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
35
# We need to import both shutil and rmtree as we export the later on posix
36
# and need the former on windows
38
from shutil import rmtree
41
# We need to import both tempfile and mkdtemp as we export the later on posix
42
# and need the former on windows
44
from tempfile import mkdtemp
52
from breezy.i18n import gettext
73
# On win32, O_BINARY is used to indicate the file should
74
# be opened in binary mode, rather than text mode.
75
# On other platforms, O_BINARY doesn't exist, because
76
# they always open in binary mode, so it is okay to
77
# OR with 0 on those platforms.
78
# O_NOINHERIT and O_TEXT exists only on win32 too.
79
O_BINARY = getattr(os, 'O_BINARY', 0)
80
O_TEXT = getattr(os, 'O_TEXT', 0)
81
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
84
class UnsupportedTimezoneFormat(errors.BzrError):
86
_fmt = ('Unsupported timezone format "%(timezone)s", '
87
'options are "utc", "original", "local".')
89
def __init__(self, timezone):
90
self.timezone = timezone
93
def get_unicode_argv():
97
user_encoding = get_user_encoding()
98
return [a.decode(user_encoding) for a in sys.argv[1:]]
99
except UnicodeDecodeError:
100
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
101
"application locale.").format(a, user_encoding))
104
def make_readonly(filename):
105
"""Make a filename read-only."""
106
mod = os.lstat(filename).st_mode
107
if not stat.S_ISLNK(mod):
109
chmod_if_possible(filename, mod)
112
def make_writable(filename):
113
mod = os.lstat(filename).st_mode
114
if not stat.S_ISLNK(mod):
116
chmod_if_possible(filename, mod)
119
def chmod_if_possible(filename, mode):
120
# Set file mode if that can be safely done.
121
# Sometimes even on unix the filesystem won't allow it - see
122
# https://bugs.launchpad.net/bzr/+bug/606537
124
# It is probably faster to just do the chmod, rather than
125
# doing a stat, and then trying to compare
126
os.chmod(filename, mode)
127
except (IOError, OSError) as e:
128
# Permission/access denied seems to commonly happen on smbfs; there's
129
# probably no point warning about it.
130
# <https://bugs.launchpad.net/bzr/+bug/606537>
131
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
132
trace.mutter("ignore error on chmod of %r: %r" % (
138
def minimum_path_selection(paths):
139
"""Return the smallset subset of paths which are outside paths.
141
:param paths: A container (and hence not None) of paths.
142
:return: A set of paths sufficient to include everything in paths via
143
is_inside, drawn from the paths parameter.
149
if isinstance(path, bytes):
150
return path.split(b'/')
152
return path.split('/')
153
sorted_paths = sorted(list(paths), key=sort_key)
155
search_paths = [sorted_paths[0]]
156
for path in sorted_paths[1:]:
157
if not is_inside(search_paths[-1], path):
158
# This path is unique, add it
159
search_paths.append(path)
161
return set(search_paths)
168
"""Return a quoted filename filename
170
This previously used backslash quoting, but that works poorly on
172
# TODO: I'm not really sure this is the best format either.x
174
if _QUOTE_RE is None:
175
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
177
if _QUOTE_RE.search(f):
183
_directory_kind = 'directory'
187
"""Return the current umask"""
188
# Assume that people aren't messing with the umask while running
189
# XXX: This is not thread safe, but there is no way to get the
190
# umask without setting it
198
_directory_kind: "/",
200
'tree-reference': '+',
204
def kind_marker(kind):
206
return _kind_marker_map[kind]
208
# Slightly faster than using .get(, '') when the common case is that
213
lexists = getattr(os.path, 'lexists', None)
217
stat = getattr(os, 'lstat', os.stat)
221
if e.errno == errno.ENOENT:
224
raise errors.BzrError(
225
gettext("lstat/stat of ({0!r}): {1!r}").format(f, e))
228
def fancy_rename(old, new, rename_func, unlink_func):
229
"""A fancy rename, when you don't have atomic rename.
231
:param old: The old path, to rename from
232
:param new: The new path, to rename to
233
:param rename_func: The potentially non-atomic rename function
234
:param unlink_func: A way to delete the target file if the full rename
237
# sftp rename doesn't allow overwriting, so play tricks:
238
base = os.path.basename(new)
239
dirname = os.path.dirname(new)
240
# callers use different encodings for the paths so the following MUST
241
# respect that. We rely on python upcasting to unicode if new is unicode
242
# and keeping a str if not.
243
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
244
os.getpid(), rand_chars(10))
245
tmp_name = pathjoin(dirname, tmp_name)
247
# Rename the file out of the way, but keep track if it didn't exist
248
# We don't want to grab just any exception
249
# something like EACCES should prevent us from continuing
250
# The downside is that the rename_func has to throw an exception
251
# with an errno = ENOENT, or NoSuchFile
254
rename_func(new, tmp_name)
255
except (errors.NoSuchFile,):
258
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
259
# function raises an IOError with errno is None when a rename fails.
260
# This then gets caught here.
261
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
263
except Exception as e:
264
if (getattr(e, 'errno', None) is None
265
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
272
# This may throw an exception, in which case success will
274
rename_func(old, new)
276
except (IOError, OSError) as e:
277
# source and target may be aliases of each other (e.g. on a
278
# case-insensitive filesystem), so we may have accidentally renamed
279
# source by when we tried to rename target
280
if (file_existed and e.errno in (None, errno.ENOENT)
281
and old.lower() == new.lower()):
282
# source and target are the same file on a case-insensitive
283
# filesystem, so we don't generate an exception
289
# If the file used to exist, rename it back into place
290
# otherwise just delete it from the tmp location
292
unlink_func(tmp_name)
294
rename_func(tmp_name, new)
297
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
298
# choke on a Unicode string containing a relative path if
299
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
301
def _posix_abspath(path):
302
# jam 20060426 rather than encoding to fsencoding
303
# copy posixpath.abspath, but use os.getcwdu instead
304
if not posixpath.isabs(path):
305
path = posixpath.join(getcwd(), path)
306
return _posix_normpath(path)
309
def _posix_realpath(path):
310
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
313
def _posix_normpath(path):
314
path = posixpath.normpath(path)
315
# Bug 861008: posixpath.normpath() returns a path normalized according to
316
# the POSIX standard, which stipulates (for compatibility reasons) that two
317
# leading slashes must not be simplified to one, and only if there are 3 or
318
# more should they be simplified as one. So we treat the leading 2 slashes
319
# as a special case here by simply removing the first slash, as we consider
320
# that breaking POSIX compatibility for this obscure feature is acceptable.
321
# This is not a paranoid precaution, as we notably get paths like this when
322
# the repo is hosted at the root of the filesystem, i.e. in "/".
323
if path.startswith('//'):
328
def _posix_path_from_environ(key):
329
"""Get unicode path from `key` in environment or None if not present
331
Note that posix systems use arbitrary byte strings for filesystem objects,
332
so a path that raises BadFilenameEncoding here may still be accessible.
334
val = os.environ.get(key, None)
335
if PY3 or val is None:
338
return val.decode(_fs_enc)
339
except UnicodeDecodeError:
340
# GZ 2011-12-12:Ideally want to include `key` in the exception message
341
raise errors.BadFilenameEncoding(val, _fs_enc)
344
def _posix_get_home_dir():
345
"""Get the home directory of the current user as a unicode path"""
346
path = posixpath.expanduser("~")
348
return path.decode(_fs_enc)
349
except AttributeError:
351
except UnicodeDecodeError:
352
raise errors.BadFilenameEncoding(path, _fs_enc)
355
def _posix_getuser_unicode():
356
"""Get username from environment or password database as unicode"""
357
name = getpass.getuser()
360
user_encoding = get_user_encoding()
362
return name.decode(user_encoding)
363
except UnicodeDecodeError:
364
raise errors.BzrError("Encoding of username %r is unsupported by %s "
365
"application locale." % (name, user_encoding))
368
def _win32_fixdrive(path):
369
"""Force drive letters to be consistent.
