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,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
27
from bzrlib.lazy_import import lazy_import
28
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
45
from tempfile import (
58
from bzrlib.symbol_versioning import (
63
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
65
if sys.version_info < (2, 5):
66
import md5 as _mod_md5
68
import sha as _mod_sha
78
from bzrlib import symbol_versioning
81
# Cross platform wall-clock time functionality with decent resolution.
82
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
83
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
84
# synchronized with ``time.time()``, this is only meant to be used to find
85
# delta times by subtracting from another call to this function.
86
timer_func = time.time
87
if sys.platform == 'win32':
88
timer_func = time.clock
90
# On win32, O_BINARY is used to indicate the file should
91
# be opened in binary mode, rather than text mode.
92
# On other platforms, O_BINARY doesn't exist, because
93
# they always open in binary mode, so it is okay to
94
# OR with 0 on those platforms.
95
# O_NOINHERIT and O_TEXT exists only on win32 too.
96
O_BINARY = getattr(os, 'O_BINARY', 0)
97
O_TEXT = getattr(os, 'O_TEXT', 0)
98
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
101
def get_unicode_argv():
103
user_encoding = get_user_encoding()
104
return [a.decode(user_encoding) for a in sys.argv[1:]]
105
except UnicodeDecodeError:
106
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
110
def make_readonly(filename):
111
"""Make a filename read-only."""
112
mod = os.lstat(filename).st_mode
113
if not stat.S_ISLNK(mod):
115
os.chmod(filename, mod)
118
def make_writable(filename):
119
mod = os.lstat(filename).st_mode
120
if not stat.S_ISLNK(mod):
122
os.chmod(filename, mod)
125
def minimum_path_selection(paths):
126
"""Return the smallset subset of paths which are outside paths.
128
:param paths: A container (and hence not None) of paths.
129
:return: A set of paths sufficient to include everything in paths via
130
is_inside, drawn from the paths parameter.
136
return path.split('/')
137
sorted_paths = sorted(list(paths), key=sort_key)
139
search_paths = [sorted_paths[0]]
140
for path in sorted_paths[1:]:
141
if not is_inside(search_paths[-1], path):
142
# This path is unique, add it
143
search_paths.append(path)
145
return set(search_paths)
152
"""Return a quoted filename filename
154
This previously used backslash quoting, but that works poorly on
156
# TODO: I'm not really sure this is the best format either.x
158
if _QUOTE_RE is None:
159
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
161
if _QUOTE_RE.search(f):
167
_directory_kind = 'directory'
170
"""Return the current umask"""
171
# Assume that people aren't messing with the umask while running
172
# XXX: This is not thread safe, but there is no way to get the
173
# umask without setting it
181
_directory_kind: "/",
183
'tree-reference': '+',
187
def kind_marker(kind):
189
return _kind_marker_map[kind]
191
# Slightly faster than using .get(, '') when the common case is that
196
lexists = getattr(os.path, 'lexists', None)
200
stat = getattr(os, 'lstat', os.stat)
204
if e.errno == errno.ENOENT:
207
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
210
def fancy_rename(old, new, rename_func, unlink_func):
211
"""A fancy rename, when you don't have atomic rename.
213
:param old: The old path, to rename from
214
:param new: The new path, to rename to
215
:param rename_func: The potentially non-atomic rename function
216
:param unlink_func: A way to delete the target file if the full rename
219
# sftp rename doesn't allow overwriting, so play tricks:
220
base = os.path.basename(new)
221
dirname = os.path.dirname(new)
222
# callers use different encodings for the paths so the following MUST
223
# respect that. We rely on python upcasting to unicode if new is unicode
224
# and keeping a str if not.
225
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
226
os.getpid(), rand_chars(10))
227
tmp_name = pathjoin(dirname, tmp_name)
229
# Rename the file out of the way, but keep track if it didn't exist
230
# We don't want to grab just any exception
231
# something like EACCES should prevent us from continuing
232
# The downside is that the rename_func has to throw an exception
233
# with an errno = ENOENT, or NoSuchFile
236
rename_func(new, tmp_name)
237
except (errors.NoSuchFile,), e:
240
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
241
# function raises an IOError with errno is None when a rename fails.
242
# This then gets caught here.
243
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
246
if (getattr(e, 'errno', None) is None
247
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
256
# This may throw an exception, in which case success will
258
rename_func(old, new)
260
except (IOError, OSError), e:
261
# source and target may be aliases of each other (e.g. on a
262
# case-insensitive filesystem), so we may have accidentally renamed
263
# source by when we tried to rename target
264
failure_exc = sys.exc_info()
265
if (file_existed and e.errno in (None, errno.ENOENT)
266
and old.lower() == new.lower()):
267
# source and target are the same file on a case-insensitive
268
# filesystem, so we don't generate an exception
272
# If the file used to exist, rename it back into place
273
# otherwise just delete it from the tmp location
275
unlink_func(tmp_name)
277
rename_func(tmp_name, new)
278
if failure_exc is not None:
279
raise failure_exc[0], failure_exc[1], failure_exc[2]
282
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
283
# choke on a Unicode string containing a relative path if
284
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
286
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
287
def _posix_abspath(path):
288
# jam 20060426 rather than encoding to fsencoding
289
# copy posixpath.abspath, but use os.getcwdu instead
290
if not posixpath.isabs(path):
291
path = posixpath.join(getcwd(), path)
292
return posixpath.normpath(path)
295
def _posix_realpath(path):
296
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
299
def _win32_fixdrive(path):
300
"""Force drive letters to be consistent.
302
win32 is inconsistent whether it returns lower or upper case
303
and even if it was consistent the user might type the other
304
so we force it to uppercase
305
running python.exe under cmd.exe return capital C:\\
306
running win32 python inside a cygwin shell returns lowercase c:\\
308
drive, path = _nt_splitdrive(path)
309
return drive.upper() + path
312
def _win32_abspath(path):
313
# Real _nt_abspath doesn't have a problem with a unicode cwd
314
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
317
def _win98_abspath(path):
318
"""Return the absolute version of a path.
319
Windows 98 safe implementation (python reimplementation
320
of Win32 API function GetFullPathNameW)
325
# \\HOST\path => //HOST/path
326
# //HOST/path => //HOST/path
327
# path => C:/cwd/path
330
# check for absolute path
331
drive = _nt_splitdrive(path)[0]
332
if drive == '' and path[:2] not in('//','\\\\'):
334
# we cannot simply os.path.join cwd and path
335
# because os.path.join('C:','/path') produce '/path'
336
# and this is incorrect
337
if path[:1] in ('/','\\'):
338
cwd = _nt_splitdrive(cwd)[0]
340
path = cwd + '\\' + path
341
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
344
def _win32_realpath(path):
345
# Real _nt_realpath doesn't have a problem with a unicode cwd
346
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
349
def _win32_pathjoin(*args):
350
return _nt_join(*args).replace('\\', '/')
353
def _win32_normpath(path):
354
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
358
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
361
def _win32_mkdtemp(*args, **kwargs):
362
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
365
def _add_rename_error_details(e, old, new):
366
new_e = OSError(e.errno, "failed to rename %s to %s: %s"
367
% (old, new, e.strerror))
369
new_e.to_filename = new
373
def _win32_rename(old, new):
374
"""We expect to be able to atomically replace 'new' with old.
