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 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.
1160
if len(base) < MIN_ABS_PATHLENGTH:
1161
# must have space for e.g. a drive letter
1162
raise ValueError('%r is too short to calculate a relative path'
1170
if len(head) <= len(base) and head != base:
1171
raise errors.PathNotChild(rp, base)
1174
head, tail = split(head)
1179
return pathjoin(*reversed(s))
1184
def _cicp_canonical_relpath(base, path):
1185
"""Return the canonical path relative to base.
1187
Like relpath, but on case-insensitive-case-preserving file-systems, this
1188
will return the relpath as stored on the file-system rather than in the
1189
case specified in the input string, for all existing portions of the path.
1191
This will cause O(N) behaviour if called for every path in a tree; if you
1192
have a number of paths to convert, you should use canonical_relpaths().
1194
# TODO: it should be possible to optimize this for Windows by using the
1195
# win32 API FindFiles function to look for the specified name - but using
1196
# os.listdir() still gives us the correct, platform agnostic semantics in
1199
rel = relpath(base, path)
1200
# '.' will have been turned into ''
1204
abs_base = abspath(base)
1206
_listdir = os.listdir
1208
# use an explicit iterator so we can easily consume the rest on early exit.
1209
bit_iter = iter(rel.split('/'))
1210
for bit in bit_iter:
1213
next_entries = _listdir(current)
1214
except OSError: # enoent, eperm, etc
1215
# We can't find this in the filesystem, so just append the
1217
current = pathjoin(current, bit, *list(bit_iter))
1219
for look in next_entries:
1220
if lbit == look.lower():
1221
current = pathjoin(current, look)
1224
# got to the end, nothing matched, so we just return the
1225
# non-existing bits as they were specified (the filename may be
1226
# the target of a move, for example).
1227
current = pathjoin(current, bit, *list(bit_iter))
1229
return current[len(abs_base):].lstrip('/')
1231
# XXX - TODO - we need better detection/integration of case-insensitive
1232
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1233
# filesystems), for example, so could probably benefit from the same basic
1234
# support there. For now though, only Windows and OSX get that support, and
1235
# they get it for *all* file-systems!
1236
if sys.platform in ('win32', 'darwin'):
1237
canonical_relpath = _cicp_canonical_relpath
1239
canonical_relpath = relpath
1241
def canonical_relpaths(base, paths):
1242
"""Create an iterable to canonicalize a sequence of relative paths.
1244
The intent is for this implementation to use a cache, vastly speeding
1245
up multiple transformations in the same directory.
1247
# but for now, we haven't optimized...
1248
return [canonical_relpath(base, p) for p in paths]
1250
def safe_unicode(unicode_or_utf8_string):
1251
"""Coerce unicode_or_utf8_string into unicode.
1253
If it is unicode, it is returned.
1254
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1255
wrapped in a BzrBadParameterNotUnicode exception.
1257
if isinstance(unicode_or_utf8_string, unicode):
1258
return unicode_or_utf8_string
1260
return unicode_or_utf8_string.decode('utf8')
1261
except UnicodeDecodeError:
1262
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1265
def safe_utf8(unicode_or_utf8_string):
1266
"""Coerce unicode_or_utf8_string to a utf8 string.
1268
If it is a str, it is returned.
1269
If it is Unicode, it is encoded into a utf-8 string.
1271
if isinstance(unicode_or_utf8_string, str):
1272
# TODO: jam 20070209 This is overkill, and probably has an impact on
1273
# performance if we are dealing with lots of apis that want a
1276
# Make sure it is a valid utf-8 string
1277
unicode_or_utf8_string.decode('utf-8')
1278
except UnicodeDecodeError:
1279
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1280
return unicode_or_utf8_string
1281
return unicode_or_utf8_string.encode('utf-8')
1284
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1285
' Revision id generators should be creating utf8'
1289
def safe_revision_id(unicode_or_utf8_string, warn=True):
1290
"""Revision ids should now be utf8, but at one point they were unicode.
