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()
352
user_encoding = get_user_encoding()
354
return name.decode(user_encoding)
355
except UnicodeDecodeError:
356
raise errors.BzrError("Encoding of username %r is unsupported by %s "
357
"application locale." % (name, user_encoding))
360
def _win32_fixdrive(path):
361
"""Force drive letters to be consistent.
363
win32 is inconsistent whether it returns lower or upper case
364
and even if it was consistent the user might type the other
365
so we force it to uppercase
366
running python.exe under cmd.exe return capital C:\\
367
running win32 python inside a cygwin shell returns lowercase c:\\
369
drive, path = ntpath.splitdrive(path)
370
return drive.upper() + path
373
def _win32_abspath(path):
374
# Real ntpath.abspath doesn't have a problem with a unicode cwd
375
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
378
def _win32_realpath(path):
379
# Real ntpath.realpath doesn't have a problem with a unicode cwd
380
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
383
def _win32_pathjoin(*args):
384
return ntpath.join(*args).replace('\\', '/')
387
def _win32_normpath(path):
388
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
392
return _win32_fixdrive(_getcwd().replace('\\', '/'))
395
def _win32_mkdtemp(*args, **kwargs):
396
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
399
def _win32_rename(old, new):
400
"""We expect to be able to atomically replace 'new' with old.
402
On win32, if new exists, it must be moved out of the way first,
406
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
408
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
409
# If we try to rename a non-existant file onto cwd, we get
410
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
411
# if the old path doesn't exist, sometimes we get EACCES
412
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
418
return unicodedata.normalize('NFC', _getcwd())
421
def _rename_wrap_exception(rename_func):
422
"""Adds extra information to any exceptions that come from rename().
424
The exception has an updated message and 'old_filename' and 'new_filename'
428
def _rename_wrapper(old, new):
430
rename_func(old, new)
432
detailed_error = OSError(e.errno, e.strerror +
433
" [occurred when renaming '%s' to '%s']" %
435
detailed_error.old_filename = old
436
detailed_error.new_filename = new
439
return _rename_wrapper
442
if sys.version_info > (3,):
448
# Default rename wraps os.rename()
449
rename = _rename_wrap_exception(os.rename)
451
# Default is to just use the python builtins, but these can be rebound on
452
# particular platforms.
453
abspath = _posix_abspath
454
realpath = _posix_realpath
455
pathjoin = os.path.join
456
normpath = _posix_normpath
457
path_from_environ = _posix_path_from_environ
458
_get_home_dir = _posix_get_home_dir
459
getuser_unicode = _posix_getuser_unicode
461
dirname = os.path.dirname
462
basename = os.path.basename
463
split = os.path.split
464
splitext = os.path.splitext
465
# These were already lazily imported into local scope
466
# mkdtemp = tempfile.mkdtemp
467
# rmtree = shutil.rmtree
475
MIN_ABS_PATHLENGTH = 1
478
if sys.platform == 'win32':
479
abspath = _win32_abspath
480
realpath = _win32_realpath
481
pathjoin = _win32_pathjoin
482
normpath = _win32_normpath
483
getcwd = _win32_getcwd
484
mkdtemp = _win32_mkdtemp
485
rename = _rename_wrap_exception(_win32_rename)
487
from . import _walkdirs_win32
491
lstat = _walkdirs_win32.lstat
492
fstat = _walkdirs_win32.fstat
493
wrap_stat = _walkdirs_win32.wrap_stat
495
MIN_ABS_PATHLENGTH = 3
497
def _win32_delete_readonly(function, path, excinfo):
498
"""Error handler for shutil.rmtree function [for win32]
499
Helps to remove files and dirs marked as read-only.
501
exception = excinfo[1]
502
if function in (os.remove, os.rmdir) \
503
and isinstance(exception, OSError) \
504
and exception.errno == errno.EACCES:
510
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
511
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
512
return shutil.rmtree(path, ignore_errors, onerror)
514
f = win32utils.get_unicode_argv # special function or None
517
path_from_environ = win32utils.get_environ_unicode
518
_get_home_dir = win32utils.get_home_location
519
getuser_unicode = win32utils.get_user_name
521
elif sys.platform == 'darwin':
525
def get_terminal_encoding(trace=False):
526
"""Find the best encoding for printing to the screen.
528
This attempts to check both sys.stdout and sys.stdin to see
529
what encoding they are in, and if that fails it falls back to
530
osutils.get_user_encoding().
531
The problem is that on Windows, locale.getpreferredencoding()
532
is not the same encoding as that used by the console:
533
http://mail.python.org/pipermail/python-list/2003-May/162357.html
535
On my standard US Windows XP, the preferred encoding is
536
cp1252, but the console is cp437
538
:param trace: If True trace the selected encoding via mutter().
540
from .trace import mutter
541
output_encoding = getattr(sys.stdout, 'encoding', None)
542
if not output_encoding:
543
input_encoding = getattr(sys.stdin, 'encoding', None)
544
if not input_encoding:
545
output_encoding = get_user_encoding()
547
mutter('encoding stdout as osutils.get_user_encoding() %r',
550
output_encoding = input_encoding
552
mutter('encoding stdout as sys.stdin encoding %r',
556
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
557
if output_encoding == 'cp0':
558
# invalid encoding (cp0 means 'no codepage' on Windows)
559
output_encoding = get_user_encoding()
561
mutter('cp0 is invalid encoding.'
562
' encoding stdout as osutils.get_user_encoding() %r',
566
codecs.lookup(output_encoding)
568
sys.stderr.write('brz: warning:'
569
' unknown terminal encoding %s.\n'
570
' Using encoding %s instead.\n'
571
% (output_encoding, get_user_encoding())
573
output_encoding = get_user_encoding()
575
return output_encoding
578
def normalizepath(f):
579
if getattr(os.path, 'realpath', None) is not None:
583
[p,e] = os.path.split(f)
584
if e == "" or e == "." or e == "..":
587
return pathjoin(F(p), e)
591
"""True if f is an accessible directory."""
593
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
599
"""True if f is a regular file."""
601
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
606
"""True if f is a symlink."""
608
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
612
def is_inside(dir, fname):
613
"""True if fname is inside dir.
615
The parameters should typically be passed to osutils.normpath first, so
616
that . and .. and repeated slashes are eliminated, and the separators
617
are canonical for the platform.
619
The empty string as a dir name is taken as top-of-tree and matches
622
# XXX: Most callers of this can actually do something smarter by
623
# looking at the inventory
633
return fname.startswith(dir)
636
def is_inside_any(dir_list, fname):
637
"""True if fname is inside any of given dirs."""
638
for dirname in dir_list:
639
if is_inside(dirname, fname):
644
def is_inside_or_parent_of_any(dir_list, fname):
645
"""True if fname is a child or a parent of any of the given files."""
646
for dirname in dir_list:
647
if is_inside(dirname, fname) or is_inside(fname, dirname):
652
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
653
report_activity=None, direction='read'):
654
"""Copy contents of one file to another.
