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
# Cross platform wall-clock time functionality with decent resolution.
74
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
75
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
76
# synchronized with ``time.time()``, this is only meant to be used to find
77
# delta times by subtracting from another call to this function.
78
timer_func = time.time
79
if sys.platform == 'win32':
80
timer_func = time.clock
82
# On win32, O_BINARY is used to indicate the file should
83
# be opened in binary mode, rather than text mode.
84
# On other platforms, O_BINARY doesn't exist, because
85
# they always open in binary mode, so it is okay to
86
# OR with 0 on those platforms.
87
# O_NOINHERIT and O_TEXT exists only on win32 too.
88
O_BINARY = getattr(os, 'O_BINARY', 0)
89
O_TEXT = getattr(os, 'O_TEXT', 0)
90
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
93
class UnsupportedTimezoneFormat(errors.BzrError):
95
_fmt = ('Unsupported timezone format "%(timezone)s", '
96
'options are "utc", "original", "local".')
98
def __init__(self, timezone):
99
self.timezone = timezone
102
def get_unicode_argv():
106
user_encoding = get_user_encoding()
107
return [a.decode(user_encoding) for a in sys.argv[1:]]
108
except UnicodeDecodeError:
109
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
110
"application locale.").format(a, user_encoding))
113
def make_readonly(filename):
114
"""Make a filename read-only."""
115
mod = os.lstat(filename).st_mode
116
if not stat.S_ISLNK(mod):
118
chmod_if_possible(filename, mod)
121
def make_writable(filename):
122
mod = os.lstat(filename).st_mode
123
if not stat.S_ISLNK(mod):
125
chmod_if_possible(filename, mod)
128
def chmod_if_possible(filename, mode):
129
# Set file mode if that can be safely done.
130
# Sometimes even on unix the filesystem won't allow it - see
131
# https://bugs.launchpad.net/bzr/+bug/606537
133
# It is probably faster to just do the chmod, rather than
134
# doing a stat, and then trying to compare
135
os.chmod(filename, mode)
136
except (IOError, OSError) as e:
137
# Permission/access denied seems to commonly happen on smbfs; there's
138
# probably no point warning about it.
139
# <https://bugs.launchpad.net/bzr/+bug/606537>
140
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
141
trace.mutter("ignore error on chmod of %r: %r" % (
147
def minimum_path_selection(paths):
148
"""Return the smallset subset of paths which are outside paths.
150
:param paths: A container (and hence not None) of paths.
151
:return: A set of paths sufficient to include everything in paths via
152
is_inside, drawn from the paths parameter.
158
return path.split('/')
159
sorted_paths = sorted(list(paths), key=sort_key)
161
search_paths = [sorted_paths[0]]
162
for path in sorted_paths[1:]:
163
if not is_inside(search_paths[-1], path):
164
# This path is unique, add it
165
search_paths.append(path)
167
return set(search_paths)
174
"""Return a quoted filename filename
176
This previously used backslash quoting, but that works poorly on
178
# TODO: I'm not really sure this is the best format either.x
180
if _QUOTE_RE is None:
181
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
183
if _QUOTE_RE.search(f):
189
_directory_kind = 'directory'
192
"""Return the current umask"""
193
# Assume that people aren't messing with the umask while running
194
# XXX: This is not thread safe, but there is no way to get the
195
# umask without setting it
203
_directory_kind: "/",
205
'tree-reference': '+',
209
def kind_marker(kind):
211
return _kind_marker_map[kind]
213
# Slightly faster than using .get(, '') when the common case is that
218
lexists = getattr(os.path, 'lexists', None)
222
stat = getattr(os, 'lstat', os.stat)
226
if e.errno == errno.ENOENT:
229
raise errors.BzrError(gettext("lstat/stat of ({0!r}): {1!r}").format(f, e))
232
def fancy_rename(old, new, rename_func, unlink_func):
233
"""A fancy rename, when you don't have atomic rename.
235
:param old: The old path, to rename from
236
:param new: The new path, to rename to
237
:param rename_func: The potentially non-atomic rename function
238
:param unlink_func: A way to delete the target file if the full rename
241
# sftp rename doesn't allow overwriting, so play tricks:
242
base = os.path.basename(new)
243
dirname = os.path.dirname(new)
244
# callers use different encodings for the paths so the following MUST
245
# respect that. We rely on python upcasting to unicode if new is unicode
246
# and keeping a str if not.
247
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
248
os.getpid(), rand_chars(10))
249
tmp_name = pathjoin(dirname, tmp_name)
251
# Rename the file out of the way, but keep track if it didn't exist
252
# We don't want to grab just any exception
253
# something like EACCES should prevent us from continuing
254
# The downside is that the rename_func has to throw an exception
255
# with an errno = ENOENT, or NoSuchFile
258
rename_func(new, tmp_name)
259
except (errors.NoSuchFile,) as e:
262
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
263
# function raises an IOError with errno is None when a rename fails.
264
# This then gets caught here.
265
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
267
except Exception as e:
268
if (getattr(e, 'errno', None) is None
269
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
276
# This may throw an exception, in which case success will
278
rename_func(old, new)
280
except (IOError, OSError) as e:
281
# source and target may be aliases of each other (e.g. on a
282
# case-insensitive filesystem), so we may have accidentally renamed
283
# source by when we tried to rename target
284
if (file_existed and e.errno in (None, errno.ENOENT)
285
and old.lower() == new.lower()):
286
# source and target are the same file on a case-insensitive
287
# filesystem, so we don't generate an exception
293
# If the file used to exist, rename it back into place
294
# otherwise just delete it from the tmp location
296
unlink_func(tmp_name)
298
rename_func(tmp_name, new)
301
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
302
# choke on a Unicode string containing a relative path if
303
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
305
def _posix_abspath(path):
306
# jam 20060426 rather than encoding to fsencoding
307
# copy posixpath.abspath, but use os.getcwdu instead
308
if not posixpath.isabs(path):
309
path = posixpath.join(getcwd(), path)
310
return _posix_normpath(path)
313
def _posix_realpath(path):
314
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
317
def _posix_normpath(path):
318
path = posixpath.normpath(path)
319
# Bug 861008: posixpath.normpath() returns a path normalized according to
320
# the POSIX standard, which stipulates (for compatibility reasons) that two
321
# leading slashes must not be simplified to one, and only if there are 3 or
322
# more should they be simplified as one. So we treat the leading 2 slashes
323
# as a special case here by simply removing the first slash, as we consider
324
# that breaking POSIX compatibility for this obscure feature is acceptable.
325
# This is not a paranoid precaution, as we notably get paths like this when
326
# the repo is hosted at the root of the filesystem, i.e. in "/".
327
if path.startswith('//'):
332
def _posix_path_from_environ(key):
333
"""Get unicode path from `key` in environment or None if not present
335
Note that posix systems use arbitrary byte strings for filesystem objects,
336
so a path that raises BadFilenameEncoding here may still be accessible.
338
val = os.environ.get(key, None)
339
if PY3 or val is None:
342
return val.decode(_fs_enc)
343
except UnicodeDecodeError:
344
# GZ 2011-12-12:Ideally want to include `key` in the exception message
345
raise errors.BadFilenameEncoding(val, _fs_enc)
348
def _posix_get_home_dir():
349
"""Get the home directory of the current user as a unicode path"""
350
path = posixpath.expanduser("~")
352
return path.decode(_fs_enc)
353
except AttributeError:
355
except UnicodeDecodeError:
356
raise errors.BadFilenameEncoding(path, _fs_enc)
359
def _posix_getuser_unicode():
360
"""Get username from environment or password database as unicode"""
361
name = getpass.getuser()
364
user_encoding = get_user_encoding()
366
return name.decode(user_encoding)
367
except UnicodeDecodeError:
368
raise errors.BzrError("Encoding of username %r is unsupported by %s "
369
"application locale." % (name, user_encoding))
372
def _win32_fixdrive(path):
373
"""Force drive letters to be consistent.