1292
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1294
:param warn: Functions that are sanitizing user data can set warn=False
1295
:return: None or a utf8 revision id.
1297
if (unicode_or_utf8_string is None
1298
or unicode_or_utf8_string.__class__ == str):
1299
return unicode_or_utf8_string
1301
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1303
return cache_utf8.encode(unicode_or_utf8_string)
1306
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1307
' generators should be creating utf8 file ids.')
1310
def safe_file_id(unicode_or_utf8_string, warn=True):
1311
"""File ids should now be utf8, but at one point they were unicode.
1313
This is the same as safe_utf8, except it uses the cached encode functions
1314
to save a little bit of performance.
1316
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1318
:param warn: Functions that are sanitizing user data can set warn=False
1319
:return: None or a utf8 file id.
1321
if (unicode_or_utf8_string is None
1322
or unicode_or_utf8_string.__class__ == str):
1323
return unicode_or_utf8_string
1325
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1327
return cache_utf8.encode(unicode_or_utf8_string)
1330
_platform_normalizes_filenames = False
1331
if sys.platform == 'darwin':
1332
_platform_normalizes_filenames = True
1335
def normalizes_filenames():
1336
"""Return True if this platform normalizes unicode filenames.
1338
Mac OSX does, Windows/Linux do not.
1340
return _platform_normalizes_filenames
1343
def _accessible_normalized_filename(path):
1344
"""Get the unicode normalized path, and if you can access the file.
1346
On platforms where the system normalizes filenames (Mac OSX),
1347
you can access a file by any path which will normalize correctly.
1348
On platforms where the system does not normalize filenames
1349
(Windows, Linux), you have to access a file by its exact path.
1351
Internally, bzr only supports NFC normalization, since that is
1352
the standard for XML documents.
1354
So return the normalized path, and a flag indicating if the file
1355
can be accessed by that path.
1358
return unicodedata.normalize('NFC', unicode(path)), True
1361
def _inaccessible_normalized_filename(path):
1362
__doc__ = _accessible_normalized_filename.__doc__
1364
normalized = unicodedata.normalize('NFC', unicode(path))
1365
return normalized, normalized == path
1368
if _platform_normalizes_filenames:
1369
normalized_filename = _accessible_normalized_filename
1371
normalized_filename = _inaccessible_normalized_filename
1374
def set_signal_handler(signum, handler, restart_syscall=True):
1375
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1376
on platforms that support that.
1378
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1379
automatically restart (by calling `signal.siginterrupt(signum,
1380
False)`). May be ignored if the feature is not available on this
1381
platform or Python version.
1385
siginterrupt = signal.siginterrupt
1387
# This python implementation doesn't provide signal support, hence no
1390
except AttributeError:
1391
# siginterrupt doesn't exist on this platform, or for this version
1393
siginterrupt = lambda signum, flag: None
1395
def sig_handler(*args):
1396
# Python resets the siginterrupt flag when a signal is
1397
# received. <http://bugs.python.org/issue8354>
1398
# As a workaround for some cases, set it back the way we want it.
1399
siginterrupt(signum, False)
1400
# Now run the handler function passed to set_signal_handler.
1403
sig_handler = handler
1404
old_handler = signal.signal(signum, sig_handler)
1406
siginterrupt(signum, False)
1410
default_terminal_width = 80
1411
"""The default terminal width for ttys.
1413
This is defined so that higher levels can share a common fallback value when
1414
terminal_width() returns None.
1418
def terminal_width():
1419
"""Return terminal width.
1421
None is returned if the width can't established precisely.
