '
1415
'File id generators should be creating utf8 file ids.')
1418
_platform_normalizes_filenames = False
1419
if sys.platform == 'darwin':
1420
_platform_normalizes_filenames = True
1423
def normalizes_filenames():
1424
"""Return True if this platform normalizes unicode filenames.
1428
return _platform_normalizes_filenames
1431
def _accessible_normalized_filename(path):
1432
"""Get the unicode normalized path, and if you can access the file.
1434
On platforms where the system normalizes filenames (Mac OSX),
1435
you can access a file by any path which will normalize correctly.
1436
On platforms where the system does not normalize filenames
1437
(everything else), you have to access a file by its exact path.
1439
Internally, bzr only supports NFC normalization, since that is
1440
the standard for XML documents.
1442
So return the normalized path, and a flag indicating if the file
1443
can be accessed by that path.
1446
return unicodedata.normalize('NFC', unicode(path)), True
1449
def _inaccessible_normalized_filename(path):
1450
__doc__ = _accessible_normalized_filename.__doc__
1452
normalized = unicodedata.normalize('NFC', unicode(path))
1453
return normalized, normalized == path
1456
if _platform_normalizes_filenames:
1457
normalized_filename = _accessible_normalized_filename
1459
normalized_filename = _inaccessible_normalized_filename
1462
def set_signal_handler(signum, handler, restart_syscall=True):
1463
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1464
on platforms that support that.
1466
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1467
automatically restart (by calling `signal.siginterrupt(signum,
1468
False)`). May be ignored if the feature is not available on this
1469
platform or Python version.
1473
siginterrupt = signal.siginterrupt
1475
# This python implementation doesn't provide signal support, hence no
1478
except AttributeError:
1479
# siginterrupt doesn't exist on this platform, or for this version
1481
siginterrupt = lambda signum, flag: None
1483
def sig_handler(*args):
1484
# Python resets the siginterrupt flag when a signal is
1485
# received. <http://bugs.python.org/issue8354>
1486
# As a workaround for some cases, set it back the way we want it.
1487
siginterrupt(signum, False)
1488
# Now run the handler function passed to set_signal_handler.
1491
sig_handler = handler
1492
old_handler = signal.signal(signum, sig_handler)
1494
siginterrupt(signum, False)
1498
default_terminal_width = 80
1499
"""The default terminal width for ttys.
1501
This is defined so that higher levels can share a common fallback value when
1502
terminal_width() returns None.
1505
# Keep some state so that terminal_width can detect if _terminal_size has
1506
# returned a different size since the process started. See docstring and
1507
# comments of terminal_width for details.
1508
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1509
_terminal_size_state = 'no_data'
1510
_first_terminal_size = None
1512
def terminal_width():
1513
"""Return terminal width.
1515
None is returned if the width can't established precisely.
1518
- if BRZ_COLUMNS is set, returns its value
1519
- if there is no controlling terminal, returns None
1520
- query the OS, if the queried size has changed since the last query,
1522
- if COLUMNS is set, returns its value,
1523
- if the OS has a value (even though it's never changed), return its value.
1525
From there, we need to query the OS to get the size of the controlling
1528
On Unices we query the OS by:
1529
- get termios.TIOCGWINSZ
1530
- if an error occurs or a negative value is obtained, returns None
1532
On Windows we query the OS by:
1533
- win32utils.get_console_size() decides,
1534
- returns None on error (provided default value)
1536
# Note to implementors: if changing the rules for determining the width,
1537
# make sure you've considered the behaviour in these cases:
1538
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1539
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1541
# - (add more interesting cases here, if you find any)
1542
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1543
# but we don't want to register a signal handler because it is impossible
1544
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1545
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1546
# time so we can notice if the reported size has changed, which should have
1549
# If BRZ_COLUMNS is set, take it, user is always right
1550
# Except if they specified 0 in which case, impose no limit here
1552
width = int(os.environ['BRZ_COLUMNS'])
1553
except (KeyError, ValueError):
1555
if width is not None:
1561
isatty = getattr(sys.stdout, 'isatty', None)
1562
if isatty is None or not isatty():
1563
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
1567
width, height = os_size = _terminal_size(None, None)
1568
global _first_terminal_size, _terminal_size_state
1569
if _terminal_size_state == 'no_data':
1570
_first_terminal_size = os_size
1571
_terminal_size_state = 'unchanged'
1572
elif (_terminal_size_state == 'unchanged' and
1573
_first_terminal_size != os_size):
1574
_terminal_size_state = 'changed'