375
win32 is inconsistent whether it returns lower or upper case
376
and even if it was consistent the user might type the other
377
so we force it to uppercase
378
running python.exe under cmd.exe return capital C:\\
379
running win32 python inside a cygwin shell returns lowercase c:\\
381
drive, path = ntpath.splitdrive(path)
382
return drive.upper() + path
385
def _win32_abspath(path):
386
# Real ntpath.abspath doesn't have a problem with a unicode cwd
387
return _win32_fixdrive(ntpath.abspath(path).replace('\\', '/'))
390
def _win32_realpath(path):
391
# Real ntpath.realpath doesn't have a problem with a unicode cwd
392
return _win32_fixdrive(ntpath.realpath(path).replace('\\', '/'))
395
def _win32_pathjoin(*args):
396
return ntpath.join(*args).replace('\\', '/')
399
def _win32_normpath(path):
400
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
404
return _win32_fixdrive(_getcwd().replace('\\', '/'))
407
def _win32_mkdtemp(*args, **kwargs):
408
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
411
def _win32_rename(old, new):
412
"""We expect to be able to atomically replace 'new' with old.
414
On win32, if new exists, it must be moved out of the way first,
418
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
420
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
421
# If we try to rename a non-existant file onto cwd, we get
422
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
423
# if the old path doesn't exist, sometimes we get EACCES
424
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
430
return unicodedata.normalize('NFC', _getcwd())
433
def _rename_wrap_exception(rename_func):
434
"""Adds extra information to any exceptions that come from rename().
436
The exception has an updated message and 'old_filename' and 'new_filename'
440
def _rename_wrapper(old, new):
442
rename_func(old, new)
444
detailed_error = OSError(e.errno, e.strerror +
445
" [occurred when renaming '%s' to '%s']" %
447
detailed_error.old_filename = old
448
detailed_error.new_filename = new
451
return _rename_wrapper
454
if sys.version_info > (3,):
460
# Default rename wraps os.rename()
461
rename = _rename_wrap_exception(os.rename)
463
# Default is to just use the python builtins, but these can be rebound on
464
# particular platforms.
465
abspath = _posix_abspath
466
realpath = _posix_realpath
467
pathjoin = os.path.join
468
normpath = _posix_normpath
469
path_from_environ = _posix_path_from_environ
470
_get_home_dir = _posix_get_home_dir
471
getuser_unicode = _posix_getuser_unicode
473
dirname = os.path.dirname
474
basename = os.path.basename
475
split = os.path.split
476
splitext = os.path.splitext
477
# These were already lazily imported into local scope
478
# mkdtemp = tempfile.mkdtemp
479
# rmtree = shutil.rmtree
487
MIN_ABS_PATHLENGTH = 1
490
if sys.platform == 'win32':
491
abspath = _win32_abspath
492
realpath = _win32_realpath
493
pathjoin = _win32_pathjoin
494
normpath = _win32_normpath
495
getcwd = _win32_getcwd
496
mkdtemp = _win32_mkdtemp
497
rename = _rename_wrap_exception(_win32_rename)
499
from . import _walkdirs_win32
503
lstat = _walkdirs_win32.lstat
504
fstat = _walkdirs_win32.fstat
505
wrap_stat = _walkdirs_win32.wrap_stat
507
MIN_ABS_PATHLENGTH = 3
509
def _win32_delete_readonly(function, path, excinfo):
510
"""Error handler for shutil.rmtree function [for win32]
511
Helps to remove files and dirs marked as read-only.
513
exception = excinfo[1]
514
if function in (os.remove, os.rmdir) \
515
and isinstance(exception, OSError) \
516
and exception.errno == errno.EACCES:
522
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
523
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
524
return shutil.rmtree(path, ignore_errors, onerror)
526
f = win32utils.get_unicode_argv # special function or None
529
path_from_environ = win32utils.get_environ_unicode
530
_get_home_dir = win32utils.get_home_location
531
getuser_unicode = win32utils.get_user_name
533
elif sys.platform == 'darwin':
537
def get_terminal_encoding(trace=False):
538
"""Find the best encoding for printing to the screen.
540
This attempts to check both sys.stdout and sys.stdin to see
541
what encoding they are in, and if that fails it falls back to
542
osutils.get_user_encoding().
543
The problem is that on Windows, locale.getpreferredencoding()
544
is not the same encoding as that used by the console:
545
http://mail.python.org/pipermail/python-list/2003-May/162357.html
547
On my standard US Windows XP, the preferred encoding is
548
cp1252, but the console is cp437
550
:param trace: If True trace the selected encoding via mutter().
552
from .trace import mutter
553
output_encoding = getattr(sys.stdout, 'encoding', None)
554
if not output_encoding:
555
input_encoding = getattr(sys.stdin, 'encoding', None)
556
if not input_encoding:
557
output_encoding = get_user_encoding()
559
mutter('encoding stdout as osutils.get_user_encoding() %r',
562
output_encoding = input_encoding
564
mutter('encoding stdout as sys.stdin encoding %r',
568
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
569
if output_encoding == 'cp0':
570
# invalid encoding (cp0 means 'no codepage' on Windows)
571
output_encoding = get_user_encoding()
573
mutter('cp0 is invalid encoding.'
574
' encoding stdout as osutils.get_user_encoding() %r',
578
codecs.lookup(output_encoding)
580
sys.stderr.write('brz: warning:'
581
' unknown terminal encoding %s.\n'
582
' Using encoding %s instead.\n'
583
% (output_encoding, get_user_encoding())
585
output_encoding = get_user_encoding()
587
return output_encoding
590
def normalizepath(f):
591
if getattr(os.path, 'realpath', None) is not None:
595
[p, e] = os.path.split(f)
596
if e == "" or e == "." or e == "..":
599
return pathjoin(F(p), e)
603
"""True if f is an accessible directory."""
605
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
611
"""True if f is a regular file."""
613
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
618
"""True if f is a symlink."""
620
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
624
def is_inside(dir, fname):
625
"""True if fname is inside dir.
627
The parameters should typically be passed to osutils.normpath first, so
628
that . and .. and repeated slashes are eliminated, and the separators
629
are canonical for the platform.
631
The empty string as a dir name is taken as top-of-tree and matches
634
# XXX: Most callers of this can actually do something smarter by
635
# looking at the inventory
642
if not dir.endswith('/'):
645
return fname.startswith(dir)
648
def is_inside_any(dir_list, fname):
649
"""True if fname is inside any of given dirs."""
650
for dirname in dir_list:
651
if is_inside(dirname, fname):
656
def is_inside_or_parent_of_any(dir_list, fname):
657
"""True if fname is a child or a parent of any of the given files."""
658
for dirname in dir_list:
659
if is_inside(dirname, fname) or is_inside(fname, dirname):
664
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
665
report_activity=None, direction='read'):
666
"""Copy contents of one file to another.
668
The read_length can either be -1 to read to end-of-file (EOF) or
669
it can specify the maximum number of bytes to read.
671
The buff_size represents the maximum size for each read operation
672
performed on from_file.
674
:param report_activity: Call this as bytes are read, see
675
Transport._report_activity
676
:param direction: Will be passed to report_activity
678
:return: The number of bytes copied.
682
# read specified number of bytes
684
while read_length > 0:
685
num_bytes_to_read = min(read_length, buff_size)
687
block = from_file.read(num_bytes_to_read)
691
if report_activity is not None:
692
report_activity(len(block), direction)
695
actual_bytes_read = len(block)
696
read_length -= actual_bytes_read
697
length += actual_bytes_read
701
block = from_file.read(buff_size)
705
if report_activity is not None:
706
report_activity(len(block), direction)
712
def pump_string_file(bytes, file_handle, segment_size=None):
713
"""Write bytes to file_handle in many smaller writes.
715
:param bytes: The string to write.
716
:param file_handle: The file to write to.
