1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005-2011 Canonical Ltd
5
3
# This program is free software; you can redistribute it and/or modify
6
4
# it under the terms of the GNU General Public License as published by
7
5
# the Free Software Foundation; either version 2 of the License, or
8
6
# (at your option) any later version.
10
8
# This program is distributed in the hope that it will be useful,
11
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
11
# GNU General Public License for more details.
15
13
# You should have received a copy of the GNU General Public License
16
14
# along with this program; if not, write to the Free Software
17
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
import os, types, re, time, errno, sys
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
22
from errors import bailout, BzrError
23
from trace import mutter
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
27
from datetime import datetime
32
# We need to import both shutil and rmtree as we export the later on posix
33
# and need the former on windows
35
from shutil import rmtree
38
# We need to import both tempfile and mkdtemp as we export the later on posix
39
# and need the former on windows
41
from tempfile import mkdtemp
51
from bzrlib.i18n import gettext
54
from bzrlib.symbol_versioning import (
66
from bzrlib import symbol_versioning
69
# Cross platform wall-clock time functionality with decent resolution.
70
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
71
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
72
# synchronized with ``time.time()``, this is only meant to be used to find
73
# delta times by subtracting from another call to this function.
74
timer_func = time.time
75
if sys.platform == 'win32':
76
timer_func = time.clock
78
# On win32, O_BINARY is used to indicate the file should
79
# be opened in binary mode, rather than text mode.
80
# On other platforms, O_BINARY doesn't exist, because
81
# they always open in binary mode, so it is okay to
82
# OR with 0 on those platforms.
83
# O_NOINHERIT and O_TEXT exists only on win32 too.
84
O_BINARY = getattr(os, 'O_BINARY', 0)
85
O_TEXT = getattr(os, 'O_TEXT', 0)
86
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
89
def get_unicode_argv():
91
user_encoding = get_user_encoding()
92
return [a.decode(user_encoding) for a in sys.argv[1:]]
93
except UnicodeDecodeError:
94
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
95
"application locale.").format(a, user_encoding))
26
98
def make_readonly(filename):
27
99
"""Make a filename read-only."""
28
# TODO: probably needs to be fixed for windows
29
mod = os.stat(filename).st_mode
31
os.chmod(filename, mod)
100
mod = os.lstat(filename).st_mode
101
if not stat.S_ISLNK(mod):
103
chmod_if_possible(filename, mod)
34
106
def make_writable(filename):
35
mod = os.stat(filename).st_mode
37
os.chmod(filename, mod)
40
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/_~-])')
107
mod = os.lstat(filename).st_mode
108
if not stat.S_ISLNK(mod):
110
chmod_if_possible(filename, mod)
113
def chmod_if_possible(filename, mode):
114
# Set file mode if that can be safely done.
115
# Sometimes even on unix the filesystem won't allow it - see
116
# https://bugs.launchpad.net/bzr/+bug/606537
118
# It is probably faster to just do the chmod, rather than
119
# doing a stat, and then trying to compare
120
os.chmod(filename, mode)
121
except (IOError, OSError),e:
122
# Permission/access denied seems to commonly happen on smbfs; there's
123
# probably no point warning about it.
124
# <https://bugs.launchpad.net/bzr/+bug/606537>
125
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
126
trace.mutter("ignore error on chmod of %r: %r" % (
132
def minimum_path_selection(paths):
133
"""Return the smallset subset of paths which are outside paths.
135
:param paths: A container (and hence not None) of paths.
136
:return: A set of paths sufficient to include everything in paths via
137
is_inside, drawn from the paths parameter.
143
return path.split('/')
144
sorted_paths = sorted(list(paths), key=sort_key)
146
search_paths = [sorted_paths[0]]
147
for path in sorted_paths[1:]:
148
if not is_inside(search_paths[-1], path):
149
# This path is unique, add it
150
search_paths.append(path)
152
return set(search_paths)
42
"""Return shell-quoted filename"""
43
## We could be a bit more terse by using double-quotes etc
44
f = _QUOTE_RE.sub(r'\\\1', f)
51
mode = os.lstat(f)[ST_MODE]
159
"""Return a quoted filename filename
161
This previously used backslash quoting, but that works poorly on
163
# TODO: I'm not really sure this is the best format either.x
165
if _QUOTE_RE is None:
166
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
168
if _QUOTE_RE.search(f):
59
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
174
_directory_kind = 'directory'
177
"""Return the current umask"""
178
# Assume that people aren't messing with the umask while running
179
# XXX: This is not thread safe, but there is no way to get the
180
# umask without setting it
188
_directory_kind: "/",
190
'tree-reference': '+',
62
194
def kind_marker(kind):
196
return _kind_marker_map[kind]
198
# Slightly faster than using .get(, '') when the common case is that
65
elif kind == 'directory':
67
elif kind == 'symlink':
70
raise BzrError('invalid file kind %r' % kind)
203
lexists = getattr(os.path, 'lexists', None)
207
stat = getattr(os, 'lstat', os.stat)
211
if e.errno == errno.ENOENT:
214
raise errors.BzrError(gettext("lstat/stat of ({0!r}): {1!r}").format(f, e))
217
def fancy_rename(old, new, rename_func, unlink_func):
218
"""A fancy rename, when you don't have atomic rename.
220
:param old: The old path, to rename from
221
:param new: The new path, to rename to
222
:param rename_func: The potentially non-atomic rename function
223
:param unlink_func: A way to delete the target file if the full rename
226
# sftp rename doesn't allow overwriting, so play tricks:
227
base = os.path.basename(new)
228
dirname = os.path.dirname(new)
229
# callers use different encodings for the paths so the following MUST
230
# respect that. We rely on python upcasting to unicode if new is unicode
231
# and keeping a str if not.
232
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
233
os.getpid(), rand_chars(10))
234
tmp_name = pathjoin(dirname, tmp_name)
236
# Rename the file out of the way, but keep track if it didn't exist
237
# We don't want to grab just any exception
238
# something like EACCES should prevent us from continuing
239
# The downside is that the rename_func has to throw an exception
240
# with an errno = ENOENT, or NoSuchFile
243
rename_func(new, tmp_name)
244
except (errors.NoSuchFile,), e:
247
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
248
# function raises an IOError with errno is None when a rename fails.
249
# This then gets caught here.
250
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
253
if (getattr(e, 'errno', None) is None
254
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
263
# This may throw an exception, in which case success will
265
rename_func(old, new)
267
except (IOError, OSError), e:
268
# source and target may be aliases of each other (e.g. on a
269
# case-insensitive filesystem), so we may have accidentally renamed
270
# source by when we tried to rename target
271
failure_exc = sys.exc_info()
272
if (file_existed and e.errno in (None, errno.ENOENT)
273
and old.lower() == new.lower()):
274
# source and target are the same file on a case-insensitive
275
# filesystem, so we don't generate an exception
279
# If the file used to exist, rename it back into place
280
# otherwise just delete it from the tmp location
282
unlink_func(tmp_name)
284
rename_func(tmp_name, new)
285
if failure_exc is not None:
287
raise failure_exc[0], failure_exc[1], failure_exc[2]
292
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
293
# choke on a Unicode string containing a relative path if
294
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
296
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
297
def _posix_abspath(path):
298
# jam 20060426 rather than encoding to fsencoding
299
# copy posixpath.abspath, but use os.getcwdu instead
300
if not posixpath.isabs(path):
301
path = posixpath.join(getcwd(), path)
302
return _posix_normpath(path)
305
def _posix_realpath(path):
306
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
309
def _posix_normpath(path):
310
path = posixpath.normpath(path)
311
# Bug 861008: posixpath.normpath() returns a path normalized according to
312
# the POSIX standard, which stipulates (for compatibility reasons) that two
313
# leading slashes must not be simplified to one, and only if there are 3 or
314
# more should they be simplified as one. So we treat the leading 2 slashes
315
# as a special case here by simply removing the first slash, as we consider
316
# that breaking POSIX compatibility for this obscure feature is acceptable.
317
# This is not a paranoid precaution, as we notably get paths like this when
318
# the repo is hosted at the root of the filesystem, i.e. in "/".
319
if path.startswith('//'):
324
def _win32_fixdrive(path):
325
"""Force drive letters to be consistent.
