1
# Copyright (C) 2005-2010 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
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
27
from datetime import datetime
29
from ntpath import (abspath as _nt_abspath,
31
normpath as _nt_normpath,
32
realpath as _nt_realpath,
33
splitdrive as _nt_splitdrive,
43
from tempfile import (
56
from bzrlib.symbol_versioning import (
61
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
63
if sys.version_info < (2, 5):
64
import md5 as _mod_md5
66
import sha as _mod_sha
76
from bzrlib import symbol_versioning
79
# Cross platform wall-clock time functionality with decent resolution.
80
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
81
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
82
# synchronized with ``time.time()``, this is only meant to be used to find
83
# delta times by subtracting from another call to this function.
84
timer_func = time.time
85
if sys.platform == 'win32':
86
timer_func = time.clock
88
# On win32, O_BINARY is used to indicate the file should
89
# be opened in binary mode, rather than text mode.
90
# On other platforms, O_BINARY doesn't exist, because
91
# they always open in binary mode, so it is okay to
92
# OR with 0 on those platforms.
93
# O_NOINHERIT and O_TEXT exists only on win32 too.
94
O_BINARY = getattr(os, 'O_BINARY', 0)
95
O_TEXT = getattr(os, 'O_TEXT', 0)
96
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
99
def get_unicode_argv():
101
user_encoding = get_user_encoding()
102
return [a.decode(user_encoding) for a in sys.argv[1:]]
103
except UnicodeDecodeError:
104
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
108
def make_readonly(filename):
109
"""Make a filename read-only."""
110
mod = os.lstat(filename).st_mode
111
if not stat.S_ISLNK(mod):
113
os.chmod(filename, mod)
116
def make_writable(filename):
117
mod = os.lstat(filename).st_mode
118
if not stat.S_ISLNK(mod):
120
os.chmod(filename, mod)
123
def minimum_path_selection(paths):
124
"""Return the smallset subset of paths which are outside paths.
126
:param paths: A container (and hence not None) of paths.
127
:return: A set of paths sufficient to include everything in paths via
128
is_inside, drawn from the paths parameter.
134
return path.split('/')
135
sorted_paths = sorted(list(paths), key=sort_key)
137
search_paths = [sorted_paths[0]]
138
for path in sorted_paths[1:]:
139
if not is_inside(search_paths[-1], path):
140
# This path is unique, add it
141
search_paths.append(path)
143
return set(search_paths)
150
"""Return a quoted filename filename
152
This previously used backslash quoting, but that works poorly on
154
# TODO: I'm not really sure this is the best format either.x
156
if _QUOTE_RE is None:
157
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
159
if _QUOTE_RE.search(f):
165
_directory_kind = 'directory'
168
"""Return the current umask"""
169
# Assume that people aren't messing with the umask while running
170
# XXX: This is not thread safe, but there is no way to get the
171
# umask without setting it
179
_directory_kind: "/",
181
'tree-reference': '+',
185
def kind_marker(kind):
187
return _kind_marker_map[kind]
189
# Slightly faster than using .get(, '') when the common case is that
194
lexists = getattr(os.path, 'lexists', None)
198
stat = getattr(os, 'lstat', os.stat)
202
if e.errno == errno.ENOENT:
205
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
208
def fancy_rename(old, new, rename_func, unlink_func):
209
"""A fancy rename, when you don't have atomic rename.
211
:param old: The old path, to rename from
212
:param new: The new path, to rename to
213
:param rename_func: The potentially non-atomic rename function
214
:param unlink_func: A way to delete the target file if the full rename
217
# sftp rename doesn't allow overwriting, so play tricks:
218
base = os.path.basename(new)
219
dirname = os.path.dirname(new)
220
# callers use different encodings for the paths so the following MUST
221
# respect that. We rely on python upcasting to unicode if new is unicode
222
# and keeping a str if not.
223
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
224
os.getpid(), rand_chars(10))
225
tmp_name = pathjoin(dirname, tmp_name)
227
# Rename the file out of the way, but keep track if it didn't exist
228
# We don't want to grab just any exception
229
# something like EACCES should prevent us from continuing
230
# The downside is that the rename_func has to throw an exception
231
# with an errno = ENOENT, or NoSuchFile
234
rename_func(new, tmp_name)
235
except (errors.NoSuchFile,), e:
238
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
239
# function raises an IOError with errno is None when a rename fails.
240
# This then gets caught here.
241
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
244
if (getattr(e, 'errno', None) is None
245
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
254
# This may throw an exception, in which case success will
256
rename_func(old, new)
258
except (IOError, OSError), e:
259
# source and target may be aliases of each other (e.g. on a
260
# case-insensitive filesystem), so we may have accidentally renamed
261
# source by when we tried to rename target
262
failure_exc = sys.exc_info()
263
if (file_existed and e.errno in (None, errno.ENOENT)
264
and old.lower() == new.lower()):
265
# source and target are the same file on a case-insensitive
266
# filesystem, so we don't generate an exception
270
# If the file used to exist, rename it back into place
271
# otherwise just delete it from the tmp location
273
unlink_func(tmp_name)
275
rename_func(tmp_name, new)
276
if failure_exc is not None:
277
raise failure_exc[0], failure_exc[1], failure_exc[2]
280
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
281
# choke on a Unicode string containing a relative path if
282
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
284
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
285
def _posix_abspath(path):
286
# jam 20060426 rather than encoding to fsencoding
287
# copy posixpath.abspath, but use os.getcwdu instead
288
if not posixpath.isabs(path):
289
path = posixpath.join(getcwd(), path)
290
return posixpath.normpath(path)
293
def _posix_realpath(path):
294
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
297
def _win32_fixdrive(path):
298
"""Force drive letters to be consistent.
300
win32 is inconsistent whether it returns lower or upper case
301
and even if it was consistent the user might type the other
302
so we force it to uppercase
303
running python.exe under cmd.exe return capital C:\\
304
running win32 python inside a cygwin shell returns lowercase c:\\
306
drive, path = _nt_splitdrive(path)
307
return drive.upper() + path
310
def _win32_abspath(path):
311
# Real _nt_abspath doesn't have a problem with a unicode cwd
312
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
315
def _win98_abspath(path):
316
"""Return the absolute version of a path.
317
Windows 98 safe implementation (python reimplementation
318
of Win32 API function GetFullPathNameW)
323
# \\HOST\path => //HOST/path
324
# //HOST/path => //HOST/path
325
# path => C:/cwd/path
328
# check for absolute path
329
drive = _nt_splitdrive(path)[0]
330
if drive == '' and path[:2] not in('//','\\\\'):
332
# we cannot simply os.path.join cwd and path
333
# because os.path.join('C:','/path') produce '/path'
334
# and this is incorrect
335
if path[:1] in ('/','\\'):
336
cwd = _nt_splitdrive(cwd)[0]
338
path = cwd + '\\' + path
339
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
342
def _win32_realpath(path):
343
# Real _nt_realpath doesn't have a problem with a unicode cwd
344
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
347
def _win32_pathjoin(*args):
348
return _nt_join(*args).replace('\\', '/')
351
def _win32_normpath(path):
352
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
356
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
359
def _win32_mkdtemp(*args, **kwargs):
360
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
363
def _add_rename_error_details(e, old, new):
364
new_e = OSError(e.errno, "failed to rename %s to %s: %s"
365
% (old, new, e.strerror))
367
new_e.to_filename = new
371
def _win32_rename(old, new):
372
"""We expect to be able to atomically replace 'new' with old.