376
On win32, if new exists, it must be moved out of the way first,
380
fancy_rename(old, new, rename_func=_wrapped_rename, unlink_func=os.unlink)
382
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
383
# If we try to rename a non-existant file onto cwd, we get
384
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
385
# if the old path doesn't exist, sometimes we get EACCES
386
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
391
def _wrapped_rename(old, new):
392
"""Rename a file or directory"""
395
except (IOError, OSError), e:
396
# this is eventually called by all rename-like functions, so should
398
raise _add_rename_error_details(e, old, new)
402
return unicodedata.normalize('NFC', os.getcwdu())
405
# Default is to just use the python builtins, but these can be rebound on
406
# particular platforms.
407
abspath = _posix_abspath
408
realpath = _posix_realpath
409
pathjoin = os.path.join
410
normpath = os.path.normpath
411
rename = _wrapped_rename # overridden below on win32
413
dirname = os.path.dirname
414
basename = os.path.basename
415
split = os.path.split
416
splitext = os.path.splitext
417
# These were already imported into local scope
418
# mkdtemp = tempfile.mkdtemp
419
# rmtree = shutil.rmtree
421
MIN_ABS_PATHLENGTH = 1
424
if sys.platform == 'win32':
425
if win32utils.winver == 'Windows 98':
426
abspath = _win98_abspath
428
abspath = _win32_abspath
429
realpath = _win32_realpath
430
pathjoin = _win32_pathjoin
431
normpath = _win32_normpath
432
getcwd = _win32_getcwd
433
mkdtemp = _win32_mkdtemp
434
rename = _win32_rename
436
MIN_ABS_PATHLENGTH = 3
438
def _win32_delete_readonly(function, path, excinfo):
439
"""Error handler for shutil.rmtree function [for win32]
440
Helps to remove files and dirs marked as read-only.
442
exception = excinfo[1]
443
if function in (os.remove, os.rmdir) \
444
and isinstance(exception, OSError) \
445
and exception.errno == errno.EACCES:
451
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
452
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
453
return shutil.rmtree(path, ignore_errors, onerror)
455
f = win32utils.get_unicode_argv # special function or None
459
elif sys.platform == 'darwin':
463
def get_terminal_encoding():
464
"""Find the best encoding for printing to the screen.
466
This attempts to check both sys.stdout and sys.stdin to see
467
what encoding they are in, and if that fails it falls back to
468
osutils.get_user_encoding().
469
The problem is that on Windows, locale.getpreferredencoding()
470
is not the same encoding as that used by the console:
471
http://mail.python.org/pipermail/python-list/2003-May/162357.html
473
On my standard US Windows XP, the preferred encoding is
474
cp1252, but the console is cp437
476
from bzrlib.trace import mutter
477
output_encoding = getattr(sys.stdout, 'encoding', None)
478
if not output_encoding:
479
input_encoding = getattr(sys.stdin, 'encoding', None)
480
if not input_encoding:
481
output_encoding = get_user_encoding()
482
mutter('encoding stdout as osutils.get_user_encoding() %r',
485
output_encoding = input_encoding
486
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
488
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
489
if output_encoding == 'cp0':
490
# invalid encoding (cp0 means 'no codepage' on Windows)
491
output_encoding = get_user_encoding()
492
mutter('cp0 is invalid encoding.'
493
' encoding stdout as osutils.get_user_encoding() %r',
497
codecs.lookup(output_encoding)
499
sys.stderr.write('bzr: warning:'
500
' unknown terminal encoding %s.\n'
501
' Using encoding %s instead.\n'
502
% (output_encoding, get_user_encoding())
504
output_encoding = get_user_encoding()
506
return output_encoding
509
def normalizepath(f):
510
if getattr(os.path, 'realpath', None) is not None:
514
[p,e] = os.path.split(f)
515
if e == "" or e == "." or e == "..":
518
return pathjoin(F(p), e)
522
"""True if f is an accessible directory."""
524
return S_ISDIR(os.lstat(f)[ST_MODE])
530
"""True if f is a regular file."""
532
return S_ISREG(os.lstat(f)[ST_MODE])
537
"""True if f is a symlink."""
539
return S_ISLNK(os.lstat(f)[ST_MODE])
543
def is_inside(dir, fname):
544
"""True if fname is inside dir.
546
The parameters should typically be passed to osutils.normpath first, so
547
that . and .. and repeated slashes are eliminated, and the separators
548
are canonical for the platform.
550
The empty string as a dir name is taken as top-of-tree and matches
553
# XXX: Most callers of this can actually do something smarter by
554
# looking at the inventory
564
return fname.startswith(dir)
567
def is_inside_any(dir_list, fname):
568
"""True if fname is inside any of given dirs."""
569
for dirname in dir_list:
570
if is_inside(dirname, fname):
575
def is_inside_or_parent_of_any(dir_list, fname):
576
"""True if fname is a child or a parent of any of the given files."""
577
for dirname in dir_list:
578
if is_inside(dirname, fname) or is_inside(fname, dirname):
583
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
584
report_activity=None, direction='read'):
585
"""Copy contents of one file to another.
587
The read_length can either be -1 to read to end-of-file (EOF) or
588
it can specify the maximum number of bytes to read.
590
The buff_size represents the maximum size for each read operation
591
performed on from_file.
593
:param report_activity: Call this as bytes are read, see
594
Transport._report_activity
595
:param direction: Will be passed to report_activity
597
:return: The number of bytes copied.
601
# read specified number of bytes
603
while read_length > 0:
604
num_bytes_to_read = min(read_length, buff_size)
606
block = from_file.read(num_bytes_to_read)
610
if report_activity is not None:
611
report_activity(len(block), direction)
614
actual_bytes_read = len(block)
615
read_length -= actual_bytes_read
616
length += actual_bytes_read
620
block = from_file.read(buff_size)
624
if report_activity is not None:
625
report_activity(len(block), direction)
631
def pump_string_file(bytes, file_handle, segment_size=None):
632
"""Write bytes to file_handle in many smaller writes.
634
:param bytes: The string to write.
635
:param file_handle: The file to write to.
637
# Write data in chunks rather than all at once, because very large
638
# writes fail on some platforms (e.g. Windows with SMB mounted
641
segment_size = 5242880 # 5MB
642
segments = range(len(bytes) / segment_size + 1)
643
write = file_handle.write
644
for segment_index in segments:
645
segment = buffer(bytes, segment_index * segment_size, segment_size)
649
def file_iterator(input_file, readsize=32768):
651
b = input_file.read(readsize)
658
"""Calculate the hexdigest of an open file.
660
The file cursor should be already at the start.
672
def size_sha_file(f):
673
"""Calculate the size and hexdigest of an open file.
675
The file cursor should be already at the start and
676
the caller is responsible for closing the file afterwards.