327
win32 is inconsistent whether it returns lower or upper case
328
and even if it was consistent the user might type the other
329
so we force it to uppercase
330
running python.exe under cmd.exe return capital C:\\
331
running win32 python inside a cygwin shell returns lowercase c:\\
333
drive, path = ntpath.splitdrive(path)
334
return drive.upper() + path
337
def _win32_abspath(path):
338
# Real ntpath.abspath doesn't have a problem with a unicode cwd
339
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
342
def _win98_abspath(path):
343
"""Return the absolute version of a path.
344
Windows 98 safe implementation (python reimplementation
345
of Win32 API function GetFullPathNameW)
350
# \\HOST\path => //HOST/path
351
# //HOST/path => //HOST/path
352
# path => C:/cwd/path
355
# check for absolute path
356
drive = ntpath.splitdrive(path)[0]
357
if drive == '' and path[:2] not in('//','\\\\'):
359
# we cannot simply os.path.join cwd and path
360
# because os.path.join('C:','/path') produce '/path'
361
# and this is incorrect
362
if path[:1] in ('/','\\'):
363
cwd = ntpath.splitdrive(cwd)[0]
365
path = cwd + '\\' + path
366
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
369
def _win32_realpath(path):
370
# Real ntpath.realpath doesn't have a problem with a unicode cwd
371
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
374
def _win32_pathjoin(*args):
375
return ntpath.join(*args).replace('\\', '/')
378
def _win32_normpath(path):
379
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
383
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
386
def _win32_mkdtemp(*args, **kwargs):
387
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
390
def _win32_rename(old, new):
391
"""We expect to be able to atomically replace 'new' with old.
393
On win32, if new exists, it must be moved out of the way first,
397
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
399
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
400
# If we try to rename a non-existant file onto cwd, we get
401
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
402
# if the old path doesn't exist, sometimes we get EACCES
403
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
409
return unicodedata.normalize('NFC', os.getcwdu())
412
# Default is to just use the python builtins, but these can be rebound on
413
# particular platforms.
414
abspath = _posix_abspath
415
realpath = _posix_realpath
416
pathjoin = os.path.join
417
normpath = _posix_normpath
420
dirname = os.path.dirname
421
basename = os.path.basename
422
split = os.path.split
423
splitext = os.path.splitext
424
# These were already lazily imported into local scope
425
# mkdtemp = tempfile.mkdtemp
426
# rmtree = shutil.rmtree
434
MIN_ABS_PATHLENGTH = 1
437
if sys.platform == 'win32':
438
if win32utils.winver == 'Windows 98':
439
abspath = _win98_abspath
441
abspath = _win32_abspath
442
realpath = _win32_realpath
443
pathjoin = _win32_pathjoin
444
normpath = _win32_normpath
445
getcwd = _win32_getcwd
446
mkdtemp = _win32_mkdtemp
447
rename = _win32_rename
449
from bzrlib import _walkdirs_win32
453
lstat = _walkdirs_win32.lstat
454
fstat = _walkdirs_win32.fstat
455
wrap_stat = _walkdirs_win32.wrap_stat
457
MIN_ABS_PATHLENGTH = 3
459
def _win32_delete_readonly(function, path, excinfo):
460
"""Error handler for shutil.rmtree function [for win32]
461
Helps to remove files and dirs marked as read-only.
463
exception = excinfo[1]
464
if function in (os.remove, os.rmdir) \
465
and isinstance(exception, OSError) \
466
and exception.errno == errno.EACCES:
472
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
473
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
474
return shutil.rmtree(path, ignore_errors, onerror)
476
f = win32utils.get_unicode_argv # special function or None
480
elif sys.platform == 'darwin':
484
def get_terminal_encoding(trace=False):
485
"""Find the best encoding for printing to the screen.
487
This attempts to check both sys.stdout and sys.stdin to see
488
what encoding they are in, and if that fails it falls back to
489
osutils.get_user_encoding().
490
The problem is that on Windows, locale.getpreferredencoding()
491
is not the same encoding as that used by the console:
492
http://mail.python.org/pipermail/python-list/2003-May/162357.html
494
On my standard US Windows XP, the preferred encoding is
495
cp1252, but the console is cp437
497
:param trace: If True trace the selected encoding via mutter().
499
from bzrlib.trace import mutter
500
output_encoding = getattr(sys.stdout, 'encoding', None)
501
if not output_encoding:
502
input_encoding = getattr(sys.stdin, 'encoding', None)
503
if not input_encoding:
504
output_encoding = get_user_encoding()
506
mutter('encoding stdout as osutils.get_user_encoding() %r',
509
output_encoding = input_encoding
511
mutter('encoding stdout as sys.stdin encoding %r',
515
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
516
if output_encoding == 'cp0':
517
# invalid encoding (cp0 means 'no codepage' on Windows)
518
output_encoding = get_user_encoding()
520
mutter('cp0 is invalid encoding.'
521
' encoding stdout as osutils.get_user_encoding() %r',
525
codecs.lookup(output_encoding)
527
sys.stderr.write('bzr: warning:'
528
' unknown terminal encoding %s.\n'
529
' Using encoding %s instead.\n'
530
% (output_encoding, get_user_encoding())
532
output_encoding = get_user_encoding()
534
return output_encoding
537
def normalizepath(f):
538
if getattr(os.path, 'realpath', None) is not None:
542
[p,e] = os.path.split(f)
543
if e == "" or e == "." or e == "..":
546
return pathjoin(F(p), e)
75
550
"""True if f is an accessible directory."""
77
return S_ISDIR(os.lstat(f)[ST_MODE])
552
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
84
558
"""True if f is a regular file."""
86
return S_ISREG(os.lstat(f)[ST_MODE])
560
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
565
"""True if f is a symlink."""
567
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
91
571
def is_inside(dir, fname):
92
572
"""True if fname is inside dir.
574
The parameters should typically be passed to osutils.normpath first, so
575
that . and .. and repeated slashes are eliminated, and the separators
576
are canonical for the platform.
578
The empty string as a dir name is taken as top-of-tree and matches
94
return os.path.commonprefix([dir, fname]) == dir
581
# XXX: Most callers of this can actually do something smarter by
582
# looking at the inventory
592
return fname.startswith(dir)
97
595
def is_inside_any(dir_list, fname):
98
596
"""True if fname is inside any of given dirs."""
99
# quick scan for perfect match
100
if fname in dir_list:
103
597
for dirname in dir_list:
104
598
if is_inside(dirname, fname):
603
def is_inside_or_parent_of_any(dir_list, fname):
604
"""True if fname is a child or a parent of any of the given files."""
605
for dirname in dir_list:
606
if is_inside(dirname, fname) or is_inside(fname, dirname):
611
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
612
report_activity=None, direction='read'):
613
"""Copy contents of one file to another.
615
The read_length can either be -1 to read to end-of-file (EOF) or
616
it can specify the maximum number of bytes to read.
618
The buff_size represents the maximum size for each read operation
619
performed on from_file.
621
:param report_activity: Call this as bytes are read, see
622
Transport._report_activity
623
:param direction: Will be passed to report_activity
625
:return: The number of bytes copied.
629
# read specified number of bytes
631
while read_length > 0:
632
num_bytes_to_read = min(read_length, buff_size)
634
block = from_file.read(num_bytes_to_read)
638
if report_activity is not None:
639
report_activity(len(block), direction)
642
actual_bytes_read = len(block)
643
read_length -= actual_bytes_read
644
length += actual_bytes_read
110
def pumpfile(fromfile, tofile):
111
"""Copy contents of one file to another."""