374
On win32, if new exists, it must be moved out of the way first,
378
fancy_rename(old, new, rename_func=_wrapped_rename, unlink_func=os.unlink)
380
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
381
# If we try to rename a non-existant file onto cwd, we get
382
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
383
# if the old path doesn't exist, sometimes we get EACCES
384
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
389
def _wrapped_rename(old, new):
390
"""Rename a file or directory"""
393
except (IOError, OSError), e:
394
# this is eventually called by all rename-like functions, so should
396
raise _add_rename_error_details(e, old, new)
400
return unicodedata.normalize('NFC', os.getcwdu())
403
# Default is to just use the python builtins, but these can be rebound on
404
# particular platforms.
405
abspath = _posix_abspath
406
realpath = _posix_realpath
407
pathjoin = os.path.join
408
normpath = os.path.normpath
409
rename = _wrapped_rename # overridden below on win32
411
dirname = os.path.dirname
412
basename = os.path.basename
413
split = os.path.split
414
splitext = os.path.splitext
415
# These were already imported into local scope
416
# mkdtemp = tempfile.mkdtemp
417
# rmtree = shutil.rmtree
419
MIN_ABS_PATHLENGTH = 1
422
if sys.platform == 'win32':
423
if win32utils.winver == 'Windows 98':
424
abspath = _win98_abspath
426
abspath = _win32_abspath
427
realpath = _win32_realpath
428
pathjoin = _win32_pathjoin
429
normpath = _win32_normpath
430
getcwd = _win32_getcwd
431
mkdtemp = _win32_mkdtemp
432
rename = _win32_rename
434
MIN_ABS_PATHLENGTH = 3
436
def _win32_delete_readonly(function, path, excinfo):
437
"""Error handler for shutil.rmtree function [for win32]
438
Helps to remove files and dirs marked as read-only.
440
exception = excinfo[1]
441
if function in (os.remove, os.rmdir) \
442
and isinstance(exception, OSError) \
443
and exception.errno == errno.EACCES:
449
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
450
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
451
return shutil.rmtree(path, ignore_errors, onerror)
453
f = win32utils.get_unicode_argv # special function or None
457
elif sys.platform == 'darwin':
461
def get_terminal_encoding():
462
"""Find the best encoding for printing to the screen.
464
This attempts to check both sys.stdout and sys.stdin to see
465
what encoding they are in, and if that fails it falls back to
466
osutils.get_user_encoding().
467
The problem is that on Windows, locale.getpreferredencoding()
468
is not the same encoding as that used by the console:
469
http://mail.python.org/pipermail/python-list/2003-May/162357.html
471
On my standard US Windows XP, the preferred encoding is
472
cp1252, but the console is cp437
474
from bzrlib.trace import mutter
475
output_encoding = getattr(sys.stdout, 'encoding', None)
476
if not output_encoding:
477
input_encoding = getattr(sys.stdin, 'encoding', None)
478
if not input_encoding:
479
output_encoding = get_user_encoding()
480
mutter('encoding stdout as osutils.get_user_encoding() %r',
483
output_encoding = input_encoding
484
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
486
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
487
if output_encoding == 'cp0':
488
# invalid encoding (cp0 means 'no codepage' on Windows)
489
output_encoding = get_user_encoding()
490
mutter('cp0 is invalid encoding.'
491
' encoding stdout as osutils.get_user_encoding() %r',
495
codecs.lookup(output_encoding)
497
sys.stderr.write('bzr: warning:'
498
' unknown terminal encoding %s.\n'
499
' Using encoding %s instead.\n'
500
% (output_encoding, get_user_encoding())
502
output_encoding = get_user_encoding()
504
return output_encoding
507
def normalizepath(f):
508
if getattr(os.path, 'realpath', None) is not None:
512
[p,e] = os.path.split(f)
513
if e == "" or e == "." or e == "..":
516
return pathjoin(F(p), e)
520
"""True if f is an accessible directory."""
522
return S_ISDIR(os.lstat(f)[ST_MODE])
528
"""True if f is a regular file."""
530
return S_ISREG(os.lstat(f)[ST_MODE])
535
"""True if f is a symlink."""
537
return S_ISLNK(os.lstat(f)[ST_MODE])
541
def is_inside(dir, fname):
542
"""True if fname is inside dir.
544
The parameters should typically be passed to osutils.normpath first, so
545
that . and .. and repeated slashes are eliminated, and the separators
546
are canonical for the platform.
548
The empty string as a dir name is taken as top-of-tree and matches
551
# XXX: Most callers of this can actually do something smarter by
552
# looking at the inventory
562
return fname.startswith(dir)
565
def is_inside_any(dir_list, fname):
566
"""True if fname is inside any of given dirs."""
567
for dirname in dir_list:
568
if is_inside(dirname, fname):
573
def is_inside_or_parent_of_any(dir_list, fname):
574
"""True if fname is a child or a parent of any of the given files."""
575
for dirname in dir_list:
576
if is_inside(dirname, fname) or is_inside(fname, dirname):
581
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
582
report_activity=None, direction='read'):
583
"""Copy contents of one file to another.
585
The read_length can either be -1 to read to end-of-file (EOF) or
586
it can specify the maximum number of bytes to read.
588
The buff_size represents the maximum size for each read operation
589
performed on from_file.
591
:param report_activity: Call this as bytes are read, see
592
Transport._report_activity
593
:param direction: Will be passed to report_activity
595
:return: The number of bytes copied.
599
# read specified number of bytes
601
while read_length > 0:
602
num_bytes_to_read = min(read_length, buff_size)
604
block = from_file.read(num_bytes_to_read)
608
if report_activity is not None:
609
report_activity(len(block), direction)
612
actual_bytes_read = len(block)
613
read_length -= actual_bytes_read
614
length += actual_bytes_read
618
block = from_file.read(buff_size)
622
if report_activity is not None:
623
report_activity(len(block), direction)
629
def pump_string_file(bytes, file_handle, segment_size=None):
630
"""Write bytes to file_handle in many smaller writes.
632
:param bytes: The string to write.
633
:param file_handle: The file to write to.
635
# Write data in chunks rather than all at once, because very large
636
# writes fail on some platforms (e.g. Windows with SMB mounted
639
segment_size = 5242880 # 5MB
640
segments = range(len(bytes) / segment_size + 1)
641
write = file_handle.write
642
for segment_index in segments:
643
segment = buffer(bytes, segment_index * segment_size, segment_size)
647
def file_iterator(input_file, readsize=32768):
649
b = input_file.read(readsize)
656
"""Calculate the hexdigest of an open file.
658
The file cursor should be already at the start.
670
def size_sha_file(f):
671
"""Calculate the size and hexdigest of an open file.
673
The file cursor should be already at the start and
674
the caller is responsible for closing the file afterwards.