1424
- if BZR_COLUMNS is set, returns its value
1425
- if there is no controlling terminal, returns None
1426
- if COLUMNS is set, returns its value,
1428
From there, we need to query the OS to get the size of the controlling
1432
- get termios.TIOCGWINSZ
1433
- if an error occurs or a negative value is obtained, returns None
1437
- win32utils.get_console_size() decides,
1438
- returns None on error (provided default value)
1441
# If BZR_COLUMNS is set, take it, user is always right
1443
return int(os.environ['BZR_COLUMNS'])
1444
except (KeyError, ValueError):
1447
isatty = getattr(sys.stdout, 'isatty', None)
1448
if isatty is None or not isatty():
1449
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1452
# If COLUMNS is set, take it, the terminal knows better (even inside a
1453
# given terminal, the application can decide to set COLUMNS to a lower
1454
# value (splitted screen) or a bigger value (scroll bars))
1456
return int(os.environ['COLUMNS'])
1457
except (KeyError, ValueError):
1460
width, height = _terminal_size(None, None)
1462
# Consider invalid values as meaning no width
1468
def _win32_terminal_size(width, height):
1469
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1470
return width, height
1473
def _ioctl_terminal_size(width, height):
1475
import struct, fcntl, termios
1476
s = struct.pack('HHHH', 0, 0, 0, 0)
1477
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1478
height, width = struct.unpack('HHHH', x)[0:2]
1479
except (IOError, AttributeError):
1481
return width, height
1483
_terminal_size = None
1484
"""Returns the terminal size as (width, height).
1486
:param width: Default value for width.
1487
:param height: Default value for height.
1489
This is defined specifically for each OS and query the size of the controlling
1490
terminal. If any error occurs, the provided default values should be returned.
1492
if sys.platform == 'win32':
1493
_terminal_size = _win32_terminal_size
1495
_terminal_size = _ioctl_terminal_size
1498
def _terminal_size_changed(signum, frame):
1499
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1500
width, height = _terminal_size(None, None)
1501
if width is not None:
1502
os.environ['COLUMNS'] = str(width)
1505
_registered_sigwinch = False
1506
def watch_sigwinch():
1507
"""Register for SIGWINCH, once and only once.
1509
Do nothing if the signal module is not available.
1511
global _registered_sigwinch
1512
if not _registered_sigwinch:
1515
if getattr(signal, "SIGWINCH", None) is not None:
1516
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1518
# python doesn't provide signal support, nothing we can do about it
1520
_registered_sigwinch = True
1523
def supports_executable():
1524
return sys.platform != "win32"
1527
def supports_posix_readonly():
1528
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1530
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1531
directory controls creation/deletion, etc.
1533
And under win32, readonly means that the directory itself cannot be
1534
deleted. The contents of a readonly directory can be changed, unlike POSIX
1535
where files in readonly directories cannot be added, deleted or renamed.
1537
return sys.platform != "win32"
1540
def set_or_unset_env(env_variable, value):
1541
"""Modify the environment, setting or removing the env_variable.
1543
:param env_variable: The environment variable in question
1544
:param value: The value to set the environment to. If None, then
1545
the variable will be removed.
1546
:return: The original value of the environment variable.
1548
orig_val = os.environ.get(env_variable)
1550
if orig_val is not None:
1551
del os.environ[env_variable]
1553
if isinstance(value, unicode):
1554
value = value.encode(get_user_encoding())
1555
os.environ[env_variable] = value
1559
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1562
def check_legal_path(path):
1563
"""Check whether the supplied path is legal.
1564
This is only required on Windows, so we don't test on other platforms
1567
if sys.platform != "win32":
1569
if _validWin32PathRE.match(path) is None:
1570
raise errors.IllegalPath(path)
1573
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1575
def _is_error_enotdir(e):
1576
"""Check if this exception represents ENOTDIR.
1578
Unfortunately, python is very inconsistent about the exception
1579
here. The cases are:
1580
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1581
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1582
which is the windows error code.
1583
3) Windows, Python2.5 uses errno == EINVAL and
1584
winerror == ERROR_DIRECTORY
1586
:param e: An Exception object (expected to be OSError with an errno
1587
attribute, but we should be able to cope with anything)
1588
:return: True if this represents an ENOTDIR error. False otherwise.
1590
en = getattr(e, 'errno', None)
1591
if (en == errno.ENOTDIR
1592
or (sys.platform == 'win32'
1593
and (en == _WIN32_ERROR_DIRECTORY
1594
or (en == errno.EINVAL
1595
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1601
def walkdirs(top, prefix=""):
1602
"""Yield data about all the directories in a tree.