687
return size, s.hexdigest()
690
def sha_file_by_name(fname):
691
"""Calculate the SHA1 of a file by reading the full text"""
693
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
696
b = os.read(f, 1<<16)
704
def sha_strings(strings, _factory=sha):
705
"""Return the sha-1 of concatenation of strings"""
707
map(s.update, strings)
711
def sha_string(f, _factory=sha):
712
return _factory(f).hexdigest()
715
def fingerprint_file(f):
717
return {'size': len(b),
718
'sha1': sha(b).hexdigest()}
721
def compare_files(a, b):
722
"""Returns true if equal in contents"""
733
def local_time_offset(t=None):
734
"""Return offset of local zone from GMT, either at present or at time t."""
737
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
738
return offset.days * 86400 + offset.seconds
740
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
741
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
744
def format_date(t, offset=0, timezone='original', date_fmt=None,
746
"""Return a formatted date string.
748
:param t: Seconds since the epoch.
749
:param offset: Timezone offset in seconds east of utc.
750
:param timezone: How to display the time: 'utc', 'original' for the
751
timezone specified by offset, or 'local' for the process's current
753
:param date_fmt: strftime format.
754
:param show_offset: Whether to append the timezone.
756
(date_fmt, tt, offset_str) = \
757
_format_date(t, offset, timezone, date_fmt, show_offset)
758
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
759
date_str = time.strftime(date_fmt, tt)
760
return date_str + offset_str
763
# Cache of formatted offset strings
767
def format_date_with_offset_in_original_timezone(t, offset=0,
768
_cache=_offset_cache):
769
"""Return a formatted date string in the original timezone.
771
This routine may be faster then format_date.
773
:param t: Seconds since the epoch.
774
:param offset: Timezone offset in seconds east of utc.
778
tt = time.gmtime(t + offset)
779
date_fmt = _default_format_by_weekday_num[tt[6]]
780
date_str = time.strftime(date_fmt, tt)
781
offset_str = _cache.get(offset, None)
782
if offset_str is None:
783
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
784
_cache[offset] = offset_str
785
return date_str + offset_str
788
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
790
"""Return an unicode date string formatted according to the current locale.
792
:param t: Seconds since the epoch.
793
:param offset: Timezone offset in seconds east of utc.
794
:param timezone: How to display the time: 'utc', 'original' for the
795
timezone specified by offset, or 'local' for the process's current
797
:param date_fmt: strftime format.
798
:param show_offset: Whether to append the timezone.
800
(date_fmt, tt, offset_str) = \
801
_format_date(t, offset, timezone, date_fmt, show_offset)
802
date_str = time.strftime(date_fmt, tt)
803
if not isinstance(date_str, unicode):
804
date_str = date_str.decode(get_user_encoding(), 'replace')
805
return date_str + offset_str
808
def _format_date(t, offset, timezone, date_fmt, show_offset):
809
if timezone == 'utc':
812
elif timezone == 'original':
815
tt = time.gmtime(t + offset)
816
elif timezone == 'local':
817
tt = time.localtime(t)
818
offset = local_time_offset(t)
820
raise errors.UnsupportedTimezoneFormat(timezone)
822
date_fmt = "%a %Y-%m-%d %H:%M:%S"
824
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
827
return (date_fmt, tt, offset_str)
830
def compact_date(when):
831
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
834
def format_delta(delta):
835
"""Get a nice looking string for a time delta.
837
:param delta: The time difference in seconds, can be positive or negative.
838
positive indicates time in the past, negative indicates time in the
839
future. (usually time.time() - stored_time)
840
:return: String formatted to show approximate resolution
846
direction = 'in the future'
850
if seconds < 90: # print seconds up to 90 seconds
852
return '%d second %s' % (seconds, direction,)
854
return '%d seconds %s' % (seconds, direction)
856
minutes = int(seconds / 60)
857
seconds -= 60 * minutes
862
if minutes < 90: # print minutes, seconds up to 90 minutes
864
return '%d minute, %d second%s %s' % (
865
minutes, seconds, plural_seconds, direction)
867
return '%d minutes, %d second%s %s' % (
868
minutes, seconds, plural_seconds, direction)
870
hours = int(minutes / 60)
871
minutes -= 60 * hours
878
return '%d hour, %d minute%s %s' % (hours, minutes,
879
plural_minutes, direction)
880
return '%d hours, %d minute%s %s' % (hours, minutes,
881
plural_minutes, direction)
884
"""Return size of given open file."""
885
return os.fstat(f.fileno())[ST_SIZE]
888
# Define rand_bytes based on platform.
890
# Python 2.4 and later have os.urandom,
891
# but it doesn't work on some arches
893
rand_bytes = os.urandom
894
except (NotImplementedError, AttributeError):
895
# If python doesn't have os.urandom, or it doesn't work,
896
# then try to first pull random data from /dev/urandom
898
rand_bytes = file('/dev/urandom', 'rb').read
899
# Otherwise, use this hack as a last resort
900
except (IOError, OSError):
901
# not well seeded, but better than nothing
906
s += chr(random.randint(0, 255))
911
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
913
"""Return a random string of num alphanumeric characters
915
The result only contains lowercase chars because it may be used on
916
case-insensitive filesystems.
919
for raw_byte in rand_bytes(num):
920
s += ALNUM[ord(raw_byte) % 36]
924
## TODO: We could later have path objects that remember their list
925
## decomposition (might be too tricksy though.)
928
"""Turn string into list of parts."""
929
# split on either delimiter because people might use either on
931
ps = re.split(r'[\\/]', p)
936
raise errors.BzrError("sorry, %r not allowed in path" % f)
937
elif (f == '.') or (f == ''):
946
if (f == '..') or (f is None) or (f == ''):
947
raise errors.BzrError("sorry, %r not allowed in path" % f)
951
def parent_directories(filename):
952
"""Return the list of parent directories, deepest first.
954
For example, parent_directories("a/b/c") -> ["a/b", "a"].
957
parts = splitpath(dirname(filename))
959
parents.append(joinpath(parts))
964
_extension_load_failures = []
967
def failed_to_load_extension(exception):
968
"""Handle failing to load a binary extension.
970
This should be called from the ImportError block guarding the attempt to
971
import the native extension. If this function returns, the pure-Python
972
implementation should be loaded instead::
975
>>> import bzrlib._fictional_extension_pyx
976
>>> except ImportError, e:
977
>>> bzrlib.osutils.failed_to_load_extension(e)
978
>>> import bzrlib._fictional_extension_py
980
# NB: This docstring is just an example, not a doctest, because doctest
981
# currently can't cope with the use of lazy imports in this namespace --
984
# This currently doesn't report the failure at the time it occurs, because
985
# they tend to happen very early in startup when we can't check config
986
# files etc, and also we want to report all failures but not spam the user
988
from bzrlib import trace
989
exception_str = str(exception)
990
if exception_str not in _extension_load_failures:
991
trace.mutter("failed to load compiled extension: %s" % exception_str)
992
_extension_load_failures.append(exception_str)
995
def report_extension_load_failures():
996
if not _extension_load_failures:
998
from bzrlib.config import GlobalConfig
999
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1001
# the warnings framework should by default show this only once
1002
from bzrlib.trace import warning
1004
"bzr: warning: some compiled extensions could not be loaded; "
1005
"see <https://answers.launchpad.net/bzr/+faq/703>")
1006
# we no longer show the specific missing extensions here, because it makes
1007
# the message too long and scary - see
1008
# https://bugs.launchpad.net/bzr/+bug/430529
1012
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1013
except ImportError, e:
1014
failed_to_load_extension(e)
1015
from bzrlib._chunks_to_lines_py import chunks_to_lines
1019
"""Split s into lines, but without removing the newline characters."""