685
return size, s.hexdigest()
688
def sha_file_by_name(fname):
689
"""Calculate the SHA1 of a file by reading the full text"""
691
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
694
b = os.read(f, 1<<16)
702
def sha_strings(strings, _factory=sha):
703
"""Return the sha-1 of concatenation of strings"""
705
map(s.update, strings)
709
def sha_string(f, _factory=sha):
710
return _factory(f).hexdigest()
713
def fingerprint_file(f):
715
return {'size': len(b),
716
'sha1': sha(b).hexdigest()}
719
def compare_files(a, b):
720
"""Returns true if equal in contents"""
731
def local_time_offset(t=None):
732
"""Return offset of local zone from GMT, either at present or at time t."""
735
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
736
return offset.days * 86400 + offset.seconds
738
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
739
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
742
def format_date(t, offset=0, timezone='original', date_fmt=None,
744
"""Return a formatted date string.
746
:param t: Seconds since the epoch.
747
:param offset: Timezone offset in seconds east of utc.
748
:param timezone: How to display the time: 'utc', 'original' for the
749
timezone specified by offset, or 'local' for the process's current
751
:param date_fmt: strftime format.
752
:param show_offset: Whether to append the timezone.
754
(date_fmt, tt, offset_str) = \
755
_format_date(t, offset, timezone, date_fmt, show_offset)
756
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
757
date_str = time.strftime(date_fmt, tt)
758
return date_str + offset_str
761
# Cache of formatted offset strings
765
def format_date_with_offset_in_original_timezone(t, offset=0,
766
_cache=_offset_cache):
767
"""Return a formatted date string in the original timezone.
769
This routine may be faster then format_date.
771
:param t: Seconds since the epoch.
772
:param offset: Timezone offset in seconds east of utc.
776
tt = time.gmtime(t + offset)
777
date_fmt = _default_format_by_weekday_num[tt[6]]
778
date_str = time.strftime(date_fmt, tt)
779
offset_str = _cache.get(offset, None)
780
if offset_str is None:
781
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
782
_cache[offset] = offset_str
783
return date_str + offset_str
786
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
788
"""Return an unicode date string formatted according to the current locale.
790
:param t: Seconds since the epoch.
791
:param offset: Timezone offset in seconds east of utc.
792
:param timezone: How to display the time: 'utc', 'original' for the
793
timezone specified by offset, or 'local' for the process's current
795
:param date_fmt: strftime format.
796
:param show_offset: Whether to append the timezone.
798
(date_fmt, tt, offset_str) = \
799
_format_date(t, offset, timezone, date_fmt, show_offset)
800
date_str = time.strftime(date_fmt, tt)
801
if not isinstance(date_str, unicode):
802
date_str = date_str.decode(get_user_encoding(), 'replace')
803
return date_str + offset_str
806
def _format_date(t, offset, timezone, date_fmt, show_offset):
807
if timezone == 'utc':
810
elif timezone == 'original':
813
tt = time.gmtime(t + offset)
814
elif timezone == 'local':
815
tt = time.localtime(t)
816
offset = local_time_offset(t)
818
raise errors.UnsupportedTimezoneFormat(timezone)
820
date_fmt = "%a %Y-%m-%d %H:%M:%S"
822
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
825
return (date_fmt, tt, offset_str)
828
def compact_date(when):
829
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
832
def format_delta(delta):
833
"""Get a nice looking string for a time delta.
835
:param delta: The time difference in seconds, can be positive or negative.
836
positive indicates time in the past, negative indicates time in the
837
future. (usually time.time() - stored_time)
838
:return: String formatted to show approximate resolution
844
direction = 'in the future'
848
if seconds < 90: # print seconds up to 90 seconds
850
return '%d second %s' % (seconds, direction,)
852
return '%d seconds %s' % (seconds, direction)
854
minutes = int(seconds / 60)
855
seconds -= 60 * minutes
860
if minutes < 90: # print minutes, seconds up to 90 minutes
862
return '%d minute, %d second%s %s' % (
863
minutes, seconds, plural_seconds, direction)
865
return '%d minutes, %d second%s %s' % (
866
minutes, seconds, plural_seconds, direction)
868
hours = int(minutes / 60)
869
minutes -= 60 * hours
876
return '%d hour, %d minute%s %s' % (hours, minutes,
877
plural_minutes, direction)
878
return '%d hours, %d minute%s %s' % (hours, minutes,
879
plural_minutes, direction)
882
"""Return size of given open file."""
883
return os.fstat(f.fileno())[ST_SIZE]
886
# Define rand_bytes based on platform.
888
# Python 2.4 and later have os.urandom,
889
# but it doesn't work on some arches
891
rand_bytes = os.urandom
892
except (NotImplementedError, AttributeError):
893
# If python doesn't have os.urandom, or it doesn't work,
894
# then try to first pull random data from /dev/urandom
896
rand_bytes = file('/dev/urandom', 'rb').read
897
# Otherwise, use this hack as a last resort
898
except (IOError, OSError):
899
# not well seeded, but better than nothing
904
s += chr(random.randint(0, 255))
909
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
911
"""Return a random string of num alphanumeric characters
913
The result only contains lowercase chars because it may be used on
914
case-insensitive filesystems.
917
for raw_byte in rand_bytes(num):
918
s += ALNUM[ord(raw_byte) % 36]
922
## TODO: We could later have path objects that remember their list
923
## decomposition (might be too tricksy though.)
926
"""Turn string into list of parts."""
927
# split on either delimiter because people might use either on
929
ps = re.split(r'[\\/]', p)
934
raise errors.BzrError("sorry, %r not allowed in path" % f)
935
elif (f == '.') or (f == ''):
944
if (f == '..') or (f is None) or (f == ''):
945
raise errors.BzrError("sorry, %r not allowed in path" % f)
949
def parent_directories(filename):
950
"""Return the list of parent directories, deepest first.
952
For example, parent_directories("a/b/c") -> ["a/b", "a"].
955
parts = splitpath(dirname(filename))
957
parents.append(joinpath(parts))
962
_extension_load_failures = []
965
def failed_to_load_extension(exception):
966
"""Handle failing to load a binary extension.
968
This should be called from the ImportError block guarding the attempt to
969
import the native extension. If this function returns, the pure-Python
970
implementation should be loaded instead::
973
>>> import bzrlib._fictional_extension_pyx
974
>>> except ImportError, e:
975
>>> bzrlib.osutils.failed_to_load_extension(e)
976
>>> import bzrlib._fictional_extension_py
978
# NB: This docstring is just an example, not a doctest, because doctest
979
# currently can't cope with the use of lazy imports in this namespace --
982
# This currently doesn't report the failure at the time it occurs, because
983
# they tend to happen very early in startup when we can't check config
984
# files etc, and also we want to report all failures but not spam the user
986
from bzrlib import trace
987
exception_str = str(exception)
988
if exception_str not in _extension_load_failures:
989
trace.mutter("failed to load compiled extension: %s" % exception_str)
990
_extension_load_failures.append(exception_str)
993
def report_extension_load_failures():
994
if not _extension_load_failures:
996
from bzrlib.config import GlobalConfig
997
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
999
# the warnings framework should by default show this only once
1000
from bzrlib.trace import warning
1002
"bzr: warning: some compiled extensions could not be loaded; "
1003
"see <https://answers.launchpad.net/bzr/+faq/703>")
1004
# we no longer show the specific missing extensions here, because it makes
1005
# the message too long and scary - see
1006
# https://bugs.launchpad.net/bzr/+bug/430529
1010
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1011
except ImportError, e:
1012
failed_to_load_extension(e)
1013
from bzrlib._chunks_to_lines_py import chunks_to_lines
1017
"""Split s into lines, but without removing the newline characters."""