1604
This yields all the data about the contents of a directory at a time.
1605
After each directory has been yielded, if the caller has mutated the list
1606
to exclude some directories, they are then not descended into.
1608
The data yielded is of the form:
1609
((directory-relpath, directory-path-from-top),
1610
[(relpath, basename, kind, lstat, path-from-top), ...]),
1611
- directory-relpath is the relative path of the directory being returned
1612
with respect to top. prefix is prepended to this.
1613
- directory-path-from-root is the path including top for this directory.
1614
It is suitable for use with os functions.
1615
- relpath is the relative path within the subtree being walked.
1616
- basename is the basename of the path
1617
- kind is the kind of the file now. If unknown then the file is not
1618
present within the tree - but it may be recorded as versioned. See
1620
- lstat is the stat data *if* the file was statted.
1621
- planned, not implemented:
1622
path_from_tree_root is the path from the root of the tree.
1624
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1625
allows one to walk a subtree but get paths that are relative to a tree
1627
:return: an iterator over the dirs.
1629
#TODO there is a bit of a smell where the results of the directory-
1630
# summary in this, and the path from the root, may not agree
1631
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1632
# potentially confusing output. We should make this more robust - but
1633
# not at a speed cost. RBC 20060731
1635
_directory = _directory_kind
1636
_listdir = os.listdir
1637
_kind_from_mode = file_kind_from_stat_mode
1638
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1640
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1641
relroot, _, _, _, top = pending.pop()
1643
relprefix = relroot + u'/'
1646
top_slash = top + u'/'
1649
append = dirblock.append
1651
names = sorted(_listdir(top))
1653
if not _is_error_enotdir(e):
1657
abspath = top_slash + name
1658
statvalue = _lstat(abspath)
1659
kind = _kind_from_mode(statvalue.st_mode)
1660
append((relprefix + name, name, kind, statvalue, abspath))
1661
yield (relroot, top), dirblock
1663
# push the user specified dirs from dirblock
1664
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1667
class DirReader(object):
1668
"""An interface for reading directories."""
1670
def top_prefix_to_starting_dir(self, top, prefix=""):
1671
"""Converts top and prefix to a starting dir entry
1673
:param top: A utf8 path
1674
:param prefix: An optional utf8 path to prefix output relative paths
1676
:return: A tuple starting with prefix, and ending with the native
1679
raise NotImplementedError(self.top_prefix_to_starting_dir)
1681
def read_dir(self, prefix, top):
1682
"""Read a specific dir.
1684
:param prefix: A utf8 prefix to be preprended to the path basenames.
1685
:param top: A natively encoded path to read.
1686
:return: A list of the directories contents. Each item contains:
1687
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1689
raise NotImplementedError(self.read_dir)
1692
_selected_dir_reader = None
1695
def _walkdirs_utf8(top, prefix=""):
1696
"""Yield data about all the directories in a tree.
1698
This yields the same information as walkdirs() only each entry is yielded
1699
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1700
are returned as exact byte-strings.
1702
:return: yields a tuple of (dir_info, [file_info])
1703
dir_info is (utf8_relpath, path-from-top)
1704
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1705
if top is an absolute path, path-from-top is also an absolute path.