112
tofile.write(fromfile.read())
116
"""Return a new UUID"""
118
return file('/proc/sys/kernel/random/uuid').readline().rstrip('\n')
120
return chomp(os.popen('uuidgen').readline())
648
block = from_file.read(buff_size)
652
if report_activity is not None:
653
report_activity(len(block), direction)
659
def pump_string_file(bytes, file_handle, segment_size=None):
660
"""Write bytes to file_handle in many smaller writes.
662
:param bytes: The string to write.
663
:param file_handle: The file to write to.
665
# Write data in chunks rather than all at once, because very large
666
# writes fail on some platforms (e.g. Windows with SMB mounted
669
segment_size = 5242880 # 5MB
670
segments = range(len(bytes) / segment_size + 1)
671
write = file_handle.write
672
for segment_index in segments:
673
segment = buffer(bytes, segment_index * segment_size, segment_size)
677
def file_iterator(input_file, readsize=32768):
679
b = input_file.read(readsize)
125
if hasattr(f, 'tell'):
686
"""Calculate the hexdigest of an open file.
688
The file cursor should be already at the start.
700
def size_sha_file(f):
701
"""Calculate the size and hexdigest of an open file.
703
The file cursor should be already at the start and
704
the caller is responsible for closing the file afterwards.
715
return size, s.hexdigest()
718
def sha_file_by_name(fname):
719
"""Calculate the SHA1 of a file by reading the full text"""
721
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
724
b = os.read(f, 1<<16)
732
def sha_strings(strings, _factory=sha):
733
"""Return the sha-1 of concatenation of strings"""
735
map(s.update, strings)
739
def sha_string(f, _factory=sha):
740
return _factory(f).hexdigest()
145
743
def fingerprint_file(f):
151
return {'size': size,
152
'sha1': s.hexdigest()}
156
"""Return per-user configuration directory.
158
By default this is ~/.bzr.conf/
160
TODO: Global option --config-dir to override this.
162
return os.path.expanduser("~/.bzr.conf")
166
"""Calculate automatic user identification.
168
Returns (realname, email).
170
Only used when none is set in the environment or the id file.
172
This previously used the FQDN as the default domain, but that can
173
be very slow on machines where DNS is broken. So now we simply
178
# XXX: Any good way to get real user name on win32?
183
w = pwd.getpwuid(uid)
184
gecos = w.pw_gecos.decode(bzrlib.user_encoding)
185
username = w.pw_name.decode(bzrlib.user_encoding)
186
comma = gecos.find(',')
190
realname = gecos[:comma]
196
realname = username = getpass.getuser().decode(bzrlib.user_encoding)
198
return realname, (username + '@' + socket.gethostname())
202
"""Return the full user id from a file or environment variable.
204
TODO: Allow taking this from a file in the branch directory too
205
for per-branch ids."""
206
v = os.environ.get('BZREMAIL')
208
return v.decode(bzrlib.user_encoding)
211
return (open(os.path.join(config_dir(), "email"))
213
.decode(bzrlib.user_encoding)
216
if e.errno != errno.ENOENT:
219
v = os.environ.get('EMAIL')
221
return v.decode(bzrlib.user_encoding)
227
"""Return email-style username.
229
Something similar to 'Martin Pool <mbp@sourcefrog.net>'
231
TODO: Check it's reasonably well-formed.
237
name, email = _auto_user_id()
239
return '%s <%s>' % (name, email)
244
_EMAIL_RE = re.compile(r'[\w+.-]+@[\w+.-]+')
246
"""Return just the email component of a username."""
249
m = _EMAIL_RE.search(e)
251
bailout("%r doesn't seem to contain a reasonable email address" % e)
254
return _auto_user_id()[1]
745
return {'size': len(b),
746
'sha1': sha(b).hexdigest()}
258
749
def compare_files(a, b):
359
bailout("sorry, %r not allowed in path" % f)
964
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
360
965
elif (f == '.') or (f == ''):
367
assert isinstance(p, list)
369
if (f == '..') or (f == None) or (f == ''):
370
bailout("sorry, %r not allowed in path" % f)
371
return os.path.join(*p)
374
def appendpath(p1, p2):
378
return os.path.join(p1, p2)
381
def extern_command(cmd, ignore_errors = False):
382
mutter('external command: %s' % `cmd`)
384
if not ignore_errors:
385
bailout('command failed')
974
if (f == '..') or (f is None) or (f == ''):
975
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
979
def parent_directories(filename):
980
"""Return the list of parent directories, deepest first.
982
For example, parent_directories("a/b/c") -> ["a/b", "a"].
985
parts = splitpath(dirname(filename))
987
parents.append(joinpath(parts))
992
_extension_load_failures = []
995
def failed_to_load_extension(exception):
996
"""Handle failing to load a binary extension.
998
This should be called from the ImportError block guarding the attempt to
999
import the native extension. If this function returns, the pure-Python
1000
implementation should be loaded instead::
1003
>>> import bzrlib._fictional_extension_pyx
1004
>>> except ImportError, e:
1005
>>> bzrlib.osutils.failed_to_load_extension(e)
1006
>>> import bzrlib._fictional_extension_py
1008
# NB: This docstring is just an example, not a doctest, because doctest
1009
# currently can't cope with the use of lazy imports in this namespace --
1012
# This currently doesn't report the failure at the time it occurs, because
1013
# they tend to happen very early in startup when we can't check config
1014
# files etc, and also we want to report all failures but not spam the user
1016
exception_str = str(exception)
1017
if exception_str not in _extension_load_failures:
1018
trace.mutter("failed to load compiled extension: %s" % exception_str)
1019
_extension_load_failures.append(exception_str)
1022
def report_extension_load_failures():
1023
if not _extension_load_failures:
1025
if config.GlobalStack().get('ignore_missing_extensions'):
1027
# the warnings framework should by default show this only once
1028
from bzrlib.trace import warning
1030
"bzr: warning: some compiled extensions could not be loaded; "
1031
"see <https://answers.launchpad.net/bzr/+faq/703>")
1032
# we no longer show the specific missing extensions here, because it makes
1033
# the message too long and scary - see
1034
# https://bugs.launchpad.net/bzr/+bug/430529
1038
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1039
except ImportError, e:
1040
failed_to_load_extension(e)
1041
from bzrlib._chunks_to_lines_py import chunks_to_lines
1045
"""Split s into lines, but without removing the newline characters."""
1046
# Trivially convert a fulltext into a 'chunked' representation, and let
1047
# chunks_to_lines do the heavy lifting.
1048
if isinstance(s, str):
1049
# chunks_to_lines only supports 8-bit strings
1050
return chunks_to_lines([s])
1052
return _split_lines(s)
1055
def _split_lines(s):
1056
"""Split s into lines, but without removing the newline characters.
1058
This supports Unicode or plain string objects.
1060
lines = s.split('\n')
1061
result = [line + '\n' for line in lines[:-1]]
1063
result.append(lines[-1])
1067
def hardlinks_good():
1068
return sys.platform not in ('win32', 'cygwin', 'darwin')
1071
def link_or_copy(src, dest):
1072
"""Hardlink a file, or copy it if it can't be hardlinked."""
1073
if not hardlinks_good():
1074
shutil.copyfile(src, dest)
1078
except (OSError, IOError), e:
1079
if e.errno != errno.EXDEV:
1081
shutil.copyfile(src, dest)
1084
def delete_any(path):
1085
"""Delete a file, symlink or directory.
1087
Will delete even if readonly.
1090
_delete_file_or_dir(path)
1091
except (OSError, IOError), e:
1092
if e.errno in (errno.EPERM, errno.EACCES):
1093
# make writable and try again
1096
except (OSError, IOError):
1098
_delete_file_or_dir(path)
1103
def _delete_file_or_dir(path):
1104
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1105
# Forgiveness than Permission (EAFP) because:
1106
# - root can damage a solaris file system by using unlink,
1107
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1108
# EACCES, OSX: EPERM) when invoked on a directory.
1109
if isdir(path): # Takes care of symlinks
1116
if getattr(os, 'symlink', None) is not None:
1122
def has_hardlinks():
1123
if getattr(os, 'link', None) is not None:
1129
def host_os_dereferences_symlinks():
1130
return (has_symlinks()
1131
and sys.platform not in ('cygwin', 'win32'))
1134
def readlink(abspath):
1135
"""Return a string representing the path to which the symbolic link points.