1576
# If the OS claims to know how wide the terminal is, and this value has
1577
# ever changed, use that.
1578
if _terminal_size_state == 'changed':
1579
if width is not None and width > 0:
1582
# If COLUMNS is set, use it.
1584
return int(os.environ['COLUMNS'])
1585
except (KeyError, ValueError):
1588
# Finally, use an unchanged size from the OS, if we have one.
1589
if _terminal_size_state == 'unchanged':
1590
if width is not None and width > 0:
1593
# The width could not be determined.
1597
def _win32_terminal_size(width, height):
1598
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1599
return width, height
1602
def _ioctl_terminal_size(width, height):
1604
import struct, fcntl, termios
1605
s = struct.pack('HHHH', 0, 0, 0, 0)
1606
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1607
height, width = struct.unpack('HHHH', x)[0:2]
1608
except (IOError, AttributeError):
1610
return width, height
1612
_terminal_size = None
1613
"""Returns the terminal size as (width, height).
1615
:param width: Default value for width.
1616
:param height: Default value for height.
1618
This is defined specifically for each OS and query the size of the controlling
1619
terminal. If any error occurs, the provided default values should be returned.
1621
if sys.platform == 'win32':
1622
_terminal_size = _win32_terminal_size
1624
_terminal_size = _ioctl_terminal_size
1627
def supports_executable():
1628
return sys.platform != "win32"
1631
def supports_posix_readonly():
1632
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1634
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1635
directory controls creation/deletion, etc.
1637
And under win32, readonly means that the directory itself cannot be
1638
deleted. The contents of a readonly directory can be changed, unlike POSIX
1639
where files in readonly directories cannot be added, deleted or renamed.
1641
return sys.platform != "win32"
1644
def set_or_unset_env(env_variable, value):
1645
"""Modify the environment, setting or removing the env_variable.
1647
:param env_variable: The environment variable in question
1648
:param value: The value to set the environment to. If None, then
1649
the variable will be removed.
1650
:return: The original value of the environment variable.
1652
orig_val = os.environ.get(env_variable)
1654
if orig_val is not None:
1655
del os.environ[env_variable]
1657
if not PY3 and isinstance(value, text_type):
1658
value = value.encode(get_user_encoding())
1659
os.environ[env_variable] = value
1663
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1666
def check_legal_path(path):
1667
"""Check whether the supplied path is legal.
1668
This is only required on Windows, so we don't test on other platforms
1671
if sys.platform != "win32":
1673
if _validWin32PathRE.match(path) is None:
1674
raise errors.IllegalPath(path)
1677
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1679
def _is_error_enotdir(e):
1680
"""Check if this exception represents ENOTDIR.
1682
Unfortunately, python is very inconsistent about the exception
1683
here. The cases are:
1684
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1685
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1686
which is the windows error code.
1687
3) Windows, Python2.5 uses errno == EINVAL and
1688
winerror == ERROR_DIRECTORY
1690
:param e: An Exception object (expected to be OSError with an errno
1691
attribute, but we should be able to cope with anything)
1692
:return: True if this represents an ENOTDIR error. False otherwise.