1706
path-from-top might be unicode or utf8, but it is the correct path to
1707
pass to os functions to affect the file in question. (such as os.lstat)
1709
global _selected_dir_reader
1710
if _selected_dir_reader is None:
1711
fs_encoding = _fs_enc.upper()
1712
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1713
# Win98 doesn't have unicode apis like FindFirstFileW
1714
# TODO: We possibly could support Win98 by falling back to the
1715
# original FindFirstFile, and using TCHAR instead of WCHAR,
1716
# but that gets a bit tricky, and requires custom compiling
1719
from bzrlib._walkdirs_win32 import Win32ReadDir
1720
_selected_dir_reader = Win32ReadDir()
1723
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1724
# ANSI_X3.4-1968 is a form of ASCII
1726
from bzrlib._readdir_pyx import UTF8DirReader
1727
_selected_dir_reader = UTF8DirReader()
1728
except ImportError, e:
1729
failed_to_load_extension(e)
1732
if _selected_dir_reader is None:
1733
# Fallback to the python version
1734
_selected_dir_reader = UnicodeDirReader()
1736
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1737
# But we don't actually uses 1-3 in pending, so set them to None
1738
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1739
read_dir = _selected_dir_reader.read_dir
1740
_directory = _directory_kind
1742
relroot, _, _, _, top = pending[-1].pop()
1745
dirblock = sorted(read_dir(relroot, top))
1746
yield (relroot, top), dirblock
1747
# push the user specified dirs from dirblock
1748
next = [d for d in reversed(dirblock) if d[2] == _directory]
1750
pending.append(next)
1753
class UnicodeDirReader(DirReader):
1754
"""A dir reader for non-utf8 file systems, which transcodes."""
1756
__slots__ = ['_utf8_encode']
1759
self._utf8_encode = codecs.getencoder('utf8')
1761
def top_prefix_to_starting_dir(self, top, prefix=""):
1762
"""See DirReader.top_prefix_to_starting_dir."""
1763
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1765
def read_dir(self, prefix, top):
1766
"""Read a single directory from a non-utf8 file system.
1768
top, and the abspath element in the output are unicode, all other paths
1769
are utf8. Local disk IO is done via unicode calls to listdir etc.
1771
This is currently the fallback code path when the filesystem encoding is
1772
not UTF-8. It may be better to implement an alternative so that we can
1773
safely handle paths that are not properly decodable in the current
1776
See DirReader.read_dir for details.
1778
_utf8_encode = self._utf8_encode
1780
_listdir = os.listdir
1781
_kind_from_mode = file_kind_from_stat_mode
1784
relprefix = prefix + '/'
1787
top_slash = top + u'/'
1790
append = dirblock.append
1791
for name in sorted(_listdir(top)):
1793
name_utf8 = _utf8_encode(name)[0]
1794
except UnicodeDecodeError:
1795
raise errors.BadFilenameEncoding(
1796
_utf8_encode(relprefix)[0] + name, _fs_enc)
1797
abspath = top_slash + name
1798
statvalue = _lstat(abspath)
1799
kind = _kind_from_mode(statvalue.st_mode)
1800
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1804
def copy_tree(from_path, to_path, handlers={}):
1805
"""Copy all of the entries in from_path into to_path.
1807
:param from_path: The base directory to copy.
1808
:param to_path: The target directory. If it does not exist, it will
1810
:param handlers: A dictionary of functions, which takes a source and
1811
destinations for files, directories, etc.
1812
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1813
'file', 'directory', and 'symlink' should always exist.
1814
If they are missing, they will be replaced with 'os.mkdir()',
1815
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1817
# Now, just copy the existing cached tree to the new location
1818
# We use a cheap trick here.
1819
# Absolute paths are prefixed with the first parameter
1820
# relative paths are prefixed with the second.
1821
# So we can get both the source and target returned
1822
# without any extra work.
1824
def copy_dir(source, dest):
1827
def copy_link(source, dest):
1828
"""Copy the contents of a symlink"""
1829
link_to = os.readlink(source)
1830
os.symlink(link_to, dest)
1832
real_handlers = {'file':shutil.copy2,
1833
'symlink':copy_link,
1834
'directory':copy_dir,
1836
real_handlers.update(handlers)
1838
if not os.path.exists(to_path):
1839
real_handlers['directory'](from_path, to_path)
1841
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1842
for relpath, name, kind, st, abspath in entries:
1843
real_handlers[kind](abspath, relpath)
1846
def copy_ownership_from_path(dst, src=None):
1847
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1849
If src is None, the containing directory is used as source. If chown
1850
fails, the error is ignored and a warning is printed.
