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 bzrlib.errors import BzrError
23
from bzrlib.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
31
# We need to import both shutil and rmtree as we export the later on posix
32
# and need the former on windows
34
from shutil import rmtree
37
# We need to import both tempfile and mkdtemp as we export the later on posix
38
# and need the former on windows
40
from tempfile import mkdtemp
51
from bzrlib.symbol_versioning import (
56
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
58
if sys.version_info < (2, 5):
59
import md5 as _mod_md5
61
import sha as _mod_sha
71
from bzrlib import symbol_versioning
74
# Cross platform wall-clock time functionality with decent resolution.
75
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
76
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
77
# synchronized with ``time.time()``, this is only meant to be used to find
78
# delta times by subtracting from another call to this function.
79
timer_func = time.time
80
if sys.platform == 'win32':
81
timer_func = time.clock
83
# On win32, O_BINARY is used to indicate the file should
84
# be opened in binary mode, rather than text mode.
85
# On other platforms, O_BINARY doesn't exist, because
86
# they always open in binary mode, so it is okay to
87
# OR with 0 on those platforms.
88
# O_NOINHERIT and O_TEXT exists only on win32 too.
89
O_BINARY = getattr(os, 'O_BINARY', 0)
90
O_TEXT = getattr(os, 'O_TEXT', 0)
91
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
94
def get_unicode_argv():
96
user_encoding = get_user_encoding()
97
return [a.decode(user_encoding) for a in sys.argv[1:]]
98
except UnicodeDecodeError:
99
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
26
103
def make_readonly(filename):
27
104
"""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)
105
mod = os.lstat(filename).st_mode
106
if not stat.S_ISLNK(mod):
108
os.chmod(filename, mod)
34
111
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.,:/_~-])')
112
mod = os.lstat(filename).st_mode
113
if not stat.S_ISLNK(mod):
115
os.chmod(filename, mod)
118
def minimum_path_selection(paths):
119
"""Return the smallset subset of paths which are outside paths.
121
:param paths: A container (and hence not None) of paths.
122
:return: A set of paths sufficient to include everything in paths via
123
is_inside, drawn from the paths parameter.
129
return path.split('/')
130
sorted_paths = sorted(list(paths), key=sort_key)
132
search_paths = [sorted_paths[0]]
133
for path in sorted_paths[1:]:
134
if not is_inside(search_paths[-1], path):
135
# This path is unique, add it
136
search_paths.append(path)
138
return set(search_paths)
42
145
"""Return a quoted filename filename
44
147
This previously used backslash quoting, but that works poorly on
46
149
# TODO: I'm not really sure this is the best format either.x
151
if _QUOTE_RE is None:
152
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
47
154
if _QUOTE_RE.search(f):
48
155
return '"' + f + '"'
54
mode = os.lstat(f)[ST_MODE]
62
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
160
_directory_kind = 'directory'
163
"""Return the current umask"""
164
# Assume that people aren't messing with the umask while running
165
# XXX: This is not thread safe, but there is no way to get the
166
# umask without setting it
174
_directory_kind: "/",
176
'tree-reference': '+',
65
180
def kind_marker(kind):
182
return _kind_marker_map[kind]
184
# Slightly faster than using .get(, '') when the common case is that
68
elif kind == 'directory':
70
elif kind == 'symlink':
73
raise BzrError('invalid file kind %r' % kind)
78
"""Copy a file to a backup.
80
Backups are named in GNU-style, with a ~ suffix.
82
If the file is already a backup, it's not copied.
95
outf = file(bfn, 'wb')
101
def rename(path_from, path_to):
102
"""Basically the same as os.rename() just special for win32"""
103
if sys.platform == 'win32':
189
lexists = getattr(os.path, 'lexists', None)
193
stat = getattr(os, 'lstat', os.stat)
106
196
except OSError, e:
107
if e.errno != e.ENOENT:
109
os.rename(path_from, path_to)
197
if e.errno == errno.ENOENT:
200
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
203
def fancy_rename(old, new, rename_func, unlink_func):
204
"""A fancy rename, when you don't have atomic rename.
206
:param old: The old path, to rename from
207
:param new: The new path, to rename to
208
:param rename_func: The potentially non-atomic rename function
209
:param unlink_func: A way to delete the target file if the full rename
212
# sftp rename doesn't allow overwriting, so play tricks:
213
base = os.path.basename(new)
214
dirname = os.path.dirname(new)
215
# callers use different encodings for the paths so the following MUST
216
# respect that. We rely on python upcasting to unicode if new is unicode
217
# and keeping a str if not.
218
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
219
os.getpid(), rand_chars(10))
220
tmp_name = pathjoin(dirname, tmp_name)
222
# Rename the file out of the way, but keep track if it didn't exist
223
# We don't want to grab just any exception
224
# something like EACCES should prevent us from continuing
225
# The downside is that the rename_func has to throw an exception
226
# with an errno = ENOENT, or NoSuchFile
229
rename_func(new, tmp_name)
230
except (errors.NoSuchFile,), e:
233
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
234
# function raises an IOError with errno is None when a rename fails.
235
# This then gets caught here.
236
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
239
if (getattr(e, 'errno', None) is None
240
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
249
# This may throw an exception, in which case success will
251
rename_func(old, new)
253
except (IOError, OSError), e:
254
# source and target may be aliases of each other (e.g. on a
255
# case-insensitive filesystem), so we may have accidentally renamed
256
# source by when we tried to rename target
257
failure_exc = sys.exc_info()
258
if (file_existed and e.errno in (None, errno.ENOENT)
259
and old.lower() == new.lower()):
260
# source and target are the same file on a case-insensitive
261
# filesystem, so we don't generate an exception
265
# If the file used to exist, rename it back into place
266
# otherwise just delete it from the tmp location
268
unlink_func(tmp_name)
270
rename_func(tmp_name, new)
271
if failure_exc is not None:
272
raise failure_exc[0], failure_exc[1], failure_exc[2]
275
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
276
# choke on a Unicode string containing a relative path if
277
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
279
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
280
def _posix_abspath(path):
281
# jam 20060426 rather than encoding to fsencoding
282
# copy posixpath.abspath, but use os.getcwdu instead
283
if not posixpath.isabs(path):
284
path = posixpath.join(getcwd(), path)
285
return posixpath.normpath(path)
288
def _posix_realpath(path):
289
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
292
def _win32_fixdrive(path):
293
"""Force drive letters to be consistent.