1018
# Trivially convert a fulltext into a 'chunked' representation, and let
1019
# chunks_to_lines do the heavy lifting.
1020
if isinstance(s, str):
1021
# chunks_to_lines only supports 8-bit strings
1022
return chunks_to_lines([s])
1024
return _split_lines(s)
1027
def _split_lines(s):
1028
"""Split s into lines, but without removing the newline characters.
1030
This supports Unicode or plain string objects.
1032
lines = s.split('\n')
1033
result = [line + '\n' for line in lines[:-1]]
1035
result.append(lines[-1])
1039
def hardlinks_good():
1040
return sys.platform not in ('win32', 'cygwin', 'darwin')
1043
def link_or_copy(src, dest):
1044
"""Hardlink a file, or copy it if it can't be hardlinked."""
1045
if not hardlinks_good():
1046
shutil.copyfile(src, dest)
1050
except (OSError, IOError), e:
1051
if e.errno != errno.EXDEV:
1053
shutil.copyfile(src, dest)
1056
def delete_any(path):
1057
"""Delete a file, symlink or directory.
1059
Will delete even if readonly.
1062
_delete_file_or_dir(path)
1063
except (OSError, IOError), e:
1064
if e.errno in (errno.EPERM, errno.EACCES):
1065
# make writable and try again
1068
except (OSError, IOError):
1070
_delete_file_or_dir(path)
1075
def _delete_file_or_dir(path):
1076
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1077
# Forgiveness than Permission (EAFP) because:
1078
# - root can damage a solaris file system by using unlink,
1079
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1080
# EACCES, OSX: EPERM) when invoked on a directory.
1081
if isdir(path): # Takes care of symlinks
1088
if getattr(os, 'symlink', None) is not None:
1094
def has_hardlinks():
1095
if getattr(os, 'link', None) is not None:
1101
def host_os_dereferences_symlinks():
1102
return (has_symlinks()
1103
and sys.platform not in ('cygwin', 'win32'))
1106
def readlink(abspath):
1107
"""Return a string representing the path to which the symbolic link points.
1109
:param abspath: The link absolute unicode path.
1111
This his guaranteed to return the symbolic link in unicode in all python
1114
link = abspath.encode(_fs_enc)
1115
target = os.readlink(link)
1116
target = target.decode(_fs_enc)
1120
def contains_whitespace(s):
1121
"""True if there are any whitespace characters in s."""
1122
# string.whitespace can include '\xa0' in certain locales, because it is
1123
# considered "non-breaking-space" as part of ISO-8859-1. But it
1124
# 1) Isn't a breaking whitespace
1125
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1127
# 3) '\xa0' isn't unicode safe since it is >128.
1129
# This should *not* be a unicode set of characters in case the source
1130
# string is not a Unicode string. We can auto-up-cast the characters since
1131
# they are ascii, but we don't want to auto-up-cast the string in case it
1133
for ch in ' \t\n\r\v\f':
1140
def contains_linebreaks(s):
1141
"""True if there is any vertical whitespace in s."""
1149
def relpath(base, path):
1150
"""Return path relative to base, or raise PathNotChild exception.
1152
The path may be either an absolute path or a path relative to the
1153
current working directory.
1155
os.path.commonprefix (python2.4) has a bad bug that it works just
1156
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1157
avoids that problem.
1159
NOTE: `base` should not have a trailing slash otherwise you'll get
1160
PathNotChild exceptions regardless of `path`.
1163
if len(base) < MIN_ABS_PATHLENGTH:
1164
# must have space for e.g. a drive letter
1165
raise ValueError('%r is too short to calculate a relative path'
1173
if len(head) <= len(base) and head != base:
1174
raise errors.PathNotChild(rp, base)
1177
head, tail = split(head)
1182
return pathjoin(*reversed(s))
1187
def _cicp_canonical_relpath(base, path):
1188
"""Return the canonical path relative to base.
1190
Like relpath, but on case-insensitive-case-preserving file-systems, this
1191
will return the relpath as stored on the file-system rather than in the
1192
case specified in the input string, for all existing portions of the path.
1194
This will cause O(N) behaviour if called for every path in a tree; if you
1195
have a number of paths to convert, you should use canonical_relpaths().
1197
# TODO: it should be possible to optimize this for Windows by using the
1198
# win32 API FindFiles function to look for the specified name - but using
1199
# os.listdir() still gives us the correct, platform agnostic semantics in
1202
rel = relpath(base, path)
1203
# '.' will have been turned into ''
1207
abs_base = abspath(base)
1209
_listdir = os.listdir
1211
# use an explicit iterator so we can easily consume the rest on early exit.
1212
bit_iter = iter(rel.split('/'))
1213
for bit in bit_iter:
1216
next_entries = _listdir(current)
1217
except OSError: # enoent, eperm, etc
1218
# We can't find this in the filesystem, so just append the
1220
current = pathjoin(current, bit, *list(bit_iter))
1222
for look in next_entries:
1223
if lbit == look.lower():
1224
current = pathjoin(current, look)
1227
# got to the end, nothing matched, so we just return the
1228
# non-existing bits as they were specified (the filename may be
1229
# the target of a move, for example).
1230
current = pathjoin(current, bit, *list(bit_iter))
1232
return current[len(abs_base):].lstrip('/')
1234
# XXX - TODO - we need better detection/integration of case-insensitive
1235
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1236
# filesystems), for example, so could probably benefit from the same basic
1237
# support there. For now though, only Windows and OSX get that support, and
1238
# they get it for *all* file-systems!
1239
if sys.platform in ('win32', 'darwin'):
1240
canonical_relpath = _cicp_canonical_relpath
1242
canonical_relpath = relpath
1244
def canonical_relpaths(base, paths):
1245
"""Create an iterable to canonicalize a sequence of relative paths.
1247
The intent is for this implementation to use a cache, vastly speeding
1248
up multiple transformations in the same directory.
1250
# but for now, we haven't optimized...
1251
return [canonical_relpath(base, p) for p in paths]
1253
def safe_unicode(unicode_or_utf8_string):
1254
"""Coerce unicode_or_utf8_string into unicode.
1256
If it is unicode, it is returned.
1257
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1258
wrapped in a BzrBadParameterNotUnicode exception.
1260
if isinstance(unicode_or_utf8_string, unicode):
1261
return unicode_or_utf8_string
1263
return unicode_or_utf8_string.decode('utf8')
1264
except UnicodeDecodeError:
1265
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1268
def safe_utf8(unicode_or_utf8_string):
1269
"""Coerce unicode_or_utf8_string to a utf8 string.