1694
en = getattr(e, 'errno', None)
1695
if (en == errno.ENOTDIR
1696
or (sys.platform == 'win32'
1697
and (en == _WIN32_ERROR_DIRECTORY
1698
or (en == errno.EINVAL
1699
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1705
def walkdirs(top, prefix=""):
1706
"""Yield data about all the directories in a tree.
1708
This yields all the data about the contents of a directory at a time.
1709
After each directory has been yielded, if the caller has mutated the list
1710
to exclude some directories, they are then not descended into.
1712
The data yielded is of the form:
1713
((directory-relpath, directory-path-from-top),
1714
[(relpath, basename, kind, lstat, path-from-top), ...]),
1715
- directory-relpath is the relative path of the directory being returned
1716
with respect to top. prefix is prepended to this.
1717
- directory-path-from-root is the path including top for this directory.
1718
It is suitable for use with os functions.
1719
- relpath is the relative path within the subtree being walked.
1720
- basename is the basename of the path
1721
- kind is the kind of the file now. If unknown then the file is not
1722
present within the tree - but it may be recorded as versioned. See
1724
- lstat is the stat data *if* the file was statted.
1725
- planned, not implemented:
1726
path_from_tree_root is the path from the root of the tree.
1728
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1729
allows one to walk a subtree but get paths that are relative to a tree
1731
:return: an iterator over the dirs.
1733
#TODO there is a bit of a smell where the results of the directory-
1734
# summary in this, and the path from the root, may not agree
1735
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1736
# potentially confusing output. We should make this more robust - but
1737
# not at a speed cost. RBC 20060731
1739
_directory = _directory_kind
1740
_listdir = os.listdir
1741
_kind_from_mode = file_kind_from_stat_mode
1742
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1744
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1745
relroot, _, _, _, top = pending.pop()
1747
relprefix = relroot + u'/'
1750
top_slash = top + u'/'
1753
append = dirblock.append
1755
names = sorted(map(decode_filename, _listdir(top)))
1756
except OSError as e:
1757
if not _is_error_enotdir(e):
1761
abspath = top_slash + name
1762
statvalue = _lstat(abspath)
1763
kind = _kind_from_mode(statvalue.st_mode)
1764
append((relprefix + name, name, kind, statvalue, abspath))
1765
yield (relroot, top), dirblock
1767
# push the user specified dirs from dirblock
1768
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1771
class DirReader(object):
1772
"""An interface for reading directories."""
1774
def top_prefix_to_starting_dir(self, top, prefix=""):
1775
"""Converts top and prefix to a starting dir entry
1777
:param top: A utf8 path
1778
:param prefix: An optional utf8 path to prefix output relative paths
1780
:return: A tuple starting with prefix, and ending with the native
1783
raise NotImplementedError(self.top_prefix_to_starting_dir)
1785
def read_dir(self, prefix, top):
1786
"""Read a specific dir.
1788
:param prefix: A utf8 prefix to be preprended to the path basenames.
1789
:param top: A natively encoded path to read.
1790
:return: A list of the directories contents. Each item contains:
1791
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1793
raise NotImplementedError(self.read_dir)
1796
_selected_dir_reader = None
1799
def _walkdirs_utf8(top, prefix=""):
1800
"""Yield data about all the directories in a tree.
1802
This yields the same information as walkdirs() only each entry is yielded
1803
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1804
are returned as exact byte-strings.
1806
:return: yields a tuple of (dir_info, [file_info])
1807
dir_info is (utf8_relpath, path-from-top)
1808
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1809
if top is an absolute path, path-from-top is also an absolute path.