295
win32 is inconsistent whether it returns lower or upper case
296
and even if it was consistent the user might type the other
297
so we force it to uppercase
298
running python.exe under cmd.exe return capital C:\\
299
running win32 python inside a cygwin shell returns lowercase c:\\
301
drive, path = ntpath.splitdrive(path)
302
return drive.upper() + path
305
def _win32_abspath(path):
306
# Real ntpath.abspath doesn't have a problem with a unicode cwd
307
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
310
def _win98_abspath(path):
311
"""Return the absolute version of a path.
312
Windows 98 safe implementation (python reimplementation
313
of Win32 API function GetFullPathNameW)
318
# \\HOST\path => //HOST/path
319
# //HOST/path => //HOST/path
320
# path => C:/cwd/path
323
# check for absolute path
324
drive = ntpath.splitdrive(path)[0]
325
if drive == '' and path[:2] not in('//','\\\\'):
327
# we cannot simply os.path.join cwd and path
328
# because os.path.join('C:','/path') produce '/path'
329
# and this is incorrect
330
if path[:1] in ('/','\\'):
331
cwd = ntpath.splitdrive(cwd)[0]
333
path = cwd + '\\' + path
334
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
337
def _win32_realpath(path):
338
# Real ntpath.realpath doesn't have a problem with a unicode cwd
339
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
342
def _win32_pathjoin(*args):
343
return ntpath.join(*args).replace('\\', '/')
346
def _win32_normpath(path):
347
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
351
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
354
def _win32_mkdtemp(*args, **kwargs):
355
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
358
def _win32_rename(old, new):
359
"""We expect to be able to atomically replace 'new' with old.
361
On win32, if new exists, it must be moved out of the way first,
365
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
367
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
368
# If we try to rename a non-existant file onto cwd, we get
369
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
370
# if the old path doesn't exist, sometimes we get EACCES
371
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
377
return unicodedata.normalize('NFC', os.getcwdu())
380
# Default is to just use the python builtins, but these can be rebound on
381
# particular platforms.
382
abspath = _posix_abspath
383
realpath = _posix_realpath
384
pathjoin = os.path.join
385
normpath = os.path.normpath
388
dirname = os.path.dirname
389
basename = os.path.basename
390
split = os.path.split
391
splitext = os.path.splitext
392
# These were already lazily imported into local scope
393
# mkdtemp = tempfile.mkdtemp
394
# rmtree = shutil.rmtree
396
MIN_ABS_PATHLENGTH = 1
399
if sys.platform == 'win32':
400
if win32utils.winver == 'Windows 98':
401
abspath = _win98_abspath
403
abspath = _win32_abspath
404
realpath = _win32_realpath
405
pathjoin = _win32_pathjoin
406
normpath = _win32_normpath
407
getcwd = _win32_getcwd
408
mkdtemp = _win32_mkdtemp
409
rename = _win32_rename
411
MIN_ABS_PATHLENGTH = 3
413
def _win32_delete_readonly(function, path, excinfo):
414
"""Error handler for shutil.rmtree function [for win32]
415
Helps to remove files and dirs marked as read-only.
417
exception = excinfo[1]
418
if function in (os.remove, os.rmdir) \
419
and isinstance(exception, OSError) \
420
and exception.errno == errno.EACCES:
426
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
427
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
428
return shutil.rmtree(path, ignore_errors, onerror)
430
f = win32utils.get_unicode_argv # special function or None
434
elif sys.platform == 'darwin':
438
def get_terminal_encoding(trace=False):
439
"""Find the best encoding for printing to the screen.
441
This attempts to check both sys.stdout and sys.stdin to see
442
what encoding they are in, and if that fails it falls back to
443
osutils.get_user_encoding().
444
The problem is that on Windows, locale.getpreferredencoding()
445
is not the same encoding as that used by the console:
446
http://mail.python.org/pipermail/python-list/2003-May/162357.html
448
On my standard US Windows XP, the preferred encoding is
449
cp1252, but the console is cp437
451
:param trace: If True trace the selected encoding via mutter().
453
from bzrlib.trace import mutter
454
output_encoding = getattr(sys.stdout, 'encoding', None)
455
if not output_encoding:
456
input_encoding = getattr(sys.stdin, 'encoding', None)
457
if not input_encoding:
458
output_encoding = get_user_encoding()
460
mutter('encoding stdout as osutils.get_user_encoding() %r',
463
output_encoding = input_encoding
465
mutter('encoding stdout as sys.stdin encoding %r',
469
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
470
if output_encoding == 'cp0':
471
# invalid encoding (cp0 means 'no codepage' on Windows)
472
output_encoding = get_user_encoding()
474
mutter('cp0 is invalid encoding.'
475
' encoding stdout as osutils.get_user_encoding() %r',
479
codecs.lookup(output_encoding)
481
sys.stderr.write('bzr: warning:'
482
' unknown terminal encoding %s.\n'
483
' Using encoding %s instead.\n'
484
% (output_encoding, get_user_encoding())
486
output_encoding = get_user_encoding()
488
return output_encoding
491
def normalizepath(f):
492
if getattr(os.path, 'realpath', None) is not None:
496
[p,e] = os.path.split(f)
497
if e == "" or e == "." or e == "..":
500
return pathjoin(F(p), e)
116
504
"""True if f is an accessible directory."""
118
return S_ISDIR(os.lstat(f)[ST_MODE])
506
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
125
512
"""True if f is a regular file."""
127
return S_ISREG(os.lstat(f)[ST_MODE])
514
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
519
"""True if f is a symlink."""
521
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
132
525
def is_inside(dir, fname):
133
526
"""True if fname is inside dir.
528
The parameters should typically be passed to osutils.normpath first, so
529
that . and .. and repeated slashes are eliminated, and the separators
530
are canonical for the platform.
532
The empty string as a dir name is taken as top-of-tree and matches
135
return os.path.commonprefix([dir, fname]) == dir
535
# XXX: Most callers of this can actually do something smarter by
536
# looking at the inventory
546
return fname.startswith(dir)
138
549
def is_inside_any(dir_list, fname):
139
550
"""True if fname is inside any of given dirs."""
140
# quick scan for perfect match
141
if fname in dir_list:
144
551
for dirname in dir_list:
145
552
if is_inside(dirname, fname):
557
def is_inside_or_parent_of_any(dir_list, fname):
558
"""True if fname is a child or a parent of any of the given files."""
559
for dirname in dir_list:
560
if is_inside(dirname, fname) or is_inside(fname, dirname):
565
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
566
report_activity=None, direction='read'):
567
"""Copy contents of one file to another.
569
The read_length can either be -1 to read to end-of-file (EOF) or
570
it can specify the maximum number of bytes to read.
572
The buff_size represents the maximum size for each read operation
573
performed on from_file.
575
:param report_activity: Call this as bytes are read, see
576
Transport._report_activity
577
:param direction: Will be passed to report_activity
579
:return: The number of bytes copied.