1271
If it is a str, it is returned.
1272
If it is Unicode, it is encoded into a utf-8 string.
1274
if isinstance(unicode_or_utf8_string, str):
1275
# TODO: jam 20070209 This is overkill, and probably has an impact on
1276
# performance if we are dealing with lots of apis that want a
1279
# Make sure it is a valid utf-8 string
1280
unicode_or_utf8_string.decode('utf-8')
1281
except UnicodeDecodeError:
1282
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1283
return unicode_or_utf8_string
1284
return unicode_or_utf8_string.encode('utf-8')
1287
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1288
' Revision id generators should be creating utf8'
1292
def safe_revision_id(unicode_or_utf8_string, warn=True):
1293
"""Revision ids should now be utf8, but at one point they were unicode.
1295
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1297
:param warn: Functions that are sanitizing user data can set warn=False
1298
:return: None or a utf8 revision id.
1300
if (unicode_or_utf8_string is None
1301
or unicode_or_utf8_string.__class__ == str):
1302
return unicode_or_utf8_string
1304
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1306
return cache_utf8.encode(unicode_or_utf8_string)
1309
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1310
' generators should be creating utf8 file ids.')
1313
def safe_file_id(unicode_or_utf8_string, warn=True):
1314
"""File ids should now be utf8, but at one point they were unicode.
1316
This is the same as safe_utf8, except it uses the cached encode functions
1317
to save a little bit of performance.
1319
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1321
:param warn: Functions that are sanitizing user data can set warn=False
1322
:return: None or a utf8 file id.
1324
if (unicode_or_utf8_string is None
1325
or unicode_or_utf8_string.__class__ == str):
1326
return unicode_or_utf8_string
1328
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1330
return cache_utf8.encode(unicode_or_utf8_string)
1333
_platform_normalizes_filenames = False
1334
if sys.platform == 'darwin':
1335
_platform_normalizes_filenames = True
1338
def normalizes_filenames():
1339
"""Return True if this platform normalizes unicode filenames.
1341
Mac OSX does, Windows/Linux do not.
1343
return _platform_normalizes_filenames
1346
def _accessible_normalized_filename(path):
1347
"""Get the unicode normalized path, and if you can access the file.
1349
On platforms where the system normalizes filenames (Mac OSX),
1350
you can access a file by any path which will normalize correctly.
1351
On platforms where the system does not normalize filenames
1352
(Windows, Linux), you have to access a file by its exact path.
1354
Internally, bzr only supports NFC normalization, since that is
1355
the standard for XML documents.
1357
So return the normalized path, and a flag indicating if the file
1358
can be accessed by that path.
1361
return unicodedata.normalize('NFC', unicode(path)), True
1364
def _inaccessible_normalized_filename(path):
1365
__doc__ = _accessible_normalized_filename.__doc__
1367
normalized = unicodedata.normalize('NFC', unicode(path))
1368
return normalized, normalized == path
1371
if _platform_normalizes_filenames:
1372
normalized_filename = _accessible_normalized_filename
1374
normalized_filename = _inaccessible_normalized_filename
1377
def set_signal_handler(signum, handler, restart_syscall=True):
1378
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1379
on platforms that support that.
1381
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1382
automatically restart (by calling `signal.siginterrupt(signum,
1383
False)`). May be ignored if the feature is not available on this
1384
platform or Python version.
1388
siginterrupt = signal.siginterrupt
1390
# This python implementation doesn't provide signal support, hence no
1393
except AttributeError:
1394
# siginterrupt doesn't exist on this platform, or for this version
1396
siginterrupt = lambda signum, flag: None
1398
def sig_handler(*args):
1399
# Python resets the siginterrupt flag when a signal is
1400
# received. <http://bugs.python.org/issue8354>
1401
# As a workaround for some cases, set it back the way we want it.
1402
siginterrupt(signum, False)
1403
# Now run the handler function passed to set_signal_handler.
1406
sig_handler = handler
1407
old_handler = signal.signal(signum, sig_handler)
1409
siginterrupt(signum, False)
1413
default_terminal_width = 80
1414
"""The default terminal width for ttys.
1416
This is defined so that higher levels can share a common fallback value when
1417
terminal_width() returns None.
1421
def terminal_width():
1422
"""Return terminal width.
1424
None is returned if the width can't established precisely.
1427
- if BZR_COLUMNS is set, returns its value
1428
- if there is no controlling terminal, returns None
1429
- if COLUMNS is set, returns its value,
1431
From there, we need to query the OS to get the size of the controlling
1435
- get termios.TIOCGWINSZ
1436
- if an error occurs or a negative value is obtained, returns None
1440
- win32utils.get_console_size() decides,
1441
- returns None on error (provided default value)
1444
# If BZR_COLUMNS is set, take it, user is always right
1446
return int(os.environ['BZR_COLUMNS'])
1447
except (KeyError, ValueError):
1450
isatty = getattr(sys.stdout, 'isatty', None)
1451
if isatty is None or not isatty():
1452
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1455
# If COLUMNS is set, take it, the terminal knows better (even inside a
1456
# given terminal, the application can decide to set COLUMNS to a lower
1457
# value (splitted screen) or a bigger value (scroll bars))
1459
return int(os.environ['COLUMNS'])
1460
except (KeyError, ValueError):
1463
width, height = _terminal_size(None, None)
1465
# Consider invalid values as meaning no width
1471
def _win32_terminal_size(width, height):
1472
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1473
return width, height
1476
def _ioctl_terminal_size(width, height):
1478
import struct, fcntl, termios
1479
s = struct.pack('HHHH', 0, 0, 0, 0)
1480
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1481
height, width = struct.unpack('HHHH', x)[0:2]
1482
except (IOError, AttributeError):
1484
return width, height
1486
_terminal_size = None
1487
"""Returns the terminal size as (width, height).
1489
:param width: Default value for width.
1490
:param height: Default value for height.
1492
This is defined specifically for each OS and query the size of the controlling
1493
terminal. If any error occurs, the provided default values should be returned.
1495
if sys.platform == 'win32':
1496
_terminal_size = _win32_terminal_size
1498
_terminal_size = _ioctl_terminal_size
1501
def _terminal_size_changed(signum, frame):
1502
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1503
width, height = _terminal_size(None, None)
1504
if width is not None:
1505
os.environ['COLUMNS'] = str(width)
1508
_registered_sigwinch = False
1509
def watch_sigwinch():
1510
"""Register for SIGWINCH, once and only once.
1512
Do nothing if the signal module is not available.
1514
global _registered_sigwinch
1515
if not _registered_sigwinch:
1518
if getattr(signal, "SIGWINCH", None) is not None:
1519
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1521
# python doesn't provide signal support, nothing we can do about it
1523
_registered_sigwinch = True
1526
def supports_executable():
1527
return sys.platform != "win32"
1530
def supports_posix_readonly():
1531
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1533
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1534
directory controls creation/deletion, etc.
1536
And under win32, readonly means that the directory itself cannot be
1537
deleted. The contents of a readonly directory can be changed, unlike POSIX
1538
where files in readonly directories cannot be added, deleted or renamed.