1852
chown = getattr(os, 'chown', None)
1857
src = os.path.dirname(dst)
1863
chown(dst, s.st_uid, s.st_gid)
1865
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1868
def path_prefix_key(path):
1869
"""Generate a prefix-order path key for path.
1871
This can be used to sort paths in the same way that walkdirs does.
1873
return (dirname(path) , path)
1876
def compare_paths_prefix_order(path_a, path_b):
1877
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1878
key_a = path_prefix_key(path_a)
1879
key_b = path_prefix_key(path_b)
1880
return cmp(key_a, key_b)
1883
_cached_user_encoding = None
1886
def get_user_encoding(use_cache=True):
1887
"""Find out what the preferred user encoding is.
1889
This is generally the encoding that is used for command line parameters
1890
and file contents. This may be different from the terminal encoding
1891
or the filesystem encoding.
1893
:param use_cache: Enable cache for detected encoding.
1894
(This parameter is turned on by default,
1895
and required only for selftesting)
1897
:return: A string defining the preferred user encoding
1899
global _cached_user_encoding
1900
if _cached_user_encoding is not None and use_cache:
1901
return _cached_user_encoding
1903
if sys.platform == 'darwin':
1904
# python locale.getpreferredencoding() always return
1905
# 'mac-roman' on darwin. That's a lie.
1906
sys.platform = 'posix'
1908
if os.environ.get('LANG', None) is None:
1909
# If LANG is not set, we end up with 'ascii', which is bad
1910
# ('mac-roman' is more than ascii), so we set a default which
1911
# will give us UTF-8 (which appears to work in all cases on
1912
# OSX). Users are still free to override LANG of course, as
1913
# long as it give us something meaningful. This work-around
1914
# *may* not be needed with python 3k and/or OSX 10.5, but will
1915
# work with them too -- vila 20080908
1916
os.environ['LANG'] = 'en_US.UTF-8'
1919
sys.platform = 'darwin'
1924
user_encoding = locale.getpreferredencoding()
1925
except locale.Error, e:
1926
sys.stderr.write('bzr: warning: %s\n'
1927
' Could not determine what text encoding to use.\n'
1928
' This error usually means your Python interpreter\n'
1929
' doesn\'t support the locale set by $LANG (%s)\n'
1930
" Continuing with ascii encoding.\n"
1931
% (e, os.environ.get('LANG')))
1932
user_encoding = 'ascii'
1934
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1935
# treat that as ASCII, and not support printing unicode characters to the
1938
# For python scripts run under vim, we get '', so also treat that as ASCII
1939
if user_encoding in (None, 'cp0', ''):
1940
user_encoding = 'ascii'
1944
codecs.lookup(user_encoding)
1946
sys.stderr.write('bzr: warning:'
1947
' unknown encoding %s.'
1948
' Continuing with ascii encoding.\n'
1951
user_encoding = 'ascii'
1954
_cached_user_encoding = user_encoding
1956
return user_encoding
1959
def get_host_name():
1960
"""Return the current unicode host name.
1962
This is meant to be used in place of socket.gethostname() because that
1963
behaves inconsistently on different platforms.
1965
if sys.platform == "win32":
1967
return win32utils.get_host_name()
1970
return socket.gethostname().decode(get_user_encoding())
1973
# We must not read/write any more than 64k at a time from/to a socket so we
1974
# don't risk "no buffer space available" errors on some platforms. Windows in
1975
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1977
MAX_SOCKET_CHUNK = 64 * 1024
1979
def read_bytes_from_socket(sock, report_activity=None,
1980
max_read_size=MAX_SOCKET_CHUNK):
1981
"""Read up to max_read_size of bytes from sock and notify of progress.
1983
Translates "Connection reset by peer" into file-like EOF (return an
1984
empty string rather than raise an error), and repeats the recv if
1985
interrupted by a signal.