656
The read_length can either be -1 to read to end-of-file (EOF) or
657
it can specify the maximum number of bytes to read.
659
The buff_size represents the maximum size for each read operation
660
performed on from_file.
662
:param report_activity: Call this as bytes are read, see
663
Transport._report_activity
664
:param direction: Will be passed to report_activity
666
:return: The number of bytes copied.
670
# read specified number of bytes
672
while read_length > 0:
673
num_bytes_to_read = min(read_length, buff_size)
675
block = from_file.read(num_bytes_to_read)
679
if report_activity is not None:
680
report_activity(len(block), direction)
683
actual_bytes_read = len(block)
684
read_length -= actual_bytes_read
685
length += actual_bytes_read
689
block = from_file.read(buff_size)
693
if report_activity is not None:
694
report_activity(len(block), direction)
700
def pump_string_file(bytes, file_handle, segment_size=None):
701
"""Write bytes to file_handle in many smaller writes.
703
:param bytes: The string to write.
704
:param file_handle: The file to write to.
706
# Write data in chunks rather than all at once, because very large
707
# writes fail on some platforms (e.g. Windows with SMB mounted
710
segment_size = 5242880 # 5MB
711
offsets = range(0, len(bytes), segment_size)
712
view = memoryview(bytes)
713
write = file_handle.write
714
for offset in offsets:
715
write(view[offset:offset+segment_size])
718
def file_iterator(input_file, readsize=32768):
720
b = input_file.read(readsize)
727
"""Calculate the hexdigest of an open file.
729
The file cursor should be already at the start.
741
def size_sha_file(f):
742
"""Calculate the size and hexdigest of an open file.
744
The file cursor should be already at the start and
745
the caller is responsible for closing the file afterwards.
756
return size, s.hexdigest()
759
def sha_file_by_name(fname):
760
"""Calculate the SHA1 of a file by reading the full text"""
762
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
765
b = os.read(f, 1<<16)
773
def sha_strings(strings, _factory=sha):
774
"""Return the sha-1 of concatenation of strings"""
776
for string in strings:
781
def sha_string(f, _factory=sha):
782
return _factory(f).hexdigest()
785
def fingerprint_file(f):
787
return {'size': len(b),
788
'sha1': sha(b).hexdigest()}
791
def compare_files(a, b):
792
"""Returns true if equal in contents"""
803
def gmtime(seconds=None):
804
"""Convert seconds since the Epoch to a time tuple expressing UTC (a.k.a.
805
GMT). When 'seconds' is not passed in, convert the current time instead.
806
Handy replacement for time.gmtime() buggy on Windows and 32-bit platforms.
809
seconds = time.time()
810
return (datetime(1970, 1, 1) + timedelta(seconds=seconds)).timetuple()
813
def local_time_offset(t=None):
814
"""Return offset of local zone from GMT, either at present or at time t."""
817
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
818
return offset.days * 86400 + offset.seconds
820
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
821
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
824
def format_date(t, offset=0, timezone='original', date_fmt=None,
826
"""Return a formatted date string.
828
:param t: Seconds since the epoch.
829
:param offset: Timezone offset in seconds east of utc.
830
:param timezone: How to display the time: 'utc', 'original' for the
831
timezone specified by offset, or 'local' for the process's current
833
:param date_fmt: strftime format.
834
:param show_offset: Whether to append the timezone.
836
(date_fmt, tt, offset_str) = \
837
_format_date(t, offset, timezone, date_fmt, show_offset)
838
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
839
date_str = time.strftime(date_fmt, tt)
840
return date_str + offset_str
843
# Cache of formatted offset strings
847
def format_date_with_offset_in_original_timezone(t, offset=0,
848
_cache=_offset_cache):
849
"""Return a formatted date string in the original timezone.
851
This routine may be faster then format_date.
853
:param t: Seconds since the epoch.
854
:param offset: Timezone offset in seconds east of utc.
858
tt = gmtime(t + offset)
859
date_fmt = _default_format_by_weekday_num[tt[6]]
860
date_str = time.strftime(date_fmt, tt)
861
offset_str = _cache.get(offset, None)
862
if offset_str is None:
863
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
864
_cache[offset] = offset_str
865
return date_str + offset_str
868
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
870
"""Return an unicode date string formatted according to the current locale.
872
:param t: Seconds since the epoch.
873
:param offset: Timezone offset in seconds east of utc.
874
:param timezone: How to display the time: 'utc', 'original' for the
875
timezone specified by offset, or 'local' for the process's current
877
:param date_fmt: strftime format.
878
:param show_offset: Whether to append the timezone.
880
(date_fmt, tt, offset_str) = \
881
_format_date(t, offset, timezone, date_fmt, show_offset)
882
date_str = time.strftime(date_fmt, tt)
883
if not isinstance(date_str, text_type):
884
date_str = date_str.decode(get_user_encoding(), 'replace')
885
return date_str + offset_str
888
def _format_date(t, offset, timezone, date_fmt, show_offset):
889
if timezone == 'utc':
892
elif timezone == 'original':
895
tt = gmtime(t + offset)
896
elif timezone == 'local':
897
tt = time.localtime(t)
898
offset = local_time_offset(t)
900
raise errors.UnsupportedTimezoneFormat(timezone)
902
date_fmt = "%a %Y-%m-%d %H:%M:%S"
904
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
907
return (date_fmt, tt, offset_str)
910
def compact_date(when):
911
return time.strftime('%Y%m%d%H%M%S', gmtime(when))
914
def format_delta(delta):
915
"""Get a nice looking string for a time delta.
917
:param delta: The time difference in seconds, can be positive or negative.
918
positive indicates time in the past, negative indicates time in the
919
future. (usually time.time() - stored_time)
920
:return: String formatted to show approximate resolution
926
direction = 'in the future'
930
if seconds < 90: # print seconds up to 90 seconds
932
return '%d second %s' % (seconds, direction,)
934
return '%d seconds %s' % (seconds, direction)
936
minutes = int(seconds / 60)
937
seconds -= 60 * minutes
942
if minutes < 90: # print minutes, seconds up to 90 minutes
944
return '%d minute, %d second%s %s' % (
945
minutes, seconds, plural_seconds, direction)
947
return '%d minutes, %d second%s %s' % (
948
minutes, seconds, plural_seconds, direction)
950
hours = int(minutes / 60)
951
minutes -= 60 * hours
958
return '%d hour, %d minute%s %s' % (hours, minutes,
959
plural_minutes, direction)
960
return '%d hours, %d minute%s %s' % (hours, minutes,
961
plural_minutes, direction)
964
"""Return size of given open file."""
965
return os.fstat(f.fileno())[stat.ST_SIZE]
968
# Alias os.urandom to support platforms (which?) without /dev/urandom and
969
# override if it doesn't work. Avoid checking on windows where there is
970
# significant initialisation cost that can be avoided for some bzr calls.
972
rand_bytes = os.urandom
974
if rand_bytes.__module__ != "nt":
977
except NotImplementedError:
978
# not well seeded, but better than nothing
983
s += chr(random.randint(0, 255))
988
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
990
"""Return a random string of num alphanumeric characters
992
The result only contains lowercase chars because it may be used on
993
case-insensitive filesystems.