1540
return sys.platform != "win32"
1543
def set_or_unset_env(env_variable, value):
1544
"""Modify the environment, setting or removing the env_variable.
1546
:param env_variable: The environment variable in question
1547
:param value: The value to set the environment to. If None, then
1548
the variable will be removed.
1549
:return: The original value of the environment variable.
1551
orig_val = os.environ.get(env_variable)
1553
if orig_val is not None:
1554
del os.environ[env_variable]
1556
if isinstance(value, unicode):
1557
value = value.encode(get_user_encoding())
1558
os.environ[env_variable] = value
1562
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1565
def check_legal_path(path):
1566
"""Check whether the supplied path is legal.
1567
This is only required on Windows, so we don't test on other platforms
1570
if sys.platform != "win32":
1572
if _validWin32PathRE.match(path) is None:
1573
raise errors.IllegalPath(path)
1576
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1578
def _is_error_enotdir(e):
1579
"""Check if this exception represents ENOTDIR.
1581
Unfortunately, python is very inconsistent about the exception
1582
here. The cases are:
1583
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1584
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1585
which is the windows error code.
1586
3) Windows, Python2.5 uses errno == EINVAL and
1587
winerror == ERROR_DIRECTORY
1589
:param e: An Exception object (expected to be OSError with an errno
1590
attribute, but we should be able to cope with anything)
1591
:return: True if this represents an ENOTDIR error. False otherwise.
1593
en = getattr(e, 'errno', None)
1594
if (en == errno.ENOTDIR
1595
or (sys.platform == 'win32'
1596
and (en == _WIN32_ERROR_DIRECTORY
1597
or (en == errno.EINVAL
1598
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1604
def walkdirs(top, prefix=""):
1605
"""Yield data about all the directories in a tree.
1607
This yields all the data about the contents of a directory at a time.
1608
After each directory has been yielded, if the caller has mutated the list
1609
to exclude some directories, they are then not descended into.
1611
The data yielded is of the form:
1612
((directory-relpath, directory-path-from-top),
1613
[(relpath, basename, kind, lstat, path-from-top), ...]),
1614
- directory-relpath is the relative path of the directory being returned
1615
with respect to top. prefix is prepended to this.
1616
- directory-path-from-root is the path including top for this directory.
1617
It is suitable for use with os functions.
1618
- relpath is the relative path within the subtree being walked.
1619
- basename is the basename of the path
1620
- kind is the kind of the file now. If unknown then the file is not
1621
present within the tree - but it may be recorded as versioned. See
1623
- lstat is the stat data *if* the file was statted.
1624
- planned, not implemented:
1625
path_from_tree_root is the path from the root of the tree.
1627
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1628
allows one to walk a subtree but get paths that are relative to a tree
1630
:return: an iterator over the dirs.
1632
#TODO there is a bit of a smell where the results of the directory-
1633
# summary in this, and the path from the root, may not agree
1634
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1635
# potentially confusing output. We should make this more robust - but
1636
# not at a speed cost. RBC 20060731
1638
_directory = _directory_kind
1639
_listdir = os.listdir
1640
_kind_from_mode = file_kind_from_stat_mode
1641
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1643
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1644
relroot, _, _, _, top = pending.pop()
1646
relprefix = relroot + u'/'
1649
top_slash = top + u'/'
1652
append = dirblock.append
1654
names = sorted(_listdir(top))
1656
if not _is_error_enotdir(e):
1660
abspath = top_slash + name
1661
statvalue = _lstat(abspath)
1662
kind = _kind_from_mode(statvalue.st_mode)
1663
append((relprefix + name, name, kind, statvalue, abspath))
1664
yield (relroot, top), dirblock
1666
# push the user specified dirs from dirblock
1667
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1670
class DirReader(object):
1671
"""An interface for reading directories."""
1673
def top_prefix_to_starting_dir(self, top, prefix=""):
1674
"""Converts top and prefix to a starting dir entry
1676
:param top: A utf8 path
1677
:param prefix: An optional utf8 path to prefix output relative paths
1679
:return: A tuple starting with prefix, and ending with the native
1682
raise NotImplementedError(self.top_prefix_to_starting_dir)
1684
def read_dir(self, prefix, top):
1685
"""Read a specific dir.
1687
:param prefix: A utf8 prefix to be preprended to the path basenames.
1688
:param top: A natively encoded path to read.
1689
:return: A list of the directories contents. Each item contains:
1690
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1692
raise NotImplementedError(self.read_dir)
1695
_selected_dir_reader = None
1698
def _walkdirs_utf8(top, prefix=""):
1699
"""Yield data about all the directories in a tree.
1701
This yields the same information as walkdirs() only each entry is yielded
1702
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1703
are returned as exact byte-strings.
1705
:return: yields a tuple of (dir_info, [file_info])
1706
dir_info is (utf8_relpath, path-from-top)
1707
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1708
if top is an absolute path, path-from-top is also an absolute path.
1709
path-from-top might be unicode or utf8, but it is the correct path to
1710
pass to os functions to affect the file in question. (such as os.lstat)
1712
global _selected_dir_reader
1713
if _selected_dir_reader is None:
1714
fs_encoding = _fs_enc.upper()
1715
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1716
# Win98 doesn't have unicode apis like FindFirstFileW
1717
# TODO: We possibly could support Win98 by falling back to the
1718
# original FindFirstFile, and using TCHAR instead of WCHAR,
1719
# but that gets a bit tricky, and requires custom compiling
1722
from bzrlib._walkdirs_win32 import Win32ReadDir
1723
_selected_dir_reader = Win32ReadDir()
1726
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1727
# ANSI_X3.4-1968 is a form of ASCII
1729
from bzrlib._readdir_pyx import UTF8DirReader
1730
_selected_dir_reader = UTF8DirReader()
1731
except ImportError, e:
1732
failed_to_load_extension(e)
1735
if _selected_dir_reader is None:
1736
# Fallback to the python version
1737
_selected_dir_reader = UnicodeDirReader()
1739
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1740
# But we don't actually uses 1-3 in pending, so set them to None
1741
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1742
read_dir = _selected_dir_reader.read_dir
1743
_directory = _directory_kind
1745
relroot, _, _, _, top = pending[-1].pop()
1748
dirblock = sorted(read_dir(relroot, top))
1749
yield (relroot, top), dirblock
1750
# push the user specified dirs from dirblock
1751
next = [d for d in reversed(dirblock) if d[2] == _directory]
1753
pending.append(next)
1756
class UnicodeDirReader(DirReader):
1757
"""A dir reader for non-utf8 file systems, which transcodes."""
1759
__slots__ = ['_utf8_encode']
1762
self._utf8_encode = codecs.getencoder('utf8')
1764
def top_prefix_to_starting_dir(self, top, prefix=""):
1765
"""See DirReader.top_prefix_to_starting_dir."""
1766
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1768
def read_dir(self, prefix, top):
1769
"""Read a single directory from a non-utf8 file system.
1771
top, and the abspath element in the output are unicode, all other paths
1772
are utf8. Local disk IO is done via unicode calls to listdir etc.