583
# read specified number of bytes
585
while read_length > 0:
586
num_bytes_to_read = min(read_length, buff_size)
588
block = from_file.read(num_bytes_to_read)
592
if report_activity is not None:
593
report_activity(len(block), direction)
596
actual_bytes_read = len(block)
597
read_length -= actual_bytes_read
598
length += actual_bytes_read
151
def pumpfile(fromfile, tofile):
152
"""Copy contents of one file to another."""
153
tofile.write(fromfile.read())
157
"""Return a new UUID"""
159
return file('/proc/sys/kernel/random/uuid').readline().rstrip('\n')
161
return chomp(os.popen('uuidgen').readline())
602
block = from_file.read(buff_size)
606
if report_activity is not None:
607
report_activity(len(block), direction)
613
def pump_string_file(bytes, file_handle, segment_size=None):
614
"""Write bytes to file_handle in many smaller writes.
616
:param bytes: The string to write.
617
:param file_handle: The file to write to.
619
# Write data in chunks rather than all at once, because very large
620
# writes fail on some platforms (e.g. Windows with SMB mounted
623
segment_size = 5242880 # 5MB
624
segments = range(len(bytes) / segment_size + 1)
625
write = file_handle.write
626
for segment_index in segments:
627
segment = buffer(bytes, segment_index * segment_size, segment_size)
631
def file_iterator(input_file, readsize=32768):
633
b = input_file.read(readsize)
166
if hasattr(f, 'tell'):
640
"""Calculate the hexdigest of an open file.
642
The file cursor should be already at the start.
654
def size_sha_file(f):
655
"""Calculate the size and hexdigest of an open file.
657
The file cursor should be already at the start and
658
the caller is responsible for closing the file afterwards.
669
return size, s.hexdigest()
672
def sha_file_by_name(fname):
673
"""Calculate the SHA1 of a file by reading the full text"""
675
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
678
b = os.read(f, 1<<16)
686
def sha_strings(strings, _factory=sha):
687
"""Return the sha-1 of concatenation of strings"""
689
map(s.update, strings)
693
def sha_string(f, _factory=sha):
694
return _factory(f).hexdigest()
186
697
def fingerprint_file(f):
192
return {'size': size,
193
'sha1': s.hexdigest()}
197
"""Return per-user configuration directory.
199
By default this is ~/.bzr.conf/
201
TODO: Global option --config-dir to override this.
203
return os.path.expanduser("~/.bzr.conf")
207
"""Calculate automatic user identification.
209
Returns (realname, email).
211
Only used when none is set in the environment or the id file.
213
This previously used the FQDN as the default domain, but that can
214
be very slow on machines where DNS is broken. So now we simply
219
# XXX: Any good way to get real user name on win32?
224
w = pwd.getpwuid(uid)
225
gecos = w.pw_gecos.decode(bzrlib.user_encoding)
226
username = w.pw_name.decode(bzrlib.user_encoding)
227
comma = gecos.find(',')
231
realname = gecos[:comma]
237
realname = username = getpass.getuser().decode(bzrlib.user_encoding)
239
return realname, (username + '@' + socket.gethostname())
243
"""Return the full user id from a file or environment variable.
245
TODO: Allow taking this from a file in the branch directory too
246
for per-branch ids."""
247
v = os.environ.get('BZREMAIL')
249
return v.decode(bzrlib.user_encoding)
252
return (open(os.path.join(config_dir(), "email"))
254
.decode(bzrlib.user_encoding)
257
if e.errno != errno.ENOENT:
260
v = os.environ.get('EMAIL')
262
return v.decode(bzrlib.user_encoding)
268
"""Return email-style username.
270
Something similar to 'Martin Pool <mbp@sourcefrog.net>'
272
TODO: Check it's reasonably well-formed.
278
name, email = _auto_user_id()
280
return '%s <%s>' % (name, email)
285
_EMAIL_RE = re.compile(r'[\w+.-]+@[\w+.-]+')
287
"""Return just the email component of a username."""
290
m = _EMAIL_RE.search(e)
292
raise BzrError("%r doesn't seem to contain a reasonable email address" % e)
295
return _auto_user_id()[1]
699
return {'size': len(b),
700
'sha1': sha(b).hexdigest()}
299
703
def compare_files(a, b):
400
raise BzrError("sorry, %r not allowed in path" % f)
918
raise errors.BzrError("sorry, %r not allowed in path" % f)
401
919
elif (f == '.') or (f == ''):
408
assert isinstance(p, list)
410
if (f == '..') or (f == None) or (f == ''):
411
raise BzrError("sorry, %r not allowed in path" % f)
412
return os.path.join(*p)
415
def appendpath(p1, p2):
419
return os.path.join(p1, p2)
422
def extern_command(cmd, ignore_errors = False):
423
mutter('external command: %s' % `cmd`)
425
if not ignore_errors:
426
raise BzrError('command failed')
429
def _read_config_value(name):
430
"""Read a config value from the file ~/.bzr.conf/<name>
431
Return None if the file does not exist"""
433
f = file(os.path.join(config_dir(), name), "r")
434
return f.read().decode(bzrlib.user_encoding).rstrip("\r\n")
436
if e.errno == errno.ENOENT:
928
if (f == '..') or (f is None) or (f == ''):
929
raise errors.BzrError("sorry, %r not allowed in path" % f)
933
def parent_directories(filename):
934
"""Return the list of parent directories, deepest first.
936
For example, parent_directories("a/b/c") -> ["a/b", "a"].
939
parts = splitpath(dirname(filename))
941
parents.append(joinpath(parts))
946
_extension_load_failures = []
949
def failed_to_load_extension(exception):
950
"""Handle failing to load a binary extension.
952
This should be called from the ImportError block guarding the attempt to
953
import the native extension. If this function returns, the pure-Python
954
implementation should be loaded instead::
957
>>> import bzrlib._fictional_extension_pyx
958
>>> except ImportError, e:
959
>>> bzrlib.osutils.failed_to_load_extension(e)
960
>>> import bzrlib._fictional_extension_py
962
# NB: This docstring is just an example, not a doctest, because doctest
963
# currently can't cope with the use of lazy imports in this namespace --
966
# This currently doesn't report the failure at the time it occurs, because
967
# they tend to happen very early in startup when we can't check config
968
# files etc, and also we want to report all failures but not spam the user
970
exception_str = str(exception)
971
if exception_str not in _extension_load_failures:
972
trace.mutter("failed to load compiled extension: %s" % exception_str)
973
_extension_load_failures.append(exception_str)
976
def report_extension_load_failures():
977
if not _extension_load_failures:
979
from bzrlib.config import GlobalConfig
980
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
982
# the warnings framework should by default show this only once
983
from bzrlib.trace import warning
985
"bzr: warning: some compiled extensions could not be loaded; "
986
"see <https://answers.launchpad.net/bzr/+faq/703>")
987
# we no longer show the specific missing extensions here, because it makes
988
# the message too long and scary - see
989
# https://bugs.launchpad.net/bzr/+bug/430529
993
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
994
except ImportError, e:
995
failed_to_load_extension(e)
996
from bzrlib._chunks_to_lines_py import chunks_to_lines
1000
"""Split s into lines, but without removing the newline characters."""