996
for raw_byte in rand_bytes(num):
997
s += ALNUM[ord(raw_byte) % 36]
1001
## TODO: We could later have path objects that remember their list
1002
## decomposition (might be too tricksy though.)
1005
"""Turn string into list of parts."""
1006
# split on either delimiter because people might use either on
1008
ps = re.split(r'[\\/]', p)
1013
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1014
elif (f == '.') or (f == ''):
1023
if (f == '..') or (f is None) or (f == ''):
1024
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1028
def parent_directories(filename):
1029
"""Return the list of parent directories, deepest first.
1031
For example, parent_directories("a/b/c") -> ["a/b", "a"].
1034
parts = splitpath(dirname(filename))
1036
parents.append(joinpath(parts))
1041
_extension_load_failures = []
1044
def failed_to_load_extension(exception):
1045
"""Handle failing to load a binary extension.
1047
This should be called from the ImportError block guarding the attempt to
1048
import the native extension. If this function returns, the pure-Python
1049
implementation should be loaded instead::
1052
>>> import breezy._fictional_extension_pyx
1053
>>> except ImportError, e:
1054
>>> breezy.osutils.failed_to_load_extension(e)
1055
>>> import breezy._fictional_extension_py
1057
# NB: This docstring is just an example, not a doctest, because doctest
1058
# currently can't cope with the use of lazy imports in this namespace --
1061
# This currently doesn't report the failure at the time it occurs, because
1062
# they tend to happen very early in startup when we can't check config
1063
# files etc, and also we want to report all failures but not spam the user
1065
exception_str = str(exception)
1066
if exception_str not in _extension_load_failures:
1067
trace.mutter("failed to load compiled extension: %s" % exception_str)
1068
_extension_load_failures.append(exception_str)
1071
def report_extension_load_failures():
1072
if not _extension_load_failures:
1074
if config.GlobalStack().get('ignore_missing_extensions'):
1076
# the warnings framework should by default show this only once
1077
from .trace import warning
1079
"brz: warning: some compiled extensions could not be loaded; "
1080
"see <https://answers.launchpad.net/bzr/+faq/703>")
1081
# we no longer show the specific missing extensions here, because it makes
1082
# the message too long and scary - see
1083
# https://bugs.launchpad.net/bzr/+bug/430529
1087
from ._chunks_to_lines_pyx import chunks_to_lines
1088
except ImportError as e:
1089
failed_to_load_extension(e)
1090
from ._chunks_to_lines_py import chunks_to_lines
1094
"""Split s into lines, but without removing the newline characters."""
1095
# Trivially convert a fulltext into a 'chunked' representation, and let
1096
# chunks_to_lines do the heavy lifting.
1097
if isinstance(s, str):
1098
# chunks_to_lines only supports 8-bit strings
1099
return chunks_to_lines([s])
1101
return _split_lines(s)
1104
def _split_lines(s):
1105
"""Split s into lines, but without removing the newline characters.
1107
This supports Unicode or plain string objects.
1109
lines = s.split('\n')
1110
result = [line + '\n' for line in lines[:-1]]
1112
result.append(lines[-1])
1116
def hardlinks_good():
1117
return sys.platform not in ('win32', 'cygwin', 'darwin')
1120
def link_or_copy(src, dest):
1121
"""Hardlink a file, or copy it if it can't be hardlinked."""
1122
if not hardlinks_good():
1123
shutil.copyfile(src, dest)
1127
except (OSError, IOError) as e:
1128
if e.errno != errno.EXDEV:
1130
shutil.copyfile(src, dest)
1133
def delete_any(path):
1134
"""Delete a file, symlink or directory.
1136
Will delete even if readonly.
1139
_delete_file_or_dir(path)
1140
except (OSError, IOError) as e:
1141
if e.errno in (errno.EPERM, errno.EACCES):
1142
# make writable and try again
1145
except (OSError, IOError):
1147
_delete_file_or_dir(path)
1152
def _delete_file_or_dir(path):
1153
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1154
# Forgiveness than Permission (EAFP) because:
1155
# - root can damage a solaris file system by using unlink,
1156
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1157
# EACCES, OSX: EPERM) when invoked on a directory.
1158
if isdir(path): # Takes care of symlinks
1165
if getattr(os, 'symlink', None) is not None:
1171
def has_hardlinks():
1172
if getattr(os, 'link', None) is not None:
1178
def host_os_dereferences_symlinks():
1179
return (has_symlinks()
1180
and sys.platform not in ('cygwin', 'win32'))
1183
def readlink(abspath):
1184
"""Return a string representing the path to which the symbolic link points.
1186
:param abspath: The link absolute unicode path.
1188
This his guaranteed to return the symbolic link in unicode in all python
1191
link = abspath.encode(_fs_enc)
1192
target = os.readlink(link)
1193
target = target.decode(_fs_enc)
1197
def contains_whitespace(s):
1198
"""True if there are any whitespace characters in s."""
1199
# string.whitespace can include '\xa0' in certain locales, because it is
1200
# considered "non-breaking-space" as part of ISO-8859-1. But it
1201
# 1) Isn't a breaking whitespace
1202
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1204
# 3) '\xa0' isn't unicode safe since it is >128.
1206
# This should *not* be a unicode set of characters in case the source
1207
# string is not a Unicode string. We can auto-up-cast the characters since
1208
# they are ascii, but we don't want to auto-up-cast the string in case it
1210
for ch in ' \t\n\r\v\f':
1217
def contains_linebreaks(s):
1218
"""True if there is any vertical whitespace in s."""
1226
def relpath(base, path):
1227
"""Return path relative to base, or raise PathNotChild exception.
1229
The path may be either an absolute path or a path relative to the
1230
current working directory.
1232
os.path.commonprefix (python2.4) has a bad bug that it works just
1233
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1234
avoids that problem.
1236
NOTE: `base` should not have a trailing slash otherwise you'll get
1237
PathNotChild exceptions regardless of `path`.
1240
if len(base) < MIN_ABS_PATHLENGTH:
1241
# must have space for e.g. a drive letter
1242
raise ValueError(gettext('%r is too short to calculate a relative path')
1250
if len(head) <= len(base) and head != base:
1251
raise errors.PathNotChild(rp, base)
1254
head, tail = split(head)
1259
return pathjoin(*reversed(s))
1264
def _cicp_canonical_relpath(base, path):
1265
"""Return the canonical path relative to base.
1267
Like relpath, but on case-insensitive-case-preserving file-systems, this
1268
will return the relpath as stored on the file-system rather than in the
1269
case specified in the input string, for all existing portions of the path.
1271
This will cause O(N) behaviour if called for every path in a tree; if you
1272
have a number of paths to convert, you should use canonical_relpaths().