1774
This is currently the fallback code path when the filesystem encoding is
1775
not UTF-8. It may be better to implement an alternative so that we can
1776
safely handle paths that are not properly decodable in the current
1779
See DirReader.read_dir for details.
1781
_utf8_encode = self._utf8_encode
1783
_listdir = os.listdir
1784
_kind_from_mode = file_kind_from_stat_mode
1787
relprefix = prefix + '/'
1790
top_slash = top + u'/'
1793
append = dirblock.append
1794
for name in sorted(_listdir(top)):
1796
name_utf8 = _utf8_encode(name)[0]
1797
except UnicodeDecodeError:
1798
raise errors.BadFilenameEncoding(
1799
_utf8_encode(relprefix)[0] + name, _fs_enc)
1800
abspath = top_slash + name
1801
statvalue = _lstat(abspath)
1802
kind = _kind_from_mode(statvalue.st_mode)
1803
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1807
def copy_tree(from_path, to_path, handlers={}):
1808
"""Copy all of the entries in from_path into to_path.
1810
:param from_path: The base directory to copy.
1811
:param to_path: The target directory. If it does not exist, it will
1813
:param handlers: A dictionary of functions, which takes a source and
1814
destinations for files, directories, etc.
1815
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1816
'file', 'directory', and 'symlink' should always exist.
1817
If they are missing, they will be replaced with 'os.mkdir()',
1818
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1820
# Now, just copy the existing cached tree to the new location
1821
# We use a cheap trick here.
1822
# Absolute paths are prefixed with the first parameter
1823
# relative paths are prefixed with the second.
1824
# So we can get both the source and target returned
1825
# without any extra work.
1827
def copy_dir(source, dest):
1830
def copy_link(source, dest):
1831
"""Copy the contents of a symlink"""
1832
link_to = os.readlink(source)
1833
os.symlink(link_to, dest)
1835
real_handlers = {'file':shutil.copy2,
1836
'symlink':copy_link,
1837
'directory':copy_dir,
1839
real_handlers.update(handlers)
1841
if not os.path.exists(to_path):
1842
real_handlers['directory'](from_path, to_path)
1844
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1845
for relpath, name, kind, st, abspath in entries:
1846
real_handlers[kind](abspath, relpath)
1849
def copy_ownership_from_path(dst, src=None):
1850
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1852
If src is None, the containing directory is used as source. If chown
1853
fails, the error is ignored and a warning is printed.
1855
chown = getattr(os, 'chown', None)
1860
src = os.path.dirname(dst)
1866
chown(dst, s.st_uid, s.st_gid)
1868
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1871
def path_prefix_key(path):
1872
"""Generate a prefix-order path key for path.
1874
This can be used to sort paths in the same way that walkdirs does.
1876
return (dirname(path) , path)
1879
def compare_paths_prefix_order(path_a, path_b):
1880
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1881
key_a = path_prefix_key(path_a)
1882
key_b = path_prefix_key(path_b)
1883
return cmp(key_a, key_b)
1886
_cached_user_encoding = None
1889
def get_user_encoding(use_cache=True):
1890
"""Find out what the preferred user encoding is.
1892
This is generally the encoding that is used for command line parameters
1893
and file contents. This may be different from the terminal encoding
1894
or the filesystem encoding.
1896
:param use_cache: Enable cache for detected encoding.
1897
(This parameter is turned on by default,
1898
and required only for selftesting)
1900
:return: A string defining the preferred user encoding
1902
global _cached_user_encoding
1903
if _cached_user_encoding is not None and use_cache:
1904
return _cached_user_encoding
1906
if sys.platform == 'darwin':
1907
# python locale.getpreferredencoding() always return
1908
# 'mac-roman' on darwin. That's a lie.
1909
sys.platform = 'posix'
1911
if os.environ.get('LANG', None) is None:
1912
# If LANG is not set, we end up with 'ascii', which is bad
1913
# ('mac-roman' is more than ascii), so we set a default which
1914
# will give us UTF-8 (which appears to work in all cases on
1915
# OSX). Users are still free to override LANG of course, as
1916
# long as it give us something meaningful. This work-around
1917
# *may* not be needed with python 3k and/or OSX 10.5, but will
1918
# work with them too -- vila 20080908
1919
os.environ['LANG'] = 'en_US.UTF-8'
1922
sys.platform = 'darwin'
1927
user_encoding = locale.getpreferredencoding()
1928
except locale.Error, e:
1929
sys.stderr.write('bzr: warning: %s\n'
1930
' Could not determine what text encoding to use.\n'
1931
' This error usually means your Python interpreter\n'
1932
' doesn\'t support the locale set by $LANG (%s)\n'
1933
" Continuing with ascii encoding.\n"
1934
% (e, os.environ.get('LANG')))
1935
user_encoding = 'ascii'
1937
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1938
# treat that as ASCII, and not support printing unicode characters to the
1941
# For python scripts run under vim, we get '', so also treat that as ASCII
1942
if user_encoding in (None, 'cp0', ''):
1943
user_encoding = 'ascii'
1947
codecs.lookup(user_encoding)
1949
sys.stderr.write('bzr: warning:'
1950
' unknown encoding %s.'
1951
' Continuing with ascii encoding.\n'
1954
user_encoding = 'ascii'
1957
_cached_user_encoding = user_encoding
1959
return user_encoding
1962
def get_host_name():
1963
"""Return the current unicode host name.
1965
This is meant to be used in place of socket.gethostname() because that
1966
behaves inconsistently on different platforms.
1968
if sys.platform == "win32":
1970
return win32utils.get_host_name()
1973
return socket.gethostname().decode(get_user_encoding())
1976
# We must not read/write any more than 64k at a time from/to a socket so we
1977
# don't risk "no buffer space available" errors on some platforms. Windows in
1978
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1980
MAX_SOCKET_CHUNK = 64 * 1024
1982
def read_bytes_from_socket(sock, report_activity=None,
1983
max_read_size=MAX_SOCKET_CHUNK):
1984
"""Read up to max_read_size of bytes from sock and notify of progress.
1986
Translates "Connection reset by peer" into file-like EOF (return an
1987
empty string rather than raise an error), and repeats the recv if
1988
interrupted by a signal.
1992
bytes = sock.recv(max_read_size)
1993
except socket.error, e:
1995
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1996
# The connection was closed by the other side. Callers expect
1997
# an empty string to signal end-of-stream.
1999
elif eno == errno.EINTR:
2000
# Retry the interrupted recv.
2004
if report_activity is not None:
2005
report_activity(len(bytes), 'read')
2009
def recv_all(socket, count):
2010
"""Receive an exact number of bytes.
2012
Regular Socket.recv() may return less than the requested number of bytes,
2013
depending on what's in the OS buffer. MSG_WAITALL is not available
2014
on all platforms, but this should work everywhere. This will return
2015
less than the requested amount if the remote end closes.
2017
This isn't optimized and is intended mostly for use in testing.
2020
while len(b) < count:
2021
new = read_bytes_from_socket(socket, None, count - len(b))
2028
def send_all(sock, bytes, report_activity=None):
2029
"""Send all bytes on a socket.