1137
:param abspath: The link absolute unicode path.
1139
This his guaranteed to return the symbolic link in unicode in all python
1142
link = abspath.encode(_fs_enc)
1143
target = os.readlink(link)
1144
target = target.decode(_fs_enc)
1148
def contains_whitespace(s):
1149
"""True if there are any whitespace characters in s."""
1150
# string.whitespace can include '\xa0' in certain locales, because it is
1151
# considered "non-breaking-space" as part of ISO-8859-1. But it
1152
# 1) Isn't a breaking whitespace
1153
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1155
# 3) '\xa0' isn't unicode safe since it is >128.
1157
# This should *not* be a unicode set of characters in case the source
1158
# string is not a Unicode string. We can auto-up-cast the characters since
1159
# they are ascii, but we don't want to auto-up-cast the string in case it
1161
for ch in ' \t\n\r\v\f':
1168
def contains_linebreaks(s):
1169
"""True if there is any vertical whitespace in s."""
1177
def relpath(base, path):
1178
"""Return path relative to base, or raise PathNotChild exception.
1180
The path may be either an absolute path or a path relative to the
1181
current working directory.
1183
os.path.commonprefix (python2.4) has a bad bug that it works just
1184
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1185
avoids that problem.
1187
NOTE: `base` should not have a trailing slash otherwise you'll get
1188
PathNotChild exceptions regardless of `path`.
1191
if len(base) < MIN_ABS_PATHLENGTH:
1192
# must have space for e.g. a drive letter
1193
raise ValueError(gettext('%r is too short to calculate a relative path')
1201
if len(head) <= len(base) and head != base:
1202
raise errors.PathNotChild(rp, base)
1205
head, tail = split(head)
1210
return pathjoin(*reversed(s))
1215
def _cicp_canonical_relpath(base, path):
1216
"""Return the canonical path relative to base.
1218
Like relpath, but on case-insensitive-case-preserving file-systems, this
1219
will return the relpath as stored on the file-system rather than in the
1220
case specified in the input string, for all existing portions of the path.
1222
This will cause O(N) behaviour if called for every path in a tree; if you
1223
have a number of paths to convert, you should use canonical_relpaths().
1225
# TODO: it should be possible to optimize this for Windows by using the
1226
# win32 API FindFiles function to look for the specified name - but using
1227
# os.listdir() still gives us the correct, platform agnostic semantics in
1230
rel = relpath(base, path)
1231
# '.' will have been turned into ''
1235
abs_base = abspath(base)
1237
_listdir = os.listdir
1239
# use an explicit iterator so we can easily consume the rest on early exit.
1240
bit_iter = iter(rel.split('/'))
1241
for bit in bit_iter:
1244
next_entries = _listdir(current)
1245
except OSError: # enoent, eperm, etc
1246
# We can't find this in the filesystem, so just append the
1248
current = pathjoin(current, bit, *list(bit_iter))
1250
for look in next_entries:
1251
if lbit == look.lower():
1252
current = pathjoin(current, look)
1255
# got to the end, nothing matched, so we just return the
1256
# non-existing bits as they were specified (the filename may be
1257
# the target of a move, for example).
1258
current = pathjoin(current, bit, *list(bit_iter))
1260
return current[len(abs_base):].lstrip('/')
1262
# XXX - TODO - we need better detection/integration of case-insensitive
1263
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1264
# filesystems), for example, so could probably benefit from the same basic
1265
# support there. For now though, only Windows and OSX get that support, and
1266
# they get it for *all* file-systems!
1267
if sys.platform in ('win32', 'darwin'):
1268
canonical_relpath = _cicp_canonical_relpath
1270
canonical_relpath = relpath
1272
def canonical_relpaths(base, paths):
1273
"""Create an iterable to canonicalize a sequence of relative paths.
1275
The intent is for this implementation to use a cache, vastly speeding
1276
up multiple transformations in the same directory.
1278
# but for now, we haven't optimized...
1279
return [canonical_relpath(base, p) for p in paths]
1282
def decode_filename(filename):
1283
"""Decode the filename using the filesystem encoding
1285
If it is unicode, it is returned.
1286
Otherwise it is decoded from the the filesystem's encoding. If decoding
1287
fails, a errors.BadFilenameEncoding exception is raised.
1289
if type(filename) is unicode:
1292
return filename.decode(_fs_enc)
1293
except UnicodeDecodeError:
1294
raise errors.BadFilenameEncoding(filename, _fs_enc)
1297
def safe_unicode(unicode_or_utf8_string):
1298
"""Coerce unicode_or_utf8_string into unicode.
1300
If it is unicode, it is returned.
1301
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1302
wrapped in a BzrBadParameterNotUnicode exception.
1304
if isinstance(unicode_or_utf8_string, unicode):
1305
return unicode_or_utf8_string
1307
return unicode_or_utf8_string.decode('utf8')
1308
except UnicodeDecodeError:
1309
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1312
def safe_utf8(unicode_or_utf8_string):
1313
"""Coerce unicode_or_utf8_string to a utf8 string.
1315
If it is a str, it is returned.
1316
If it is Unicode, it is encoded into a utf-8 string.
1318
if isinstance(unicode_or_utf8_string, str):
1319
# TODO: jam 20070209 This is overkill, and probably has an impact on
1320
# performance if we are dealing with lots of apis that want a
1323
# Make sure it is a valid utf-8 string
1324
unicode_or_utf8_string.decode('utf-8')
1325
except UnicodeDecodeError:
1326
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1327
return unicode_or_utf8_string
1328
return unicode_or_utf8_string.encode('utf-8')
1331
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1332
' Revision id generators should be creating utf8'
1336
def safe_revision_id(unicode_or_utf8_string, warn=True):
1337
"""Revision ids should now be utf8, but at one point they were unicode.
1339
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1341
:param warn: Functions that are sanitizing user data can set warn=False
1342
:return: None or a utf8 revision id.
1344
if (unicode_or_utf8_string is None
1345
or unicode_or_utf8_string.__class__ == str):
1346
return unicode_or_utf8_string
1348
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1350
return cache_utf8.encode(unicode_or_utf8_string)
1353
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1354
' generators should be creating utf8 file ids.')
1357
def safe_file_id(unicode_or_utf8_string, warn=True):
1358
"""File ids should now be utf8, but at one point they were unicode.
1360
This is the same as safe_utf8, except it uses the cached encode functions
1361
to save a little bit of performance.
1363
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1365
:param warn: Functions that are sanitizing user data can set warn=False
1366
:return: None or a utf8 file id.
1368
if (unicode_or_utf8_string is None
1369
or unicode_or_utf8_string.__class__ == str):
1370
return unicode_or_utf8_string
1372
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1374
return cache_utf8.encode(unicode_or_utf8_string)
1377
_platform_normalizes_filenames = False
1378
if sys.platform == 'darwin':
1379
_platform_normalizes_filenames = True
1382
def normalizes_filenames():
1383
"""Return True if this platform normalizes unicode filenames.
1387
return _platform_normalizes_filenames
1390
def _accessible_normalized_filename(path):
1391
"""Get the unicode normalized path, and if you can access the file.
1393
On platforms where the system normalizes filenames (Mac OSX),
1394
you can access a file by any path which will normalize correctly.
1395
On platforms where the system does not normalize filenames
1396
(everything else), you have to access a file by its exact path.
1398
Internally, bzr only supports NFC normalization, since that is
1399
the standard for XML documents.
1401
So return the normalized path, and a flag indicating if the file
1402
can be accessed by that path.