1274
# TODO: it should be possible to optimize this for Windows by using the
1275
# win32 API FindFiles function to look for the specified name - but using
1276
# os.listdir() still gives us the correct, platform agnostic semantics in
1279
rel = relpath(base, path)
1280
# '.' will have been turned into ''
1284
abs_base = abspath(base)
1286
_listdir = os.listdir
1288
# use an explicit iterator so we can easily consume the rest on early exit.
1289
bit_iter = iter(rel.split('/'))
1290
for bit in bit_iter:
1293
next_entries = _listdir(current)
1294
except OSError: # enoent, eperm, etc
1295
# We can't find this in the filesystem, so just append the
1297
current = pathjoin(current, bit, *list(bit_iter))
1299
for look in next_entries:
1300
if lbit == look.lower():
1301
current = pathjoin(current, look)
1304
# got to the end, nothing matched, so we just return the
1305
# non-existing bits as they were specified (the filename may be
1306
# the target of a move, for example).
1307
current = pathjoin(current, bit, *list(bit_iter))
1309
return current[len(abs_base):].lstrip('/')
1311
# XXX - TODO - we need better detection/integration of case-insensitive
1312
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1313
# filesystems), for example, so could probably benefit from the same basic
1314
# support there. For now though, only Windows and OSX get that support, and
1315
# they get it for *all* file-systems!
1316
if sys.platform in ('win32', 'darwin'):
1317
canonical_relpath = _cicp_canonical_relpath
1319
canonical_relpath = relpath
1321
def canonical_relpaths(base, paths):
1322
"""Create an iterable to canonicalize a sequence of relative paths.
1324
The intent is for this implementation to use a cache, vastly speeding
1325
up multiple transformations in the same directory.
1327
# but for now, we haven't optimized...
1328
return [canonical_relpath(base, p) for p in paths]
1331
def decode_filename(filename):
1332
"""Decode the filename using the filesystem encoding
1334
If it is unicode, it is returned.
1335
Otherwise it is decoded from the the filesystem's encoding. If decoding
1336
fails, a errors.BadFilenameEncoding exception is raised.
1338
if isinstance(filename, text_type):
1341
return filename.decode(_fs_enc)
1342
except UnicodeDecodeError:
1343
raise errors.BadFilenameEncoding(filename, _fs_enc)
1346
def safe_unicode(unicode_or_utf8_string):
1347
"""Coerce unicode_or_utf8_string into unicode.
1349
If it is unicode, it is returned.
1350
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1351
wrapped in a BzrBadParameterNotUnicode exception.
1353
if isinstance(unicode_or_utf8_string, text_type):
1354
return unicode_or_utf8_string
1356
return unicode_or_utf8_string.decode('utf8')
1357
except UnicodeDecodeError:
1358
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1361
def safe_utf8(unicode_or_utf8_string):
1362
"""Coerce unicode_or_utf8_string to a utf8 string.
1364
If it is a str, it is returned.
1365
If it is Unicode, it is encoded into a utf-8 string.
1367
if isinstance(unicode_or_utf8_string, str):
1368
# TODO: jam 20070209 This is overkill, and probably has an impact on
1369
# performance if we are dealing with lots of apis that want a
1372
# Make sure it is a valid utf-8 string
1373
unicode_or_utf8_string.decode('utf-8')
1374
except UnicodeDecodeError:
1375
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1376
return unicode_or_utf8_string
1377
return unicode_or_utf8_string.encode('utf-8')
1380
def safe_revision_id(unicode_or_utf8_string):
1381
"""Revision ids should now be utf8, but at one point they were unicode.
1383
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1385
:return: None or a utf8 revision id.
1387
if (unicode_or_utf8_string is None
1388
or unicode_or_utf8_string.__class__ == str):
1389
return unicode_or_utf8_string
1390
raise TypeError('Unicode revision ids are no longer supported. '
1391
'Revision id generators should be creating utf8 revision '
1395
def safe_file_id(unicode_or_utf8_string):
1396
"""File ids should now be utf8, but at one point they were unicode.
1398
This is the same as safe_utf8, except it uses the cached encode functions
1399
to save a little bit of performance.
1401
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1403
:return: None or a utf8 file id.
1405
if (unicode_or_utf8_string is None
1406
or unicode_or_utf8_string.__class__ == str):
1407
return unicode_or_utf8_string
1408
raise TypeError('Unicode file ids are no longer supported. '
1409
'File id generators should be creating utf8 file ids.')
1412
_platform_normalizes_filenames = False
1413
if sys.platform == 'darwin':
1414
_platform_normalizes_filenames = True
1417
def normalizes_filenames():
1418
"""Return True if this platform normalizes unicode filenames.
1422
return _platform_normalizes_filenames
1425
def _accessible_normalized_filename(path):
1426
"""Get the unicode normalized path, and if you can access the file.
1428
On platforms where the system normalizes filenames (Mac OSX),
1429
you can access a file by any path which will normalize correctly.
1430
On platforms where the system does not normalize filenames
1431
(everything else), you have to access a file by its exact path.
1433
Internally, bzr only supports NFC normalization, since that is
1434
the standard for XML documents.
1436
So return the normalized path, and a flag indicating if the file
1437
can be accessed by that path.
1440
return unicodedata.normalize('NFC', unicode(path)), True
1443
def _inaccessible_normalized_filename(path):
1444
__doc__ = _accessible_normalized_filename.__doc__
1446
normalized = unicodedata.normalize('NFC', unicode(path))
1447
return normalized, normalized == path
1450
if _platform_normalizes_filenames:
1451
normalized_filename = _accessible_normalized_filename
1453
normalized_filename = _inaccessible_normalized_filename
1456
def set_signal_handler(signum, handler, restart_syscall=True):
1457
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1458
on platforms that support that.
1460
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1461
automatically restart (by calling `signal.siginterrupt(signum,
1462
False)`). May be ignored if the feature is not available on this
1463
platform or Python version.
1467
siginterrupt = signal.siginterrupt
1469
# This python implementation doesn't provide signal support, hence no
1472
except AttributeError:
1473
# siginterrupt doesn't exist on this platform, or for this version
1475
siginterrupt = lambda signum, flag: None
1477
def sig_handler(*args):
1478
# Python resets the siginterrupt flag when a signal is
1479
# received. <http://bugs.python.org/issue8354>
1480
# As a workaround for some cases, set it back the way we want it.
1481
siginterrupt(signum, False)
1482
# Now run the handler function passed to set_signal_handler.
1485
sig_handler = handler
1486
old_handler = signal.signal(signum, sig_handler)
1488
siginterrupt(signum, False)
1492
default_terminal_width = 80
1493
"""The default terminal width for ttys.
1495
This is defined so that higher levels can share a common fallback value when
1496
terminal_width() returns None.
1499
# Keep some state so that terminal_width can detect if _terminal_size has
1500
# returned a different size since the process started. See docstring and
1501
# comments of terminal_width for details.
1502
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1503
_terminal_size_state = 'no_data'
1504
_first_terminal_size = None
1506
def terminal_width():
1507
"""Return terminal width.