371
win32 is inconsistent whether it returns lower or upper case
372
and even if it was consistent the user might type the other
373
so we force it to uppercase
374
running python.exe under cmd.exe return capital C:\\
375
running win32 python inside a cygwin shell returns lowercase c:\\
377
drive, path = ntpath.splitdrive(path)
378
return drive.upper() + path
381
def _win32_abspath(path):
382
# Real ntpath.abspath doesn't have a problem with a unicode cwd
383
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
386
def _win32_realpath(path):
387
# Real ntpath.realpath doesn't have a problem with a unicode cwd
388
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
391
def _win32_pathjoin(*args):
392
return ntpath.join(*args).replace('\\', '/')
395
def _win32_normpath(path):
396
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
400
return _win32_fixdrive(_getcwd().replace('\\', '/'))
403
def _win32_mkdtemp(*args, **kwargs):
404
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
407
def _win32_rename(old, new):
408
"""We expect to be able to atomically replace 'new' with old.
410
On win32, if new exists, it must be moved out of the way first,
414
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
416
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
417
# If we try to rename a non-existant file onto cwd, we get
418
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
419
# if the old path doesn't exist, sometimes we get EACCES
420
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
426
return unicodedata.normalize('NFC', _getcwd())
429
def _rename_wrap_exception(rename_func):
430
"""Adds extra information to any exceptions that come from rename().
432
The exception has an updated message and 'old_filename' and 'new_filename'
436
def _rename_wrapper(old, new):
438
rename_func(old, new)
440
detailed_error = OSError(e.errno, e.strerror +
441
" [occurred when renaming '%s' to '%s']" %
443
detailed_error.old_filename = old
444
detailed_error.new_filename = new
447
return _rename_wrapper
450
if sys.version_info > (3,):
456
# Default rename wraps os.rename()
457
rename = _rename_wrap_exception(os.rename)
459
# Default is to just use the python builtins, but these can be rebound on
460
# particular platforms.
461
abspath = _posix_abspath
462
realpath = _posix_realpath
463
pathjoin = os.path.join
464
normpath = _posix_normpath
465
path_from_environ = _posix_path_from_environ
466
_get_home_dir = _posix_get_home_dir
467
getuser_unicode = _posix_getuser_unicode
469
dirname = os.path.dirname
470
basename = os.path.basename
471
split = os.path.split
472
splitext = os.path.splitext
473
# These were already lazily imported into local scope
474
# mkdtemp = tempfile.mkdtemp
475
# rmtree = shutil.rmtree
484
MIN_ABS_PATHLENGTH = 1
487
if sys.platform == 'win32':
488
abspath = _win32_abspath
489
realpath = _win32_realpath
490
pathjoin = _win32_pathjoin
491
normpath = _win32_normpath
492
getcwd = _win32_getcwd
493
mkdtemp = _win32_mkdtemp
494
rename = _rename_wrap_exception(_win32_rename)
496
from . import _walkdirs_win32
500
lstat = _walkdirs_win32.lstat
501
fstat = _walkdirs_win32.fstat
502
wrap_stat = _walkdirs_win32.wrap_stat
504
MIN_ABS_PATHLENGTH = 3
506
def _win32_delete_readonly(function, path, excinfo):
507
"""Error handler for shutil.rmtree function [for win32]
508
Helps to remove files and dirs marked as read-only.
510
exception = excinfo[1]
511
if function in (os.remove, os.rmdir) \
512
and isinstance(exception, OSError) \
513
and exception.errno == errno.EACCES:
519
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
520
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
521
return shutil.rmtree(path, ignore_errors, onerror)
523
f = win32utils.get_unicode_argv # special function or None
526
path_from_environ = win32utils.get_environ_unicode
527
_get_home_dir = win32utils.get_home_location
528
getuser_unicode = win32utils.get_user_name
530
elif sys.platform == 'darwin':
534
def get_terminal_encoding(trace=False):
535
"""Find the best encoding for printing to the screen.
537
This attempts to check both sys.stdout and sys.stdin to see
538
what encoding they are in, and if that fails it falls back to
539
osutils.get_user_encoding().
540
The problem is that on Windows, locale.getpreferredencoding()
541
is not the same encoding as that used by the console:
542
http://mail.python.org/pipermail/python-list/2003-May/162357.html
544
On my standard US Windows XP, the preferred encoding is
545
cp1252, but the console is cp437
547
:param trace: If True trace the selected encoding via mutter().
549
from .trace import mutter
550
output_encoding = getattr(sys.stdout, 'encoding', None)
551
if not output_encoding:
552
input_encoding = getattr(sys.stdin, 'encoding', None)
553
if not input_encoding:
554
output_encoding = get_user_encoding()
556
mutter('encoding stdout as osutils.get_user_encoding() %r',
559
output_encoding = input_encoding
561
mutter('encoding stdout as sys.stdin encoding %r',
565
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
566
if output_encoding == 'cp0':
567
# invalid encoding (cp0 means 'no codepage' on Windows)
568
output_encoding = get_user_encoding()
570
mutter('cp0 is invalid encoding.'
571
' encoding stdout as osutils.get_user_encoding() %r',
575
codecs.lookup(output_encoding)
577
sys.stderr.write('brz: warning:'
578
' unknown terminal encoding %s.\n'
579
' Using encoding %s instead.\n'
580
% (output_encoding, get_user_encoding())
582
output_encoding = get_user_encoding()
584
return output_encoding
587
def normalizepath(f):
588
if getattr(os.path, 'realpath', None) is not None:
592
[p, e] = os.path.split(f)
593
if e == "" or e == "." or e == "..":
596
return pathjoin(F(p), e)
600
"""True if f is an accessible directory."""
602
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
608
"""True if f is a regular file."""
610
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
616
"""True if f is a symlink."""
618
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
623
def is_inside(dir, fname):
624
"""True if fname is inside dir.
626
The parameters should typically be passed to osutils.normpath first, so
627
that . and .. and repeated slashes are eliminated, and the separators
628
are canonical for the platform.
630
The empty string as a dir name is taken as top-of-tree and matches
633
# XXX: Most callers of this can actually do something smarter by
634
# looking at the inventory
641
if isinstance(dir, bytes):
642
if not dir.endswith(b'/'):
645
if not dir.endswith('/'):
648
return fname.startswith(dir)
651
def is_inside_any(dir_list, fname):
652
"""True if fname is inside any of given dirs."""
653
for dirname in dir_list:
654
if is_inside(dirname, fname):
659
def is_inside_or_parent_of_any(dir_list, fname):
660
"""True if fname is a child or a parent of any of the given files."""
661
for dirname in dir_list:
662
if is_inside(dirname, fname) or is_inside(fname, dirname):
667
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
668
report_activity=None, direction='read'):
669
"""Copy contents of one file to another.
671
The read_length can either be -1 to read to end-of-file (EOF) or
672
it can specify the maximum number of bytes to read.
674
The buff_size represents the maximum size for each read operation
675
performed on from_file.
677
:param report_activity: Call this as bytes are read, see
678
Transport._report_activity
679
:param direction: Will be passed to report_activity
681
:return: The number of bytes copied.
685
# read specified number of bytes
687
while read_length > 0:
688
num_bytes_to_read = min(read_length, buff_size)
690
block = from_file.read(num_bytes_to_read)
694
if report_activity is not None:
695
report_activity(len(block), direction)
698
actual_bytes_read = len(block)
699
read_length -= actual_bytes_read
700
length += actual_bytes_read
704
block = from_file.read(buff_size)
708
if report_activity is not None:
709
report_activity(len(block), direction)
715
def pump_string_file(bytes, file_handle, segment_size=None):
716
"""Write bytes to file_handle in many smaller writes.
718
:param bytes: The string to write.
719
:param file_handle: The file to write to.