1810
path-from-top might be unicode or utf8, but it is the correct path to
1811
pass to os functions to affect the file in question. (such as os.lstat)
1813
global _selected_dir_reader
1814
if _selected_dir_reader is None:
1815
if sys.platform == "win32":
1817
from ._walkdirs_win32 import Win32ReadDir
1818
_selected_dir_reader = Win32ReadDir()
1821
elif _fs_enc in ('utf-8', 'ascii'):
1823
from ._readdir_pyx import UTF8DirReader
1824
_selected_dir_reader = UTF8DirReader()
1825
except ImportError as e:
1826
failed_to_load_extension(e)
1829
if _selected_dir_reader is None:
1830
# Fallback to the python version
1831
_selected_dir_reader = UnicodeDirReader()
1833
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1834
# But we don't actually uses 1-3 in pending, so set them to None
1835
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1836
read_dir = _selected_dir_reader.read_dir
1837
_directory = _directory_kind
1839
relroot, _, _, _, top = pending[-1].pop()
1842
dirblock = sorted(read_dir(relroot, top))
1843
yield (relroot, top), dirblock
1844
# push the user specified dirs from dirblock
1845
next = [d for d in reversed(dirblock) if d[2] == _directory]
1847
pending.append(next)
1850
class UnicodeDirReader(DirReader):
1851
"""A dir reader for non-utf8 file systems, which transcodes."""
1853
__slots__ = ['_utf8_encode']
1856
self._utf8_encode = codecs.getencoder('utf8')
1858
def top_prefix_to_starting_dir(self, top, prefix=""):
1859
"""See DirReader.top_prefix_to_starting_dir."""
1860
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1862
def read_dir(self, prefix, top):
1863
"""Read a single directory from a non-utf8 file system.
1865
top, and the abspath element in the output are unicode, all other paths
1866
are utf8. Local disk IO is done via unicode calls to listdir etc.
1868
This is currently the fallback code path when the filesystem encoding is
1869
not UTF-8. It may be better to implement an alternative so that we can
1870
safely handle paths that are not properly decodable in the current
1873
See DirReader.read_dir for details.
1875
_utf8_encode = self._utf8_encode
1877
_listdir = os.listdir
1878
_kind_from_mode = file_kind_from_stat_mode
1881
relprefix = prefix + '/'
1884
top_slash = top + u'/'
1887
append = dirblock.append
1888
for name in sorted(_listdir(top)):
1890
name_utf8 = _utf8_encode(name)[0]
1891
except UnicodeDecodeError:
1892
raise errors.BadFilenameEncoding(
1893
_utf8_encode(relprefix)[0] + name, _fs_enc)
1894
abspath = top_slash + name
1895
statvalue = _lstat(abspath)
1896
kind = _kind_from_mode(statvalue.st_mode)
1897
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1901
def copy_tree(from_path, to_path, handlers={}):
1902
"""Copy all of the entries in from_path into to_path.
1904
:param from_path: The base directory to copy.
1905
:param to_path: The target directory. If it does not exist, it will
1907
:param handlers: A dictionary of functions, which takes a source and
1908
destinations for files, directories, etc.
1909
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1910
'file', 'directory', and 'symlink' should always exist.
1911
If they are missing, they will be replaced with 'os.mkdir()',
1912
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1914
# Now, just copy the existing cached tree to the new location
1915
# We use a cheap trick here.
1916
# Absolute paths are prefixed with the first parameter
1917
# relative paths are prefixed with the second.
1918
# So we can get both the source and target returned
1919
# without any extra work.
1921
def copy_dir(source, dest):
1924
def copy_link(source, dest):
1925
"""Copy the contents of a symlink"""
1926
link_to = os.readlink(source)
1927
os.symlink(link_to, dest)
1929
real_handlers = {'file':shutil.copy2,
1930
'symlink':copy_link,
1931
'directory':copy_dir,
1933
real_handlers.update(handlers)
1935
if not os.path.exists(to_path):
1936
real_handlers['directory'](from_path, to_path)
1938
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1939
for relpath, name, kind, st, abspath in entries:
1940
real_handlers[kind](abspath, relpath)
1943
def copy_ownership_from_path(dst, src=None):
1944
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1946
If src is None, the containing directory is used as source. If chown
1947
fails, the error is ignored and a warning is printed.