1001
# Trivially convert a fulltext into a 'chunked' representation, and let
1002
# chunks_to_lines do the heavy lifting.
1003
if isinstance(s, str):
1004
# chunks_to_lines only supports 8-bit strings
1005
return chunks_to_lines([s])
1007
return _split_lines(s)
1010
def _split_lines(s):
1011
"""Split s into lines, but without removing the newline characters.
1013
This supports Unicode or plain string objects.
1015
lines = s.split('\n')
1016
result = [line + '\n' for line in lines[:-1]]
1018
result.append(lines[-1])
1022
def hardlinks_good():
1023
return sys.platform not in ('win32', 'cygwin', 'darwin')
1026
def link_or_copy(src, dest):
1027
"""Hardlink a file, or copy it if it can't be hardlinked."""
1028
if not hardlinks_good():
1029
shutil.copyfile(src, dest)
1033
except (OSError, IOError), e:
1034
if e.errno != errno.EXDEV:
1036
shutil.copyfile(src, dest)
1039
def delete_any(path):
1040
"""Delete a file, symlink or directory.
1042
Will delete even if readonly.
1045
_delete_file_or_dir(path)
1046
except (OSError, IOError), e:
1047
if e.errno in (errno.EPERM, errno.EACCES):
1048
# make writable and try again
1051
except (OSError, IOError):
1053
_delete_file_or_dir(path)
1058
def _delete_file_or_dir(path):
1059
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1060
# Forgiveness than Permission (EAFP) because:
1061
# - root can damage a solaris file system by using unlink,
1062
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1063
# EACCES, OSX: EPERM) when invoked on a directory.
1064
if isdir(path): # Takes care of symlinks
1071
if getattr(os, 'symlink', None) is not None:
1077
def has_hardlinks():
1078
if getattr(os, 'link', None) is not None:
1084
def host_os_dereferences_symlinks():
1085
return (has_symlinks()
1086
and sys.platform not in ('cygwin', 'win32'))
1089
def readlink(abspath):
1090
"""Return a string representing the path to which the symbolic link points.
1092
:param abspath: The link absolute unicode path.
1094
This his guaranteed to return the symbolic link in unicode in all python
1097
link = abspath.encode(_fs_enc)
1098
target = os.readlink(link)
1099
target = target.decode(_fs_enc)
1103
def contains_whitespace(s):
1104
"""True if there are any whitespace characters in s."""
1105
# string.whitespace can include '\xa0' in certain locales, because it is
1106
# considered "non-breaking-space" as part of ISO-8859-1. But it
1107
# 1) Isn't a breaking whitespace
1108
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1110
# 3) '\xa0' isn't unicode safe since it is >128.
1112
# This should *not* be a unicode set of characters in case the source
1113
# string is not a Unicode string. We can auto-up-cast the characters since
1114
# they are ascii, but we don't want to auto-up-cast the string in case it
1116
for ch in ' \t\n\r\v\f':
1123
def contains_linebreaks(s):
1124
"""True if there is any vertical whitespace in s."""
1132
def relpath(base, path):
1133
"""Return path relative to base, or raise PathNotChild exception.
1135
The path may be either an absolute path or a path relative to the
1136
current working directory.
1138
os.path.commonprefix (python2.4) has a bad bug that it works just
1139
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1140
avoids that problem.
1142
NOTE: `base` should not have a trailing slash otherwise you'll get
1143
PathNotChild exceptions regardless of `path`.
1146
if len(base) < MIN_ABS_PATHLENGTH:
1147
# must have space for e.g. a drive letter
1148
raise ValueError('%r is too short to calculate a relative path'
1156
if len(head) <= len(base) and head != base:
1157
raise errors.PathNotChild(rp, base)
1160
head, tail = split(head)
1165
return pathjoin(*reversed(s))
1170
def _cicp_canonical_relpath(base, path):
1171
"""Return the canonical path relative to base.
1173
Like relpath, but on case-insensitive-case-preserving file-systems, this
1174
will return the relpath as stored on the file-system rather than in the
1175
case specified in the input string, for all existing portions of the path.
1177
This will cause O(N) behaviour if called for every path in a tree; if you
1178
have a number of paths to convert, you should use canonical_relpaths().
1180
# TODO: it should be possible to optimize this for Windows by using the
1181
# win32 API FindFiles function to look for the specified name - but using
1182
# os.listdir() still gives us the correct, platform agnostic semantics in
1185
rel = relpath(base, path)
1186
# '.' will have been turned into ''
1190
abs_base = abspath(base)
1192
_listdir = os.listdir
1194
# use an explicit iterator so we can easily consume the rest on early exit.
1195
bit_iter = iter(rel.split('/'))
1196
for bit in bit_iter:
1199
next_entries = _listdir(current)
1200
except OSError: # enoent, eperm, etc
1201
# We can't find this in the filesystem, so just append the
1203
current = pathjoin(current, bit, *list(bit_iter))
1205
for look in next_entries:
1206
if lbit == look.lower():
1207
current = pathjoin(current, look)
1210
# got to the end, nothing matched, so we just return the
1211
# non-existing bits as they were specified (the filename may be
1212
# the target of a move, for example).
1213
current = pathjoin(current, bit, *list(bit_iter))
1215
return current[len(abs_base):].lstrip('/')
1217
# XXX - TODO - we need better detection/integration of case-insensitive
1218
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1219
# filesystems), for example, so could probably benefit from the same basic
1220
# support there. For now though, only Windows and OSX get that support, and
1221
# they get it for *all* file-systems!
1222
if sys.platform in ('win32', 'darwin'):
1223
canonical_relpath = _cicp_canonical_relpath
1225
canonical_relpath = relpath
1227
def canonical_relpaths(base, paths):
1228
"""Create an iterable to canonicalize a sequence of relative paths.
1230
The intent is for this implementation to use a cache, vastly speeding
1231
up multiple transformations in the same directory.
1233
# but for now, we haven't optimized...
1234
return [canonical_relpath(base, p) for p in paths]
1237
def decode_filename(filename):
1238
"""Decode the filename using the filesystem encoding
1240
If it is unicode, it is returned.