1509
None is returned if the width can't established precisely.
1512
- if BRZ_COLUMNS is set, returns its value
1513
- if there is no controlling terminal, returns None
1514
- query the OS, if the queried size has changed since the last query,
1516
- if COLUMNS is set, returns its value,
1517
- if the OS has a value (even though it's never changed), return its value.
1519
From there, we need to query the OS to get the size of the controlling
1522
On Unices we query the OS by:
1523
- get termios.TIOCGWINSZ
1524
- if an error occurs or a negative value is obtained, returns None
1526
On Windows we query the OS by:
1527
- win32utils.get_console_size() decides,
1528
- returns None on error (provided default value)
1530
# Note to implementors: if changing the rules for determining the width,
1531
# make sure you've considered the behaviour in these cases:
1532
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1533
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1535
# - (add more interesting cases here, if you find any)
1536
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1537
# but we don't want to register a signal handler because it is impossible
1538
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1539
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1540
# time so we can notice if the reported size has changed, which should have
1543
# If BRZ_COLUMNS is set, take it, user is always right
1544
# Except if they specified 0 in which case, impose no limit here
1546
width = int(os.environ['BRZ_COLUMNS'])
1547
except (KeyError, ValueError):
1549
if width is not None:
1555
isatty = getattr(sys.stdout, 'isatty', None)
1556
if isatty is None or not isatty():
1557
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
1561
width, height = os_size = _terminal_size(None, None)
1562
global _first_terminal_size, _terminal_size_state
1563
if _terminal_size_state == 'no_data':
1564
_first_terminal_size = os_size
1565
_terminal_size_state = 'unchanged'
1566
elif (_terminal_size_state == 'unchanged' and
1567
_first_terminal_size != os_size):
1568
_terminal_size_state = 'changed'
1570
# If the OS claims to know how wide the terminal is, and this value has
1571
# ever changed, use that.
1572
if _terminal_size_state == 'changed':
1573
if width is not None and width > 0:
1576
# If COLUMNS is set, use it.
1578
return int(os.environ['COLUMNS'])
1579
except (KeyError, ValueError):
1582
# Finally, use an unchanged size from the OS, if we have one.
1583
if _terminal_size_state == 'unchanged':
1584
if width is not None and width > 0:
1587
# The width could not be determined.
1591
def _win32_terminal_size(width, height):
1592
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1593
return width, height
1596
def _ioctl_terminal_size(width, height):
1598
import struct, fcntl, termios
1599
s = struct.pack('HHHH', 0, 0, 0, 0)
1600
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1601
height, width = struct.unpack('HHHH', x)[0:2]
1602
except (IOError, AttributeError):
1604
return width, height
1606
_terminal_size = None
1607
"""Returns the terminal size as (width, height).
1609
:param width: Default value for width.
1610
:param height: Default value for height.
1612
This is defined specifically for each OS and query the size of the controlling
1613
terminal. If any error occurs, the provided default values should be returned.
1615
if sys.platform == 'win32':
1616
_terminal_size = _win32_terminal_size
1618
_terminal_size = _ioctl_terminal_size
1621
def supports_executable():
1622
return sys.platform != "win32"
1625
def supports_posix_readonly():
1626
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1628
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1629
directory controls creation/deletion, etc.
1631
And under win32, readonly means that the directory itself cannot be
1632
deleted. The contents of a readonly directory can be changed, unlike POSIX
1633
where files in readonly directories cannot be added, deleted or renamed.
1635
return sys.platform != "win32"
1638
def set_or_unset_env(env_variable, value):
1639
"""Modify the environment, setting or removing the env_variable.
1641
:param env_variable: The environment variable in question
1642
:param value: The value to set the environment to. If None, then
1643
the variable will be removed.
1644
:return: The original value of the environment variable.
1646
orig_val = os.environ.get(env_variable)
1648
if orig_val is not None:
1649
del os.environ[env_variable]
1651
if not PY3 and isinstance(value, text_type):
1652
value = value.encode(get_user_encoding())
1653
os.environ[env_variable] = value
1657
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1660
def check_legal_path(path):
1661
"""Check whether the supplied path is legal.
1662
This is only required on Windows, so we don't test on other platforms
1665
if sys.platform != "win32":
1667
if _validWin32PathRE.match(path) is None:
1668
raise errors.IllegalPath(path)
1671
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1673
def _is_error_enotdir(e):
1674
"""Check if this exception represents ENOTDIR.
1676
Unfortunately, python is very inconsistent about the exception
1677
here. The cases are:
1678
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1679
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1680
which is the windows error code.
1681
3) Windows, Python2.5 uses errno == EINVAL and
1682
winerror == ERROR_DIRECTORY
1684
:param e: An Exception object (expected to be OSError with an errno
1685
attribute, but we should be able to cope with anything)
1686
:return: True if this represents an ENOTDIR error. False otherwise.
1688
en = getattr(e, 'errno', None)
1689
if (en == errno.ENOTDIR
1690
or (sys.platform == 'win32'
1691
and (en == _WIN32_ERROR_DIRECTORY
1692
or (en == errno.EINVAL
1693
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1699
def walkdirs(top, prefix=""):
1700
"""Yield data about all the directories in a tree.
1702
This yields all the data about the contents of a directory at a time.
1703
After each directory has been yielded, if the caller has mutated the list
1704
to exclude some directories, they are then not descended into.
1706
The data yielded is of the form:
1707
((directory-relpath, directory-path-from-top),
1708
[(relpath, basename, kind, lstat, path-from-top), ...]),
1709
- directory-relpath is the relative path of the directory being returned
1710
with respect to top. prefix is prepended to this.
1711
- directory-path-from-root is the path including top for this directory.
1712
It is suitable for use with os functions.
1713
- relpath is the relative path within the subtree being walked.
1714
- basename is the basename of the path
1715
- kind is the kind of the file now. If unknown then the file is not
1716
present within the tree - but it may be recorded as versioned. See
1718
- lstat is the stat data *if* the file was statted.
1719
- planned, not implemented:
1720
path_from_tree_root is the path from the root of the tree.
1722
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1723
allows one to walk a subtree but get paths that are relative to a tree
1725
:return: an iterator over the dirs.
1727
#TODO there is a bit of a smell where the results of the directory-
1728
# summary in this, and the path from the root, may not agree
1729
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1730
# potentially confusing output. We should make this more robust - but
1731
# not at a speed cost. RBC 20060731
1733
_directory = _directory_kind
1734
_listdir = os.listdir
1735
_kind_from_mode = file_kind_from_stat_mode
1736
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1738
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1739
relroot, _, _, _, top = pending.pop()
1741
relprefix = relroot + u'/'
1744
top_slash = top + u'/'
1747
append = dirblock.append
1749
names = sorted(map(decode_filename, _listdir(top)))
1750
except OSError as e:
1751
if not _is_error_enotdir(e):
1755
abspath = top_slash + name
1756
statvalue = _lstat(abspath)
1757
kind = _kind_from_mode(statvalue.st_mode)
1758
append((relprefix + name, name, kind, statvalue, abspath))
1759
yield (relroot, top), dirblock
1761
# push the user specified dirs from dirblock
1762
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1765
class DirReader(object):
1766
"""An interface for reading directories."""