1949
chown = getattr(os, 'chown', None)
1954
src = os.path.dirname(dst)
1960
chown(dst, s.st_uid, s.st_gid)
1961
except OSError as e:
1963
'Unable to copy ownership from "%s" to "%s". '
1964
'You may want to set it manually.', src, dst)
1965
trace.log_exception_quietly()
1968
def path_prefix_key(path):
1969
"""Generate a prefix-order path key for path.
1971
This can be used to sort paths in the same way that walkdirs does.
1973
return (dirname(path) , path)
1976
def compare_paths_prefix_order(path_a, path_b):
1977
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1978
key_a = path_prefix_key(path_a)
1979
key_b = path_prefix_key(path_b)
1980
return cmp(key_a, key_b)
1983
_cached_user_encoding = None
1986
def get_user_encoding():
1987
"""Find out what the preferred user encoding is.
1989
This is generally the encoding that is used for command line parameters
1990
and file contents. This may be different from the terminal encoding
1991
or the filesystem encoding.
1993
:return: A string defining the preferred user encoding
1995
global _cached_user_encoding
1996
if _cached_user_encoding is not None:
1997
return _cached_user_encoding
1999
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2000
# Use the existing locale settings and call nl_langinfo directly
2001
# rather than going through getpreferredencoding. This avoids
2002
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2003
# possibility of the setlocale call throwing an error.
2004
user_encoding = locale.nl_langinfo(locale.CODESET)
2006
# GZ 2011-12-19: On windows could call GetACP directly instead.
2007
user_encoding = locale.getpreferredencoding(False)
2010
user_encoding = codecs.lookup(user_encoding).name
2012
if user_encoding not in ("", "cp0"):
2013
sys.stderr.write('brz: warning:'
2014
' unknown encoding %s.'
2015
' Continuing with ascii encoding.\n'
2018
user_encoding = 'ascii'
2020
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2021
if user_encoding == 'ascii':
2022
if sys.platform == 'darwin':
2023
# OSX is special-cased in Python to have a UTF-8 filesystem
2024
# encoding and previously had LANG set here if not present.
2025
user_encoding = 'utf-8'
2026
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2027
# for some other posix platforms as well.
2029
_cached_user_encoding = user_encoding
2030
return user_encoding
2033
def get_diff_header_encoding():
2034
return get_terminal_encoding()
2037
def get_host_name():
2038
"""Return the current unicode host name.
2040
This is meant to be used in place of socket.gethostname() because that
2041
behaves inconsistently on different platforms.
2043
if sys.platform == "win32":
2044
return win32utils.get_host_name()
2048
return socket.gethostname()
2049
return socket.gethostname().decode(get_user_encoding())
2052
# We must not read/write any more than 64k at a time from/to a socket so we
2053
# don't risk "no buffer space available" errors on some platforms. Windows in
2054
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2056
MAX_SOCKET_CHUNK = 64 * 1024
2058
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2059
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2060
_eno = getattr(errno, _eno, None)
2061
if _eno is not None:
2062
_end_of_stream_errors.append(_eno)
2066
def read_bytes_from_socket(sock, report_activity=None,
2067
max_read_size=MAX_SOCKET_CHUNK):
2068
"""Read up to max_read_size of bytes from sock and notify of progress.
2070
Translates "Connection reset by peer" into file-like EOF (return an
2071
empty string rather than raise an error), and repeats the recv if
2072
interrupted by a signal.
2076
bytes = sock.recv(max_read_size)
2077
except socket.error as e:
2079
if eno in _end_of_stream_errors:
2080
# The connection was closed by the other side. Callers expect
2081
# an empty string to signal end-of-stream.
2083
elif eno == errno.EINTR:
2084
# Retry the interrupted recv.
2088
if report_activity is not None:
2089
report_activity(len(bytes), 'read')
2093
def recv_all(socket, count):
2094
"""Receive an exact number of bytes.