721
# Write data in chunks rather than all at once, because very large
722
# writes fail on some platforms (e.g. Windows with SMB mounted
725
segment_size = 5242880 # 5MB
726
offsets = range(0, len(bytes), segment_size)
727
view = memoryview(bytes)
728
write = file_handle.write
729
for offset in offsets:
730
write(view[offset:offset + segment_size])
733
def file_iterator(input_file, readsize=32768):
735
b = input_file.read(readsize)
741
# GZ 2017-09-16: Makes sense in general for hexdigest() result to be text, but
742
# used as bytes through most interfaces so encode with this wrapper.
744
def _hexdigest(hashobj):
745
return hashobj.hexdigest().encode()
747
def _hexdigest(hashobj):
748
return hashobj.hexdigest()
752
"""Calculate the hexdigest of an open file.
754
The file cursor should be already at the start.
766
def size_sha_file(f):
767
"""Calculate the size and hexdigest of an open file.
769
The file cursor should be already at the start and
770
the caller is responsible for closing the file afterwards.
781
return size, _hexdigest(s)
784
def sha_file_by_name(fname):
785
"""Calculate the SHA1 of a file by reading the full text"""
787
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
790
b = os.read(f, 1 << 16)
798
def sha_strings(strings, _factory=sha):
799
"""Return the sha-1 of concatenation of strings"""
801
for string in strings:
806
def sha_string(f, _factory=sha):
807
# GZ 2017-09-16: Dodgy if factory is ever not sha, probably shouldn't be.
808
return _hexdigest(_factory(f))
811
def fingerprint_file(f):
813
return {'size': len(b),
814
'sha1': _hexdigest(sha(b))}
817
def compare_files(a, b):
818
"""Returns true if equal in contents"""
829
def local_time_offset(t=None):
830
"""Return offset of local zone from GMT, either at present or at time t."""
833
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
834
return offset.days * 86400 + offset.seconds
837
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
838
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
841
def format_date(t, offset=0, timezone='original', date_fmt=None,
843
"""Return a formatted date string.
845
:param t: Seconds since the epoch.
846
:param offset: Timezone offset in seconds east of utc.
847
:param timezone: How to display the time: 'utc', 'original' for the
848
timezone specified by offset, or 'local' for the process's current
850
:param date_fmt: strftime format.
851
:param show_offset: Whether to append the timezone.
853
(date_fmt, tt, offset_str) = \
854
_format_date(t, offset, timezone, date_fmt, show_offset)
855
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
856
date_str = time.strftime(date_fmt, tt)
857
return date_str + offset_str
860
# Cache of formatted offset strings
864
def format_date_with_offset_in_original_timezone(t, offset=0,
865
_cache=_offset_cache):
866
"""Return a formatted date string in the original timezone.
868
This routine may be faster then format_date.
870
:param t: Seconds since the epoch.
871
:param offset: Timezone offset in seconds east of utc.
875
tt = time.gmtime(t + offset)
876
date_fmt = _default_format_by_weekday_num[tt[6]]
877
date_str = time.strftime(date_fmt, tt)
878
offset_str = _cache.get(offset, None)
879
if offset_str is None:
880
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
881
_cache[offset] = offset_str
882
return date_str + offset_str
885
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
887
"""Return an unicode date string formatted according to the current locale.
889
:param t: Seconds since the epoch.
890
:param offset: Timezone offset in seconds east of utc.
891
:param timezone: How to display the time: 'utc', 'original' for the
892
timezone specified by offset, or 'local' for the process's current
894
:param date_fmt: strftime format.
895
:param show_offset: Whether to append the timezone.
897
(date_fmt, tt, offset_str) = \
898
_format_date(t, offset, timezone, date_fmt, show_offset)
899
date_str = time.strftime(date_fmt, tt)
900
if not isinstance(date_str, text_type):
901
date_str = date_str.decode(get_user_encoding(), 'replace')
902
return date_str + offset_str
905
def _format_date(t, offset, timezone, date_fmt, show_offset):
906
if timezone == 'utc':
909
elif timezone == 'original':
912
tt = time.gmtime(t + offset)
913
elif timezone == 'local':
914
tt = time.localtime(t)
915
offset = local_time_offset(t)
917
raise UnsupportedTimezoneFormat(timezone)
919
date_fmt = "%a %Y-%m-%d %H:%M:%S"
921
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
924
return (date_fmt, tt, offset_str)
927
def compact_date(when):
928
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
931
def format_delta(delta):
932
"""Get a nice looking string for a time delta.
934
:param delta: The time difference in seconds, can be positive or negative.
935
positive indicates time in the past, negative indicates time in the
936
future. (usually time.time() - stored_time)
937
:return: String formatted to show approximate resolution
943
direction = 'in the future'
947
if seconds < 90: # print seconds up to 90 seconds
949
return '%d second %s' % (seconds, direction,)
951
return '%d seconds %s' % (seconds, direction)
953
minutes = int(seconds / 60)
954
seconds -= 60 * minutes
959
if minutes < 90: # print minutes, seconds up to 90 minutes
961
return '%d minute, %d second%s %s' % (
962
minutes, seconds, plural_seconds, direction)
964
return '%d minutes, %d second%s %s' % (
965
minutes, seconds, plural_seconds, direction)
967
hours = int(minutes / 60)
968
minutes -= 60 * hours
975
return '%d hour, %d minute%s %s' % (hours, minutes,
976
plural_minutes, direction)
977
return '%d hours, %d minute%s %s' % (hours, minutes,
978
plural_minutes, direction)
982
"""Return size of given open file."""
983
return os.fstat(f.fileno())[stat.ST_SIZE]
986
# Alias os.urandom to support platforms (which?) without /dev/urandom and
987
# override if it doesn't work. Avoid checking on windows where there is
988
# significant initialisation cost that can be avoided for some bzr calls.
990
rand_bytes = os.urandom
992
if rand_bytes.__module__ != "nt":
995
except NotImplementedError:
996
# not well seeded, but better than nothing
1001
s += chr(random.randint(0, 255))
1006
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
1009
def rand_chars(num):
1010
"""Return a random string of num alphanumeric characters
1012
The result only contains lowercase chars because it may be used on
1013
case-insensitive filesystems.
1016
for raw_byte in rand_bytes(num):
1018
s += ALNUM[ord(raw_byte) % 36]
1020
s += ALNUM[raw_byte % 36]
1024
# TODO: We could later have path objects that remember their list
1025
# decomposition (might be too tricksy though.)
1028
"""Turn string into list of parts."""
1029
if os.path.sep == '\\':
1030
# split on either delimiter because people might use either on
1032
if isinstance(p, bytes):
1033
ps = re.split(b'[\\\\/]', p)
1035
ps = re.split(r'[\\/]', p)
1037
if isinstance(p, bytes):
1044
if f in ('..', b'..'):
1045
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1046
elif f in ('.', '', b'.', b''):
1055
if (f == '..') or (f is None) or (f == ''):
1056
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1060
def parent_directories(filename):
1061
"""Return the list of parent directories, deepest first.
1063
For example, parent_directories("a/b/c") -> ["a/b", "a"].
1066
parts = splitpath(dirname(filename))
1068
parents.append(joinpath(parts))
1073
_extension_load_failures = []
1076
def failed_to_load_extension(exception):
1077
"""Handle failing to load a binary extension.
1079
This should be called from the ImportError block guarding the attempt to
1080
import the native extension. If this function returns, the pure-Python
1081
implementation should be loaded instead::
1084
>>> import breezy._fictional_extension_pyx
1085
>>> except ImportError, e:
1086
>>> breezy.osutils.failed_to_load_extension(e)
1087
>>> import breezy._fictional_extension_py
1089
# NB: This docstring is just an example, not a doctest, because doctest
1090
# currently can't cope with the use of lazy imports in this namespace --
1093
# This currently doesn't report the failure at the time it occurs, because
1094
# they tend to happen very early in startup when we can't check config
1095
# files etc, and also we want to report all failures but not spam the user
1097
exception_str = str(exception)
1098
if exception_str not in _extension_load_failures:
1099
trace.mutter("failed to load compiled extension: %s" % exception_str)
1100
_extension_load_failures.append(exception_str)
1103
def report_extension_load_failures():
1104
if not _extension_load_failures:
1106
if config.GlobalConfig().suppress_warning('missing_extensions'):
1108
# the warnings framework should by default show this only once
1109
from .trace import warning
1111
"brz: warning: some compiled extensions could not be loaded; "
1112
"see ``brz help missing-extensions``")
1113
# we no longer show the specific missing extensions here, because it makes
1114
# the message too long and scary - see
1115
# https://bugs.launchpad.net/bzr/+bug/430529
1119
from ._chunks_to_lines_pyx import chunks_to_lines
1120
except ImportError as e:
1121
failed_to_load_extension(e)
1122
from ._chunks_to_lines_py import chunks_to_lines
1126
"""Split s into lines, but without removing the newline characters."""