1241
Otherwise it is decoded from the the filesystem's encoding. If decoding
1242
fails, a errors.BadFilenameEncoding exception is raised.
1244
if type(filename) is unicode:
1247
return filename.decode(_fs_enc)
1248
except UnicodeDecodeError:
1249
raise errors.BadFilenameEncoding(filename, _fs_enc)
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.
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
(everything else), 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.
1419
# Keep some state so that terminal_width can detect if _terminal_size has
1420
# returned a different size since the process started. See docstring and
1421
# comments of terminal_width for details.
1422
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1423
_terminal_size_state = 'no_data'
1424
_first_terminal_size = None
1426
def terminal_width():
1427
"""Return terminal width.
1429
None is returned if the width can't established precisely.
1432
- if BZR_COLUMNS is set, returns its value
1433
- if there is no controlling terminal, returns None
1434
- query the OS, if the queried size has changed since the last query,
1436
- if COLUMNS is set, returns its value,
1437
- if the OS has a value (even though it's never changed), return its value.
1439
From there, we need to query the OS to get the size of the controlling
1442
On Unices we query the OS by:
1443
- get termios.TIOCGWINSZ
1444
- if an error occurs or a negative value is obtained, returns None
1446
On Windows we query the OS by:
1447
- win32utils.get_console_size() decides,
1448
- returns None on error (provided default value)
1450
# Note to implementors: if changing the rules for determining the width,
1451
# make sure you've considered the behaviour in these cases:
1452
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1453
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1455
# - (add more interesting cases here, if you find any)
1456
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1457
# but we don't want to register a signal handler because it is impossible
1458
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1459
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1460
# time so we can notice if the reported size has changed, which should have
1463
# If BZR_COLUMNS is set, take it, user is always right
1464
# Except if they specified 0 in which case, impose no limit here
1466
width = int(os.environ['BZR_COLUMNS'])
1467
except (KeyError, ValueError):
1469
if width is not None:
442
"""Return a sequence of possible editor binaries for the current platform"""
443
e = _read_config_value("editor")
447
if os.name == "windows":
449
elif os.name == "posix":
451
yield os.environ["EDITOR"]
456
def _run_editor(filename):
457
"""Try to execute an editor to edit the commit message. Returns True on success,
459
for e in _get_editor():
460
x = os.spawnvp(os.P_WAIT, e, (e, filename))
467
raise BzrError("Could not start any editor. Please specify $EDITOR or use ~/.bzr.conf/editor")
1475
isatty = getattr(sys.stdout, 'isatty', None)
1476
if isatty is None or not isatty():
1477
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1481
width, height = os_size = _terminal_size(None, None)
1482
global _first_terminal_size, _terminal_size_state
1483
if _terminal_size_state == 'no_data':
1484
_first_terminal_size = os_size
1485
_terminal_size_state = 'unchanged'
1486
elif (_terminal_size_state == 'unchanged' and
1487
_first_terminal_size != os_size):
1488
_terminal_size_state = 'changed'
1490
# If the OS claims to know how wide the terminal is, and this value has
1491
# ever changed, use that.
1492
if _terminal_size_state == 'changed':
1493
if width is not None and width > 0:
1496
# If COLUMNS is set, use it.
1498
return int(os.environ['COLUMNS'])
1499
except (KeyError, ValueError):
1502
# Finally, use an unchanged size from the OS, if we have one.
1503
if _terminal_size_state == 'unchanged':
1504
if width is not None and width > 0:
1507
# The width could not be determined.
1511
def _win32_terminal_size(width, height):
1512
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1513
return width, height
1516
def _ioctl_terminal_size(width, height):
1518
import struct, fcntl, termios
1519
s = struct.pack('HHHH', 0, 0, 0, 0)
1520
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1521
height, width = struct.unpack('HHHH', x)[0:2]
1522
except (IOError, AttributeError):
1524
return width, height
1526
_terminal_size = None
1527
"""Returns the terminal size as (width, height).
1529
:param width: Default value for width.
1530
:param height: Default value for height.
1532
This is defined specifically for each OS and query the size of the controlling
1533
terminal. If any error occurs, the provided default values should be returned.
1535
if sys.platform == 'win32':
1536
_terminal_size = _win32_terminal_size
1538
_terminal_size = _ioctl_terminal_size
1541
def supports_executable():
1542
return sys.platform != "win32"
1545
def supports_posix_readonly():
1546
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1548
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1549
directory controls creation/deletion, etc.
1551
And under win32, readonly means that the directory itself cannot be
1552
deleted. The contents of a readonly directory can be changed, unlike POSIX
1553
where files in readonly directories cannot be added, deleted or renamed.
1555
return sys.platform != "win32"
1558
def set_or_unset_env(env_variable, value):
1559
"""Modify the environment, setting or removing the env_variable.
1561
:param env_variable: The environment variable in question
1562
:param value: The value to set the environment to. If None, then
1563
the variable will be removed.
1564
:return: The original value of the environment variable.
1566
orig_val = os.environ.get(env_variable)
1568
if orig_val is not None:
1569
del os.environ[env_variable]
1571
if isinstance(value, unicode):
1572
value = value.encode(get_user_encoding())
1573
os.environ[env_variable] = value
1577
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1580
def check_legal_path(path):
1581
"""Check whether the supplied path is legal.
1582
This is only required on Windows, so we don't test on other platforms
1585
if sys.platform != "win32":
1587
if _validWin32PathRE.match(path) is None:
1588
raise errors.IllegalPath(path)
1591
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1593
def _is_error_enotdir(e):
1594
"""Check if this exception represents ENOTDIR.
1596
Unfortunately, python is very inconsistent about the exception
1597
here. The cases are:
1598
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1599
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1600
which is the windows error code.
1601
3) Windows, Python2.5 uses errno == EINVAL and
1602
winerror == ERROR_DIRECTORY
1604
:param e: An Exception object (expected to be OSError with an errno
1605
attribute, but we should be able to cope with anything)
1606
:return: True if this represents an ENOTDIR error. False otherwise.