1405
return unicodedata.normalize('NFC', unicode(path)), True
1408
def _inaccessible_normalized_filename(path):
1409
__doc__ = _accessible_normalized_filename.__doc__
1411
normalized = unicodedata.normalize('NFC', unicode(path))
1412
return normalized, normalized == path
1415
if _platform_normalizes_filenames:
1416
normalized_filename = _accessible_normalized_filename
1418
normalized_filename = _inaccessible_normalized_filename
1421
def set_signal_handler(signum, handler, restart_syscall=True):
1422
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1423
on platforms that support that.
1425
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1426
automatically restart (by calling `signal.siginterrupt(signum,
1427
False)`). May be ignored if the feature is not available on this
1428
platform or Python version.
1432
siginterrupt = signal.siginterrupt
1434
# This python implementation doesn't provide signal support, hence no
1437
except AttributeError:
1438
# siginterrupt doesn't exist on this platform, or for this version
1440
siginterrupt = lambda signum, flag: None
1442
def sig_handler(*args):
1443
# Python resets the siginterrupt flag when a signal is
1444
# received. <http://bugs.python.org/issue8354>
1445
# As a workaround for some cases, set it back the way we want it.
1446
siginterrupt(signum, False)
1447
# Now run the handler function passed to set_signal_handler.
1450
sig_handler = handler
1451
old_handler = signal.signal(signum, sig_handler)
1453
siginterrupt(signum, False)
1457
default_terminal_width = 80
1458
"""The default terminal width for ttys.
1460
This is defined so that higher levels can share a common fallback value when
1461
terminal_width() returns None.
1464
# Keep some state so that terminal_width can detect if _terminal_size has
1465
# returned a different size since the process started. See docstring and
1466
# comments of terminal_width for details.
1467
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1468
_terminal_size_state = 'no_data'
1469
_first_terminal_size = None
1471
def terminal_width():
1472
"""Return terminal width.
1474
None is returned if the width can't established precisely.
1477
- if BZR_COLUMNS is set, returns its value
1478
- if there is no controlling terminal, returns None
1479
- query the OS, if the queried size has changed since the last query,
1481
- if COLUMNS is set, returns its value,
1482
- if the OS has a value (even though it's never changed), return its value.
1484
From there, we need to query the OS to get the size of the controlling
1487
On Unices we query the OS by:
1488
- get termios.TIOCGWINSZ
1489
- if an error occurs or a negative value is obtained, returns None
1491
On Windows we query the OS by:
1492
- win32utils.get_console_size() decides,
1493
- returns None on error (provided default value)
1495
# Note to implementors: if changing the rules for determining the width,
1496
# make sure you've considered the behaviour in these cases:
1497
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1498
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1500
# - (add more interesting cases here, if you find any)
1501
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1502
# but we don't want to register a signal handler because it is impossible
1503
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1504
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1505
# time so we can notice if the reported size has changed, which should have
1508
# If BZR_COLUMNS is set, take it, user is always right
1509
# Except if they specified 0 in which case, impose no limit here
1511
width = int(os.environ['BZR_COLUMNS'])
1512
except (KeyError, ValueError):
1514
if width is not None:
1520
isatty = getattr(sys.stdout, 'isatty', None)
1521
if isatty is None or not isatty():
1522
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1526
width, height = os_size = _terminal_size(None, None)
1527
global _first_terminal_size, _terminal_size_state
1528
if _terminal_size_state == 'no_data':
1529
_first_terminal_size = os_size
1530
_terminal_size_state = 'unchanged'
1531
elif (_terminal_size_state == 'unchanged' and
1532
_first_terminal_size != os_size):
1533
_terminal_size_state = 'changed'
1535
# If the OS claims to know how wide the terminal is, and this value has
1536
# ever changed, use that.
1537
if _terminal_size_state == 'changed':
1538
if width is not None and width > 0:
1541
# If COLUMNS is set, use it.
1543
return int(os.environ['COLUMNS'])
1544
except (KeyError, ValueError):
1547
# Finally, use an unchanged size from the OS, if we have one.
1548
if _terminal_size_state == 'unchanged':
1549
if width is not None and width > 0:
1552
# The width could not be determined.
1556
def _win32_terminal_size(width, height):
1557
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1558
return width, height
1561
def _ioctl_terminal_size(width, height):
1563
import struct, fcntl, termios
1564
s = struct.pack('HHHH', 0, 0, 0, 0)
1565
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1566
height, width = struct.unpack('HHHH', x)[0:2]
1567
except (IOError, AttributeError):
1569
return width, height
1571
_terminal_size = None
1572
"""Returns the terminal size as (width, height).
1574
:param width: Default value for width.
1575
:param height: Default value for height.
1577
This is defined specifically for each OS and query the size of the controlling
1578
terminal. If any error occurs, the provided default values should be returned.
1580
if sys.platform == 'win32':
1581
_terminal_size = _win32_terminal_size
1583
_terminal_size = _ioctl_terminal_size
1586
def supports_executable():
1587
return sys.platform != "win32"
1590
def supports_posix_readonly():
1591
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1593
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1594
directory controls creation/deletion, etc.
1596
And under win32, readonly means that the directory itself cannot be
1597
deleted. The contents of a readonly directory can be changed, unlike POSIX
1598
where files in readonly directories cannot be added, deleted or renamed.
1600
return sys.platform != "win32"
1603
def set_or_unset_env(env_variable, value):
1604
"""Modify the environment, setting or removing the env_variable.
1606
:param env_variable: The environment variable in question
1607
:param value: The value to set the environment to. If None, then
1608
the variable will be removed.
1609
:return: The original value of the environment variable.
1611
orig_val = os.environ.get(env_variable)
1613
if orig_val is not None:
1614
del os.environ[env_variable]
1616
if isinstance(value, unicode):
1617
value = value.encode(get_user_encoding())
1618
os.environ[env_variable] = value
1622
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1625
def check_legal_path(path):
1626
"""Check whether the supplied path is legal.
1627
This is only required on Windows, so we don't test on other platforms
1630
if sys.platform != "win32":
1632
if _validWin32PathRE.match(path) is None:
1633
raise errors.IllegalPath(path)
1636
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1638
def _is_error_enotdir(e):
1639
"""Check if this exception represents ENOTDIR.
1641
Unfortunately, python is very inconsistent about the exception
1642
here. The cases are:
1643
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1644
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1645
which is the windows error code.
1646
3) Windows, Python2.5 uses errno == EINVAL and
1647
winerror == ERROR_DIRECTORY
1649
:param e: An Exception object (expected to be OSError with an errno
1650
attribute, but we should be able to cope with anything)
1651
:return: True if this represents an ENOTDIR error. False otherwise.
1653
en = getattr(e, 'errno', None)
1654
if (en == errno.ENOTDIR
1655
or (sys.platform == 'win32'
1656
and (en == _WIN32_ERROR_DIRECTORY
1657
or (en == errno.EINVAL
1658
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1664
def walkdirs(top, prefix=""):
1665
"""Yield data about all the directories in a tree.
1667
This yields all the data about the contents of a directory at a time.
1668
After each directory has been yielded, if the caller has mutated the list
1669
to exclude some directories, they are then not descended into.
1671
The data yielded is of the form:
1672
((directory-relpath, directory-path-from-top),
1673
[(relpath, basename, kind, lstat, path-from-top), ...]),
1674
- directory-relpath is the relative path of the directory being returned
1675
with respect to top. prefix is prepended to this.
1676
- directory-path-from-root is the path including top for this directory.
1677
It is suitable for use with os functions.
1678
- relpath is the relative path within the subtree being walked.
1679
- basename is the basename of the path
1680
- kind is the kind of the file now. If unknown then the file is not
1681
present within the tree - but it may be recorded as versioned. See
1683
- lstat is the stat data *if* the file was statted.
1684
- planned, not implemented:
1685
path_from_tree_root is the path from the root of the tree.
1687
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1688
allows one to walk a subtree but get paths that are relative to a tree
1690
:return: an iterator over the dirs.