2096
Regular Socket.recv() may return less than the requested number of bytes,
2097
depending on what's in the OS buffer. MSG_WAITALL is not available
2098
on all platforms, but this should work everywhere. This will return
2099
less than the requested amount if the remote end closes.
2101
This isn't optimized and is intended mostly for use in testing.
2104
while len(b) < count:
2105
new = read_bytes_from_socket(socket, None, count - len(b))
2112
def send_all(sock, bytes, report_activity=None):
2113
"""Send all bytes on a socket.
2115
Breaks large blocks in smaller chunks to avoid buffering limitations on
2116
some platforms, and catches EINTR which may be thrown if the send is
2117
interrupted by a signal.
2119
This is preferred to socket.sendall(), because it avoids portability bugs
2120
and provides activity reporting.
2122
:param report_activity: Call this as bytes are read, see
2123
Transport._report_activity
2126
byte_count = len(bytes)
2127
view = memoryview(bytes)
2128
while sent_total < byte_count:
2130
sent = sock.send(view[sent_total:sent_total+MAX_SOCKET_CHUNK])
2131
except (socket.error, IOError) as e:
2132
if e.args[0] in _end_of_stream_errors:
2133
raise errors.ConnectionReset(
2134
"Error trying to write to socket", e)
2135
if e.args[0] != errno.EINTR:
2139
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2142
if report_activity is not None:
2143
report_activity(sent, 'write')
2146
def connect_socket(address):
2147
# Slight variation of the socket.create_connection() function (provided by
2148
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2149
# provide it for previous python versions. Also, we don't use the timeout
2150
# parameter (provided by the python implementation) so we don't implement
2152
err = socket.error('getaddrinfo returns an empty list')
2153
host, port = address
2154
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2155
af, socktype, proto, canonname, sa = res
2158
sock = socket.socket(af, socktype, proto)
2162
except socket.error as err:
2163
# 'err' is now the most recent error
2164
if sock is not None:
2169
def dereference_path(path):
2170
"""Determine the real path to a file.
2172
All parent elements are dereferenced. But the file itself is not
2174
:param path: The original path. May be absolute or relative.
2175
:return: the real path *to* the file
2177
parent, base = os.path.split(path)
2178
# The pathjoin for '.' is a workaround for Python bug #1213894.
2179
# (initial path components aren't dereferenced)
2180
return pathjoin(realpath(pathjoin('.', parent)), base)
2183
def supports_mapi():
2184
"""Return True if we can use MAPI to launch a mail client."""
2185
return sys.platform == "win32"
2188
def resource_string(package, resource_name):
2189
"""Load a resource from a package and return it as a string.
2191
Note: Only packages that start with breezy are currently supported.
2193
This is designed to be a lightweight implementation of resource
2194
loading in a way which is API compatible with the same API from
2196
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2197
If and when pkg_resources becomes a standard library, this routine
2200
# Check package name is within breezy
2201
if package == "breezy":
2202
resource_relpath = resource_name
2203
elif package.startswith("breezy."):
2204
package = package[len("breezy."):].replace('.', os.sep)
2205
resource_relpath = pathjoin(package, resource_name)
2207
raise errors.BzrError('resource package %s not in breezy' % package)
2209
# Map the resource to a file and read its contents
2210
base = dirname(breezy.__file__)
2211
if getattr(sys, 'frozen', None): # bzr.exe
2212
base = abspath(pathjoin(base, '..', '..'))