1608
en = getattr(e, 'errno', None)
1609
if (en == errno.ENOTDIR
1610
or (sys.platform == 'win32'
1611
and (en == _WIN32_ERROR_DIRECTORY
1612
or (en == errno.EINVAL
1613
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
471
def get_text_message(infotext, ignoreline = "default"):
474
if ignoreline == "default":
475
ignoreline = "-- This line and the following will be ignored --"
478
tmp_fileno, msgfilename = tempfile.mkstemp()
479
msgfile = os.close(tmp_fileno)
480
if infotext is not None and infotext != "":
482
msgfile = file(msgfilename, "w")
483
msgfile.write("\n\n%s\n\n%s" % (ignoreline, infotext))
488
if not _run_editor(msgfilename):
493
lastline, nlines = 0, 0
494
for line in file(msgfilename, "r"):
495
stripped_line = line.strip()
496
# strip empty line before the log message starts
498
if stripped_line != "":
502
# check for the ignore line only if there
503
# is additional information at the end
504
if hasinfo and stripped_line == ignoreline:
507
# keep track of the last line that had some content
508
if stripped_line != "":
514
# delete empty lines at the end
516
# add a newline at the end, if needed
517
if not msg[-1].endswith("\n"):
518
return "%s%s" % ("".join(msg), "\n")
1619
def walkdirs(top, prefix=""):
1620
"""Yield data about all the directories in a tree.
1622
This yields all the data about the contents of a directory at a time.
1623
After each directory has been yielded, if the caller has mutated the list
1624
to exclude some directories, they are then not descended into.
1626
The data yielded is of the form:
1627
((directory-relpath, directory-path-from-top),
1628
[(relpath, basename, kind, lstat, path-from-top), ...]),
1629
- directory-relpath is the relative path of the directory being returned
1630
with respect to top. prefix is prepended to this.
1631
- directory-path-from-root is the path including top for this directory.
1632
It is suitable for use with os functions.
1633
- relpath is the relative path within the subtree being walked.
1634
- basename is the basename of the path
1635
- kind is the kind of the file now. If unknown then the file is not
1636
present within the tree - but it may be recorded as versioned. See
1638
- lstat is the stat data *if* the file was statted.
1639
- planned, not implemented:
1640
path_from_tree_root is the path from the root of the tree.
1642
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1643
allows one to walk a subtree but get paths that are relative to a tree
1645
:return: an iterator over the dirs.
1647
#TODO there is a bit of a smell where the results of the directory-
1648
# summary in this, and the path from the root, may not agree
1649
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1650
# potentially confusing output. We should make this more robust - but
1651
# not at a speed cost. RBC 20060731
1653
_directory = _directory_kind
1654
_listdir = os.listdir
1655
_kind_from_mode = file_kind_from_stat_mode
1656
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1658
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1659
relroot, _, _, _, top = pending.pop()
1661
relprefix = relroot + u'/'
1664
top_slash = top + u'/'
1667
append = dirblock.append
1669
names = sorted(map(decode_filename, _listdir(top)))
1671
if not _is_error_enotdir(e):
1675
abspath = top_slash + name
1676
statvalue = _lstat(abspath)
1677
kind = _kind_from_mode(statvalue.st_mode)
1678
append((relprefix + name, name, kind, statvalue, abspath))
1679
yield (relroot, top), dirblock
1681
# push the user specified dirs from dirblock
1682
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1685
class DirReader(object):
1686
"""An interface for reading directories."""
1688
def top_prefix_to_starting_dir(self, top, prefix=""):
1689
"""Converts top and prefix to a starting dir entry
1691
:param top: A utf8 path
1692
:param prefix: An optional utf8 path to prefix output relative paths
1694
:return: A tuple starting with prefix, and ending with the native
1697
raise NotImplementedError(self.top_prefix_to_starting_dir)
1699
def read_dir(self, prefix, top):
1700
"""Read a specific dir.
1702
:param prefix: A utf8 prefix to be preprended to the path basenames.
1703
:param top: A natively encoded path to read.
1704
:return: A list of the directories contents. Each item contains:
1705
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1707
raise NotImplementedError(self.read_dir)
1710
_selected_dir_reader = None
1713
def _walkdirs_utf8(top, prefix=""):
1714
"""Yield data about all the directories in a tree.
1716
This yields the same information as walkdirs() only each entry is yielded
1717
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1718
are returned as exact byte-strings.
1720
:return: yields a tuple of (dir_info, [file_info])
1721
dir_info is (utf8_relpath, path-from-top)
1722
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1723
if top is an absolute path, path-from-top is also an absolute path.
1724
path-from-top might be unicode or utf8, but it is the correct path to
1725
pass to os functions to affect the file in question. (such as os.lstat)
1727
global _selected_dir_reader
1728
if _selected_dir_reader is None:
1729
fs_encoding = _fs_enc.upper()
1730
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1731
# Win98 doesn't have unicode apis like FindFirstFileW
1732
# TODO: We possibly could support Win98 by falling back to the
1733
# original FindFirstFile, and using TCHAR instead of WCHAR,
1734
# but that gets a bit tricky, and requires custom compiling
1737
from bzrlib._walkdirs_win32 import Win32ReadDir
1738
_selected_dir_reader = Win32ReadDir()
1741
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1742
# ANSI_X3.4-1968 is a form of ASCII
1744
from bzrlib._readdir_pyx import UTF8DirReader
1745
_selected_dir_reader = UTF8DirReader()
1746
except ImportError, e:
1747
failed_to_load_extension(e)
1750
if _selected_dir_reader is None:
1751
# Fallback to the python version
1752
_selected_dir_reader = UnicodeDirReader()
1754
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1755
# But we don't actually uses 1-3 in pending, so set them to None
1756
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1757
read_dir = _selected_dir_reader.read_dir
1758
_directory = _directory_kind
1760
relroot, _, _, _, top = pending[-1].pop()
1763
dirblock = sorted(read_dir(relroot, top))
1764
yield (relroot, top), dirblock
1765
# push the user specified dirs from dirblock
1766
next = [d for d in reversed(dirblock) if d[2] == _directory]
1768
pending.append(next)
1771
class UnicodeDirReader(DirReader):
1772
"""A dir reader for non-utf8 file systems, which transcodes."""
1774
__slots__ = ['_utf8_encode']
1777
self._utf8_encode = codecs.getencoder('utf8')
1779
def top_prefix_to_starting_dir(self, top, prefix=""):
1780
"""See DirReader.top_prefix_to_starting_dir."""
1781
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1783
def read_dir(self, prefix, top):
1784
"""Read a single directory from a non-utf8 file system.
1786
top, and the abspath element in the output are unicode, all other paths
1787
are utf8. Local disk IO is done via unicode calls to listdir etc.
1789
This is currently the fallback code path when the filesystem encoding is
1790
not UTF-8. It may be better to implement an alternative so that we can
1791
safely handle paths that are not properly decodable in the current
1794
See DirReader.read_dir for details.