1692
#TODO there is a bit of a smell where the results of the directory-
1693
# summary in this, and the path from the root, may not agree
1694
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1695
# potentially confusing output. We should make this more robust - but
1696
# not at a speed cost. RBC 20060731
1698
_directory = _directory_kind
1699
_listdir = os.listdir
1700
_kind_from_mode = file_kind_from_stat_mode
1701
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1703
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1704
relroot, _, _, _, top = pending.pop()
1706
relprefix = relroot + u'/'
1709
top_slash = top + u'/'
1712
append = dirblock.append
1714
names = sorted(map(decode_filename, _listdir(top)))
1716
if not _is_error_enotdir(e):
1720
abspath = top_slash + name
1721
statvalue = _lstat(abspath)
1722
kind = _kind_from_mode(statvalue.st_mode)
1723
append((relprefix + name, name, kind, statvalue, abspath))
1724
yield (relroot, top), dirblock
1726
# push the user specified dirs from dirblock
1727
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1730
class DirReader(object):
1731
"""An interface for reading directories."""
1733
def top_prefix_to_starting_dir(self, top, prefix=""):
1734
"""Converts top and prefix to a starting dir entry
1736
:param top: A utf8 path
1737
:param prefix: An optional utf8 path to prefix output relative paths
1739
:return: A tuple starting with prefix, and ending with the native
1742
raise NotImplementedError(self.top_prefix_to_starting_dir)
1744
def read_dir(self, prefix, top):
1745
"""Read a specific dir.
1747
:param prefix: A utf8 prefix to be preprended to the path basenames.
1748
:param top: A natively encoded path to read.
1749
:return: A list of the directories contents. Each item contains:
1750
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1752
raise NotImplementedError(self.read_dir)
1755
_selected_dir_reader = None
1758
def _walkdirs_utf8(top, prefix=""):
1759
"""Yield data about all the directories in a tree.
1761
This yields the same information as walkdirs() only each entry is yielded
1762
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1763
are returned as exact byte-strings.
1765
:return: yields a tuple of (dir_info, [file_info])
1766
dir_info is (utf8_relpath, path-from-top)
1767
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1768
if top is an absolute path, path-from-top is also an absolute path.
1769
path-from-top might be unicode or utf8, but it is the correct path to
1770
pass to os functions to affect the file in question. (such as os.lstat)
1772
global _selected_dir_reader
1773
if _selected_dir_reader is None:
1774
fs_encoding = _fs_enc.upper()
1775
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1776
# Win98 doesn't have unicode apis like FindFirstFileW
1777
# TODO: We possibly could support Win98 by falling back to the
1778
# original FindFirstFile, and using TCHAR instead of WCHAR,
1779
# but that gets a bit tricky, and requires custom compiling
1782
from bzrlib._walkdirs_win32 import Win32ReadDir
1783
_selected_dir_reader = Win32ReadDir()
1786
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1787
# ANSI_X3.4-1968 is a form of ASCII
1789
from bzrlib._readdir_pyx import UTF8DirReader
1790
_selected_dir_reader = UTF8DirReader()
1791
except ImportError, e:
1792
failed_to_load_extension(e)
1795
if _selected_dir_reader is None:
1796
# Fallback to the python version
1797
_selected_dir_reader = UnicodeDirReader()
1799
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1800
# But we don't actually uses 1-3 in pending, so set them to None
1801
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1802
read_dir = _selected_dir_reader.read_dir
1803
_directory = _directory_kind
1805
relroot, _, _, _, top = pending[-1].pop()
1808
dirblock = sorted(read_dir(relroot, top))
1809
yield (relroot, top), dirblock
1810
# push the user specified dirs from dirblock
1811
next = [d for d in reversed(dirblock) if d[2] == _directory]
1813
pending.append(next)
1816
class UnicodeDirReader(DirReader):
1817
"""A dir reader for non-utf8 file systems, which transcodes."""
1819
__slots__ = ['_utf8_encode']
1822
self._utf8_encode = codecs.getencoder('utf8')
1824
def top_prefix_to_starting_dir(self, top, prefix=""):
1825
"""See DirReader.top_prefix_to_starting_dir."""
1826
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1828
def read_dir(self, prefix, top):
1829
"""Read a single directory from a non-utf8 file system.
1831
top, and the abspath element in the output are unicode, all other paths
1832
are utf8. Local disk IO is done via unicode calls to listdir etc.
1834
This is currently the fallback code path when the filesystem encoding is
1835
not UTF-8. It may be better to implement an alternative so that we can
1836
safely handle paths that are not properly decodable in the current
1839
See DirReader.read_dir for details.
1841
_utf8_encode = self._utf8_encode
1843
_listdir = os.listdir
1844
_kind_from_mode = file_kind_from_stat_mode
1847
relprefix = prefix + '/'
1850
top_slash = top + u'/'
1853
append = dirblock.append
1854
for name in sorted(_listdir(top)):
1856
name_utf8 = _utf8_encode(name)[0]
1857
except UnicodeDecodeError:
1858
raise errors.BadFilenameEncoding(
1859
_utf8_encode(relprefix)[0] + name, _fs_enc)
1860
abspath = top_slash + name
1861
statvalue = _lstat(abspath)
1862
kind = _kind_from_mode(statvalue.st_mode)
1863
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1867
def copy_tree(from_path, to_path, handlers={}):
1868
"""Copy all of the entries in from_path into to_path.
1870
:param from_path: The base directory to copy.
1871
:param to_path: The target directory. If it does not exist, it will
1873
:param handlers: A dictionary of functions, which takes a source and
1874
destinations for files, directories, etc.
1875
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1876
'file', 'directory', and 'symlink' should always exist.
1877
If they are missing, they will be replaced with 'os.mkdir()',
1878
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1880
# Now, just copy the existing cached tree to the new location
1881
# We use a cheap trick here.
1882
# Absolute paths are prefixed with the first parameter
1883
# relative paths are prefixed with the second.
1884
# So we can get both the source and target returned
1885
# without any extra work.
1887
def copy_dir(source, dest):
1890
def copy_link(source, dest):
1891
"""Copy the contents of a symlink"""
1892
link_to = os.readlink(source)
1893
os.symlink(link_to, dest)
1895
real_handlers = {'file':shutil.copy2,
1896
'symlink':copy_link,
1897
'directory':copy_dir,
1899
real_handlers.update(handlers)
1901
if not os.path.exists(to_path):
1902
real_handlers['directory'](from_path, to_path)
1904
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1905
for relpath, name, kind, st, abspath in entries:
1906
real_handlers[kind](abspath, relpath)
1909
def copy_ownership_from_path(dst, src=None):
1910
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1912
If src is None, the containing directory is used as source. If chown
1913
fails, the error is ignored and a warning is printed.
1915
chown = getattr(os, 'chown', None)
1920
src = os.path.dirname(dst)
1926
chown(dst, s.st_uid, s.st_gid)
1929
'Unable to copy ownership from "%s" to "%s". '
1930
'You may want to set it manually.', src, dst)
1931
trace.log_exception_quietly()
1934
def path_prefix_key(path):
1935
"""Generate a prefix-order path key for path.
1937
This can be used to sort paths in the same way that walkdirs does.
1939
return (dirname(path) , path)
1942
def compare_paths_prefix_order(path_a, path_b):
1943
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1944
key_a = path_prefix_key(path_a)
1945
key_b = path_prefix_key(path_b)
1946
return cmp(key_a, key_b)
1949
_cached_user_encoding = None
1952
def get_user_encoding(use_cache=True):
1953
"""Find out what the preferred user encoding is.
1955
This is generally the encoding that is used for command line parameters
1956
and file contents. This may be different from the terminal encoding
1957
or the filesystem encoding.
1959
:param use_cache: Enable cache for detected encoding.