1127
# Trivially convert a fulltext into a 'chunked' representation, and let
1128
# chunks_to_lines do the heavy lifting.
1129
if isinstance(s, bytes):
1130
# chunks_to_lines only supports 8-bit strings
1131
return chunks_to_lines([s])
1133
return _split_lines(s)
1136
def _split_lines(s):
1137
"""Split s into lines, but without removing the newline characters.
1139
This supports Unicode or plain string objects.
1141
nl = b'\n' if isinstance(s, bytes) else u'\n'
1143
result = [line + nl for line in lines[:-1]]
1145
result.append(lines[-1])
1149
def hardlinks_good():
1150
return sys.platform not in ('win32', 'cygwin', 'darwin')
1153
def link_or_copy(src, dest):
1154
"""Hardlink a file, or copy it if it can't be hardlinked."""
1155
if not hardlinks_good():
1156
shutil.copyfile(src, dest)
1160
except (OSError, IOError) as e:
1161
if e.errno != errno.EXDEV:
1163
shutil.copyfile(src, dest)
1166
def delete_any(path):
1167
"""Delete a file, symlink or directory.
1169
Will delete even if readonly.
1172
_delete_file_or_dir(path)
1173
except (OSError, IOError) as e:
1174
if e.errno in (errno.EPERM, errno.EACCES):
1175
# make writable and try again
1178
except (OSError, IOError):
1180
_delete_file_or_dir(path)
1185
def _delete_file_or_dir(path):
1186
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1187
# Forgiveness than Permission (EAFP) because:
1188
# - root can damage a solaris file system by using unlink,
1189
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1190
# EACCES, OSX: EPERM) when invoked on a directory.
1191
if isdir(path): # Takes care of symlinks
1198
if getattr(os, 'symlink', None) is not None:
1204
def has_hardlinks():
1205
if getattr(os, 'link', None) is not None:
1211
def host_os_dereferences_symlinks():
1212
return (has_symlinks()
1213
and sys.platform not in ('cygwin', 'win32'))
1216
def readlink(abspath):
1217
"""Return a string representing the path to which the symbolic link points.
1219
:param abspath: The link absolute unicode path.
1221
This his guaranteed to return the symbolic link in unicode in all python
1224
link = abspath.encode(_fs_enc)
1225
target = os.readlink(link)
1226
target = target.decode(_fs_enc)
1230
def contains_whitespace(s):
1231
"""True if there are any whitespace characters in s."""
1232
# string.whitespace can include '\xa0' in certain locales, because it is
1233
# considered "non-breaking-space" as part of ISO-8859-1. But it
1234
# 1) Isn't a breaking whitespace
1235
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1237
# 3) '\xa0' isn't unicode safe since it is >128.
1239
if isinstance(s, str):
1242
ws = (b' ', b'\t', b'\n', b'\r', b'\v', b'\f')
1250
def contains_linebreaks(s):
1251
"""True if there is any vertical whitespace in s."""
1259
def relpath(base, path):
1260
"""Return path relative to base, or raise PathNotChild exception.
1262
The path may be either an absolute path or a path relative to the
1263
current working directory.
1265
os.path.commonprefix (python2.4) has a bad bug that it works just
1266
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1267
avoids that problem.
1269
NOTE: `base` should not have a trailing slash otherwise you'll get
1270
PathNotChild exceptions regardless of `path`.
1273
if len(base) < MIN_ABS_PATHLENGTH:
1274
# must have space for e.g. a drive letter
1275
raise ValueError(gettext('%r is too short to calculate a relative path')
1283
if len(head) <= len(base) and head != base:
1284
raise errors.PathNotChild(rp, base)
1287
head, tail = split(head)
1292
return pathjoin(*reversed(s))
1297
def _cicp_canonical_relpath(base, path):
1298
"""Return the canonical path relative to base.
1300
Like relpath, but on case-insensitive-case-preserving file-systems, this
1301
will return the relpath as stored on the file-system rather than in the
1302
case specified in the input string, for all existing portions of the path.
1304
This will cause O(N) behaviour if called for every path in a tree; if you
1305
have a number of paths to convert, you should use canonical_relpaths().
1307
# TODO: it should be possible to optimize this for Windows by using the
1308
# win32 API FindFiles function to look for the specified name - but using
1309
# os.listdir() still gives us the correct, platform agnostic semantics in
1312
rel = relpath(base, path)
1313
# '.' will have been turned into ''
1317
abs_base = abspath(base)
1319
_listdir = os.listdir
1321
# use an explicit iterator so we can easily consume the rest on early exit.
1322
bit_iter = iter(rel.split('/'))
1323
for bit in bit_iter:
1326
next_entries = _listdir(current)
1327
except OSError: # enoent, eperm, etc
1328
# We can't find this in the filesystem, so just append the
1330
current = pathjoin(current, bit, *list(bit_iter))
1332
for look in next_entries:
1333
if lbit == look.lower():
1334
current = pathjoin(current, look)
1337
# got to the end, nothing matched, so we just return the
1338
# non-existing bits as they were specified (the filename may be
1339
# the target of a move, for example).
1340
current = pathjoin(current, bit, *list(bit_iter))
1342
return current[len(abs_base):].lstrip('/')
1345
# XXX - TODO - we need better detection/integration of case-insensitive
1346
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1347
# filesystems), for example, so could probably benefit from the same basic
1348
# support there. For now though, only Windows and OSX get that support, and
1349
# they get it for *all* file-systems!
1350
if sys.platform in ('win32', 'darwin'):
1351
canonical_relpath = _cicp_canonical_relpath
1353
canonical_relpath = relpath
1356
def canonical_relpaths(base, paths):
1357
"""Create an iterable to canonicalize a sequence of relative paths.
1359
The intent is for this implementation to use a cache, vastly speeding
1360
up multiple transformations in the same directory.
1362
# but for now, we haven't optimized...
1363
return [canonical_relpath(base, p) for p in paths]
1366
def decode_filename(filename):
1367
"""Decode the filename using the filesystem encoding
1369
If it is unicode, it is returned.
1370
Otherwise it is decoded from the the filesystem's encoding. If decoding
1371
fails, a errors.BadFilenameEncoding exception is raised.
1373
if isinstance(filename, text_type):
1376
return filename.decode(_fs_enc)
1377
except UnicodeDecodeError:
1378
raise errors.BadFilenameEncoding(filename, _fs_enc)
1381
def safe_unicode(unicode_or_utf8_string):
1382
"""Coerce unicode_or_utf8_string into unicode.
1384
If it is unicode, it is returned.
1385
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1386
wrapped in a BzrBadParameterNotUnicode exception.
1388
if isinstance(unicode_or_utf8_string, text_type):
1389
return unicode_or_utf8_string
1391
return unicode_or_utf8_string.decode('utf8')
1392
except UnicodeDecodeError:
1393
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1396
def safe_utf8(unicode_or_utf8_string):
1397
"""Coerce unicode_or_utf8_string to a utf8 string.
1399
If it is a str, it is returned.
1400
If it is Unicode, it is encoded into a utf-8 string.
1402
if isinstance(unicode_or_utf8_string, bytes):
1403
# TODO: jam 20070209 This is overkill, and probably has an impact on
1404
# performance if we are dealing with lots of apis that want a
1407
# Make sure it is a valid utf-8 string
1408
unicode_or_utf8_string.decode('utf-8')
1409
except UnicodeDecodeError:
1410
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1411
return unicode_or_utf8_string
1412
return unicode_or_utf8_string.encode('utf-8')
1415
def safe_revision_id(unicode_or_utf8_string):
1416
"""Revision ids should now be utf8, but at one point they were unicode.