1768
def top_prefix_to_starting_dir(self, top, prefix=""):
1769
"""Converts top and prefix to a starting dir entry
1771
:param top: A utf8 path
1772
:param prefix: An optional utf8 path to prefix output relative paths
1774
:return: A tuple starting with prefix, and ending with the native
1777
raise NotImplementedError(self.top_prefix_to_starting_dir)
1779
def read_dir(self, prefix, top):
1780
"""Read a specific dir.
1782
:param prefix: A utf8 prefix to be preprended to the path basenames.
1783
:param top: A natively encoded path to read.
1784
:return: A list of the directories contents. Each item contains:
1785
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1787
raise NotImplementedError(self.read_dir)
1790
_selected_dir_reader = None
1793
def _walkdirs_utf8(top, prefix=""):
1794
"""Yield data about all the directories in a tree.
1796
This yields the same information as walkdirs() only each entry is yielded
1797
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1798
are returned as exact byte-strings.
1800
:return: yields a tuple of (dir_info, [file_info])
1801
dir_info is (utf8_relpath, path-from-top)
1802
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1803
if top is an absolute path, path-from-top is also an absolute path.
1804
path-from-top might be unicode or utf8, but it is the correct path to
1805
pass to os functions to affect the file in question. (such as os.lstat)
1807
global _selected_dir_reader
1808
if _selected_dir_reader is None:
1809
if sys.platform == "win32":
1811
from ._walkdirs_win32 import Win32ReadDir
1812
_selected_dir_reader = Win32ReadDir()
1815
elif _fs_enc in ('utf-8', 'ascii'):
1817
from ._readdir_pyx import UTF8DirReader
1818
_selected_dir_reader = UTF8DirReader()
1819
except ImportError as e:
1820
failed_to_load_extension(e)
1823
if _selected_dir_reader is None:
1824
# Fallback to the python version
1825
_selected_dir_reader = UnicodeDirReader()
1827
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1828
# But we don't actually uses 1-3 in pending, so set them to None
1829
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1830
read_dir = _selected_dir_reader.read_dir
1831
_directory = _directory_kind
1833
relroot, _, _, _, top = pending[-1].pop()
1836
dirblock = sorted(read_dir(relroot, top))
1837
yield (relroot, top), dirblock
1838
# push the user specified dirs from dirblock
1839
next = [d for d in reversed(dirblock) if d[2] == _directory]
1841
pending.append(next)
1844
class UnicodeDirReader(DirReader):
1845
"""A dir reader for non-utf8 file systems, which transcodes."""
1847
__slots__ = ['_utf8_encode']
1850
self._utf8_encode = codecs.getencoder('utf8')
1852
def top_prefix_to_starting_dir(self, top, prefix=""):
1853
"""See DirReader.top_prefix_to_starting_dir."""
1854
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1856
def read_dir(self, prefix, top):
1857
"""Read a single directory from a non-utf8 file system.
1859
top, and the abspath element in the output are unicode, all other paths
1860
are utf8. Local disk IO is done via unicode calls to listdir etc.
1862
This is currently the fallback code path when the filesystem encoding is
1863
not UTF-8. It may be better to implement an alternative so that we can
1864
safely handle paths that are not properly decodable in the current
1867
See DirReader.read_dir for details.
1869
_utf8_encode = self._utf8_encode
1871
_listdir = os.listdir
1872
_kind_from_mode = file_kind_from_stat_mode
1875
relprefix = prefix + '/'
1878
top_slash = top + u'/'
1881
append = dirblock.append
1882
for name in sorted(_listdir(top)):
1884
name_utf8 = _utf8_encode(name)[0]
1885
except UnicodeDecodeError:
1886
raise errors.BadFilenameEncoding(
1887
_utf8_encode(relprefix)[0] + name, _fs_enc)
1888
abspath = top_slash + name
1889
statvalue = _lstat(abspath)
1890
kind = _kind_from_mode(statvalue.st_mode)
1891
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1895
def copy_tree(from_path, to_path, handlers={}):
1896
"""Copy all of the entries in from_path into to_path.
1898
:param from_path: The base directory to copy.
1899
:param to_path: The target directory. If it does not exist, it will
1901
:param handlers: A dictionary of functions, which takes a source and
1902
destinations for files, directories, etc.
1903
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1904
'file', 'directory', and 'symlink' should always exist.
1905
If they are missing, they will be replaced with 'os.mkdir()',
1906
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1908
# Now, just copy the existing cached tree to the new location
1909
# We use a cheap trick here.
1910
# Absolute paths are prefixed with the first parameter
1911
# relative paths are prefixed with the second.
1912
# So we can get both the source and target returned
1913
# without any extra work.
1915
def copy_dir(source, dest):
1918
def copy_link(source, dest):
1919
"""Copy the contents of a symlink"""
1920
link_to = os.readlink(source)
1921
os.symlink(link_to, dest)
1923
real_handlers = {'file':shutil.copy2,
1924
'symlink':copy_link,
1925
'directory':copy_dir,
1927
real_handlers.update(handlers)
1929
if not os.path.exists(to_path):
1930
real_handlers['directory'](from_path, to_path)
1932
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1933
for relpath, name, kind, st, abspath in entries:
1934
real_handlers[kind](abspath, relpath)
1937
def copy_ownership_from_path(dst, src=None):
1938
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1940
If src is None, the containing directory is used as source. If chown
1941
fails, the error is ignored and a warning is printed.
1943
chown = getattr(os, 'chown', None)
1948
src = os.path.dirname(dst)
1954
chown(dst, s.st_uid, s.st_gid)
1955
except OSError as e:
1957
'Unable to copy ownership from "%s" to "%s". '
1958
'You may want to set it manually.', src, dst)
1959
trace.log_exception_quietly()
1962
def path_prefix_key(path):
1963
"""Generate a prefix-order path key for path.
1965
This can be used to sort paths in the same way that walkdirs does.
1967
return (dirname(path) , path)
1970
def compare_paths_prefix_order(path_a, path_b):
1971
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1972
key_a = path_prefix_key(path_a)
1973
key_b = path_prefix_key(path_b)
1974
return cmp(key_a, key_b)
1977
_cached_user_encoding = None
1980
def get_user_encoding():
1981
"""Find out what the preferred user encoding is.
1983
This is generally the encoding that is used for command line parameters
1984
and file contents. This may be different from the terminal encoding
1985
or the filesystem encoding.