1960
(This parameter is turned on by default,
1961
and required only for selftesting)
1963
:return: A string defining the preferred user encoding
1965
global _cached_user_encoding
1966
if _cached_user_encoding is not None and use_cache:
1967
return _cached_user_encoding
1969
if sys.platform == 'darwin':
1970
# python locale.getpreferredencoding() always return
1971
# 'mac-roman' on darwin. That's a lie.
1972
sys.platform = 'posix'
1974
if os.environ.get('LANG', None) is None:
1975
# If LANG is not set, we end up with 'ascii', which is bad
1976
# ('mac-roman' is more than ascii), so we set a default which
1977
# will give us UTF-8 (which appears to work in all cases on
1978
# OSX). Users are still free to override LANG of course, as
1979
# long as it give us something meaningful. This work-around
1980
# *may* not be needed with python 3k and/or OSX 10.5, but will
1981
# work with them too -- vila 20080908
1982
os.environ['LANG'] = 'en_US.UTF-8'
1985
sys.platform = 'darwin'
1990
user_encoding = locale.getpreferredencoding()
1991
except locale.Error, e:
1992
sys.stderr.write('bzr: warning: %s\n'
1993
' Could not determine what text encoding to use.\n'
1994
' This error usually means your Python interpreter\n'
1995
' doesn\'t support the locale set by $LANG (%s)\n'
1996
" Continuing with ascii encoding.\n"
1997
% (e, os.environ.get('LANG')))
1998
user_encoding = 'ascii'
2000
# Windows returns 'cp0' to indicate there is no code page. So we'll just
2001
# treat that as ASCII, and not support printing unicode characters to the
2004
# For python scripts run under vim, we get '', so also treat that as ASCII
2005
if user_encoding in (None, 'cp0', ''):
2006
user_encoding = 'ascii'
2010
codecs.lookup(user_encoding)
2012
sys.stderr.write('bzr: warning:'
2013
' unknown encoding %s.'
2014
' Continuing with ascii encoding.\n'
2017
user_encoding = 'ascii'
2020
_cached_user_encoding = user_encoding
2022
return user_encoding
2025
def get_diff_header_encoding():
2026
return get_terminal_encoding()
2029
_message_encoding = None
2032
def get_message_encoding():
2033
"""Return the encoding used for messages
2035
While the message encoding is a general setting it should usually only be
2036
needed for decoding system error strings such as from OSError instances.
2038
global _message_encoding
2039
if _message_encoding is None:
2040
if os.name == "posix":
2042
# This is a process-global setting that can change, but should in
2043
# general just get set once at process startup then be constant.
2044
_message_encoding = locale.getlocale(locale.LC_MESSAGES)[1]
2046
# On windows want the result of GetACP() which this boils down to.
2047
_message_encoding = get_user_encoding()
2048
return _message_encoding or "ascii"
2051
def get_host_name():
2052
"""Return the current unicode host name.
2054
This is meant to be used in place of socket.gethostname() because that
2055
behaves inconsistently on different platforms.
2057
if sys.platform == "win32":
2059
return win32utils.get_host_name()
2062
return socket.gethostname().decode(get_user_encoding())
2065
# We must not read/write any more than 64k at a time from/to a socket so we
2066
# don't risk "no buffer space available" errors on some platforms. Windows in
2067
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2069
MAX_SOCKET_CHUNK = 64 * 1024
2071
_end_of_stream_errors = [errno.ECONNRESET]
2072
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2073
_eno = getattr(errno, _eno, None)
2074
if _eno is not None:
2075
_end_of_stream_errors.append(_eno)
2079
def read_bytes_from_socket(sock, report_activity=None,
2080
max_read_size=MAX_SOCKET_CHUNK):
2081
"""Read up to max_read_size of bytes from sock and notify of progress.
2083
Translates "Connection reset by peer" into file-like EOF (return an
2084
empty string rather than raise an error), and repeats the recv if
2085
interrupted by a signal.
2089
bytes = sock.recv(max_read_size)
2090
except socket.error, e:
2092
if eno in _end_of_stream_errors:
2093
# The connection was closed by the other side. Callers expect
2094
# an empty string to signal end-of-stream.
2096
elif eno == errno.EINTR:
2097
# Retry the interrupted recv.
2101
if report_activity is not None:
2102
report_activity(len(bytes), 'read')
2106
def recv_all(socket, count):
2107
"""Receive an exact number of bytes.
2109
Regular Socket.recv() may return less than the requested number of bytes,
2110
depending on what's in the OS buffer. MSG_WAITALL is not available
2111
on all platforms, but this should work everywhere. This will return
2112
less than the requested amount if the remote end closes.
2114
This isn't optimized and is intended mostly for use in testing.
2117
while len(b) < count:
2118
new = read_bytes_from_socket(socket, None, count - len(b))
2125
def send_all(sock, bytes, report_activity=None):
2126
"""Send all bytes on a socket.
2128
Breaks large blocks in smaller chunks to avoid buffering limitations on
2129
some platforms, and catches EINTR which may be thrown if the send is
2130
interrupted by a signal.
2132
This is preferred to socket.sendall(), because it avoids portability bugs
2133
and provides activity reporting.
2135
:param report_activity: Call this as bytes are read, see
2136
Transport._report_activity
2139
byte_count = len(bytes)
2140
while sent_total < byte_count:
2142
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2143
except socket.error, e:
2144
if e.args[0] != errno.EINTR:
2148
report_activity(sent, 'write')
2151
def connect_socket(address):
2152
# Slight variation of the socket.create_connection() function (provided by
2153
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2154
# provide it for previous python versions. Also, we don't use the timeout
2155
# parameter (provided by the python implementation) so we don't implement
2157
err = socket.error('getaddrinfo returns an empty list')
2158
host, port = address
2159
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2160
af, socktype, proto, canonname, sa = res
2163
sock = socket.socket(af, socktype, proto)
2167
except socket.error, err:
2168
# 'err' is now the most recent error
2169
if sock is not None:
2174
def dereference_path(path):
2175
"""Determine the real path to a file.
2177
All parent elements are dereferenced. But the file itself is not
2179
:param path: The original path. May be absolute or relative.
2180
:return: the real path *to* the file
2182
parent, base = os.path.split(path)
2183
# The pathjoin for '.' is a workaround for Python bug #1213894.
2184
# (initial path components aren't dereferenced)
2185
return pathjoin(realpath(pathjoin('.', parent)), base)
2188
def supports_mapi():
2189
"""Return True if we can use MAPI to launch a mail client."""
2190
return sys.platform == "win32"
2193
def resource_string(package, resource_name):
2194
"""Load a resource from a package and return it as a string.
2196
Note: Only packages that start with bzrlib are currently supported.
2198
This is designed to be a lightweight implementation of resource
2199
loading in a way which is API compatible with the same API from
2201
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2202
If and when pkg_resources becomes a standard library, this routine
2205
# Check package name is within bzrlib
2206
if package == "bzrlib":
2207
resource_relpath = resource_name
2208
elif package.startswith("bzrlib."):
2209
package = package[len("bzrlib."):].replace('.', os.sep)
2210
resource_relpath = pathjoin(package, resource_name)
2212
raise errors.BzrError('resource package %s not in bzrlib' % package)
2214
# Map the resource to a file and read its contents
2215
base = dirname(bzrlib.__file__)
2216
if getattr(sys, 'frozen', None): # bzr.exe
2217
base = abspath(pathjoin(base, '..', '..'))
2218
f = file(pathjoin(base, resource_relpath), "rU")
2224
def file_kind_from_stat_mode_thunk(mode):
2225
global file_kind_from_stat_mode
2226
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2228
from bzrlib._readdir_pyx import UTF8DirReader
2229
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2230
except ImportError, e:
2231
# This is one time where we won't warn that an extension failed to
2232
# load. The extension is never available on Windows anyway.
2233
from bzrlib._readdir_py import (
2234
_kind_from_mode as file_kind_from_stat_mode
2236
return file_kind_from_stat_mode(mode)
2237
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2239
def file_stat(f, _lstat=os.lstat):
2244
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2245
raise errors.NoSuchFile(f)
2248
def file_kind(f, _lstat=os.lstat):
2249
stat_value = file_stat(f, _lstat)
2250
return file_kind_from_stat_mode(stat_value.st_mode)
2252
def until_no_eintr(f, *a, **kw):
2253
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2255
WARNING: you must be certain that it is safe to retry the call repeatedly
2256
if EINTR does occur. This is typically only true for low-level operations
2257
like os.read. If in any doubt, don't use this.