1020
# Trivially convert a fulltext into a 'chunked' representation, and let
1021
# chunks_to_lines do the heavy lifting.
1022
if isinstance(s, str):
1023
# chunks_to_lines only supports 8-bit strings
1024
return chunks_to_lines([s])
1026
return _split_lines(s)
1029
def _split_lines(s):
1030
"""Split s into lines, but without removing the newline characters.
1032
This supports Unicode or plain string objects.
1034
lines = s.split('\n')
1035
result = [line + '\n' for line in lines[:-1]]
1037
result.append(lines[-1])
1041
def hardlinks_good():
1042
return sys.platform not in ('win32', 'cygwin', 'darwin')
1045
def link_or_copy(src, dest):
1046
"""Hardlink a file, or copy it if it can't be hardlinked."""
1047
if not hardlinks_good():
1048
shutil.copyfile(src, dest)
1052
except (OSError, IOError), e:
1053
if e.errno != errno.EXDEV:
1055
shutil.copyfile(src, dest)
1058
def delete_any(path):
1059
"""Delete a file, symlink or directory.
1061
Will delete even if readonly.
1064
_delete_file_or_dir(path)
1065
except (OSError, IOError), e:
1066
if e.errno in (errno.EPERM, errno.EACCES):
1067
# make writable and try again
1070
except (OSError, IOError):
1072
_delete_file_or_dir(path)
1077
def _delete_file_or_dir(path):
1078
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1079
# Forgiveness than Permission (EAFP) because:
1080
# - root can damage a solaris file system by using unlink,
1081
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1082
# EACCES, OSX: EPERM) when invoked on a directory.
1083
if isdir(path): # Takes care of symlinks
1090
if getattr(os, 'symlink', None) is not None:
1096
def has_hardlinks():
1097
if getattr(os, 'link', None) is not None:
1103
def host_os_dereferences_symlinks():
1104
return (has_symlinks()
1105
and sys.platform not in ('cygwin', 'win32'))
1108
def readlink(abspath):
1109
"""Return a string representing the path to which the symbolic link points.
1111
:param abspath: The link absolute unicode path.
1113
This his guaranteed to return the symbolic link in unicode in all python
1116
link = abspath.encode(_fs_enc)
1117
target = os.readlink(link)
1118
target = target.decode(_fs_enc)
1122
def contains_whitespace(s):
1123
"""True if there are any whitespace characters in s."""
1124
# string.whitespace can include '\xa0' in certain locales, because it is
1125
# considered "non-breaking-space" as part of ISO-8859-1. But it
1126
# 1) Isn't a breaking whitespace
1127
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1129
# 3) '\xa0' isn't unicode safe since it is >128.
1131
# This should *not* be a unicode set of characters in case the source
1132
# string is not a Unicode string. We can auto-up-cast the characters since
1133
# they are ascii, but we don't want to auto-up-cast the string in case it
1135
for ch in ' \t\n\r\v\f':
1142
def contains_linebreaks(s):
1143
"""True if there is any vertical whitespace in s."""
1151
def relpath(base, path):
1152
"""Return path relative to base, or raise exception.
1154
The path may be either an absolute path or a path relative to the
1155
current working directory.
1157
os.path.commonprefix (python2.4) has a bad bug that it works just
1158
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1159
avoids that problem.
1162
if len(base) < MIN_ABS_PATHLENGTH:
1163
# must have space for e.g. a drive letter
1164
raise ValueError('%r is too short to calculate a relative path'
1172
if len(head) <= len(base) and head != base:
1173
raise errors.PathNotChild(rp, base)
1176
head, tail = split(head)
1181
return pathjoin(*reversed(s))
1186
def _cicp_canonical_relpath(base, path):
1187
"""Return the canonical path relative to base.
1189
Like relpath, but on case-insensitive-case-preserving file-systems, this
1190
will return the relpath as stored on the file-system rather than in the
1191
case specified in the input string, for all existing portions of the path.
1193
This will cause O(N) behaviour if called for every path in a tree; if you
1194
have a number of paths to convert, you should use canonical_relpaths().
1196
# TODO: it should be possible to optimize this for Windows by using the
1197
# win32 API FindFiles function to look for the specified name - but using
1198
# os.listdir() still gives us the correct, platform agnostic semantics in
1201
rel = relpath(base, path)
1202
# '.' will have been turned into ''
1206
abs_base = abspath(base)
1208
_listdir = os.listdir
1210
# use an explicit iterator so we can easily consume the rest on early exit.
1211
bit_iter = iter(rel.split('/'))
1212
for bit in bit_iter:
1215
next_entries = _listdir(current)
1216
except OSError: # enoent, eperm, etc
1217
# We can't find this in the filesystem, so just append the
1219
current = pathjoin(current, bit, *list(bit_iter))
1221
for look in next_entries:
1222
if lbit == look.lower():
1223
current = pathjoin(current, look)
1226
# got to the end, nothing matched, so we just return the
1227
# non-existing bits as they were specified (the filename may be
1228
# the target of a move, for example).
1229
current = pathjoin(current, bit, *list(bit_iter))
1231
return current[len(abs_base):].lstrip('/')
1233
# XXX - TODO - we need better detection/integration of case-insensitive
1234
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1235
# filesystems), for example, so could probably benefit from the same basic
1236
# support there. For now though, only Windows and OSX get that support, and
1237
# they get it for *all* file-systems!
1238
if sys.platform in ('win32', 'darwin'):
1239
canonical_relpath = _cicp_canonical_relpath
1241
canonical_relpath = relpath
1243
def canonical_relpaths(base, paths):
1244
"""Create an iterable to canonicalize a sequence of relative paths.
1246
The intent is for this implementation to use a cache, vastly speeding
1247
up multiple transformations in the same directory.
1249
# but for now, we haven't optimized...
1250
return [canonical_relpath(base, p) for p in paths]
1252
def safe_unicode(unicode_or_utf8_string):
1253
"""Coerce unicode_or_utf8_string into unicode.
1255
If it is unicode, it is returned.
1256
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1257
wrapped in a BzrBadParameterNotUnicode exception.
1259
if isinstance(unicode_or_utf8_string, unicode):
1260
return unicode_or_utf8_string
1262
return unicode_or_utf8_string.decode('utf8')
1263
except UnicodeDecodeError:
1264
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1267
def safe_utf8(unicode_or_utf8_string):
1268
"""Coerce unicode_or_utf8_string to a utf8 string.
1270
If it is a str, it is returned.
1271
If it is Unicode, it is encoded into a utf-8 string.