1418
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1420
:return: None or a utf8 revision id.
1422
if (unicode_or_utf8_string is None
1423
or unicode_or_utf8_string.__class__ == bytes):
1424
return unicode_or_utf8_string
1425
raise TypeError('Unicode revision ids are no longer supported. '
1426
'Revision id generators should be creating utf8 revision '
1430
def safe_file_id(unicode_or_utf8_string):
1431
"""File ids should now be utf8, but at one point they were unicode.
1433
This is the same as safe_utf8, except it uses the cached encode functions
1434
to save a little bit of performance.
1436
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1438
:return: None or a utf8 file id.
1440
if (unicode_or_utf8_string is None
1441
or unicode_or_utf8_string.__class__ == bytes):
1442
return unicode_or_utf8_string
1443
raise TypeError('Unicode file ids are no longer supported. '
1444
'File id generators should be creating utf8 file ids.')
1447
_platform_normalizes_filenames = False
1448
if sys.platform == 'darwin':
1449
_platform_normalizes_filenames = True
1452
def normalizes_filenames():
1453
"""Return True if this platform normalizes unicode filenames.
1457
return _platform_normalizes_filenames
1460
def _accessible_normalized_filename(path):
1461
"""Get the unicode normalized path, and if you can access the file.
1463
On platforms where the system normalizes filenames (Mac OSX),
1464
you can access a file by any path which will normalize correctly.
1465
On platforms where the system does not normalize filenames
1466
(everything else), you have to access a file by its exact path.
1468
Internally, bzr only supports NFC normalization, since that is
1469
the standard for XML documents.
1471
So return the normalized path, and a flag indicating if the file
1472
can be accessed by that path.
1475
if isinstance(path, bytes):
1476
path = path.decode(sys.getfilesystemencoding())
1477
return unicodedata.normalize('NFC', path), True
1480
def _inaccessible_normalized_filename(path):
1481
__doc__ = _accessible_normalized_filename.__doc__
1483
if isinstance(path, bytes):
1484
path = path.decode(sys.getfilesystemencoding())
1485
normalized = unicodedata.normalize('NFC', path)
1486
return normalized, normalized == path
1489
if _platform_normalizes_filenames:
1490
normalized_filename = _accessible_normalized_filename
1492
normalized_filename = _inaccessible_normalized_filename
1495
def set_signal_handler(signum, handler, restart_syscall=True):
1496
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1497
on platforms that support that.
1499
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1500
automatically restart (by calling `signal.siginterrupt(signum,
1501
False)`). May be ignored if the feature is not available on this
1502
platform or Python version.
1506
siginterrupt = signal.siginterrupt
1508
# This python implementation doesn't provide signal support, hence no
1511
except AttributeError:
1512
# siginterrupt doesn't exist on this platform, or for this version
1514
def siginterrupt(signum, flag): return None
1516
def sig_handler(*args):
1517
# Python resets the siginterrupt flag when a signal is
1518
# received. <http://bugs.python.org/issue8354>
1519
# As a workaround for some cases, set it back the way we want it.
1520
siginterrupt(signum, False)
1521
# Now run the handler function passed to set_signal_handler.
1524
sig_handler = handler
1525
old_handler = signal.signal(signum, sig_handler)
1527
siginterrupt(signum, False)
1531
default_terminal_width = 80
1532
"""The default terminal width for ttys.
1534
This is defined so that higher levels can share a common fallback value when
1535
terminal_width() returns None.
1538
# Keep some state so that terminal_width can detect if _terminal_size has
1539
# returned a different size since the process started. See docstring and
1540
# comments of terminal_width for details.
1541
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1542
_terminal_size_state = 'no_data'
1543
_first_terminal_size = None
1546
def terminal_width():
1547
"""Return terminal width.
1549
None is returned if the width can't established precisely.
1552
- if BRZ_COLUMNS is set, returns its value
1553
- if there is no controlling terminal, returns None
1554
- query the OS, if the queried size has changed since the last query,
1556
- if COLUMNS is set, returns its value,
1557
- if the OS has a value (even though it's never changed), return its value.
1559
From there, we need to query the OS to get the size of the controlling
1562
On Unices we query the OS by:
1563
- get termios.TIOCGWINSZ
1564
- if an error occurs or a negative value is obtained, returns None
1566
On Windows we query the OS by:
1567
- win32utils.get_console_size() decides,
1568
- returns None on error (provided default value)
1570
# Note to implementors: if changing the rules for determining the width,
1571
# make sure you've considered the behaviour in these cases:
1572
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1573
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1575
# - (add more interesting cases here, if you find any)
1576
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1577
# but we don't want to register a signal handler because it is impossible
1578
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1579
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1580
# time so we can notice if the reported size has changed, which should have
1583
# If BRZ_COLUMNS is set, take it, user is always right
1584
# Except if they specified 0 in which case, impose no limit here
1586
width = int(os.environ['BRZ_COLUMNS'])
1587
except (KeyError, ValueError):
1589
if width is not None:
1595
isatty = getattr(sys.stdout, 'isatty', None)
1596
if isatty is None or not isatty():
1597
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
1601
width, height = os_size = _terminal_size(None, None)
1602
global _first_terminal_size, _terminal_size_state
1603
if _terminal_size_state == 'no_data':
1604
_first_terminal_size = os_size
1605
_terminal_size_state = 'unchanged'
1606
elif (_terminal_size_state == 'unchanged' and
1607
_first_terminal_size != os_size):
1608
_terminal_size_state = 'changed'
1610
# If the OS claims to know how wide the terminal is, and this value has
1611
# ever changed, use that.
1612
if _terminal_size_state == 'changed':
1613
if width is not None and width > 0:
1616
# If COLUMNS is set, use it.
1618
return int(os.environ['COLUMNS'])
1619
except (KeyError, ValueError):
1622
# Finally, use an unchanged size from the OS, if we have one.
1623
if _terminal_size_state == 'unchanged':
1624
if width is not None and width > 0:
1627
# The width could not be determined.
1631
def _win32_terminal_size(width, height):
1632
width, height = win32utils.get_console_size(
1633
defaultx=width, defaulty=height)
1634
return width, height
1637
def _ioctl_terminal_size(width, height):
1642
s = struct.pack('HHHH', 0, 0, 0, 0)
1643
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1644
height, width = struct.unpack('HHHH', x)[0:2]
1645
except (IOError, AttributeError):
1647
return width, height
1650
_terminal_size = None
1651
"""Returns the terminal size as (width, height).
1653
:param width: Default value for width.
1654
:param height: Default value for height.
1656
This is defined specifically for each OS and query the size of the controlling
1657
terminal. If any error occurs, the provided default values should be returned.
1659
if sys.platform == 'win32':
1660
_terminal_size = _win32_terminal_size
1662
_terminal_size = _ioctl_terminal_size
1665
def supports_executable():
1666
return sys.platform != "win32"
1669
def supports_posix_readonly():
1670
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1672
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1673
directory controls creation/deletion, etc.
1675
And under win32, readonly means that the directory itself cannot be
1676
deleted. The contents of a readonly directory can be changed, unlike POSIX
1677
where files in readonly directories cannot be added, deleted or renamed.
1679
return sys.platform != "win32"
1682
def set_or_unset_env(env_variable, value):
1683
"""Modify the environment, setting or removing the env_variable.
1685
:param env_variable: The environment variable in question
1686
:param value: The value to set the environment to. If None, then
1687
the variable will be removed.
1688
:return: The original value of the environment variable.
1690
orig_val = os.environ.get(env_variable)
1692
if orig_val is not None:
1693
del os.environ[env_variable]
1695
if not PY3 and isinstance(value, text_type):
1696
value = value.encode(get_user_encoding())
1697
os.environ[env_variable] = value
1701
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1704
def check_legal_path(path):
1705
"""Check whether the supplied path is legal.