2259
Keep in mind that this is not a complete solution to EINTR. There is
2260
probably code in the Python standard library and other dependencies that
2261
may encounter EINTR if a signal arrives (and there is signal handler for
2262
that signal). So this function can reduce the impact for IO that bzrlib
2263
directly controls, but it is not a complete solution.
2265
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2269
except (IOError, OSError), e:
2270
if e.errno == errno.EINTR:
2275
@deprecated_function(deprecated_in((2, 2, 0)))
2276
def re_compile_checked(re_string, flags=0, where=""):
2277
"""Return a compiled re, or raise a sensible error.
2279
This should only be used when compiling user-supplied REs.
2281
:param re_string: Text form of regular expression.
2282
:param flags: eg re.IGNORECASE
2283
:param where: Message explaining to the user the context where
2284
it occurred, eg 'log search filter'.
2286
# from https://bugs.launchpad.net/bzr/+bug/251352
2288
re_obj = re.compile(re_string, flags)
2291
except errors.InvalidPattern, e:
2293
where = ' in ' + where
2294
# despite the name 'error' is a type
2295
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2299
if sys.platform == "win32":
2302
return msvcrt.getch()
2307
fd = sys.stdin.fileno()
2308
settings = termios.tcgetattr(fd)
2311
ch = sys.stdin.read(1)
2313
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2316
if sys.platform.startswith('linux'):
2317
def _local_concurrency():
2319
return os.sysconf('SC_NPROCESSORS_ONLN')
2320
except (ValueError, OSError, AttributeError):
2322
elif sys.platform == 'darwin':
2323
def _local_concurrency():
2324
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2325
stdout=subprocess.PIPE).communicate()[0]
2326
elif "bsd" in sys.platform:
2327
def _local_concurrency():
2328
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2329
stdout=subprocess.PIPE).communicate()[0]
2330
elif sys.platform == 'sunos5':
2331
def _local_concurrency():
2332
return subprocess.Popen(['psrinfo', '-p',],
2333
stdout=subprocess.PIPE).communicate()[0]
2334
elif sys.platform == "win32":
2335
def _local_concurrency():
2336
# This appears to return the number of cores.
2337
return os.environ.get('NUMBER_OF_PROCESSORS')
2339
def _local_concurrency():
2344
_cached_local_concurrency = None
2346
def local_concurrency(use_cache=True):
2347
"""Return how many processes can be run concurrently.
2349
Rely on platform specific implementations and default to 1 (one) if
2350
anything goes wrong.
2352
global _cached_local_concurrency
2354
if _cached_local_concurrency is not None and use_cache:
2355
return _cached_local_concurrency
2357
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2358
if concurrency is None:
2360
import multiprocessing
2361
concurrency = multiprocessing.cpu_count()
2362
except (ImportError, NotImplementedError):
2363
# multiprocessing is only available on Python >= 2.6
2364
# and multiprocessing.cpu_count() isn't implemented on all
2367
concurrency = _local_concurrency()
2368
except (OSError, IOError):
2371
concurrency = int(concurrency)
2372
except (TypeError, ValueError):
2375
_cached_concurrency = concurrency
2379
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2380
"""A stream writer that doesn't decode str arguments."""
2382
def __init__(self, encode, stream, errors='strict'):
2383
codecs.StreamWriter.__init__(self, stream, errors)
2384
self.encode = encode
2386
def write(self, object):
2387
if type(object) is str:
2388
self.stream.write(object)
2390
data, _ = self.encode(object, self.errors)
2391
self.stream.write(data)
2393
if sys.platform == 'win32':
2394
def open_file(filename, mode='r', bufsize=-1):
2395
"""This function is used to override the ``open`` builtin.
2397
But it uses O_NOINHERIT flag so the file handle is not inherited by
2398
child processes. Deleting or renaming a closed file opened with this
2399
function is not blocking child processes.
2401
writing = 'w' in mode
2402
appending = 'a' in mode
2403
updating = '+' in mode
2404
binary = 'b' in mode
2407
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2408
# for flags for each modes.
2418
flags |= os.O_WRONLY
2419
flags |= os.O_CREAT | os.O_TRUNC
2424
flags |= os.O_WRONLY
2425
flags |= os.O_CREAT | os.O_APPEND
2430
flags |= os.O_RDONLY
2432
return os.fdopen(os.open(filename, flags), mode, bufsize)
2437
def getuser_unicode():
2438
"""Return the username as unicode.
2441
user_encoding = get_user_encoding()
2442
username = getpass.getuser().decode(user_encoding)
2443
except UnicodeDecodeError:
2444
raise errors.BzrError("Can't decode username as %s." % \
2446
except ImportError, e:
2447
if sys.platform != 'win32':
2449
if str(e) != 'No module named pwd':
2451
# https://bugs.launchpad.net/bzr/+bug/660174
2452
# getpass.getuser() is unable to return username on Windows
2453
# if there is no USERNAME environment variable set.
2454
# That could be true if bzr is running as a service,
2455
# e.g. running `bzr serve` as a service on Windows.
2456
# We should not fail with traceback in this case.
2457
username = u'UNKNOWN'
2461
def available_backup_name(base, exists):
2462
"""Find a non-existing backup file name.
2464
This will *not* create anything, this only return a 'free' entry. This
2465
should be used for checking names in a directory below a locked
2466
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2467
Leap) and generally discouraged.
2469
:param base: The base name.
2471
:param exists: A callable returning True if the path parameter exists.
2474
name = "%s.~%d~" % (base, counter)
2477
name = "%s.~%d~" % (base, counter)
2481
def set_fd_cloexec(fd):
2482
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2483
support for this is not available.
2487
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2488
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2489
except (ImportError, AttributeError):
2490
# Either the fcntl module or specific constants are not present
2494
def find_executable_on_path(name):
2495
"""Finds an executable on the PATH.
2497
On Windows, this will try to append each extension in the PATHEXT
2498
environment variable to the name, if it cannot be found with the name
2501
:param name: The base name of the executable.
2502
:return: The path to the executable found or None.
2504
path = os.environ.get('PATH')
2507
path = path.split(os.pathsep)
2508
if sys.platform == 'win32':
2509
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2510
exts = [ext.lower() for ext in exts]
2511
base, ext = os.path.splitext(name)
2513
if ext.lower() not in exts:
2521
f = os.path.join(d, name) + ext
2522
if os.access(f, os.X_OK):
2527
def _posix_is_local_pid_dead(pid):
2528
"""True if pid doesn't correspond to live process on this machine"""
2530
# Special meaning of unix kill: just check if it's there.
2533
if e.errno == errno.ESRCH:
2534
# On this machine, and really not found: as sure as we can be
2537
elif e.errno == errno.EPERM:
2538
# exists, though not ours
2541
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2542
# Don't really know.
2545
# Exists and our process: not dead.
2548
if sys.platform == "win32":
2549
is_local_pid_dead = win32utils.is_local_pid_dead
2551
is_local_pid_dead = _posix_is_local_pid_dead
2554
def fdatasync(fileno):
2555
"""Flush file contents to disk if possible.
2557
:param fileno: Integer OS file handle.
2558
:raises TransportNotPossible: If flushing to disk is not possible.
2560
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2565
def ensure_empty_directory_exists(path, exception_class):
2566
"""Make sure a local directory exists and is empty.
2568
If it does not exist, it is created. If it exists and is not empty, an
2569
instance of exception_class is raised.
2574
if e.errno != errno.EEXIST:
2576
if os.listdir(path) != []:
2577
raise exception_class(path)
2580
def is_environment_error(evalue):
2581
"""True if exception instance is due to a process environment issue
2583
This includes OSError and IOError, but also other errors that come from
2584
the operating system or core libraries but are not subclasses of those.
2586
if isinstance(evalue, (EnvironmentError, select.error)):
2588
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
added
removed
bzrlib/osutils.py