1273
if isinstance(unicode_or_utf8_string, str):
1274
# TODO: jam 20070209 This is overkill, and probably has an impact on
1275
# performance if we are dealing with lots of apis that want a
1278
# Make sure it is a valid utf-8 string
1279
unicode_or_utf8_string.decode('utf-8')
1280
except UnicodeDecodeError:
1281
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1282
return unicode_or_utf8_string
1283
return unicode_or_utf8_string.encode('utf-8')
1286
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1287
' Revision id generators should be creating utf8'
1291
def safe_revision_id(unicode_or_utf8_string, warn=True):
1292
"""Revision ids should now be utf8, but at one point they were unicode.
1294
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1296
:param warn: Functions that are sanitizing user data can set warn=False
1297
:return: None or a utf8 revision id.
1299
if (unicode_or_utf8_string is None
1300
or unicode_or_utf8_string.__class__ == str):
1301
return unicode_or_utf8_string
1303
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1305
return cache_utf8.encode(unicode_or_utf8_string)
1308
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1309
' generators should be creating utf8 file ids.')
1312
def safe_file_id(unicode_or_utf8_string, warn=True):
1313
"""File ids should now be utf8, but at one point they were unicode.
1315
This is the same as safe_utf8, except it uses the cached encode functions
1316
to save a little bit of performance.
1318
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1320
:param warn: Functions that are sanitizing user data can set warn=False
1321
:return: None or a utf8 file id.
1323
if (unicode_or_utf8_string is None
1324
or unicode_or_utf8_string.__class__ == str):
1325
return unicode_or_utf8_string
1327
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1329
return cache_utf8.encode(unicode_or_utf8_string)
1332
_platform_normalizes_filenames = False
1333
if sys.platform == 'darwin':
1334
_platform_normalizes_filenames = True
1337
def normalizes_filenames():
1338
"""Return True if this platform normalizes unicode filenames.
1340
Mac OSX does, Windows/Linux do not.
1342
return _platform_normalizes_filenames
1345
def _accessible_normalized_filename(path):
1346
"""Get the unicode normalized path, and if you can access the file.
1348
On platforms where the system normalizes filenames (Mac OSX),
1349
you can access a file by any path which will normalize correctly.
1350
On platforms where the system does not normalize filenames
1351
(Windows, Linux), you have to access a file by its exact path.
1353
Internally, bzr only supports NFC normalization, since that is
1354
the standard for XML documents.
1356
So return the normalized path, and a flag indicating if the file
1357
can be accessed by that path.
1360
return unicodedata.normalize('NFC', unicode(path)), True
1363
def _inaccessible_normalized_filename(path):
1364
__doc__ = _accessible_normalized_filename.__doc__
1366
normalized = unicodedata.normalize('NFC', unicode(path))
1367
return normalized, normalized == path
1370
if _platform_normalizes_filenames:
1371
normalized_filename = _accessible_normalized_filename
1373
normalized_filename = _inaccessible_normalized_filename
1376
def set_signal_handler(signum, handler, restart_syscall=True):
1377
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1378
on platforms that support that.
1380
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1381
automatically restart (by calling `signal.siginterrupt(signum,
1382
False)`). May be ignored if the feature is not available on this
1383
platform or Python version.
1387
siginterrupt = signal.siginterrupt
1389
# This python implementation doesn't provide signal support, hence no
1392
except AttributeError:
1393
# siginterrupt doesn't exist on this platform, or for this version
1395
siginterrupt = lambda signum, flag: None
1397
def sig_handler(*args):
1398
# Python resets the siginterrupt flag when a signal is
1399
# received. <http://bugs.python.org/issue8354>
1400
# As a workaround for some cases, set it back the way we want it.
1401
siginterrupt(signum, False)
1402
# Now run the handler function passed to set_signal_handler.
1405
sig_handler = handler
1406
old_handler = signal.signal(signum, sig_handler)
1408
siginterrupt(signum, False)
1412
default_terminal_width = 80
1413
"""The default terminal width for ttys.
1415
This is defined so that higher levels can share a common fallback value when
1416
terminal_width() returns None.
1420
def terminal_width():
1421
"""Return terminal width.
1423
None is returned if the width can't established precisely.
1426
- if BZR_COLUMNS is set, returns its value
1427
- if there is no controlling terminal, returns None
1428
- if COLUMNS is set, returns its value,
1430
From there, we need to query the OS to get the size of the controlling
1434
- get termios.TIOCGWINSZ
1435
- if an error occurs or a negative value is obtained, returns None
1439
- win32utils.get_console_size() decides,
1440
- returns None on error (provided default value)
1443
# If BZR_COLUMNS is set, take it, user is always right
1445
return int(os.environ['BZR_COLUMNS'])
1446
except (KeyError, ValueError):
1449
isatty = getattr(sys.stdout, 'isatty', None)
1450
if isatty is None or not isatty():
1451
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1454
# If COLUMNS is set, take it, the terminal knows better (even inside a
1455
# given terminal, the application can decide to set COLUMNS to a lower
1456
# value (splitted screen) or a bigger value (scroll bars))
1458
return int(os.environ['COLUMNS'])
1459
except (KeyError, ValueError):
1462
width, height = _terminal_size(None, None)
1464
# Consider invalid values as meaning no width
1470
def _win32_terminal_size(width, height):
1471
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1472
return width, height
1475
def _ioctl_terminal_size(width, height):
1477
import struct, fcntl, termios
1478
s = struct.pack('HHHH', 0, 0, 0, 0)
1479
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1480
height, width = struct.unpack('HHHH', x)[0:2]
1481
except (IOError, AttributeError):
1483
return width, height
1485
_terminal_size = None
1486
"""Returns the terminal size as (width, height).
1488
:param width: Default value for width.
1489
:param height: Default value for height.
1491
This is defined specifically for each OS and query the size of the controlling
1492
terminal. If any error occurs, the provided default values should be returned.
1494
if sys.platform == 'win32':
1495
_terminal_size = _win32_terminal_size
1497
_terminal_size = _ioctl_terminal_size
1500
def _terminal_size_changed(signum, frame):
1501
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1502
width, height = _terminal_size(None, None)
1503
if width is not None:
1504
os.environ['COLUMNS'] = str(width)
1507
_registered_sigwinch = False
1508
def watch_sigwinch():
1509
"""Register for SIGWINCH, once and only once.
1511
Do nothing if the signal module is not available.
1513
global _registered_sigwinch
1514
if not _registered_sigwinch:
1517
if getattr(signal, "SIGWINCH", None) is not None:
1518
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1520
# python doesn't provide signal support, nothing we can do about it
1522
_registered_sigwinch = True
1525
def supports_executable():
1526
return sys.platform != "win32"
1529
def supports_posix_readonly():
1530
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1532
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1533
directory controls creation/deletion, etc.
1535
And under win32, readonly means that the directory itself cannot be
1536
deleted. The contents of a readonly directory can be changed, unlike POSIX
1537
where files in readonly directories cannot be added, deleted or renamed.