718
# Write data in chunks rather than all at once, because very large
719
# writes fail on some platforms (e.g. Windows with SMB mounted
722
segment_size = 5242880 # 5MB
723
offsets = range(0, len(bytes), segment_size)
724
view = memoryview(bytes)
725
write = file_handle.write
726
for offset in offsets:
727
write(view[offset:offset+segment_size])
730
def file_iterator(input_file, readsize=32768):
732
b = input_file.read(readsize)
738
# GZ 2017-09-16: Makes sense in general for hexdigest() result to be text, but
739
# used as bytes through most interfaces so encode with this wrapper.
741
def _hexdigest(hashobj):
742
return hashobj.hexdigest().encode()
744
def _hexdigest(hashobj):
745
return hashobj.hexdigest()
749
"""Calculate the hexdigest of an open file.
751
The file cursor should be already at the start.
763
def size_sha_file(f):
764
"""Calculate the size and hexdigest of an open file.
766
The file cursor should be already at the start and
767
the caller is responsible for closing the file afterwards.
778
return size, _hexdigest(s)
781
def sha_file_by_name(fname):
782
"""Calculate the SHA1 of a file by reading the full text"""
784
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
787
b = os.read(f, 1<<16)
795
def sha_strings(strings, _factory=sha):
796
"""Return the sha-1 of concatenation of strings"""
798
for string in strings:
803
def sha_string(f, _factory=sha):
804
# GZ 2017-09-16: Dodgy if factory is ever not sha, probably shouldn't be.
805
return _hexdigest(_factory(f))
808
def fingerprint_file(f):
810
return {'size': len(b),
811
'sha1': _hexdigest(sha(b))}
814
def compare_files(a, b):
815
"""Returns true if equal in contents"""
826
def local_time_offset(t=None):
827
"""Return offset of local zone from GMT, either at present or at time t."""
830
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
831
return offset.days * 86400 + offset.seconds
833
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
834
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
837
def format_date(t, offset=0, timezone='original', date_fmt=None,
839
"""Return a formatted date string.
841
:param t: Seconds since the epoch.
842
:param offset: Timezone offset in seconds east of utc.
843
:param timezone: How to display the time: 'utc', 'original' for the
844
timezone specified by offset, or 'local' for the process's current
846
:param date_fmt: strftime format.
847
:param show_offset: Whether to append the timezone.
849
(date_fmt, tt, offset_str) = \
850
_format_date(t, offset, timezone, date_fmt, show_offset)
851
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
852
date_str = time.strftime(date_fmt, tt)
853
return date_str + offset_str
856
# Cache of formatted offset strings
860
def format_date_with_offset_in_original_timezone(t, offset=0,
861
_cache=_offset_cache):
862
"""Return a formatted date string in the original timezone.
864
This routine may be faster then format_date.
866
:param t: Seconds since the epoch.
867
:param offset: Timezone offset in seconds east of utc.
871
tt = time.gmtime(t + offset)
872
date_fmt = _default_format_by_weekday_num[tt[6]]
873
date_str = time.strftime(date_fmt, tt)
874
offset_str = _cache.get(offset, None)
875
if offset_str is None:
876
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
877
_cache[offset] = offset_str
878
return date_str + offset_str
881
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
883
"""Return an unicode date string formatted according to the current locale.
885
:param t: Seconds since the epoch.
886
:param offset: Timezone offset in seconds east of utc.
887
:param timezone: How to display the time: 'utc', 'original' for the
888
timezone specified by offset, or 'local' for the process's current
890
:param date_fmt: strftime format.
891
:param show_offset: Whether to append the timezone.
893
(date_fmt, tt, offset_str) = \
894
_format_date(t, offset, timezone, date_fmt, show_offset)
895
date_str = time.strftime(date_fmt, tt)
896
if not isinstance(date_str, text_type):
897
date_str = date_str.decode(get_user_encoding(), 'replace')
898
return date_str + offset_str
901
def _format_date(t, offset, timezone, date_fmt, show_offset):
902
if timezone == 'utc':
905
elif timezone == 'original':
908
tt = time.gmtime(t + offset)
909
elif timezone == 'local':
910
tt = time.localtime(t)
911
offset = local_time_offset(t)
913
raise UnsupportedTimezoneFormat(timezone)
915
date_fmt = "%a %Y-%m-%d %H:%M:%S"
917
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
920
return (date_fmt, tt, offset_str)
923
def compact_date(when):
924
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
927
def format_delta(delta):
928
"""Get a nice looking string for a time delta.
930
:param delta: The time difference in seconds, can be positive or negative.
931
positive indicates time in the past, negative indicates time in the
932
future. (usually time.time() - stored_time)
933
:return: String formatted to show approximate resolution
939
direction = 'in the future'
943
if seconds < 90: # print seconds up to 90 seconds
945
return '%d second %s' % (seconds, direction,)
947
return '%d seconds %s' % (seconds, direction)
949
minutes = int(seconds / 60)
950
seconds -= 60 * minutes
955
if minutes < 90: # print minutes, seconds up to 90 minutes
957
return '%d minute, %d second%s %s' % (
958
minutes, seconds, plural_seconds, direction)
960
return '%d minutes, %d second%s %s' % (
961
minutes, seconds, plural_seconds, direction)
963
hours = int(minutes / 60)
964
minutes -= 60 * hours
971
return '%d hour, %d minute%s %s' % (hours, minutes,
972
plural_minutes, direction)
973
return '%d hours, %d minute%s %s' % (hours, minutes,
974
plural_minutes, direction)
977
"""Return size of given open file."""
978
return os.fstat(f.fileno())[stat.ST_SIZE]
981
# Alias os.urandom to support platforms (which?) without /dev/urandom and
982
# override if it doesn't work. Avoid checking on windows where there is
983
# significant initialisation cost that can be avoided for some bzr calls.
985
rand_bytes = os.urandom
987
if rand_bytes.__module__ != "nt":
990
except NotImplementedError:
991
# not well seeded, but better than nothing
996
s += chr(random.randint(0, 255))
1001
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
1002
def rand_chars(num):
1003
"""Return a random string of num alphanumeric characters
1005
The result only contains lowercase chars because it may be used on
1006
case-insensitive filesystems.
1009
for raw_byte in rand_bytes(num):
1011
s += ALNUM[ord(raw_byte) % 36]
1013
s += ALNUM[raw_byte % 36]
1017
## TODO: We could later have path objects that remember their list
1018
## decomposition (might be too tricksy though.)
1021
"""Turn string into list of parts."""
1022
# split on either delimiter because people might use either on
1024
ps = re.split(r'[\\/]', p)
1029
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1030
elif (f == '.') or (f == ''):
1039
if (f == '..') or (f is None) or (f == ''):
1040
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
1044
def parent_directories(filename):
1045
"""Return the list of parent directories, deepest first.
1047
For example, parent_directories("a/b/c") -> ["a/b", "a"].
1050
parts = splitpath(dirname(filename))
1052
parents.append(joinpath(parts))
1057
_extension_load_failures = []
1060
def failed_to_load_extension(exception):
1061
"""Handle failing to load a binary extension.
1063
This should be called from the ImportError block guarding the attempt to
1064
import the native extension. If this function returns, the pure-Python
1065
implementation should be loaded instead::
1068
>>> import breezy._fictional_extension_pyx
1069
>>> except ImportError, e:
1070
>>> breezy.osutils.failed_to_load_extension(e)
1071
>>> import breezy._fictional_extension_py
1073
# NB: This docstring is just an example, not a doctest, because doctest
1074
# currently can't cope with the use of lazy imports in this namespace --
1077
# This currently doesn't report the failure at the time it occurs, because
1078
# they tend to happen very early in startup when we can't check config
1079
# files etc, and also we want to report all failures but not spam the user
1081
exception_str = str(exception)
1082
if exception_str not in _extension_load_failures:
1083
trace.mutter("failed to load compiled extension: %s" % exception_str)
1084
_extension_load_failures.append(exception_str)
1087
def report_extension_load_failures():
1088
if not _extension_load_failures:
1090
if config.GlobalConfig().suppress_warning('missing_extensions'):
1092
# the warnings framework should by default show this only once
1093
from .trace import warning
1095
"brz: warning: some compiled extensions could not be loaded; "
1096
"see ``brz help missing-extensions``")
1097
# we no longer show the specific missing extensions here, because it makes
1098
# the message too long and scary - see
1099
# https://bugs.launchpad.net/bzr/+bug/430529
1103
from ._chunks_to_lines_pyx import chunks_to_lines
1104
except ImportError as e:
1105
failed_to_load_extension(e)
1106
from ._chunks_to_lines_py import chunks_to_lines
1110
"""Split s into lines, but without removing the newline characters."""