1987
:return: A string defining the preferred user encoding
1989
global _cached_user_encoding
1990
if _cached_user_encoding is not None:
1991
return _cached_user_encoding
1993
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
1994
# Use the existing locale settings and call nl_langinfo directly
1995
# rather than going through getpreferredencoding. This avoids
1996
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
1997
# possibility of the setlocale call throwing an error.
1998
user_encoding = locale.nl_langinfo(locale.CODESET)
2000
# GZ 2011-12-19: On windows could call GetACP directly instead.
2001
user_encoding = locale.getpreferredencoding(False)
2004
user_encoding = codecs.lookup(user_encoding).name
2006
if user_encoding not in ("", "cp0"):
2007
sys.stderr.write('brz: warning:'
2008
' unknown encoding %s.'
2009
' Continuing with ascii encoding.\n'
2012
user_encoding = 'ascii'
2014
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2015
if user_encoding == 'ascii':
2016
if sys.platform == 'darwin':
2017
# OSX is special-cased in Python to have a UTF-8 filesystem
2018
# encoding and previously had LANG set here if not present.
2019
user_encoding = 'utf-8'
2020
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2021
# for some other posix platforms as well.
2023
_cached_user_encoding = user_encoding
2024
return user_encoding
2027
def get_diff_header_encoding():
2028
return get_terminal_encoding()
2031
def get_host_name():
2032
"""Return the current unicode host name.
2034
This is meant to be used in place of socket.gethostname() because that
2035
behaves inconsistently on different platforms.
2037
if sys.platform == "win32":
2038
return win32utils.get_host_name()
2041
return socket.gethostname().decode(get_user_encoding())
2044
# We must not read/write any more than 64k at a time from/to a socket so we
2045
# don't risk "no buffer space available" errors on some platforms. Windows in
2046
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2048
MAX_SOCKET_CHUNK = 64 * 1024
2050
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2051
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2052
_eno = getattr(errno, _eno, None)
2053
if _eno is not None:
2054
_end_of_stream_errors.append(_eno)
2058
def read_bytes_from_socket(sock, report_activity=None,
2059
max_read_size=MAX_SOCKET_CHUNK):
2060
"""Read up to max_read_size of bytes from sock and notify of progress.
2062
Translates "Connection reset by peer" into file-like EOF (return an
2063
empty string rather than raise an error), and repeats the recv if
2064
interrupted by a signal.
2068
bytes = sock.recv(max_read_size)
2069
except socket.error as e:
2071
if eno in _end_of_stream_errors:
2072
# The connection was closed by the other side. Callers expect
2073
# an empty string to signal end-of-stream.
2075
elif eno == errno.EINTR:
2076
# Retry the interrupted recv.
2080
if report_activity is not None:
2081
report_activity(len(bytes), 'read')
2085
def recv_all(socket, count):
2086
"""Receive an exact number of bytes.
2088
Regular Socket.recv() may return less than the requested number of bytes,
2089
depending on what's in the OS buffer. MSG_WAITALL is not available
2090
on all platforms, but this should work everywhere. This will return
2091
less than the requested amount if the remote end closes.
2093
This isn't optimized and is intended mostly for use in testing.
2096
while len(b) < count:
2097
new = read_bytes_from_socket(socket, None, count - len(b))
2104
def send_all(sock, bytes, report_activity=None):
2105
"""Send all bytes on a socket.
2107
Breaks large blocks in smaller chunks to avoid buffering limitations on
2108
some platforms, and catches EINTR which may be thrown if the send is
2109
interrupted by a signal.
2111
This is preferred to socket.sendall(), because it avoids portability bugs
2112
and provides activity reporting.
2114
:param report_activity: Call this as bytes are read, see
2115
Transport._report_activity
2118
byte_count = len(bytes)
2119
view = memoryview(bytes)
2120
while sent_total < byte_count:
2122
sent = sock.send(view[sent_total:sent_total+MAX_SOCKET_CHUNK])
2123
except (socket.error, IOError) as e:
2124
if e.args[0] in _end_of_stream_errors:
2125
raise errors.ConnectionReset(
2126
"Error trying to write to socket", e)
2127
if e.args[0] != errno.EINTR:
2131
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2134
if report_activity is not None:
2135
report_activity(sent, 'write')
2138
def connect_socket(address):
2139
# Slight variation of the socket.create_connection() function (provided by
2140
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2141
# provide it for previous python versions. Also, we don't use the timeout
2142
# parameter (provided by the python implementation) so we don't implement
2144
err = socket.error('getaddrinfo returns an empty list')
2145
host, port = address
2146
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2147
af, socktype, proto, canonname, sa = res
2150
sock = socket.socket(af, socktype, proto)
2154
except socket.error as err:
2155
# 'err' is now the most recent error
2156
if sock is not None:
2161
def dereference_path(path):
2162
"""Determine the real path to a file.
2164
All parent elements are dereferenced. But the file itself is not
2166
:param path: The original path. May be absolute or relative.
2167
:return: the real path *to* the file
2169
parent, base = os.path.split(path)
2170
# The pathjoin for '.' is a workaround for Python bug #1213894.
2171
# (initial path components aren't dereferenced)
2172
return pathjoin(realpath(pathjoin('.', parent)), base)
2175
def supports_mapi():
2176
"""Return True if we can use MAPI to launch a mail client."""
2177
return sys.platform == "win32"
2180
def resource_string(package, resource_name):
2181
"""Load a resource from a package and return it as a string.
2183
Note: Only packages that start with breezy are currently supported.
2185
This is designed to be a lightweight implementation of resource
2186
loading in a way which is API compatible with the same API from
2188
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2189
If and when pkg_resources becomes a standard library, this routine
2192
# Check package name is within breezy
2193
if package == "breezy":
2194
resource_relpath = resource_name
2195
elif package.startswith("breezy."):
2196
package = package[len("breezy."):].replace('.', os.sep)
2197
resource_relpath = pathjoin(package, resource_name)
2199
raise errors.BzrError('resource package %s not in breezy' % package)
2201
# Map the resource to a file and read its contents
2202
base = dirname(breezy.__file__)
2203
if getattr(sys, 'frozen', None): # bzr.exe
2204
base = abspath(pathjoin(base, '..', '..'))
2205
f = file(pathjoin(base, resource_relpath), "rU")
2211
def file_kind_from_stat_mode_thunk(mode):
2212
global file_kind_from_stat_mode
2213
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2215
from ._readdir_pyx import UTF8DirReader
2216
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2217
except ImportError as e:
2218
# This is one time where we won't warn that an extension failed to
2219
# load. The extension is never available on Windows anyway.
2220
from ._readdir_py import (
2221
_kind_from_mode as file_kind_from_stat_mode
2223
return file_kind_from_stat_mode(mode)
2224
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2226
def file_stat(f, _lstat=os.lstat):
2230
except OSError as e:
2231
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2232
raise errors.NoSuchFile(f)
2235
def file_kind(f, _lstat=os.lstat):
2236
stat_value = file_stat(f, _lstat)
2237
return file_kind_from_stat_mode(stat_value.st_mode)
2239
def until_no_eintr(f, *a, **kw):
2240
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2242
WARNING: you must be certain that it is safe to retry the call repeatedly
2243
if EINTR does occur. This is typically only true for low-level operations
2244
like os.read. If in any doubt, don't use this.