2213
f = file(pathjoin(base, resource_relpath), "rU")
2219
def file_kind_from_stat_mode_thunk(mode):
2220
global file_kind_from_stat_mode
2221
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2223
from ._readdir_pyx import UTF8DirReader
2224
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2225
except ImportError as e:
2226
# This is one time where we won't warn that an extension failed to
2227
# load. The extension is never available on Windows anyway.
2228
from ._readdir_py import (
2229
_kind_from_mode as file_kind_from_stat_mode
2231
return file_kind_from_stat_mode(mode)
2232
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2234
def file_stat(f, _lstat=os.lstat):
2238
except OSError as e:
2239
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2240
raise errors.NoSuchFile(f)
2243
def file_kind(f, _lstat=os.lstat):
2244
stat_value = file_stat(f, _lstat)
2245
return file_kind_from_stat_mode(stat_value.st_mode)
2247
def until_no_eintr(f, *a, **kw):
2248
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2250
WARNING: you must be certain that it is safe to retry the call repeatedly
2251
if EINTR does occur. This is typically only true for low-level operations
2252
like os.read. If in any doubt, don't use this.
2254
Keep in mind that this is not a complete solution to EINTR. There is
2255
probably code in the Python standard library and other dependencies that
2256
may encounter EINTR if a signal arrives (and there is signal handler for
2257
that signal). So this function can reduce the impact for IO that breezy
2258
directly controls, but it is not a complete solution.
2260
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2264
except (IOError, OSError) as e:
2265
if e.errno == errno.EINTR:
2270
if sys.platform == "win32":
2273
return msvcrt.getch()
2278
fd = sys.stdin.fileno()
2279
settings = termios.tcgetattr(fd)
2282
ch = sys.stdin.read(1)
2284
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2287
if sys.platform.startswith('linux'):
2288
def _local_concurrency():
2290
return os.sysconf('SC_NPROCESSORS_ONLN')
2291
except (ValueError, OSError, AttributeError):
2293
elif sys.platform == 'darwin':
2294
def _local_concurrency():
2295
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2296
stdout=subprocess.PIPE).communicate()[0]
2297
elif "bsd" in sys.platform:
2298
def _local_concurrency():
2299
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2300
stdout=subprocess.PIPE).communicate()[0]
2301
elif sys.platform == 'sunos5':
2302
def _local_concurrency():
2303
return subprocess.Popen(['psrinfo', '-p',],
2304
stdout=subprocess.PIPE).communicate()[0]
2305
elif sys.platform == "win32":
2306
def _local_concurrency():
2307
# This appears to return the number of cores.
2308
return os.environ.get('NUMBER_OF_PROCESSORS')
2310
def _local_concurrency():
2315
_cached_local_concurrency = None
2317
def local_concurrency(use_cache=True):
2318
"""Return how many processes can be run concurrently.
2320
Rely on platform specific implementations and default to 1 (one) if
2321
anything goes wrong.
2323
global _cached_local_concurrency
2325
if _cached_local_concurrency is not None and use_cache:
2326
return _cached_local_concurrency
2328
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2329
if concurrency is None:
2331
import multiprocessing
2332
concurrency = multiprocessing.cpu_count()
2333
except (ImportError, NotImplementedError):
2334
# multiprocessing is only available on Python >= 2.6
2335
# and multiprocessing.cpu_count() isn't implemented on all
2338
concurrency = _local_concurrency()
2339
except (OSError, IOError):
2342
concurrency = int(concurrency)
2343
except (TypeError, ValueError):
2346
_cached_concurrency = concurrency
2350
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2351
"""A stream writer that doesn't decode str arguments."""
2353
def __init__(self, encode, stream, errors='strict'):
2354
codecs.StreamWriter.__init__(self, stream, errors)
2355
self.encode = encode
2357
def write(self, object):
2358
if isinstance(object, str):
2359
self.stream.write(object)
2361
data, _ = self.encode(object, self.errors)
2362
self.stream.write(data)
2364
if sys.platform == 'win32':
2365
def open_file(filename, mode='r', bufsize=-1):
2366
"""This function is used to override the ``open`` builtin.
2368
But it uses O_NOINHERIT flag so the file handle is not inherited by
2369
child processes. Deleting or renaming a closed file opened with this
2370
function is not blocking child processes.