1111
# Trivially convert a fulltext into a 'chunked' representation, and let
1112
# chunks_to_lines do the heavy lifting.
1113
if isinstance(s, bytes):
1114
# chunks_to_lines only supports 8-bit strings
1115
return chunks_to_lines([s])
1117
return _split_lines(s)
1120
def _split_lines(s):
1121
"""Split s into lines, but without removing the newline characters.
1123
This supports Unicode or plain string objects.
1125
nl = b'\n' if isinstance(s, bytes) else u'\n'
1127
result = [line + nl for line in lines[:-1]]
1129
result.append(lines[-1])
1133
def hardlinks_good():
1134
return sys.platform not in ('win32', 'cygwin', 'darwin')
1137
def link_or_copy(src, dest):
1138
"""Hardlink a file, or copy it if it can't be hardlinked."""
1139
if not hardlinks_good():
1140
shutil.copyfile(src, dest)
1144
except (OSError, IOError) as e:
1145
if e.errno != errno.EXDEV:
1147
shutil.copyfile(src, dest)
1150
def delete_any(path):
1151
"""Delete a file, symlink or directory.
1153
Will delete even if readonly.
1156
_delete_file_or_dir(path)
1157
except (OSError, IOError) as e:
1158
if e.errno in (errno.EPERM, errno.EACCES):
1159
# make writable and try again
1162
except (OSError, IOError):
1164
_delete_file_or_dir(path)
1169
def _delete_file_or_dir(path):
1170
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1171
# Forgiveness than Permission (EAFP) because:
1172
# - root can damage a solaris file system by using unlink,
1173
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1174
# EACCES, OSX: EPERM) when invoked on a directory.
1175
if isdir(path): # Takes care of symlinks
1182
if getattr(os, 'symlink', None) is not None:
1188
def has_hardlinks():
1189
if getattr(os, 'link', None) is not None:
1195
def host_os_dereferences_symlinks():
1196
return (has_symlinks()
1197
and sys.platform not in ('cygwin', 'win32'))
1200
def readlink(abspath):
1201
"""Return a string representing the path to which the symbolic link points.
1203
:param abspath: The link absolute unicode path.
1205
This his guaranteed to return the symbolic link in unicode in all python
1208
link = abspath.encode(_fs_enc)
1209
target = os.readlink(link)
1210
target = target.decode(_fs_enc)
1214
def contains_whitespace(s):
1215
"""True if there are any whitespace characters in s."""
1216
# string.whitespace can include '\xa0' in certain locales, because it is
1217
# considered "non-breaking-space" as part of ISO-8859-1. But it
1218
# 1) Isn't a breaking whitespace
1219
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1221
# 3) '\xa0' isn't unicode safe since it is >128.
1223
if isinstance(s, str):
1226
ws = (b' ', b'\t', b'\n', b'\r', b'\v', b'\f')
1234
def contains_linebreaks(s):
1235
"""True if there is any vertical whitespace in s."""
1243
def relpath(base, path):
1244
"""Return path relative to base, or raise PathNotChild exception.
1246
The path may be either an absolute path or a path relative to the
1247
current working directory.
1249
os.path.commonprefix (python2.4) has a bad bug that it works just
1250
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1251
avoids that problem.
1253
NOTE: `base` should not have a trailing slash otherwise you'll get
1254
PathNotChild exceptions regardless of `path`.
1257
if len(base) < MIN_ABS_PATHLENGTH:
1258
# must have space for e.g. a drive letter
1259
raise ValueError(gettext('%r is too short to calculate a relative path')
1267
if len(head) <= len(base) and head != base:
1268
raise errors.PathNotChild(rp, base)
1271
head, tail = split(head)
1276
return pathjoin(*reversed(s))
1281
def _cicp_canonical_relpath(base, path):
1282
"""Return the canonical path relative to base.
1284
Like relpath, but on case-insensitive-case-preserving file-systems, this
1285
will return the relpath as stored on the file-system rather than in the
1286
case specified in the input string, for all existing portions of the path.
1288
This will cause O(N) behaviour if called for every path in a tree; if you
1289
have a number of paths to convert, you should use canonical_relpaths().
1291
# TODO: it should be possible to optimize this for Windows by using the
1292
# win32 API FindFiles function to look for the specified name - but using
1293
# os.listdir() still gives us the correct, platform agnostic semantics in
1296
rel = relpath(base, path)
1297
# '.' will have been turned into ''
1301
abs_base = abspath(base)
1303
_listdir = os.listdir
1305
# use an explicit iterator so we can easily consume the rest on early exit.
1306
bit_iter = iter(rel.split('/'))
1307
for bit in bit_iter:
1310
next_entries = _listdir(current)
1311
except OSError: # enoent, eperm, etc
1312
# We can't find this in the filesystem, so just append the
1314
current = pathjoin(current, bit, *list(bit_iter))
1316
for look in next_entries:
1317
if lbit == look.lower():
1318
current = pathjoin(current, look)
1321
# got to the end, nothing matched, so we just return the
1322
# non-existing bits as they were specified (the filename may be
1323
# the target of a move, for example).
1324
current = pathjoin(current, bit, *list(bit_iter))
1326
return current[len(abs_base):].lstrip('/')
1328
# XXX - TODO - we need better detection/integration of case-insensitive
1329
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1330
# filesystems), for example, so could probably benefit from the same basic
1331
# support there. For now though, only Windows and OSX get that support, and
1332
# they get it for *all* file-systems!
1333
if sys.platform in ('win32', 'darwin'):
1334
canonical_relpath = _cicp_canonical_relpath
1336
canonical_relpath = relpath
1338
def canonical_relpaths(base, paths):
1339
"""Create an iterable to canonicalize a sequence of relative paths.
1341
The intent is for this implementation to use a cache, vastly speeding
1342
up multiple transformations in the same directory.
1344
# but for now, we haven't optimized...
1345
return [canonical_relpath(base, p) for p in paths]
1348
def decode_filename(filename):
1349
"""Decode the filename using the filesystem encoding
1351
If it is unicode, it is returned.
1352
Otherwise it is decoded from the the filesystem's encoding. If decoding
1353
fails, a errors.BadFilenameEncoding exception is raised.
1355
if isinstance(filename, text_type):
1358
return filename.decode(_fs_enc)
1359
except UnicodeDecodeError:
1360
raise errors.BadFilenameEncoding(filename, _fs_enc)
1363
def safe_unicode(unicode_or_utf8_string):
1364
"""Coerce unicode_or_utf8_string into unicode.
1366
If it is unicode, it is returned.
1367
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1368
wrapped in a BzrBadParameterNotUnicode exception.
1370
if isinstance(unicode_or_utf8_string, text_type):
1371
return unicode_or_utf8_string
1373
return unicode_or_utf8_string.decode('utf8')
1374
except UnicodeDecodeError:
1375
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1378
def safe_utf8(unicode_or_utf8_string):
1379
"""Coerce unicode_or_utf8_string to a utf8 string.
1381
If it is a str, it is returned.
1382
If it is Unicode, it is encoded into a utf-8 string.