1989
bytes = sock.recv(max_read_size)
1990
except socket.error, e:
1992
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1993
# The connection was closed by the other side. Callers expect
1994
# an empty string to signal end-of-stream.
1996
elif eno == errno.EINTR:
1997
# Retry the interrupted recv.
2001
if report_activity is not None:
2002
report_activity(len(bytes), 'read')
2006
def recv_all(socket, count):
2007
"""Receive an exact number of bytes.
2009
Regular Socket.recv() may return less than the requested number of bytes,
2010
depending on what's in the OS buffer. MSG_WAITALL is not available
2011
on all platforms, but this should work everywhere. This will return
2012
less than the requested amount if the remote end closes.
2014
This isn't optimized and is intended mostly for use in testing.
2017
while len(b) < count:
2018
new = read_bytes_from_socket(socket, None, count - len(b))
2025
def send_all(sock, bytes, report_activity=None):
2026
"""Send all bytes on a socket.
2028
Breaks large blocks in smaller chunks to avoid buffering limitations on
2029
some platforms, and catches EINTR which may be thrown if the send is
2030
interrupted by a signal.
2032
This is preferred to socket.sendall(), because it avoids portability bugs
2033
and provides activity reporting.
2035
:param report_activity: Call this as bytes are read, see
2036
Transport._report_activity
2039
byte_count = len(bytes)
2040
while sent_total < byte_count:
2042
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2043
except socket.error, e:
2044
if e.args[0] != errno.EINTR:
2048
report_activity(sent, 'write')
2051
def dereference_path(path):
2052
"""Determine the real path to a file.
2054
All parent elements are dereferenced. But the file itself is not
2056
:param path: The original path. May be absolute or relative.
2057
:return: the real path *to* the file
2059
parent, base = os.path.split(path)
2060
# The pathjoin for '.' is a workaround for Python bug #1213894.
2061
# (initial path components aren't dereferenced)
2062
return pathjoin(realpath(pathjoin('.', parent)), base)
2065
def supports_mapi():
2066
"""Return True if we can use MAPI to launch a mail client."""
2067
return sys.platform == "win32"
2070
def resource_string(package, resource_name):
2071
"""Load a resource from a package and return it as a string.
2073
Note: Only packages that start with bzrlib are currently supported.
2075
This is designed to be a lightweight implementation of resource
2076
loading in a way which is API compatible with the same API from
2078
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2079
If and when pkg_resources becomes a standard library, this routine
2082
# Check package name is within bzrlib
2083
if package == "bzrlib":
2084
resource_relpath = resource_name
2085
elif package.startswith("bzrlib."):
2086
package = package[len("bzrlib."):].replace('.', os.sep)
2087
resource_relpath = pathjoin(package, resource_name)
2089
raise errors.BzrError('resource package %s not in bzrlib' % package)
2091
# Map the resource to a file and read its contents
2092
base = dirname(bzrlib.__file__)
2093
if getattr(sys, 'frozen', None): # bzr.exe
2094
base = abspath(pathjoin(base, '..', '..'))
2095
filename = pathjoin(base, resource_relpath)
2096
return open(filename, 'rU').read()
2099
def file_kind_from_stat_mode_thunk(mode):
2100
global file_kind_from_stat_mode
2101
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2103
from bzrlib._readdir_pyx import UTF8DirReader
2104
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2105
except ImportError, e:
2106
# This is one time where we won't warn that an extension failed to
2107
# load. The extension is never available on Windows anyway.
2108
from bzrlib._readdir_py import (
2109
_kind_from_mode as file_kind_from_stat_mode
2111
return file_kind_from_stat_mode(mode)
2112
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2115
def file_kind(f, _lstat=os.lstat):
2117
return file_kind_from_stat_mode(_lstat(f).st_mode)
2119
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2120
raise errors.NoSuchFile(f)
2124
def until_no_eintr(f, *a, **kw):
2125
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2127
WARNING: you must be certain that it is safe to retry the call repeatedly
2128
if EINTR does occur. This is typically only true for low-level operations
2129
like os.read. If in any doubt, don't use this.