1706
This is only required on Windows, so we don't test on other platforms
1709
if sys.platform != "win32":
1711
if _validWin32PathRE.match(path) is None:
1712
raise errors.IllegalPath(path)
1715
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1718
def _is_error_enotdir(e):
1719
"""Check if this exception represents ENOTDIR.
1721
Unfortunately, python is very inconsistent about the exception
1722
here. The cases are:
1723
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1724
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1725
which is the windows error code.
1726
3) Windows, Python2.5 uses errno == EINVAL and
1727
winerror == ERROR_DIRECTORY
1729
:param e: An Exception object (expected to be OSError with an errno
1730
attribute, but we should be able to cope with anything)
1731
:return: True if this represents an ENOTDIR error. False otherwise.
1733
en = getattr(e, 'errno', None)
1734
if (en == errno.ENOTDIR or
1735
(sys.platform == 'win32' and
1736
(en == _WIN32_ERROR_DIRECTORY or
1738
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1744
def walkdirs(top, prefix=""):
1745
"""Yield data about all the directories in a tree.
1747
This yields all the data about the contents of a directory at a time.
1748
After each directory has been yielded, if the caller has mutated the list
1749
to exclude some directories, they are then not descended into.
1751
The data yielded is of the form:
1752
((directory-relpath, directory-path-from-top),
1753
[(relpath, basename, kind, lstat, path-from-top), ...]),
1754
- directory-relpath is the relative path of the directory being returned
1755
with respect to top. prefix is prepended to this.
1756
- directory-path-from-root is the path including top for this directory.
1757
It is suitable for use with os functions.
1758
- relpath is the relative path within the subtree being walked.
1759
- basename is the basename of the path
1760
- kind is the kind of the file now. If unknown then the file is not
1761
present within the tree - but it may be recorded as versioned. See
1763
- lstat is the stat data *if* the file was statted.
1764
- planned, not implemented:
1765
path_from_tree_root is the path from the root of the tree.
1767
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1768
allows one to walk a subtree but get paths that are relative to a tree
1770
:return: an iterator over the dirs.
1772
# TODO there is a bit of a smell where the results of the directory-
1773
# summary in this, and the path from the root, may not agree
1774
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1775
# potentially confusing output. We should make this more robust - but
1776
# not at a speed cost. RBC 20060731
1778
_directory = _directory_kind
1779
_listdir = os.listdir
1780
_kind_from_mode = file_kind_from_stat_mode
1781
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1783
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1784
relroot, _, _, _, top = pending.pop()
1786
relprefix = relroot + u'/'
1789
top_slash = top + u'/'
1792
append = dirblock.append
1794
names = sorted(map(decode_filename, _listdir(top)))
1795
except OSError as e:
1796
if not _is_error_enotdir(e):
1800
abspath = top_slash + name
1801
statvalue = _lstat(abspath)
1802
kind = _kind_from_mode(statvalue.st_mode)
1803
append((relprefix + name, name, kind, statvalue, abspath))
1804
yield (relroot, top), dirblock
1806
# push the user specified dirs from dirblock
1807
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1810
class DirReader(object):
1811
"""An interface for reading directories."""
1813
def top_prefix_to_starting_dir(self, top, prefix=""):
1814
"""Converts top and prefix to a starting dir entry
1816
:param top: A utf8 path
1817
:param prefix: An optional utf8 path to prefix output relative paths
1819
:return: A tuple starting with prefix, and ending with the native
1822
raise NotImplementedError(self.top_prefix_to_starting_dir)
1824
def read_dir(self, prefix, top):
1825
"""Read a specific dir.
1827
:param prefix: A utf8 prefix to be preprended to the path basenames.
1828
:param top: A natively encoded path to read.
1829
:return: A list of the directories contents. Each item contains:
1830
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1832
raise NotImplementedError(self.read_dir)
1835
_selected_dir_reader = None
1838
def _walkdirs_utf8(top, prefix=""):
1839
"""Yield data about all the directories in a tree.
1841
This yields the same information as walkdirs() only each entry is yielded
1842
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1843
are returned as exact byte-strings.
1845
:return: yields a tuple of (dir_info, [file_info])
1846
dir_info is (utf8_relpath, path-from-top)
1847
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1848
if top is an absolute path, path-from-top is also an absolute path.
1849
path-from-top might be unicode or utf8, but it is the correct path to
1850
pass to os functions to affect the file in question. (such as os.lstat)
1852
global _selected_dir_reader
1853
if _selected_dir_reader is None:
1854
if sys.platform == "win32":
1856
from ._walkdirs_win32 import Win32ReadDir
1857
_selected_dir_reader = Win32ReadDir()
1860
elif _fs_enc in ('utf-8', 'ascii'):
1862
from ._readdir_pyx import UTF8DirReader
1863
_selected_dir_reader = UTF8DirReader()
1864
except ImportError as e:
1865
failed_to_load_extension(e)
1868
if _selected_dir_reader is None:
1869
# Fallback to the python version
1870
_selected_dir_reader = UnicodeDirReader()
1872
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1873
# But we don't actually uses 1-3 in pending, so set them to None
1874
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1875
read_dir = _selected_dir_reader.read_dir
1876
_directory = _directory_kind
1878
relroot, _, _, _, top = pending[-1].pop()
1881
dirblock = sorted(read_dir(relroot, top))
1882
yield (relroot, top), dirblock
1883
# push the user specified dirs from dirblock
1884
next = [d for d in reversed(dirblock) if d[2] == _directory]
1886
pending.append(next)
1889
class UnicodeDirReader(DirReader):
1890
"""A dir reader for non-utf8 file systems, which transcodes."""
1892
__slots__ = ['_utf8_encode']
1895
self._utf8_encode = codecs.getencoder('utf8')
1897
def top_prefix_to_starting_dir(self, top, prefix=""):
1898
"""See DirReader.top_prefix_to_starting_dir."""
1899
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1901
def read_dir(self, prefix, top):
1902
"""Read a single directory from a non-utf8 file system.
1904
top, and the abspath element in the output are unicode, all other paths
1905
are utf8. Local disk IO is done via unicode calls to listdir etc.
1907
This is currently the fallback code path when the filesystem encoding is
1908
not UTF-8. It may be better to implement an alternative so that we can
1909
safely handle paths that are not properly decodable in the current
1912
See DirReader.read_dir for details.
1914
_utf8_encode = self._utf8_encode
1916
def _fs_decode(s): return s.decode(_fs_enc)
1918
def _fs_encode(s): return s.encode(_fs_enc)
1920
_listdir = os.listdir
1921
_kind_from_mode = file_kind_from_stat_mode
1924
relprefix = prefix + b'/'
1927
top_slash = top + '/'
1930
append = dirblock.append
1931
for name_native in _listdir(top.encode('utf-8')):
1933
name = _fs_decode(name_native)
1934
except UnicodeDecodeError:
1935
raise errors.BadFilenameEncoding(
1936
relprefix + name_native, _fs_enc)
1937
name_utf8 = _utf8_encode(name)[0]
1938
abspath = top_slash + name
1939
statvalue = _lstat(abspath)
1940
kind = _kind_from_mode(statvalue.st_mode)
1941
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1942
return sorted(dirblock)
1945
def copy_tree(from_path, to_path, handlers={}):
1946
"""Copy all of the entries in from_path into to_path.
1948
:param from_path: The base directory to copy.
1949
:param to_path: The target directory. If it does not exist, it will
1951
:param handlers: A dictionary of functions, which takes a source and
1952
destinations for files, directories, etc.
1953
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1954
'file', 'directory', and 'symlink' should always exist.