1539
return sys.platform != "win32"
1542
def set_or_unset_env(env_variable, value):
1543
"""Modify the environment, setting or removing the env_variable.
1545
:param env_variable: The environment variable in question
1546
:param value: The value to set the environment to. If None, then
1547
the variable will be removed.
1548
:return: The original value of the environment variable.
1550
orig_val = os.environ.get(env_variable)
1552
if orig_val is not None:
1553
del os.environ[env_variable]
1555
if isinstance(value, unicode):
1556
value = value.encode(get_user_encoding())
1557
os.environ[env_variable] = value
1561
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1564
def check_legal_path(path):
1565
"""Check whether the supplied path is legal.
1566
This is only required on Windows, so we don't test on other platforms
1569
if sys.platform != "win32":
1571
if _validWin32PathRE.match(path) is None:
1572
raise errors.IllegalPath(path)
1575
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1577
def _is_error_enotdir(e):
1578
"""Check if this exception represents ENOTDIR.
1580
Unfortunately, python is very inconsistent about the exception
1581
here. The cases are:
1582
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1583
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1584
which is the windows error code.
1585
3) Windows, Python2.5 uses errno == EINVAL and
1586
winerror == ERROR_DIRECTORY
1588
:param e: An Exception object (expected to be OSError with an errno
1589
attribute, but we should be able to cope with anything)
1590
:return: True if this represents an ENOTDIR error. False otherwise.
1592
en = getattr(e, 'errno', None)
1593
if (en == errno.ENOTDIR
1594
or (sys.platform == 'win32'
1595
and (en == _WIN32_ERROR_DIRECTORY
1596
or (en == errno.EINVAL
1597
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1603
def walkdirs(top, prefix=""):
1604
"""Yield data about all the directories in a tree.
1606
This yields all the data about the contents of a directory at a time.
1607
After each directory has been yielded, if the caller has mutated the list
1608
to exclude some directories, they are then not descended into.
1610
The data yielded is of the form:
1611
((directory-relpath, directory-path-from-top),
1612
[(relpath, basename, kind, lstat, path-from-top), ...]),
1613
- directory-relpath is the relative path of the directory being returned
1614
with respect to top. prefix is prepended to this.
1615
- directory-path-from-root is the path including top for this directory.
1616
It is suitable for use with os functions.
1617
- relpath is the relative path within the subtree being walked.
1618
- basename is the basename of the path
1619
- kind is the kind of the file now. If unknown then the file is not
1620
present within the tree - but it may be recorded as versioned. See
1622
- lstat is the stat data *if* the file was statted.
1623
- planned, not implemented:
1624
path_from_tree_root is the path from the root of the tree.
1626
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1627
allows one to walk a subtree but get paths that are relative to a tree
1629
:return: an iterator over the dirs.
1631
#TODO there is a bit of a smell where the results of the directory-
1632
# summary in this, and the path from the root, may not agree
1633
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1634
# potentially confusing output. We should make this more robust - but
1635
# not at a speed cost. RBC 20060731
1637
_directory = _directory_kind
1638
_listdir = os.listdir
1639
_kind_from_mode = file_kind_from_stat_mode
1640
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1642
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1643
relroot, _, _, _, top = pending.pop()
1645
relprefix = relroot + u'/'
1648
top_slash = top + u'/'
1651
append = dirblock.append
1653
names = sorted(_listdir(top))
1655
if not _is_error_enotdir(e):
1659
abspath = top_slash + name
1660
statvalue = _lstat(abspath)
1661
kind = _kind_from_mode(statvalue.st_mode)
1662
append((relprefix + name, name, kind, statvalue, abspath))
1663
yield (relroot, top), dirblock
1665
# push the user specified dirs from dirblock
1666
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1669
class DirReader(object):
1670
"""An interface for reading directories."""
1672
def top_prefix_to_starting_dir(self, top, prefix=""):
1673
"""Converts top and prefix to a starting dir entry
1675
:param top: A utf8 path
1676
:param prefix: An optional utf8 path to prefix output relative paths
1678
:return: A tuple starting with prefix, and ending with the native
1681
raise NotImplementedError(self.top_prefix_to_starting_dir)
1683
def read_dir(self, prefix, top):
1684
"""Read a specific dir.
1686
:param prefix: A utf8 prefix to be preprended to the path basenames.
1687
:param top: A natively encoded path to read.
1688
:return: A list of the directories contents. Each item contains:
1689
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1691
raise NotImplementedError(self.read_dir)
1694
_selected_dir_reader = None
1697
def _walkdirs_utf8(top, prefix=""):
1698
"""Yield data about all the directories in a tree.
1700
This yields the same information as walkdirs() only each entry is yielded
1701
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1702
are returned as exact byte-strings.
1704
:return: yields a tuple of (dir_info, [file_info])
1705
dir_info is (utf8_relpath, path-from-top)
1706
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1707
if top is an absolute path, path-from-top is also an absolute path.
1708
path-from-top might be unicode or utf8, but it is the correct path to
1709
pass to os functions to affect the file in question. (such as os.lstat)
1711
global _selected_dir_reader
1712
if _selected_dir_reader is None:
1713
fs_encoding = _fs_enc.upper()
1714
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1715
# Win98 doesn't have unicode apis like FindFirstFileW
1716
# TODO: We possibly could support Win98 by falling back to the
1717
# original FindFirstFile, and using TCHAR instead of WCHAR,
1718
# but that gets a bit tricky, and requires custom compiling
1721
from bzrlib._walkdirs_win32 import Win32ReadDir
1722
_selected_dir_reader = Win32ReadDir()
1725
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1726
# ANSI_X3.4-1968 is a form of ASCII
1728
from bzrlib._readdir_pyx import UTF8DirReader
1729
_selected_dir_reader = UTF8DirReader()
1730
except ImportError, e:
1731
failed_to_load_extension(e)
1734
if _selected_dir_reader is None:
1735
# Fallback to the python version
1736
_selected_dir_reader = UnicodeDirReader()
1738
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1739
# But we don't actually uses 1-3 in pending, so set them to None
1740
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1741
read_dir = _selected_dir_reader.read_dir
1742
_directory = _directory_kind
1744
relroot, _, _, _, top = pending[-1].pop()
1747
dirblock = sorted(read_dir(relroot, top))
1748
yield (relroot, top), dirblock
1749
# push the user specified dirs from dirblock
1750
next = [d for d in reversed(dirblock) if d[2] == _directory]
1752
pending.append(next)
1755
class UnicodeDirReader(DirReader):
1756
"""A dir reader for non-utf8 file systems, which transcodes."""
1758
__slots__ = ['_utf8_encode']
1761
self._utf8_encode = codecs.getencoder('utf8')
1763
def top_prefix_to_starting_dir(self, top, prefix=""):
1764
"""See DirReader.top_prefix_to_starting_dir."""
1765
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1767
def read_dir(self, prefix, top):
1768
"""Read a single directory from a non-utf8 file system.
1770
top, and the abspath element in the output are unicode, all other paths
1771
are utf8. Local disk IO is done via unicode calls to listdir etc.