2131
Keep in mind that this is not a complete solution to EINTR. There is
2132
probably code in the Python standard library and other dependencies that
2133
may encounter EINTR if a signal arrives (and there is signal handler for
2134
that signal). So this function can reduce the impact for IO that bzrlib
2135
directly controls, but it is not a complete solution.
2137
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2141
except (IOError, OSError), e:
2142
if e.errno == errno.EINTR:
2147
def re_compile_checked(re_string, flags=0, where=""):
2148
"""Return a compiled re, or raise a sensible error.
2150
This should only be used when compiling user-supplied REs.
2152
:param re_string: Text form of regular expression.
2153
:param flags: eg re.IGNORECASE
2154
:param where: Message explaining to the user the context where
2155
it occurred, eg 'log search filter'.
2157
# from https://bugs.launchpad.net/bzr/+bug/251352
2159
re_obj = re.compile(re_string, flags)
2164
where = ' in ' + where
2165
# despite the name 'error' is a type
2166
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2167
% (where, re_string, e))
2170
if sys.platform == "win32":
2173
return msvcrt.getch()
2178
fd = sys.stdin.fileno()
2179
settings = termios.tcgetattr(fd)
2182
ch = sys.stdin.read(1)
2184
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2188
if sys.platform == 'linux2':
2189
def _local_concurrency():
2191
prefix = 'processor'
2192
for line in file('/proc/cpuinfo', 'rb'):
2193
if line.startswith(prefix):
2194
concurrency = int(line[line.find(':')+1:]) + 1
2196
elif sys.platform == 'darwin':
2197
def _local_concurrency():
2198
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2199
stdout=subprocess.PIPE).communicate()[0]
2200
elif sys.platform[0:7] == 'freebsd':
2201
def _local_concurrency():
2202
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2203
stdout=subprocess.PIPE).communicate()[0]
2204
elif sys.platform == 'sunos5':
2205
def _local_concurrency():
2206
return subprocess.Popen(['psrinfo', '-p',],
2207
stdout=subprocess.PIPE).communicate()[0]
2208
elif sys.platform == "win32":
2209
def _local_concurrency():
2210
# This appears to return the number of cores.
2211
return os.environ.get('NUMBER_OF_PROCESSORS')
2213
def _local_concurrency():
2218
_cached_local_concurrency = None
2220
def local_concurrency(use_cache=True):
2221
"""Return how many processes can be run concurrently.
2223
Rely on platform specific implementations and default to 1 (one) if
2224
anything goes wrong.
2226
global _cached_local_concurrency
2228
if _cached_local_concurrency is not None and use_cache:
2229
return _cached_local_concurrency
2231
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2232
if concurrency is None:
2234
concurrency = _local_concurrency()
2235
except (OSError, IOError):
2238
concurrency = int(concurrency)
2239
except (TypeError, ValueError):
2242
_cached_concurrency = concurrency
2246
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2247
"""A stream writer that doesn't decode str arguments."""
2249
def __init__(self, encode, stream, errors='strict'):
2250
codecs.StreamWriter.__init__(self, stream, errors)
2251
self.encode = encode
2253
def write(self, object):
2254
if type(object) is str:
2255
self.stream.write(object)
2257
data, _ = self.encode(object, self.errors)
2258
self.stream.write(data)
2260
if sys.platform == 'win32':
2261
def open_file(filename, mode='r', bufsize=-1):
2262
"""This function is used to override the ``open`` builtin.
2264
But it uses O_NOINHERIT flag so the file handle is not inherited by
2265
child processes. Deleting or renaming a closed file opened with this
2266
function is not blocking child processes.
2268
writing = 'w' in mode
2269
appending = 'a' in mode
2270
updating = '+' in mode
2271
binary = 'b' in mode
2274
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2275
# for flags for each modes.
2285
flags |= os.O_WRONLY
2286
flags |= os.O_CREAT | os.O_TRUNC
2291
flags |= os.O_WRONLY
2292
flags |= os.O_CREAT | os.O_APPEND
2297
flags |= os.O_RDONLY
2299
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