1796
_utf8_encode = self._utf8_encode
1798
_listdir = os.listdir
1799
_kind_from_mode = file_kind_from_stat_mode
1802
relprefix = prefix + '/'
1805
top_slash = top + u'/'
1808
append = dirblock.append
1809
for name in sorted(_listdir(top)):
1811
name_utf8 = _utf8_encode(name)[0]
1812
except UnicodeDecodeError:
1813
raise errors.BadFilenameEncoding(
1814
_utf8_encode(relprefix)[0] + name, _fs_enc)
1815
abspath = top_slash + name
1816
statvalue = _lstat(abspath)
1817
kind = _kind_from_mode(statvalue.st_mode)
1818
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1822
def copy_tree(from_path, to_path, handlers={}):
1823
"""Copy all of the entries in from_path into to_path.
1825
:param from_path: The base directory to copy.
1826
:param to_path: The target directory. If it does not exist, it will
1828
:param handlers: A dictionary of functions, which takes a source and
1829
destinations for files, directories, etc.
1830
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1831
'file', 'directory', and 'symlink' should always exist.
1832
If they are missing, they will be replaced with 'os.mkdir()',
1833
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1835
# Now, just copy the existing cached tree to the new location
1836
# We use a cheap trick here.
1837
# Absolute paths are prefixed with the first parameter
1838
# relative paths are prefixed with the second.
1839
# So we can get both the source and target returned
1840
# without any extra work.
1842
def copy_dir(source, dest):
1845
def copy_link(source, dest):
1846
"""Copy the contents of a symlink"""
1847
link_to = os.readlink(source)
1848
os.symlink(link_to, dest)
1850
real_handlers = {'file':shutil.copy2,
1851
'symlink':copy_link,
1852
'directory':copy_dir,
1854
real_handlers.update(handlers)
1856
if not os.path.exists(to_path):
1857
real_handlers['directory'](from_path, to_path)
1859
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1860
for relpath, name, kind, st, abspath in entries:
1861
real_handlers[kind](abspath, relpath)
1864
def copy_ownership_from_path(dst, src=None):
1865
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1867
If src is None, the containing directory is used as source. If chown
1868
fails, the error is ignored and a warning is printed.
1870
chown = getattr(os, 'chown', None)
1875
src = os.path.dirname(dst)
1881
chown(dst, s.st_uid, s.st_gid)
1884
'Unable to copy ownership from "%s" to "%s". '
1885
'You may want to set it manually.', src, dst)
1886
trace.log_exception_quietly()
1889
def path_prefix_key(path):
1890
"""Generate a prefix-order path key for path.
1892
This can be used to sort paths in the same way that walkdirs does.
1894
return (dirname(path) , path)
1897
def compare_paths_prefix_order(path_a, path_b):
1898
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1899
key_a = path_prefix_key(path_a)
1900
key_b = path_prefix_key(path_b)
1901
return cmp(key_a, key_b)
1904
_cached_user_encoding = None
1907
def get_user_encoding(use_cache=True):
1908
"""Find out what the preferred user encoding is.
1910
This is generally the encoding that is used for command line parameters
1911
and file contents. This may be different from the terminal encoding
1912
or the filesystem encoding.
1914
:param use_cache: Enable cache for detected encoding.
1915
(This parameter is turned on by default,
1916
and required only for selftesting)
1918
:return: A string defining the preferred user encoding
1920
global _cached_user_encoding
1921
if _cached_user_encoding is not None and use_cache:
1922
return _cached_user_encoding
1924
if sys.platform == 'darwin':
1925
# python locale.getpreferredencoding() always return
1926
# 'mac-roman' on darwin. That's a lie.
1927
sys.platform = 'posix'
1929
if os.environ.get('LANG', None) is None:
1930
# If LANG is not set, we end up with 'ascii', which is bad
1931
# ('mac-roman' is more than ascii), so we set a default which
1932
# will give us UTF-8 (which appears to work in all cases on
1933
# OSX). Users are still free to override LANG of course, as
1934
# long as it give us something meaningful. This work-around
1935
# *may* not be needed with python 3k and/or OSX 10.5, but will
1936
# work with them too -- vila 20080908
1937
os.environ['LANG'] = 'en_US.UTF-8'
1940
sys.platform = 'darwin'
1945
user_encoding = locale.getpreferredencoding()
1946
except locale.Error, e:
1947
sys.stderr.write('bzr: warning: %s\n'
1948
' Could not determine what text encoding to use.\n'
1949
' This error usually means your Python interpreter\n'
1950
' doesn\'t support the locale set by $LANG (%s)\n'
1951
" Continuing with ascii encoding.\n"
1952
% (e, os.environ.get('LANG')))
1953
user_encoding = 'ascii'
1955
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1956
# treat that as ASCII, and not support printing unicode characters to the
1959
# For python scripts run under vim, we get '', so also treat that as ASCII
1960
if user_encoding in (None, 'cp0', ''):
1961
user_encoding = 'ascii'
1965
codecs.lookup(user_encoding)
1967
sys.stderr.write('bzr: warning:'
1968
' unknown encoding %s.'
1969
' Continuing with ascii encoding.\n'
1972
user_encoding = 'ascii'
1975
_cached_user_encoding = user_encoding
1977
return user_encoding
1980
def get_diff_header_encoding():
1981
return get_terminal_encoding()
1984
def get_host_name():
1985
"""Return the current unicode host name.
1987
This is meant to be used in place of socket.gethostname() because that
1988
behaves inconsistently on different platforms.
1990
if sys.platform == "win32":
1992
return win32utils.get_host_name()
1995
return socket.gethostname().decode(get_user_encoding())
1998
# We must not read/write any more than 64k at a time from/to a socket so we
1999
# don't risk "no buffer space available" errors on some platforms. Windows in
2000
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2002
MAX_SOCKET_CHUNK = 64 * 1024
2004
_end_of_stream_errors = [errno.ECONNRESET]
2005
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2006
_eno = getattr(errno, _eno, None)
2007
if _eno is not None:
2008
_end_of_stream_errors.append(_eno)
2012
def read_bytes_from_socket(sock, report_activity=None,
2013
max_read_size=MAX_SOCKET_CHUNK):
2014
"""Read up to max_read_size of bytes from sock and notify of progress.
2016
Translates "Connection reset by peer" into file-like EOF (return an
2017
empty string rather than raise an error), and repeats the recv if
2018
interrupted by a signal.
2022
bytes = sock.recv(max_read_size)
2023
except socket.error, e:
2025
if eno in _end_of_stream_errors:
2026
# The connection was closed by the other side. Callers expect
2027
# an empty string to signal end-of-stream.
2029
elif eno == errno.EINTR:
2030
# Retry the interrupted recv.
2034
if report_activity is not None:
2035
report_activity(len(bytes), 'read')
2039
def recv_all(socket, count):
2040
"""Receive an exact number of bytes.