2031
Breaks large blocks in smaller chunks to avoid buffering limitations on
2032
some platforms, and catches EINTR which may be thrown if the send is
2033
interrupted by a signal.
2035
This is preferred to socket.sendall(), because it avoids portability bugs
2036
and provides activity reporting.
2038
:param report_activity: Call this as bytes are read, see
2039
Transport._report_activity
2042
byte_count = len(bytes)
2043
while sent_total < byte_count:
2045
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2046
except socket.error, e:
2047
if e.args[0] != errno.EINTR:
2051
report_activity(sent, 'write')
2054
def dereference_path(path):
2055
"""Determine the real path to a file.
2057
All parent elements are dereferenced. But the file itself is not
2059
:param path: The original path. May be absolute or relative.
2060
:return: the real path *to* the file
2062
parent, base = os.path.split(path)
2063
# The pathjoin for '.' is a workaround for Python bug #1213894.
2064
# (initial path components aren't dereferenced)
2065
return pathjoin(realpath(pathjoin('.', parent)), base)
2068
def supports_mapi():
2069
"""Return True if we can use MAPI to launch a mail client."""
2070
return sys.platform == "win32"
2073
def resource_string(package, resource_name):
2074
"""Load a resource from a package and return it as a string.
2076
Note: Only packages that start with bzrlib are currently supported.
2078
This is designed to be a lightweight implementation of resource
2079
loading in a way which is API compatible with the same API from
2081
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2082
If and when pkg_resources becomes a standard library, this routine
2085
# Check package name is within bzrlib
2086
if package == "bzrlib":
2087
resource_relpath = resource_name
2088
elif package.startswith("bzrlib."):
2089
package = package[len("bzrlib."):].replace('.', os.sep)
2090
resource_relpath = pathjoin(package, resource_name)
2092
raise errors.BzrError('resource package %s not in bzrlib' % package)
2094
# Map the resource to a file and read its contents
2095
base = dirname(bzrlib.__file__)
2096
if getattr(sys, 'frozen', None): # bzr.exe
2097
base = abspath(pathjoin(base, '..', '..'))
2098
filename = pathjoin(base, resource_relpath)
2099
return open(filename, 'rU').read()
2102
def file_kind_from_stat_mode_thunk(mode):
2103
global file_kind_from_stat_mode
2104
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2106
from bzrlib._readdir_pyx import UTF8DirReader
2107
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2108
except ImportError, e:
2109
# This is one time where we won't warn that an extension failed to
2110
# load. The extension is never available on Windows anyway.
2111
from bzrlib._readdir_py import (
2112
_kind_from_mode as file_kind_from_stat_mode
2114
return file_kind_from_stat_mode(mode)
2115
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2118
def file_kind(f, _lstat=os.lstat):
2120
return file_kind_from_stat_mode(_lstat(f).st_mode)
2122
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2123
raise errors.NoSuchFile(f)
2127
def until_no_eintr(f, *a, **kw):
2128
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2130
WARNING: you must be certain that it is safe to retry the call repeatedly
2131
if EINTR does occur. This is typically only true for low-level operations
2132
like os.read. If in any doubt, don't use this.
2134
Keep in mind that this is not a complete solution to EINTR. There is
2135
probably code in the Python standard library and other dependencies that
2136
may encounter EINTR if a signal arrives (and there is signal handler for
2137
that signal). So this function can reduce the impact for IO that bzrlib
2138
directly controls, but it is not a complete solution.
2140
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2144
except (IOError, OSError), e:
2145
if e.errno == errno.EINTR:
2150
def re_compile_checked(re_string, flags=0, where=""):
2151
"""Return a compiled re, or raise a sensible error.
2153
This should only be used when compiling user-supplied REs.
2155
:param re_string: Text form of regular expression.
2156
:param flags: eg re.IGNORECASE
2157
:param where: Message explaining to the user the context where
2158
it occurred, eg 'log search filter'.
2160
# from https://bugs.launchpad.net/bzr/+bug/251352
2162
re_obj = re.compile(re_string, flags)
2167
where = ' in ' + where
2168
# despite the name 'error' is a type
2169
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2170
% (where, re_string, e))
2173
if sys.platform == "win32":
2176
return msvcrt.getch()
2181
fd = sys.stdin.fileno()
2182
settings = termios.tcgetattr(fd)
2185
ch = sys.stdin.read(1)
2187
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2191
if sys.platform == 'linux2':
2192
def _local_concurrency():
2194
prefix = 'processor'
2195
for line in file('/proc/cpuinfo', 'rb'):
2196
if line.startswith(prefix):
2197
concurrency = int(line[line.find(':')+1:]) + 1
2199
elif sys.platform == 'darwin':
2200
def _local_concurrency():
2201
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2202
stdout=subprocess.PIPE).communicate()[0]
2203
elif sys.platform[0:7] == 'freebsd':
2204
def _local_concurrency():
2205
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2206
stdout=subprocess.PIPE).communicate()[0]
2207
elif sys.platform == 'sunos5':
2208
def _local_concurrency():
2209
return subprocess.Popen(['psrinfo', '-p',],
2210
stdout=subprocess.PIPE).communicate()[0]
2211
elif sys.platform == "win32":
2212
def _local_concurrency():
2213
# This appears to return the number of cores.
2214
return os.environ.get('NUMBER_OF_PROCESSORS')
2216
def _local_concurrency():
2221
_cached_local_concurrency = None
2223
def local_concurrency(use_cache=True):
2224
"""Return how many processes can be run concurrently.
2226
Rely on platform specific implementations and default to 1 (one) if
2227
anything goes wrong.
2229
global _cached_local_concurrency
2231
if _cached_local_concurrency is not None and use_cache:
2232
return _cached_local_concurrency
2234
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2235
if concurrency is None:
2237
concurrency = _local_concurrency()
2238
except (OSError, IOError):
2241
concurrency = int(concurrency)
2242
except (TypeError, ValueError):
2245
_cached_concurrency = concurrency
2249
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2250
"""A stream writer that doesn't decode str arguments."""
2252
def __init__(self, encode, stream, errors='strict'):
2253
codecs.StreamWriter.__init__(self, stream, errors)
2254
self.encode = encode
2256
def write(self, object):
2257
if type(object) is str:
2258
self.stream.write(object)
2260
data, _ = self.encode(object, self.errors)
2261
self.stream.write(data)
2263
if sys.platform == 'win32':
2264
def open_file(filename, mode='r', bufsize=-1):
2265
"""This function is used to override the ``open`` builtin.
2267
But it uses O_NOINHERIT flag so the file handle is not inherited by
2268
child processes. Deleting or renaming a closed file opened with this
2269
function is not blocking child processes.
2271
writing = 'w' in mode
2272
appending = 'a' in mode
2273
updating = '+' in mode
2274
binary = 'b' in mode
2277
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2278
# for flags for each modes.
2288
flags |= os.O_WRONLY
2289
flags |= os.O_CREAT | os.O_TRUNC
2294
flags |= os.O_WRONLY
2295
flags |= os.O_CREAT | os.O_APPEND
2300
flags |= os.O_RDONLY
2302
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