1384
if isinstance(unicode_or_utf8_string, bytes):
1385
# TODO: jam 20070209 This is overkill, and probably has an impact on
1386
# performance if we are dealing with lots of apis that want a
1389
# Make sure it is a valid utf-8 string
1390
unicode_or_utf8_string.decode('utf-8')
1391
except UnicodeDecodeError:
1392
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1393
return unicode_or_utf8_string
1394
return unicode_or_utf8_string.encode('utf-8')
1397
def safe_revision_id(unicode_or_utf8_string):
1398
"""Revision ids should now be utf8, but at one point they were unicode.
1400
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1402
:return: None or a utf8 revision id.
1404
if (unicode_or_utf8_string is None
1405
or unicode_or_utf8_string.__class__ == bytes):
1406
return unicode_or_utf8_string
1407
raise TypeError('Unicode revision ids are no longer supported. '
1408
'Revision id generators should be creating utf8 revision '
1412
def safe_file_id(unicode_or_utf8_string):
1413
"""File ids should now be utf8, but at one point they were unicode.
1415
This is the same as safe_utf8, except it uses the cached encode functions
1416
to save a little bit of performance.
1418
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1420
:return: None or a utf8 file 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 file ids are no longer supported. '
1426
'File id generators should be creating utf8 file ids.')
1429
_platform_normalizes_filenames = False
1430
if sys.platform == 'darwin':
1431
_platform_normalizes_filenames = True
1434
def normalizes_filenames():
1435
"""Return True if this platform normalizes unicode filenames.
1439
return _platform_normalizes_filenames
1442
def _accessible_normalized_filename(path):
1443
"""Get the unicode normalized path, and if you can access the file.
1445
On platforms where the system normalizes filenames (Mac OSX),
1446
you can access a file by any path which will normalize correctly.
1447
On platforms where the system does not normalize filenames
1448
(everything else), you have to access a file by its exact path.
1450
Internally, bzr only supports NFC normalization, since that is
1451
the standard for XML documents.
1453
So return the normalized path, and a flag indicating if the file
1454
can be accessed by that path.
1457
if isinstance(path, bytes):
1458
path = path.decode(sys.getfilesystemencoding())
1459
return unicodedata.normalize('NFC', path), True
1462
def _inaccessible_normalized_filename(path):
1463
__doc__ = _accessible_normalized_filename.__doc__
1465
if isinstance(path, bytes):
1466
path = path.decode(sys.getfilesystemencoding())
1467
normalized = unicodedata.normalize('NFC', path)
1468
return normalized, normalized == path
1471
if _platform_normalizes_filenames:
1472
normalized_filename = _accessible_normalized_filename
1474
normalized_filename = _inaccessible_normalized_filename
1477
def set_signal_handler(signum, handler, restart_syscall=True):
1478
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1479
on platforms that support that.
1481
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1482
automatically restart (by calling `signal.siginterrupt(signum,
1483
False)`). May be ignored if the feature is not available on this
1484
platform or Python version.
1488
siginterrupt = signal.siginterrupt
1490
# This python implementation doesn't provide signal support, hence no
1493
except AttributeError:
1494
# siginterrupt doesn't exist on this platform, or for this version
1496
siginterrupt = lambda signum, flag: None
1498
def sig_handler(*args):
1499
# Python resets the siginterrupt flag when a signal is
1500
# received. <http://bugs.python.org/issue8354>
1501
# As a workaround for some cases, set it back the way we want it.
1502
siginterrupt(signum, False)
1503
# Now run the handler function passed to set_signal_handler.
1506
sig_handler = handler
1507
old_handler = signal.signal(signum, sig_handler)
1509
siginterrupt(signum, False)
1513
default_terminal_width = 80
1514
"""The default terminal width for ttys.
1516
This is defined so that higher levels can share a common fallback value when
1517
terminal_width() returns None.
1520
# Keep some state so that terminal_width can detect if _terminal_size has
1521
# returned a different size since the process started. See docstring and
1522
# comments of terminal_width for details.
1523
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1524
_terminal_size_state = 'no_data'
1525
_first_terminal_size = None
1527
def terminal_width():
1528
"""Return terminal width.
1530
None is returned if the width can't established precisely.
1533
- if BRZ_COLUMNS is set, returns its value
1534
- if there is no controlling terminal, returns None
1535
- query the OS, if the queried size has changed since the last query,
1537
- if COLUMNS is set, returns its value,
1538
- if the OS has a value (even though it's never changed), return its value.
1540
From there, we need to query the OS to get the size of the controlling
1543
On Unices we query the OS by:
1544
- get termios.TIOCGWINSZ
1545
- if an error occurs or a negative value is obtained, returns None
1547
On Windows we query the OS by:
1548
- win32utils.get_console_size() decides,
1549
- returns None on error (provided default value)
1551
# Note to implementors: if changing the rules for determining the width,
1552
# make sure you've considered the behaviour in these cases:
1553
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1554
# - brz log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1556
# - (add more interesting cases here, if you find any)
1557
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1558
# but we don't want to register a signal handler because it is impossible
1559
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1560
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1561
# time so we can notice if the reported size has changed, which should have
1564
# If BRZ_COLUMNS is set, take it, user is always right
1565
# Except if they specified 0 in which case, impose no limit here
1567
width = int(os.environ['BRZ_COLUMNS'])
1568
except (KeyError, ValueError):
1570
if width is not None:
1576
isatty = getattr(sys.stdout, 'isatty', None)
1577
if isatty is None or not isatty():
1578
# Don't guess, setting BRZ_COLUMNS is the recommended way to override.
1582
width, height = os_size = _terminal_size(None, None)
1583
global _first_terminal_size, _terminal_size_state
1584
if _terminal_size_state == 'no_data':
1585
_first_terminal_size = os_size
1586
_terminal_size_state = 'unchanged'
1587
elif (_terminal_size_state == 'unchanged' and
1588
_first_terminal_size != os_size):
1589
_terminal_size_state = 'changed'
1591
# If the OS claims to know how wide the terminal is, and this value has
1592
# ever changed, use that.
1593
if _terminal_size_state == 'changed':
1594
if width is not None and width > 0:
1597
# If COLUMNS is set, use it.
1599
return int(os.environ['COLUMNS'])
1600
except (KeyError, ValueError):
1603
# Finally, use an unchanged size from the OS, if we have one.
1604
if _terminal_size_state == 'unchanged':
1605
if width is not None and width > 0:
1608
# The width could not be determined.
1612
def _win32_terminal_size(width, height):
1613
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1614
return width, height
1617
def _ioctl_terminal_size(width, height):
1619
import struct, fcntl, termios
1620
s = struct.pack('HHHH', 0, 0, 0, 0)
1621
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1622
height, width = struct.unpack('HHHH', x)[0:2]
1623
except (IOError, AttributeError):
1625
return width, height
1627
_terminal_size = None
1628
"""Returns the terminal size as (width, height).
1630
:param width: Default value for width.
1631
:param height: Default value for height.
1633
This is defined specifically for each OS and query the size of the controlling
1634
terminal. If any error occurs, the provided default values should be returned.
1636
if sys.platform == 'win32':
1637
_terminal_size = _win32_terminal_size
1639
_terminal_size = _ioctl_terminal_size
1642
def supports_executable():
1643
return sys.platform != "win32"
1646
def supports_posix_readonly():
1647
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1649
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1650
directory controls creation/deletion, etc.
1652
And under win32, readonly means that the directory itself cannot be
1653
deleted. The contents of a readonly directory can be changed, unlike POSIX
1654
where files in readonly directories cannot be added, deleted or renamed.
1656
return sys.platform != "win32"
1659
def set_or_unset_env(env_variable, value):
1660
"""Modify the environment, setting or removing the env_variable.
1662
:param env_variable: The environment variable in question
1663
:param value: The value to set the environment to. If None, then
1664
the variable will be removed.
1665
:return: The original value of the environment variable.
1667
orig_val = os.environ.get(env_variable)
1669
if orig_val is not None:
1670
del os.environ[env_variable]
1672
if not PY3 and isinstance(value, text_type):
1673
value = value.encode(get_user_encoding())
1674
os.environ[env_variable] = value
1678
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1681
def check_legal_path(path):
1682
"""Check whether the supplied path is legal.