1773
This is currently the fallback code path when the filesystem encoding is
1774
not UTF-8. It may be better to implement an alternative so that we can
1775
safely handle paths that are not properly decodable in the current
1778
See DirReader.read_dir for details.
1780
_utf8_encode = self._utf8_encode
1782
_listdir = os.listdir
1783
_kind_from_mode = file_kind_from_stat_mode
1786
relprefix = prefix + '/'
1789
top_slash = top + u'/'
1792
append = dirblock.append
1793
for name in sorted(_listdir(top)):
1795
name_utf8 = _utf8_encode(name)[0]
1796
except UnicodeDecodeError:
1797
raise errors.BadFilenameEncoding(
1798
_utf8_encode(relprefix)[0] + name, _fs_enc)
1799
abspath = top_slash + name
1800
statvalue = _lstat(abspath)
1801
kind = _kind_from_mode(statvalue.st_mode)
1802
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1806
def copy_tree(from_path, to_path, handlers={}):
1807
"""Copy all of the entries in from_path into to_path.
1809
:param from_path: The base directory to copy.
1810
:param to_path: The target directory. If it does not exist, it will
1812
:param handlers: A dictionary of functions, which takes a source and
1813
destinations for files, directories, etc.
1814
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1815
'file', 'directory', and 'symlink' should always exist.
1816
If they are missing, they will be replaced with 'os.mkdir()',
1817
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1819
# Now, just copy the existing cached tree to the new location
1820
# We use a cheap trick here.
1821
# Absolute paths are prefixed with the first parameter
1822
# relative paths are prefixed with the second.
1823
# So we can get both the source and target returned
1824
# without any extra work.
1826
def copy_dir(source, dest):
1829
def copy_link(source, dest):
1830
"""Copy the contents of a symlink"""
1831
link_to = os.readlink(source)
1832
os.symlink(link_to, dest)
1834
real_handlers = {'file':shutil.copy2,
1835
'symlink':copy_link,
1836
'directory':copy_dir,
1838
real_handlers.update(handlers)
1840
if not os.path.exists(to_path):
1841
real_handlers['directory'](from_path, to_path)
1843
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1844
for relpath, name, kind, st, abspath in entries:
1845
real_handlers[kind](abspath, relpath)
1848
def copy_ownership_from_path(dst, src=None):
1849
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1851
If src is None, the containing directory is used as source. If chown
1852
fails, the error is ignored and a warning is printed.
1854
chown = getattr(os, 'chown', None)
1859
src = os.path.dirname(dst)
1865
chown(dst, s.st_uid, s.st_gid)
1867
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1870
def path_prefix_key(path):
1871
"""Generate a prefix-order path key for path.
1873
This can be used to sort paths in the same way that walkdirs does.
1875
return (dirname(path) , path)
1878
def compare_paths_prefix_order(path_a, path_b):
1879
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1880
key_a = path_prefix_key(path_a)
1881
key_b = path_prefix_key(path_b)
1882
return cmp(key_a, key_b)
1885
_cached_user_encoding = None
1888
def get_user_encoding(use_cache=True):
1889
"""Find out what the preferred user encoding is.
1891
This is generally the encoding that is used for command line parameters
1892
and file contents. This may be different from the terminal encoding
1893
or the filesystem encoding.
1895
:param use_cache: Enable cache for detected encoding.
1896
(This parameter is turned on by default,
1897
and required only for selftesting)
1899
:return: A string defining the preferred user encoding
1901
global _cached_user_encoding
1902
if _cached_user_encoding is not None and use_cache:
1903
return _cached_user_encoding
1905
if sys.platform == 'darwin':
1906
# python locale.getpreferredencoding() always return
1907
# 'mac-roman' on darwin. That's a lie.
1908
sys.platform = 'posix'
1910
if os.environ.get('LANG', None) is None:
1911
# If LANG is not set, we end up with 'ascii', which is bad
1912
# ('mac-roman' is more than ascii), so we set a default which
1913
# will give us UTF-8 (which appears to work in all cases on
1914
# OSX). Users are still free to override LANG of course, as
1915
# long as it give us something meaningful. This work-around
1916
# *may* not be needed with python 3k and/or OSX 10.5, but will
1917
# work with them too -- vila 20080908
1918
os.environ['LANG'] = 'en_US.UTF-8'
1921
sys.platform = 'darwin'
1926
user_encoding = locale.getpreferredencoding()
1927
except locale.Error, e:
1928
sys.stderr.write('bzr: warning: %s\n'
1929
' Could not determine what text encoding to use.\n'
1930
' This error usually means your Python interpreter\n'
1931
' doesn\'t support the locale set by $LANG (%s)\n'
1932
" Continuing with ascii encoding.\n"
1933
% (e, os.environ.get('LANG')))
1934
user_encoding = 'ascii'
1936
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1937
# treat that as ASCII, and not support printing unicode characters to the
1940
# For python scripts run under vim, we get '', so also treat that as ASCII
1941
if user_encoding in (None, 'cp0', ''):
1942
user_encoding = 'ascii'
1946
codecs.lookup(user_encoding)
1948
sys.stderr.write('bzr: warning:'
1949
' unknown encoding %s.'
1950
' Continuing with ascii encoding.\n'
1953
user_encoding = 'ascii'
1956
_cached_user_encoding = user_encoding
1958
return user_encoding
1961
def get_host_name():
1962
"""Return the current unicode host name.
1964
This is meant to be used in place of socket.gethostname() because that
1965
behaves inconsistently on different platforms.
1967
if sys.platform == "win32":
1969
return win32utils.get_host_name()
1972
return socket.gethostname().decode(get_user_encoding())
1975
# We must not read/write any more than 64k at a time from/to a socket so we
1976
# don't risk "no buffer space available" errors on some platforms. Windows in
1977
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1979
MAX_SOCKET_CHUNK = 64 * 1024
1981
def read_bytes_from_socket(sock, report_activity=None,
1982
max_read_size=MAX_SOCKET_CHUNK):
1983
"""Read up to max_read_size of bytes from sock and notify of progress.
1985
Translates "Connection reset by peer" into file-like EOF (return an
1986
empty string rather than raise an error), and repeats the recv if
1987
interrupted by a signal.
1991
bytes = sock.recv(max_read_size)
1992
except socket.error, e:
1994
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1995
# The connection was closed by the other side. Callers expect
1996
# an empty string to signal end-of-stream.
1998
elif eno == errno.EINTR:
1999
# Retry the interrupted recv.
2003
if report_activity is not None:
2004
report_activity(len(bytes), 'read')
2008
def recv_all(socket, count):
2009
"""Receive an exact number of bytes.
2011
Regular Socket.recv() may return less than the requested number of bytes,
2012
depending on what's in the OS buffer. MSG_WAITALL is not available
2013
on all platforms, but this should work everywhere. This will return
2014
less than the requested amount if the remote end closes.
2016
This isn't optimized and is intended mostly for use in testing.
2019
while len(b) < count:
2020
new = read_bytes_from_socket(socket, None, count - len(b))
2027
def send_all(sock, bytes, report_activity=None):
2028
"""Send all bytes on a socket.