1955
If they are missing, they will be replaced with 'os.mkdir()',
1956
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1958
# Now, just copy the existing cached tree to the new location
1959
# We use a cheap trick here.
1960
# Absolute paths are prefixed with the first parameter
1961
# relative paths are prefixed with the second.
1962
# So we can get both the source and target returned
1963
# without any extra work.
1965
def copy_dir(source, dest):
1968
def copy_link(source, dest):
1969
"""Copy the contents of a symlink"""
1970
link_to = os.readlink(source)
1971
os.symlink(link_to, dest)
1973
real_handlers = {'file': shutil.copy2,
1974
'symlink': copy_link,
1975
'directory': copy_dir,
1977
real_handlers.update(handlers)
1979
if not os.path.exists(to_path):
1980
real_handlers['directory'](from_path, to_path)
1982
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1983
for relpath, name, kind, st, abspath in entries:
1984
real_handlers[kind](abspath, relpath)
1987
def copy_ownership_from_path(dst, src=None):
1988
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1990
If src is None, the containing directory is used as source. If chown
1991
fails, the error is ignored and a warning is printed.
1993
chown = getattr(os, 'chown', None)
1998
src = os.path.dirname(dst)
2004
chown(dst, s.st_uid, s.st_gid)
2007
'Unable to copy ownership from "%s" to "%s". '
2008
'You may want to set it manually.', src, dst)
2009
trace.log_exception_quietly()
2012
def path_prefix_key(path):
2013
"""Generate a prefix-order path key for path.
2015
This can be used to sort paths in the same way that walkdirs does.
2017
return (dirname(path), path)
2020
def compare_paths_prefix_order(path_a, path_b):
2021
"""Compare path_a and path_b to generate the same order walkdirs uses."""
2022
key_a = path_prefix_key(path_a)
2023
key_b = path_prefix_key(path_b)
2024
return (key_a > key_b) - (key_a < key_b)
2027
_cached_user_encoding = None
2030
def get_user_encoding():
2031
"""Find out what the preferred user encoding is.
2033
This is generally the encoding that is used for command line parameters
2034
and file contents. This may be different from the terminal encoding
2035
or the filesystem encoding.
2037
:return: A string defining the preferred user encoding
2039
global _cached_user_encoding
2040
if _cached_user_encoding is not None:
2041
return _cached_user_encoding
2043
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2044
# Use the existing locale settings and call nl_langinfo directly
2045
# rather than going through getpreferredencoding. This avoids
2046
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2047
# possibility of the setlocale call throwing an error.
2048
user_encoding = locale.nl_langinfo(locale.CODESET)
2050
# GZ 2011-12-19: On windows could call GetACP directly instead.
2051
user_encoding = locale.getpreferredencoding(False)
2054
user_encoding = codecs.lookup(user_encoding).name
2056
if user_encoding not in ("", "cp0"):
2057
sys.stderr.write('brz: warning:'
2058
' unknown encoding %s.'
2059
' Continuing with ascii encoding.\n'
2062
user_encoding = 'ascii'
2064
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2065
if user_encoding == 'ascii':
2066
if sys.platform == 'darwin':
2067
# OSX is special-cased in Python to have a UTF-8 filesystem
2068
# encoding and previously had LANG set here if not present.
2069
user_encoding = 'utf-8'
2070
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2071
# for some other posix platforms as well.
2073
_cached_user_encoding = user_encoding
2074
return user_encoding
2077
def get_diff_header_encoding():
2078
return get_terminal_encoding()
2081
def get_host_name():
2082
"""Return the current unicode host name.
2084
This is meant to be used in place of socket.gethostname() because that
2085
behaves inconsistently on different platforms.
2087
if sys.platform == "win32":
2088
return win32utils.get_host_name()
2092
return socket.gethostname()
2093
return socket.gethostname().decode(get_user_encoding())
2096
# We must not read/write any more than 64k at a time from/to a socket so we
2097
# don't risk "no buffer space available" errors on some platforms. Windows in
2098
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2100
MAX_SOCKET_CHUNK = 64 * 1024
2102
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2103
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2104
_eno = getattr(errno, _eno, None)
2105
if _eno is not None:
2106
_end_of_stream_errors.append(_eno)
2110
def read_bytes_from_socket(sock, report_activity=None,
2111
max_read_size=MAX_SOCKET_CHUNK):
2112
"""Read up to max_read_size of bytes from sock and notify of progress.
2114
Translates "Connection reset by peer" into file-like EOF (return an
2115
empty string rather than raise an error), and repeats the recv if
2116
interrupted by a signal.
2120
data = sock.recv(max_read_size)
2121
except socket.error as e:
2123
if eno in _end_of_stream_errors:
2124
# The connection was closed by the other side. Callers expect
2125
# an empty string to signal end-of-stream.
2127
elif eno == errno.EINTR:
2128
# Retry the interrupted recv.
2132
if report_activity is not None:
2133
report_activity(len(data), 'read')
2137
def recv_all(socket, count):
2138
"""Receive an exact number of bytes.
2140
Regular Socket.recv() may return less than the requested number of bytes,
2141
depending on what's in the OS buffer. MSG_WAITALL is not available
2142
on all platforms, but this should work everywhere. This will return
2143
less than the requested amount if the remote end closes.
2145
This isn't optimized and is intended mostly for use in testing.
2148
while len(b) < count:
2149
new = read_bytes_from_socket(socket, None, count - len(b))
2156
def send_all(sock, bytes, report_activity=None):
2157
"""Send all bytes on a socket.
2159
Breaks large blocks in smaller chunks to avoid buffering limitations on
2160
some platforms, and catches EINTR which may be thrown if the send is
2161
interrupted by a signal.
2163
This is preferred to socket.sendall(), because it avoids portability bugs
2164
and provides activity reporting.
2166
:param report_activity: Call this as bytes are read, see
2167
Transport._report_activity
2170
byte_count = len(bytes)
2171
view = memoryview(bytes)
2172
while sent_total < byte_count:
2174
sent = sock.send(view[sent_total:sent_total + MAX_SOCKET_CHUNK])
2175
except (socket.error, IOError) as e:
2176
if e.args[0] in _end_of_stream_errors:
2177
raise errors.ConnectionReset(
2178
"Error trying to write to socket", e)
2179
if e.args[0] != errno.EINTR:
2183
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2186
if report_activity is not None:
2187
report_activity(sent, 'write')
2190
def connect_socket(address):
2191
# Slight variation of the socket.create_connection() function (provided by
2192
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2193
# provide it for previous python versions. Also, we don't use the timeout
2194
# parameter (provided by the python implementation) so we don't implement
2196
err = socket.error('getaddrinfo returns an empty list')
2197
host, port = address
2198
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2199
af, socktype, proto, canonname, sa = res
2202
sock = socket.socket(af, socktype, proto)
2206
except socket.error as e:
2208
# 'err' is now the most recent error
2209
if sock is not None:
2214
def dereference_path(path):
2215
"""Determine the real path to a file.
2217
All parent elements are dereferenced. But the file itself is not
2219
:param path: The original path. May be absolute or relative.
2220
:return: the real path *to* the file
2222
parent, base = os.path.split(path)
2223
# The pathjoin for '.' is a workaround for Python bug #1213894.
2224
# (initial path components aren't dereferenced)
2225
return pathjoin(realpath(pathjoin('.', parent)), base)
2228
def supports_mapi():
2229
"""Return True if we can use MAPI to launch a mail client."""
2230
return sys.platform == "win32"
2233
def resource_string(package, resource_name):
2234
"""Load a resource from a package and return it as a string.
2236
Note: Only packages that start with breezy are currently supported.
2238
This is designed to be a lightweight implementation of resource
2239
loading in a way which is API compatible with the same API from
2241
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2242
If and when pkg_resources becomes a standard library, this routine
2245
# Check package name is within breezy
2246
if package == "breezy":
2247
resource_relpath = resource_name
2248
elif package.startswith("breezy."):
2249
package = package[len("breezy."):].replace('.', os.sep)
2250
resource_relpath = pathjoin(package, resource_name)
2252
raise errors.BzrError('resource package %s not in breezy' % package)
2254
# Map the resource to a file and read its contents
2255
base = dirname(breezy.__file__)
2256
if getattr(sys, 'frozen', None): # bzr.exe
2257
base = abspath(pathjoin(base, '..', '..'))