1683
This is only required on Windows, so we don't test on other platforms
1686
if sys.platform != "win32":
1688
if _validWin32PathRE.match(path) is None:
1689
raise errors.IllegalPath(path)
1692
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1694
def _is_error_enotdir(e):
1695
"""Check if this exception represents ENOTDIR.
1697
Unfortunately, python is very inconsistent about the exception
1698
here. The cases are:
1699
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1700
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1701
which is the windows error code.
1702
3) Windows, Python2.5 uses errno == EINVAL and
1703
winerror == ERROR_DIRECTORY
1705
:param e: An Exception object (expected to be OSError with an errno
1706
attribute, but we should be able to cope with anything)
1707
:return: True if this represents an ENOTDIR error. False otherwise.
1709
en = getattr(e, 'errno', None)
1710
if (en == errno.ENOTDIR
1711
or (sys.platform == 'win32'
1712
and (en == _WIN32_ERROR_DIRECTORY
1713
or (en == errno.EINVAL
1714
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1720
def walkdirs(top, prefix=""):
1721
"""Yield data about all the directories in a tree.
1723
This yields all the data about the contents of a directory at a time.
1724
After each directory has been yielded, if the caller has mutated the list
1725
to exclude some directories, they are then not descended into.
1727
The data yielded is of the form:
1728
((directory-relpath, directory-path-from-top),
1729
[(relpath, basename, kind, lstat, path-from-top), ...]),
1730
- directory-relpath is the relative path of the directory being returned
1731
with respect to top. prefix is prepended to this.
1732
- directory-path-from-root is the path including top for this directory.
1733
It is suitable for use with os functions.
1734
- relpath is the relative path within the subtree being walked.
1735
- basename is the basename of the path
1736
- kind is the kind of the file now. If unknown then the file is not
1737
present within the tree - but it may be recorded as versioned. See
1739
- lstat is the stat data *if* the file was statted.
1740
- planned, not implemented:
1741
path_from_tree_root is the path from the root of the tree.
1743
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1744
allows one to walk a subtree but get paths that are relative to a tree
1746
:return: an iterator over the dirs.
1748
#TODO there is a bit of a smell where the results of the directory-
1749
# summary in this, and the path from the root, may not agree
1750
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1751
# potentially confusing output. We should make this more robust - but
1752
# not at a speed cost. RBC 20060731
1754
_directory = _directory_kind
1755
_listdir = os.listdir
1756
_kind_from_mode = file_kind_from_stat_mode
1757
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1759
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1760
relroot, _, _, _, top = pending.pop()
1762
relprefix = relroot + u'/'
1765
top_slash = top + u'/'
1768
append = dirblock.append
1770
names = sorted(map(decode_filename, _listdir(top)))
1771
except OSError as e:
1772
if not _is_error_enotdir(e):
1776
abspath = top_slash + name
1777
statvalue = _lstat(abspath)
1778
kind = _kind_from_mode(statvalue.st_mode)
1779
append((relprefix + name, name, kind, statvalue, abspath))
1780
yield (relroot, top), dirblock
1782
# push the user specified dirs from dirblock
1783
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1786
class DirReader(object):
1787
"""An interface for reading directories."""
1789
def top_prefix_to_starting_dir(self, top, prefix=""):
1790
"""Converts top and prefix to a starting dir entry
1792
:param top: A utf8 path
1793
:param prefix: An optional utf8 path to prefix output relative paths
1795
:return: A tuple starting with prefix, and ending with the native
1798
raise NotImplementedError(self.top_prefix_to_starting_dir)
1800
def read_dir(self, prefix, top):
1801
"""Read a specific dir.
1803
:param prefix: A utf8 prefix to be preprended to the path basenames.
1804
:param top: A natively encoded path to read.
1805
:return: A list of the directories contents. Each item contains:
1806
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1808
raise NotImplementedError(self.read_dir)
1811
_selected_dir_reader = None
1814
def _walkdirs_utf8(top, prefix=""):
1815
"""Yield data about all the directories in a tree.
1817
This yields the same information as walkdirs() only each entry is yielded
1818
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1819
are returned as exact byte-strings.
1821
:return: yields a tuple of (dir_info, [file_info])
1822
dir_info is (utf8_relpath, path-from-top)
1823
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1824
if top is an absolute path, path-from-top is also an absolute path.
1825
path-from-top might be unicode or utf8, but it is the correct path to
1826
pass to os functions to affect the file in question. (such as os.lstat)
1828
global _selected_dir_reader
1829
if _selected_dir_reader is None:
1830
if sys.platform == "win32":
1832
from ._walkdirs_win32 import Win32ReadDir
1833
_selected_dir_reader = Win32ReadDir()
1836
elif _fs_enc in ('utf-8', 'ascii'):
1838
from ._readdir_pyx import UTF8DirReader
1839
_selected_dir_reader = UTF8DirReader()
1840
except ImportError as e:
1841
failed_to_load_extension(e)
1844
if _selected_dir_reader is None:
1845
# Fallback to the python version
1846
_selected_dir_reader = UnicodeDirReader()
1848
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1849
# But we don't actually uses 1-3 in pending, so set them to None
1850
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1851
read_dir = _selected_dir_reader.read_dir
1852
_directory = _directory_kind
1854
relroot, _, _, _, top = pending[-1].pop()
1857
dirblock = sorted(read_dir(relroot, top))
1858
yield (relroot, top), dirblock
1859
# push the user specified dirs from dirblock
1860
next = [d for d in reversed(dirblock) if d[2] == _directory]
1862
pending.append(next)
1865
class UnicodeDirReader(DirReader):
1866
"""A dir reader for non-utf8 file systems, which transcodes."""
1868
__slots__ = ['_utf8_encode']
1871
self._utf8_encode = codecs.getencoder('utf8')
1873
def top_prefix_to_starting_dir(self, top, prefix=""):
1874
"""See DirReader.top_prefix_to_starting_dir."""
1875
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1877
def read_dir(self, prefix, top):
1878
"""Read a single directory from a non-utf8 file system.
1880
top, and the abspath element in the output are unicode, all other paths
1881
are utf8. Local disk IO is done via unicode calls to listdir etc.
1883
This is currently the fallback code path when the filesystem encoding is
1884
not UTF-8. It may be better to implement an alternative so that we can
1885
safely handle paths that are not properly decodable in the current
1888
See DirReader.read_dir for details.
1890
_utf8_encode = self._utf8_encode
1892
_listdir = os.listdir
1893
_kind_from_mode = file_kind_from_stat_mode
1896
relprefix = prefix + b'/'
1899
top_slash = top + '/'
1902
append = dirblock.append
1903
for name in _listdir(top):
1905
name_utf8 = _utf8_encode(name)[0]
1906
except UnicodeDecodeError:
1907
raise errors.BadFilenameEncoding(
1908
_utf8_encode(relprefix)[0] + name, _fs_enc)
1909
abspath = top_slash + name
1910
statvalue = _lstat(abspath)
1911
kind = _kind_from_mode(statvalue.st_mode)
1912
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1913
return sorted(dirblock)
1916
def copy_tree(from_path, to_path, handlers={}):
1917
"""Copy all of the entries in from_path into to_path.
1919
:param from_path: The base directory to copy.
1920
:param to_path: The target directory. If it does not exist, it will
1922
:param handlers: A dictionary of functions, which takes a source and
1923
destinations for files, directories, etc.
1924
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1925
'file', 'directory', and 'symlink' should always exist.