2246
Keep in mind that this is not a complete solution to EINTR. There is
2247
probably code in the Python standard library and other dependencies that
2248
may encounter EINTR if a signal arrives (and there is signal handler for
2249
that signal). So this function can reduce the impact for IO that breezy
2250
directly controls, but it is not a complete solution.
2252
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2256
except (IOError, OSError) as e:
2257
if e.errno == errno.EINTR:
2262
if sys.platform == "win32":
2265
return msvcrt.getch()
2270
fd = sys.stdin.fileno()
2271
settings = termios.tcgetattr(fd)
2274
ch = sys.stdin.read(1)
2276
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2279
if sys.platform.startswith('linux'):
2280
def _local_concurrency():
2282
return os.sysconf('SC_NPROCESSORS_ONLN')
2283
except (ValueError, OSError, AttributeError):
2285
elif sys.platform == 'darwin':
2286
def _local_concurrency():
2287
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2288
stdout=subprocess.PIPE).communicate()[0]
2289
elif "bsd" in sys.platform:
2290
def _local_concurrency():
2291
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2292
stdout=subprocess.PIPE).communicate()[0]
2293
elif sys.platform == 'sunos5':
2294
def _local_concurrency():
2295
return subprocess.Popen(['psrinfo', '-p',],
2296
stdout=subprocess.PIPE).communicate()[0]
2297
elif sys.platform == "win32":
2298
def _local_concurrency():
2299
# This appears to return the number of cores.
2300
return os.environ.get('NUMBER_OF_PROCESSORS')
2302
def _local_concurrency():
2307
_cached_local_concurrency = None
2309
def local_concurrency(use_cache=True):
2310
"""Return how many processes can be run concurrently.
2312
Rely on platform specific implementations and default to 1 (one) if
2313
anything goes wrong.
2315
global _cached_local_concurrency
2317
if _cached_local_concurrency is not None and use_cache:
2318
return _cached_local_concurrency
2320
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2321
if concurrency is None:
2323
import multiprocessing
2324
concurrency = multiprocessing.cpu_count()
2325
except (ImportError, NotImplementedError):
2326
# multiprocessing is only available on Python >= 2.6
2327
# and multiprocessing.cpu_count() isn't implemented on all
2330
concurrency = _local_concurrency()
2331
except (OSError, IOError):
2334
concurrency = int(concurrency)
2335
except (TypeError, ValueError):
2338
_cached_concurrency = concurrency
2342
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2343
"""A stream writer that doesn't decode str arguments."""
2345
def __init__(self, encode, stream, errors='strict'):
2346
codecs.StreamWriter.__init__(self, stream, errors)
2347
self.encode = encode
2349
def write(self, object):
2350
if isinstance(object, str):
2351
self.stream.write(object)
2353
data, _ = self.encode(object, self.errors)
2354
self.stream.write(data)
2356
if sys.platform == 'win32':
2357
def open_file(filename, mode='r', bufsize=-1):
2358
"""This function is used to override the ``open`` builtin.
2360
But it uses O_NOINHERIT flag so the file handle is not inherited by
2361
child processes. Deleting or renaming a closed file opened with this
2362
function is not blocking child processes.
2364
writing = 'w' in mode
2365
appending = 'a' in mode
2366
updating = '+' in mode
2367
binary = 'b' in mode
2370
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2371
# for flags for each modes.
2381
flags |= os.O_WRONLY
2382
flags |= os.O_CREAT | os.O_TRUNC
2387
flags |= os.O_WRONLY
2388
flags |= os.O_CREAT | os.O_APPEND
2393
flags |= os.O_RDONLY
2395
return os.fdopen(os.open(filename, flags), mode, bufsize)
2400
def available_backup_name(base, exists):
2401
"""Find a non-existing backup file name.
2403
This will *not* create anything, this only return a 'free' entry. This
2404
should be used for checking names in a directory below a locked
2405
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2406
Leap) and generally discouraged.
2408
:param base: The base name.
2410
:param exists: A callable returning True if the path parameter exists.
2413
name = "%s.~%d~" % (base, counter)
2416
name = "%s.~%d~" % (base, counter)
2420
def set_fd_cloexec(fd):
2421
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2422
support for this is not available.
2426
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2427
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2428
except (ImportError, AttributeError):
2429
# Either the fcntl module or specific constants are not present
2433
def find_executable_on_path(name):
2434
"""Finds an executable on the PATH.
2436
On Windows, this will try to append each extension in the PATHEXT
2437
environment variable to the name, if it cannot be found with the name
2440
:param name: The base name of the executable.
2441
:return: The path to the executable found or None.
2443
if sys.platform == 'win32':
2444
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2445
exts = [ext.lower() for ext in exts]
2446
base, ext = os.path.splitext(name)
2448
if ext.lower() not in exts:
2454
path = os.environ.get('PATH')
2455
if path is not None:
2456
path = path.split(os.pathsep)
2459
f = os.path.join(d, name) + ext
2460
if os.access(f, os.X_OK):
2462
if sys.platform == 'win32':
2463
app_path = win32utils.get_app_path(name)
2464
if app_path != name:
2469
def _posix_is_local_pid_dead(pid):
2470
"""True if pid doesn't correspond to live process on this machine"""
2472
# Special meaning of unix kill: just check if it's there.
2474
except OSError as e:
2475
if e.errno == errno.ESRCH:
2476
# On this machine, and really not found: as sure as we can be
2479
elif e.errno == errno.EPERM:
2480
# exists, though not ours
2483
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2484
# Don't really know.
2487
# Exists and our process: not dead.
2490
if sys.platform == "win32":
2491
is_local_pid_dead = win32utils.is_local_pid_dead
2493
is_local_pid_dead = _posix_is_local_pid_dead
2495
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2496
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2497
if getattr(errno, name, None) is not None]
2500
def fdatasync(fileno):
2501
"""Flush file contents to disk if possible.
2503
:param fileno: Integer OS file handle.
2504
:raises TransportNotPossible: If flushing to disk is not possible.
2506
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2510
except IOError as e:
2511
# See bug #1075108, on some platforms fdatasync exists, but can
2512
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2513
# and reduce the chance of corruption-on-powerloss situations. It
2514
# is not a mandatory call, so it is ok to suppress failures.
2515
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2516
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2520
def ensure_empty_directory_exists(path, exception_class):
2521
"""Make sure a local directory exists and is empty.
2523
If it does not exist, it is created. If it exists and is not empty, an
2524
instance of exception_class is raised.
2528
except OSError as e:
2529
if e.errno != errno.EEXIST:
2531
if os.listdir(path) != []:
2532
raise exception_class(path)
2535
def is_environment_error(evalue):
2536
"""True if exception instance is due to a process environment issue
2538
This includes OSError and IOError, but also other errors that come from
2539
the operating system or core libraries but are not subclasses of those.
2541
if isinstance(evalue, (EnvironmentError, select.error)):
2543
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
added
removed
breezy/osutils.py