2372
writing = 'w' in mode
2373
appending = 'a' in mode
2374
updating = '+' in mode
2375
binary = 'b' in mode
2378
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2379
# for flags for each modes.
2389
flags |= os.O_WRONLY
2390
flags |= os.O_CREAT | os.O_TRUNC
2395
flags |= os.O_WRONLY
2396
flags |= os.O_CREAT | os.O_APPEND
2401
flags |= os.O_RDONLY
2403
return os.fdopen(os.open(filename, flags), mode, bufsize)
2408
def available_backup_name(base, exists):
2409
"""Find a non-existing backup file name.
2411
This will *not* create anything, this only return a 'free' entry. This
2412
should be used for checking names in a directory below a locked
2413
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2414
Leap) and generally discouraged.
2416
:param base: The base name.
2418
:param exists: A callable returning True if the path parameter exists.
2421
name = "%s.~%d~" % (base, counter)
2424
name = "%s.~%d~" % (base, counter)
2428
def set_fd_cloexec(fd):
2429
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2430
support for this is not available.
2434
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2435
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2436
except (ImportError, AttributeError):
2437
# Either the fcntl module or specific constants are not present
2441
def find_executable_on_path(name):
2442
"""Finds an executable on the PATH.
2444
On Windows, this will try to append each extension in the PATHEXT
2445
environment variable to the name, if it cannot be found with the name
2448
:param name: The base name of the executable.
2449
:return: The path to the executable found or None.
2451
if sys.platform == 'win32':
2452
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2453
exts = [ext.lower() for ext in exts]
2454
base, ext = os.path.splitext(name)
2456
if ext.lower() not in exts:
2462
path = os.environ.get('PATH')
2463
if path is not None:
2464
path = path.split(os.pathsep)
2467
f = os.path.join(d, name) + ext
2468
if os.access(f, os.X_OK):
2470
if sys.platform == 'win32':
2471
app_path = win32utils.get_app_path(name)
2472
if app_path != name:
2477
def _posix_is_local_pid_dead(pid):
2478
"""True if pid doesn't correspond to live process on this machine"""
2480
# Special meaning of unix kill: just check if it's there.
2482
except OSError as e:
2483
if e.errno == errno.ESRCH:
2484
# On this machine, and really not found: as sure as we can be
2487
elif e.errno == errno.EPERM:
2488
# exists, though not ours
2491
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2492
# Don't really know.
2495
# Exists and our process: not dead.
2498
if sys.platform == "win32":
2499
is_local_pid_dead = win32utils.is_local_pid_dead
2501
is_local_pid_dead = _posix_is_local_pid_dead
2503
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2504
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2505
if getattr(errno, name, None) is not None]
2508
def fdatasync(fileno):
2509
"""Flush file contents to disk if possible.
2511
:param fileno: Integer OS file handle.
2512
:raises TransportNotPossible: If flushing to disk is not possible.
2514
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2518
except IOError as e:
2519
# See bug #1075108, on some platforms fdatasync exists, but can
2520
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2521
# and reduce the chance of corruption-on-powerloss situations. It
2522
# is not a mandatory call, so it is ok to suppress failures.
2523
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2524
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2528
def ensure_empty_directory_exists(path, exception_class):
2529
"""Make sure a local directory exists and is empty.
2531
If it does not exist, it is created. If it exists and is not empty, an
2532
instance of exception_class is raised.
2536
except OSError as e:
2537
if e.errno != errno.EEXIST:
2539
if os.listdir(path) != []:
2540
raise exception_class(path)
2543
def is_environment_error(evalue):
2544
"""True if exception instance is due to a process environment issue
2546
This includes OSError and IOError, but also other errors that come from
2547
the operating system or core libraries but are not subclasses of those.
2549
if isinstance(evalue, (EnvironmentError, select.error)):
2551
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
added
removed
breezy/osutils.py