2258
with open(pathjoin(base, resource_relpath), "rt") as f:
2262
def file_kind_from_stat_mode_thunk(mode):
2263
global file_kind_from_stat_mode
2264
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2266
from ._readdir_pyx import UTF8DirReader
2267
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2269
# This is one time where we won't warn that an extension failed to
2270
# load. The extension is never available on Windows anyway.
2271
from ._readdir_py import (
2272
_kind_from_mode as file_kind_from_stat_mode
2274
return file_kind_from_stat_mode(mode)
2277
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2280
def file_stat(f, _lstat=os.lstat):
2284
except OSError as e:
2285
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2286
raise errors.NoSuchFile(f)
2290
def file_kind(f, _lstat=os.lstat):
2291
stat_value = file_stat(f, _lstat)
2292
return file_kind_from_stat_mode(stat_value.st_mode)
2295
def until_no_eintr(f, *a, **kw):
2296
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2298
WARNING: you must be certain that it is safe to retry the call repeatedly
2299
if EINTR does occur. This is typically only true for low-level operations
2300
like os.read. If in any doubt, don't use this.
2302
Keep in mind that this is not a complete solution to EINTR. There is
2303
probably code in the Python standard library and other dependencies that
2304
may encounter EINTR if a signal arrives (and there is signal handler for
2305
that signal). So this function can reduce the impact for IO that breezy
2306
directly controls, but it is not a complete solution.
2308
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2312
except (IOError, OSError) as e:
2313
if e.errno == errno.EINTR:
2318
if sys.platform == "win32":
2321
return msvcrt.getch()
2326
fd = sys.stdin.fileno()
2327
settings = termios.tcgetattr(fd)
2330
ch = sys.stdin.read(1)
2332
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2335
if sys.platform.startswith('linux'):
2336
def _local_concurrency():
2338
return os.sysconf('SC_NPROCESSORS_ONLN')
2339
except (ValueError, OSError, AttributeError):
2341
elif sys.platform == 'darwin':
2342
def _local_concurrency():
2343
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2344
stdout=subprocess.PIPE).communicate()[0]
2345
elif "bsd" in sys.platform:
2346
def _local_concurrency():
2347
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2348
stdout=subprocess.PIPE).communicate()[0]
2349
elif sys.platform == 'sunos5':
2350
def _local_concurrency():
2351
return subprocess.Popen(['psrinfo', '-p', ],
2352
stdout=subprocess.PIPE).communicate()[0]
2353
elif sys.platform == "win32":
2354
def _local_concurrency():
2355
# This appears to return the number of cores.
2356
return os.environ.get('NUMBER_OF_PROCESSORS')
2358
def _local_concurrency():
2363
_cached_local_concurrency = None
2366
def local_concurrency(use_cache=True):
2367
"""Return how many processes can be run concurrently.
2369
Rely on platform specific implementations and default to 1 (one) if
2370
anything goes wrong.
2372
global _cached_local_concurrency
2374
if _cached_local_concurrency is not None and use_cache:
2375
return _cached_local_concurrency
2377
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2378
if concurrency is None:
2379
import multiprocessing
2381
concurrency = multiprocessing.cpu_count()
2382
except NotImplementedError:
2383
# multiprocessing.cpu_count() isn't implemented on all platforms
2385
concurrency = _local_concurrency()
2386
except (OSError, IOError):
2389
concurrency = int(concurrency)
2390
except (TypeError, ValueError):
2393
_cached_local_concurrency = concurrency
2397
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2398
"""A stream writer that doesn't decode str arguments."""
2400
def __init__(self, encode, stream, errors='strict'):
2401
codecs.StreamWriter.__init__(self, stream, errors)
2402
self.encode = encode
2404
def write(self, object):
2405
if isinstance(object, str):
2406
self.stream.write(object)
2408
data, _ = self.encode(object, self.errors)
2409
self.stream.write(data)
2412
if sys.platform == 'win32':
2413
def open_file(filename, mode='r', bufsize=-1):
2414
"""This function is used to override the ``open`` builtin.
2416
But it uses O_NOINHERIT flag so the file handle is not inherited by
2417
child processes. Deleting or renaming a closed file opened with this
2418
function is not blocking child processes.
2420
writing = 'w' in mode
2421
appending = 'a' in mode
2422
updating = '+' in mode
2423
binary = 'b' in mode
2426
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2427
# for flags for each modes.
2437
flags |= os.O_WRONLY
2438
flags |= os.O_CREAT | os.O_TRUNC
2443
flags |= os.O_WRONLY
2444
flags |= os.O_CREAT | os.O_APPEND
2449
flags |= os.O_RDONLY
2451
return os.fdopen(os.open(filename, flags), mode, bufsize)
2456
def available_backup_name(base, exists):
2457
"""Find a non-existing backup file name.
2459
This will *not* create anything, this only return a 'free' entry. This
2460
should be used for checking names in a directory below a locked
2461
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2462
Leap) and generally discouraged.
2464
:param base: The base name.
2466
:param exists: A callable returning True if the path parameter exists.
2469
name = "%s.~%d~" % (base, counter)
2472
name = "%s.~%d~" % (base, counter)
2476
def set_fd_cloexec(fd):
2477
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2478
support for this is not available.
2482
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2483
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2484
except (ImportError, AttributeError):
2485
# Either the fcntl module or specific constants are not present
2489
def find_executable_on_path(name):
2490
"""Finds an executable on the PATH.
2492
On Windows, this will try to append each extension in the PATHEXT
2493
environment variable to the name, if it cannot be found with the name
2496
:param name: The base name of the executable.
2497
:return: The path to the executable found or None.
2499
if sys.platform == 'win32':
2500
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2501
exts = [ext.lower() for ext in exts]
2502
base, ext = os.path.splitext(name)
2504
if ext.lower() not in exts:
2510
path = os.environ.get('PATH')
2511
if path is not None:
2512
path = path.split(os.pathsep)
2515
f = os.path.join(d, name) + ext
2516
if os.access(f, os.X_OK):
2518
if sys.platform == 'win32':
2519
app_path = win32utils.get_app_path(name)
2520
if app_path != name:
2525
def _posix_is_local_pid_dead(pid):
2526
"""True if pid doesn't correspond to live process on this machine"""
2528
# Special meaning of unix kill: just check if it's there.
2530
except OSError as e:
2531
if e.errno == errno.ESRCH:
2532
# On this machine, and really not found: as sure as we can be
2535
elif e.errno == errno.EPERM:
2536
# exists, though not ours
2539
trace.mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2540
# Don't really know.
2543
# Exists and our process: not dead.
2547
if sys.platform == "win32":
2548
is_local_pid_dead = win32utils.is_local_pid_dead
2550
is_local_pid_dead = _posix_is_local_pid_dead
2552
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2553
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2554
if getattr(errno, name, None) is not None]
2557
def fdatasync(fileno):
2558
"""Flush file contents to disk if possible.
2560
:param fileno: Integer OS file handle.
2561
:raises TransportNotPossible: If flushing to disk is not possible.
2563
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2567
except IOError as e:
2568
# See bug #1075108, on some platforms fdatasync exists, but can
2569
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2570
# and reduce the chance of corruption-on-powerloss situations. It
2571
# is not a mandatory call, so it is ok to suppress failures.
2572
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2573
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2577
def ensure_empty_directory_exists(path, exception_class):
2578
"""Make sure a local directory exists and is empty.
2580
If it does not exist, it is created. If it exists and is not empty, an
2581
instance of exception_class is raised.
2585
except OSError as e:
2586
if e.errno != errno.EEXIST:
2588
if os.listdir(path) != []:
2589
raise exception_class(path)
2592
def is_environment_error(evalue):
2593
"""True if exception instance is due to a process environment issue
2595
This includes OSError and IOError, but also other errors that come from
2596
the operating system or core libraries but are not subclasses of those.
2598
if isinstance(evalue, (EnvironmentError, select.error)):
2600
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
2606
perf_counter = time.perf_counter
2608
perf_counter = time.clock
added
removed
breezy/osutils.py