1926
If they are missing, they will be replaced with 'os.mkdir()',
1927
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1929
# Now, just copy the existing cached tree to the new location
1930
# We use a cheap trick here.
1931
# Absolute paths are prefixed with the first parameter
1932
# relative paths are prefixed with the second.
1933
# So we can get both the source and target returned
1934
# without any extra work.
1936
def copy_dir(source, dest):
1939
def copy_link(source, dest):
1940
"""Copy the contents of a symlink"""
1941
link_to = os.readlink(source)
1942
os.symlink(link_to, dest)
1944
real_handlers = {'file': shutil.copy2,
1945
'symlink': copy_link,
1946
'directory': copy_dir,
1948
real_handlers.update(handlers)
1950
if not os.path.exists(to_path):
1951
real_handlers['directory'](from_path, to_path)
1953
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1954
for relpath, name, kind, st, abspath in entries:
1955
real_handlers[kind](abspath, relpath)
1958
def copy_ownership_from_path(dst, src=None):
1959
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1961
If src is None, the containing directory is used as source. If chown
1962
fails, the error is ignored and a warning is printed.
1964
chown = getattr(os, 'chown', None)
1969
src = os.path.dirname(dst)
1975
chown(dst, s.st_uid, s.st_gid)
1976
except OSError as e:
1978
'Unable to copy ownership from "%s" to "%s". '
1979
'You may want to set it manually.', src, dst)
1980
trace.log_exception_quietly()
1983
def path_prefix_key(path):
1984
"""Generate a prefix-order path key for path.
1986
This can be used to sort paths in the same way that walkdirs does.
1988
return (dirname(path), path)
1991
def compare_paths_prefix_order(path_a, path_b):
1992
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1993
key_a = path_prefix_key(path_a)
1994
key_b = path_prefix_key(path_b)
1995
return cmp(key_a, key_b)
1998
_cached_user_encoding = None
2001
def get_user_encoding():
2002
"""Find out what the preferred user encoding is.
2004
This is generally the encoding that is used for command line parameters
2005
and file contents. This may be different from the terminal encoding
2006
or the filesystem encoding.
2008
:return: A string defining the preferred user encoding
2010
global _cached_user_encoding
2011
if _cached_user_encoding is not None:
2012
return _cached_user_encoding
2014
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2015
# Use the existing locale settings and call nl_langinfo directly
2016
# rather than going through getpreferredencoding. This avoids
2017
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2018
# possibility of the setlocale call throwing an error.
2019
user_encoding = locale.nl_langinfo(locale.CODESET)
2021
# GZ 2011-12-19: On windows could call GetACP directly instead.
2022
user_encoding = locale.getpreferredencoding(False)
2025
user_encoding = codecs.lookup(user_encoding).name
2027
if user_encoding not in ("", "cp0"):
2028
sys.stderr.write('brz: warning:'
2029
' unknown encoding %s.'
2030
' Continuing with ascii encoding.\n'
2033
user_encoding = 'ascii'
2035
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2036
if user_encoding == 'ascii':
2037
if sys.platform == 'darwin':
2038
# OSX is special-cased in Python to have a UTF-8 filesystem
2039
# encoding and previously had LANG set here if not present.
2040
user_encoding = 'utf-8'
2041
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2042
# for some other posix platforms as well.
2044
_cached_user_encoding = user_encoding
2045
return user_encoding
2048
def get_diff_header_encoding():
2049
return get_terminal_encoding()
2052
def get_host_name():
2053
"""Return the current unicode host name.
2055
This is meant to be used in place of socket.gethostname() because that
2056
behaves inconsistently on different platforms.
2058
if sys.platform == "win32":
2059
return win32utils.get_host_name()
2063
return socket.gethostname()
2064
return socket.gethostname().decode(get_user_encoding())
2067
# We must not read/write any more than 64k at a time from/to a socket so we
2068
# don't risk "no buffer space available" errors on some platforms. Windows in
2069
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2071
MAX_SOCKET_CHUNK = 64 * 1024
2073
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2074
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2075
_eno = getattr(errno, _eno, None)
2076
if _eno is not None:
2077
_end_of_stream_errors.append(_eno)
2081
def read_bytes_from_socket(sock, report_activity=None,
2082
max_read_size=MAX_SOCKET_CHUNK):
2083
"""Read up to max_read_size of bytes from sock and notify of progress.
2085
Translates "Connection reset by peer" into file-like EOF (return an
2086
empty string rather than raise an error), and repeats the recv if
2087
interrupted by a signal.
2091
data = sock.recv(max_read_size)
2092
except socket.error as e:
2094
if eno in _end_of_stream_errors:
2095
# The connection was closed by the other side. Callers expect
2096
# an empty string to signal end-of-stream.
2098
elif eno == errno.EINTR:
2099
# Retry the interrupted recv.
2103
if report_activity is not None:
2104
report_activity(len(data), 'read')
2108
def recv_all(socket, count):
2109
"""Receive an exact number of bytes.
2111
Regular Socket.recv() may return less than the requested number of bytes,
2112
depending on what's in the OS buffer. MSG_WAITALL is not available
2113
on all platforms, but this should work everywhere. This will return
2114
less than the requested amount if the remote end closes.
2116
This isn't optimized and is intended mostly for use in testing.
2119
while len(b) < count:
2120
new = read_bytes_from_socket(socket, None, count - len(b))
2127
def send_all(sock, bytes, report_activity=None):
2128
"""Send all bytes on a socket.
2130
Breaks large blocks in smaller chunks to avoid buffering limitations on
2131
some platforms, and catches EINTR which may be thrown if the send is
2132
interrupted by a signal.
2134
This is preferred to socket.sendall(), because it avoids portability bugs
2135
and provides activity reporting.
2137
:param report_activity: Call this as bytes are read, see
2138
Transport._report_activity
2141
byte_count = len(bytes)
2142
view = memoryview(bytes)
2143
while sent_total < byte_count:
2145
sent = sock.send(view[sent_total:sent_total+MAX_SOCKET_CHUNK])
2146
except (socket.error, IOError) as e:
2147
if e.args[0] in _end_of_stream_errors:
2148
raise errors.ConnectionReset(
2149
"Error trying to write to socket", e)
2150
if e.args[0] != errno.EINTR:
2154
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2157
if report_activity is not None:
2158
report_activity(sent, 'write')
2161
def connect_socket(address):
2162
# Slight variation of the socket.create_connection() function (provided by
2163
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2164
# provide it for previous python versions. Also, we don't use the timeout
2165
# parameter (provided by the python implementation) so we don't implement
2167
err = socket.error('getaddrinfo returns an empty list')
2168
host, port = address
2169
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2170
af, socktype, proto, canonname, sa = res
2173
sock = socket.socket(af, socktype, proto)
2177
except socket.error as e:
2179
# 'err' is now the most recent error
2180
if sock is not None:
2185
def dereference_path(path):
2186
"""Determine the real path to a file.
2188
All parent elements are dereferenced. But the file itself is not
2190
:param path: The original path. May be absolute or relative.
2191
:return: the real path *to* the file
2193
parent, base = os.path.split(path)
2194
# The pathjoin for '.' is a workaround for Python bug #1213894.
2195
# (initial path components aren't dereferenced)
2196
return pathjoin(realpath(pathjoin('.', parent)), base)
2199
def supports_mapi():
2200
"""Return True if we can use MAPI to launch a mail client."""
2201
return sys.platform == "win32"
2204
def resource_string(package, resource_name):
2205
"""Load a resource from a package and return it as a string.
2207
Note: Only packages that start with breezy are currently supported.
2209
This is designed to be a lightweight implementation of resource
2210
loading in a way which is API compatible with the same API from
2212
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2213
If and when pkg_resources becomes a standard library, this routine
2216
# Check package name is within breezy
2217
if package == "breezy":
2218
resource_relpath = resource_name
2219
elif package.startswith("breezy."):
2220
package = package[len("breezy."):].replace('.', os.sep)
2221
resource_relpath = pathjoin(package, resource_name)
2223
raise errors.BzrError('resource package %s not in breezy' % package)
2225
# Map the resource to a file and read its contents
2226
base = dirname(breezy.__file__)
2227
if getattr(sys, 'frozen', None): # bzr.exe
2228
base = abspath(pathjoin(base, '..', '..'))