2030
Breaks large blocks in smaller chunks to avoid buffering limitations on
2031
some platforms, and catches EINTR which may be thrown if the send is
2032
interrupted by a signal.
2034
This is preferred to socket.sendall(), because it avoids portability bugs
2035
and provides activity reporting.
2037
:param report_activity: Call this as bytes are read, see
2038
Transport._report_activity
2041
byte_count = len(bytes)
2042
while sent_total < byte_count:
2044
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2045
except socket.error, e:
2046
if e.args[0] != errno.EINTR:
2050
report_activity(sent, 'write')
2053
def dereference_path(path):
2054
"""Determine the real path to a file.
2056
All parent elements are dereferenced. But the file itself is not
2058
:param path: The original path. May be absolute or relative.
2059
:return: the real path *to* the file
2061
parent, base = os.path.split(path)
2062
# The pathjoin for '.' is a workaround for Python bug #1213894.
2063
# (initial path components aren't dereferenced)
2064
return pathjoin(realpath(pathjoin('.', parent)), base)
2067
def supports_mapi():
2068
"""Return True if we can use MAPI to launch a mail client."""
2069
return sys.platform == "win32"
2072
def resource_string(package, resource_name):
2073
"""Load a resource from a package and return it as a string.
2075
Note: Only packages that start with bzrlib are currently supported.
2077
This is designed to be a lightweight implementation of resource
2078
loading in a way which is API compatible with the same API from
2080
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2081
If and when pkg_resources becomes a standard library, this routine
2084
# Check package name is within bzrlib
2085
if package == "bzrlib":
2086
resource_relpath = resource_name
2087
elif package.startswith("bzrlib."):
2088
package = package[len("bzrlib."):].replace('.', os.sep)
2089
resource_relpath = pathjoin(package, resource_name)
2091
raise errors.BzrError('resource package %s not in bzrlib' % package)
2093
# Map the resource to a file and read its contents
2094
base = dirname(bzrlib.__file__)
2095
if getattr(sys, 'frozen', None): # bzr.exe
2096
base = abspath(pathjoin(base, '..', '..'))
2097
filename = pathjoin(base, resource_relpath)
2098
return open(filename, 'rU').read()
2101
def file_kind_from_stat_mode_thunk(mode):
2102
global file_kind_from_stat_mode
2103
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2105
from bzrlib._readdir_pyx import UTF8DirReader
2106
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2107
except ImportError, e:
2108
# This is one time where we won't warn that an extension failed to
2109
# load. The extension is never available on Windows anyway.
2110
from bzrlib._readdir_py import (
2111
_kind_from_mode as file_kind_from_stat_mode
2113
return file_kind_from_stat_mode(mode)
2114
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2117
def file_kind(f, _lstat=os.lstat):
2119
return file_kind_from_stat_mode(_lstat(f).st_mode)
2121
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2122
raise errors.NoSuchFile(f)
2126
def until_no_eintr(f, *a, **kw):
2127
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2129
WARNING: you must be certain that it is safe to retry the call repeatedly
2130
if EINTR does occur. This is typically only true for low-level operations
2131
like os.read. If in any doubt, don't use this.
2133
Keep in mind that this is not a complete solution to EINTR. There is
2134
probably code in the Python standard library and other dependencies that
2135
may encounter EINTR if a signal arrives (and there is signal handler for
2136
that signal). So this function can reduce the impact for IO that bzrlib
2137
directly controls, but it is not a complete solution.
2139
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2143
except (IOError, OSError), e:
2144
if e.errno == errno.EINTR:
2149
def re_compile_checked(re_string, flags=0, where=""):
2150
"""Return a compiled re, or raise a sensible error.
2152
This should only be used when compiling user-supplied REs.
2154
:param re_string: Text form of regular expression.
2155
:param flags: eg re.IGNORECASE
2156
:param where: Message explaining to the user the context where
2157
it occurred, eg 'log search filter'.
2159
# from https://bugs.launchpad.net/bzr/+bug/251352
2161
re_obj = re.compile(re_string, flags)
2166
where = ' in ' + where
2167
# despite the name 'error' is a type
2168
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2169
% (where, re_string, e))
2172
if sys.platform == "win32":
2175
return msvcrt.getch()
2180
fd = sys.stdin.fileno()
2181
settings = termios.tcgetattr(fd)
2184
ch = sys.stdin.read(1)
2186
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2190
if sys.platform == 'linux2':
2191
def _local_concurrency():
2193
prefix = 'processor'
2194
for line in file('/proc/cpuinfo', 'rb'):
2195
if line.startswith(prefix):
2196
concurrency = int(line[line.find(':')+1:]) + 1
2198
elif sys.platform == 'darwin':
2199
def _local_concurrency():
2200
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2201
stdout=subprocess.PIPE).communicate()[0]
2202
elif sys.platform[0:7] == 'freebsd':
2203
def _local_concurrency():
2204
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2205
stdout=subprocess.PIPE).communicate()[0]
2206
elif sys.platform == 'sunos5':
2207
def _local_concurrency():
2208
return subprocess.Popen(['psrinfo', '-p',],
2209
stdout=subprocess.PIPE).communicate()[0]
2210
elif sys.platform == "win32":
2211
def _local_concurrency():
2212
# This appears to return the number of cores.
2213
return os.environ.get('NUMBER_OF_PROCESSORS')
2215
def _local_concurrency():
2220
_cached_local_concurrency = None
2222
def local_concurrency(use_cache=True):
2223
"""Return how many processes can be run concurrently.
2225
Rely on platform specific implementations and default to 1 (one) if
2226
anything goes wrong.
2228
global _cached_local_concurrency
2230
if _cached_local_concurrency is not None and use_cache:
2231
return _cached_local_concurrency
2233
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2234
if concurrency is None:
2236
concurrency = _local_concurrency()
2237
except (OSError, IOError):
2240
concurrency = int(concurrency)
2241
except (TypeError, ValueError):
2244
_cached_concurrency = concurrency
2248
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2249
"""A stream writer that doesn't decode str arguments."""
2251
def __init__(self, encode, stream, errors='strict'):
2252
codecs.StreamWriter.__init__(self, stream, errors)
2253
self.encode = encode
2255
def write(self, object):
2256
if type(object) is str:
2257
self.stream.write(object)
2259
data, _ = self.encode(object, self.errors)
2260
self.stream.write(data)
2262
if sys.platform == 'win32':
2263
def open_file(filename, mode='r', bufsize=-1):
2264
"""This function is used to override the ``open`` builtin.
2266
But it uses O_NOINHERIT flag so the file handle is not inherited by
2267
child processes. Deleting or renaming a closed file opened with this
2268
function is not blocking child processes.
2270
writing = 'w' in mode
2271
appending = 'a' in mode
2272
updating = '+' in mode
2273
binary = 'b' in mode
2276
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2277
# for flags for each modes.
2287
flags |= os.O_WRONLY
2288
flags |= os.O_CREAT | os.O_TRUNC
2293
flags |= os.O_WRONLY
2294
flags |= os.O_CREAT | os.O_APPEND
2299
flags |= os.O_RDONLY
2301
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