2229
with open(pathjoin(base, resource_relpath), "rt") as f:
2232
def file_kind_from_stat_mode_thunk(mode):
2233
global file_kind_from_stat_mode
2234
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2236
from ._readdir_pyx import UTF8DirReader
2237
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2238
except ImportError as e:
2239
# This is one time where we won't warn that an extension failed to
2240
# load. The extension is never available on Windows anyway.
2241
from ._readdir_py import (
2242
_kind_from_mode as file_kind_from_stat_mode
2244
return file_kind_from_stat_mode(mode)
2245
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2247
def file_stat(f, _lstat=os.lstat):
2251
except OSError as e:
2252
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2253
raise errors.NoSuchFile(f)
2256
def file_kind(f, _lstat=os.lstat):
2257
stat_value = file_stat(f, _lstat)
2258
return file_kind_from_stat_mode(stat_value.st_mode)
2260
def until_no_eintr(f, *a, **kw):
2261
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2263
WARNING: you must be certain that it is safe to retry the call repeatedly
2264
if EINTR does occur. This is typically only true for low-level operations
2265
like os.read. If in any doubt, don't use this.
2267
Keep in mind that this is not a complete solution to EINTR. There is
2268
probably code in the Python standard library and other dependencies that
2269
may encounter EINTR if a signal arrives (and there is signal handler for
2270
that signal). So this function can reduce the impact for IO that breezy
2271
directly controls, but it is not a complete solution.
2273
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2277
except (IOError, OSError) as e:
2278
if e.errno == errno.EINTR:
2283
if sys.platform == "win32":
2286
return msvcrt.getch()
2291
fd = sys.stdin.fileno()
2292
settings = termios.tcgetattr(fd)
2295
ch = sys.stdin.read(1)
2297
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2300
if sys.platform.startswith('linux'):
2301
def _local_concurrency():
2303
return os.sysconf('SC_NPROCESSORS_ONLN')
2304
except (ValueError, OSError, AttributeError):
2306
elif sys.platform == 'darwin':
2307
def _local_concurrency():
2308
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2309
stdout=subprocess.PIPE).communicate()[0]
2310
elif "bsd" in sys.platform:
2311
def _local_concurrency():
2312
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2313
stdout=subprocess.PIPE).communicate()[0]
2314
elif sys.platform == 'sunos5':
2315
def _local_concurrency():
2316
return subprocess.Popen(['psrinfo', '-p',],
2317
stdout=subprocess.PIPE).communicate()[0]
2318
elif sys.platform == "win32":
2319
def _local_concurrency():
2320
# This appears to return the number of cores.
2321
return os.environ.get('NUMBER_OF_PROCESSORS')
2323
def _local_concurrency():
2328
_cached_local_concurrency = None
2330
def local_concurrency(use_cache=True):
2331
"""Return how many processes can be run concurrently.
2333
Rely on platform specific implementations and default to 1 (one) if
2334
anything goes wrong.
2336
global _cached_local_concurrency
2338
if _cached_local_concurrency is not None and use_cache:
2339
return _cached_local_concurrency
2341
concurrency = os.environ.get('BRZ_CONCURRENCY', None)
2342
if concurrency is None:
2343
import multiprocessing
2345
concurrency = multiprocessing.cpu_count()
2346
except NotImplementedError:
2347
# multiprocessing.cpu_count() isn't implemented on all platforms
2349
concurrency = _local_concurrency()
2350
except (OSError, IOError):
2353
concurrency = int(concurrency)
2354
except (TypeError, ValueError):
2357
_cached_concurrency = concurrency
2361
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2362
"""A stream writer that doesn't decode str arguments."""
2364
def __init__(self, encode, stream, errors='strict'):
2365
codecs.StreamWriter.__init__(self, stream, errors)
2366
self.encode = encode
2368
def write(self, object):
2369
if isinstance(object, str):
2370
self.stream.write(object)
2372
data, _ = self.encode(object, self.errors)
2373
self.stream.write(data)
2375
if sys.platform == 'win32':
2376
def open_file(filename, mode='r', bufsize=-1):
2377
"""This function is used to override the ``open`` builtin.
2379
But it uses O_NOINHERIT flag so the file handle is not inherited by
2380
child processes. Deleting or renaming a closed file opened with this
2381
function is not blocking child processes.
2383
writing = 'w' in mode
2384
appending = 'a' in mode
2385
updating = '+' in mode
2386
binary = 'b' in mode
2389
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2390
# for flags for each modes.
2400
flags |= os.O_WRONLY
2401
flags |= os.O_CREAT | os.O_TRUNC
2406
flags |= os.O_WRONLY
2407
flags |= os.O_CREAT | os.O_APPEND
2412
flags |= os.O_RDONLY
2414
return os.fdopen(os.open(filename, flags), mode, bufsize)
2419
def available_backup_name(base, exists):
2420
"""Find a non-existing backup file name.
2422
This will *not* create anything, this only return a 'free' entry. This
2423
should be used for checking names in a directory below a locked
2424
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2425
Leap) and generally discouraged.
2427
:param base: The base name.
2429
:param exists: A callable returning True if the path parameter exists.
2432
name = "%s.~%d~" % (base, counter)
2435
name = "%s.~%d~" % (base, counter)
2439
def set_fd_cloexec(fd):
2440
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2441
support for this is not available.
2445
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2446
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2447
except (ImportError, AttributeError):
2448
# Either the fcntl module or specific constants are not present
2452
def find_executable_on_path(name):
2453
"""Finds an executable on the PATH.
2455
On Windows, this will try to append each extension in the PATHEXT
2456
environment variable to the name, if it cannot be found with the name
2459
:param name: The base name of the executable.
2460
:return: The path to the executable found or None.
2462
if sys.platform == 'win32':
2463
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2464
exts = [ext.lower() for ext in exts]
2465
base, ext = os.path.splitext(name)
2467
if ext.lower() not in exts:
2473
path = os.environ.get('PATH')
2474
if path is not None:
2475
path = path.split(os.pathsep)
2478
f = os.path.join(d, name) + ext
2479
if os.access(f, os.X_OK):
2481
if sys.platform == 'win32':
2482
app_path = win32utils.get_app_path(name)
2483
if app_path != name:
2488
def _posix_is_local_pid_dead(pid):
2489
"""True if pid doesn't correspond to live process on this machine"""
2491
# Special meaning of unix kill: just check if it's there.
2493
except OSError as e:
2494
if e.errno == errno.ESRCH:
2495
# On this machine, and really not found: as sure as we can be
2498
elif e.errno == errno.EPERM:
2499
# exists, though not ours
2502
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2503
# Don't really know.
2506
# Exists and our process: not dead.
2509
if sys.platform == "win32":
2510
is_local_pid_dead = win32utils.is_local_pid_dead
2512
is_local_pid_dead = _posix_is_local_pid_dead
2514
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2515
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2516
if getattr(errno, name, None) is not None]
2519
def fdatasync(fileno):
2520
"""Flush file contents to disk if possible.
2522
:param fileno: Integer OS file handle.
2523
:raises TransportNotPossible: If flushing to disk is not possible.
2525
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2529
except IOError as e:
2530
# See bug #1075108, on some platforms fdatasync exists, but can
2531
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2532
# and reduce the chance of corruption-on-powerloss situations. It
2533
# is not a mandatory call, so it is ok to suppress failures.
2534
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2535
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2539
def ensure_empty_directory_exists(path, exception_class):
2540
"""Make sure a local directory exists and is empty.
2542
If it does not exist, it is created. If it exists and is not empty, an
2543
instance of exception_class is raised.
2547
except OSError as e:
2548
if e.errno != errno.EEXIST:
2550
if os.listdir(path) != []:
2551
raise exception_class(path)
2554
def is_environment_error(evalue):
2555
"""True if exception instance is due to a process environment issue
2557
This includes OSError and IOError, but also other errors that come from
2558
the operating system or core libraries but are not subclasses of those.
2560
if isinstance(evalue, (EnvironmentError, select.error)):
2562
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
added
removed
breezy/osutils.py