2042
Regular Socket.recv() may return less than the requested number of bytes,
2043
depending on what's in the OS buffer. MSG_WAITALL is not available
2044
on all platforms, but this should work everywhere. This will return
2045
less than the requested amount if the remote end closes.
2047
This isn't optimized and is intended mostly for use in testing.
2050
while len(b) < count:
2051
new = read_bytes_from_socket(socket, None, count - len(b))
2058
def send_all(sock, bytes, report_activity=None):
2059
"""Send all bytes on a socket.
2061
Breaks large blocks in smaller chunks to avoid buffering limitations on
2062
some platforms, and catches EINTR which may be thrown if the send is
2063
interrupted by a signal.
2065
This is preferred to socket.sendall(), because it avoids portability bugs
2066
and provides activity reporting.
2068
:param report_activity: Call this as bytes are read, see
2069
Transport._report_activity
2072
byte_count = len(bytes)
2073
while sent_total < byte_count:
2075
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2076
except socket.error, e:
2077
if e.args[0] != errno.EINTR:
2081
report_activity(sent, 'write')
2084
def connect_socket(address):
2085
# Slight variation of the socket.create_connection() function (provided by
2086
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2087
# provide it for previous python versions. Also, we don't use the timeout
2088
# parameter (provided by the python implementation) so we don't implement
2090
err = socket.error('getaddrinfo returns an empty list')
2091
host, port = address
2092
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2093
af, socktype, proto, canonname, sa = res
2096
sock = socket.socket(af, socktype, proto)
2100
except socket.error, err:
2101
# 'err' is now the most recent error
2102
if sock is not None:
2107
def dereference_path(path):
2108
"""Determine the real path to a file.
2110
All parent elements are dereferenced. But the file itself is not
2112
:param path: The original path. May be absolute or relative.
2113
:return: the real path *to* the file
2115
parent, base = os.path.split(path)
2116
# The pathjoin for '.' is a workaround for Python bug #1213894.
2117
# (initial path components aren't dereferenced)
2118
return pathjoin(realpath(pathjoin('.', parent)), base)
2121
def supports_mapi():
2122
"""Return True if we can use MAPI to launch a mail client."""
2123
return sys.platform == "win32"
2126
def resource_string(package, resource_name):
2127
"""Load a resource from a package and return it as a string.
2129
Note: Only packages that start with bzrlib are currently supported.
2131
This is designed to be a lightweight implementation of resource
2132
loading in a way which is API compatible with the same API from
2134
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2135
If and when pkg_resources becomes a standard library, this routine
2138
# Check package name is within bzrlib
2139
if package == "bzrlib":
2140
resource_relpath = resource_name
2141
elif package.startswith("bzrlib."):
2142
package = package[len("bzrlib."):].replace('.', os.sep)
2143
resource_relpath = pathjoin(package, resource_name)
2145
raise errors.BzrError('resource package %s not in bzrlib' % package)
2147
# Map the resource to a file and read its contents
2148
base = dirname(bzrlib.__file__)
2149
if getattr(sys, 'frozen', None): # bzr.exe
2150
base = abspath(pathjoin(base, '..', '..'))
2151
f = file(pathjoin(base, resource_relpath), "rU")
522
# delete the msg file in any case
523
try: os.unlink(msgfilename)
2157
def file_kind_from_stat_mode_thunk(mode):
2158
global file_kind_from_stat_mode
2159
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2161
from bzrlib._readdir_pyx import UTF8DirReader
2162
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2163
except ImportError, e:
2164
# This is one time where we won't warn that an extension failed to
2165
# load. The extension is never available on Windows anyway.
2166
from bzrlib._readdir_py import (
2167
_kind_from_mode as file_kind_from_stat_mode
2169
return file_kind_from_stat_mode(mode)
2170
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2173
def file_kind(f, _lstat=os.lstat):
2175
return file_kind_from_stat_mode(_lstat(f).st_mode)
2177
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2178
raise errors.NoSuchFile(f)
2182
def until_no_eintr(f, *a, **kw):
2183
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2185
WARNING: you must be certain that it is safe to retry the call repeatedly
2186
if EINTR does occur. This is typically only true for low-level operations
2187
like os.read. If in any doubt, don't use this.
2189
Keep in mind that this is not a complete solution to EINTR. There is
2190
probably code in the Python standard library and other dependencies that
2191
may encounter EINTR if a signal arrives (and there is signal handler for
2192
that signal). So this function can reduce the impact for IO that bzrlib
2193
directly controls, but it is not a complete solution.
2195
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2199
except (IOError, OSError), e:
2200
if e.errno == errno.EINTR:
2205
@deprecated_function(deprecated_in((2, 2, 0)))
2206
def re_compile_checked(re_string, flags=0, where=""):
2207
"""Return a compiled re, or raise a sensible error.
2209
This should only be used when compiling user-supplied REs.
2211
:param re_string: Text form of regular expression.
2212
:param flags: eg re.IGNORECASE
2213
:param where: Message explaining to the user the context where
2214
it occurred, eg 'log search filter'.
2216
# from https://bugs.launchpad.net/bzr/+bug/251352
2218
re_obj = re.compile(re_string, flags)
2221
except errors.InvalidPattern, e:
2223
where = ' in ' + where
2224
# despite the name 'error' is a type
2225
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2229
if sys.platform == "win32":
2232
return msvcrt.getch()
2237
fd = sys.stdin.fileno()
2238
settings = termios.tcgetattr(fd)
2241
ch = sys.stdin.read(1)
2243
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2246
if sys.platform == 'linux2':
2247
def _local_concurrency():
2249
return os.sysconf('SC_NPROCESSORS_ONLN')
2250
except (ValueError, OSError, AttributeError):
2252
elif sys.platform == 'darwin':
2253
def _local_concurrency():
2254
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2255
stdout=subprocess.PIPE).communicate()[0]
2256
elif "bsd" in sys.platform:
2257
def _local_concurrency():
2258
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2259
stdout=subprocess.PIPE).communicate()[0]
2260
elif sys.platform == 'sunos5':
2261
def _local_concurrency():
2262
return subprocess.Popen(['psrinfo', '-p',],
2263
stdout=subprocess.PIPE).communicate()[0]
2264
elif sys.platform == "win32":
2265
def _local_concurrency():
2266
# This appears to return the number of cores.
2267
return os.environ.get('NUMBER_OF_PROCESSORS')
2269
def _local_concurrency():
2274
_cached_local_concurrency = None
2276
def local_concurrency(use_cache=True):
2277
"""Return how many processes can be run concurrently.
2279
Rely on platform specific implementations and default to 1 (one) if
2280
anything goes wrong.
2282
global _cached_local_concurrency
2284
if _cached_local_concurrency is not None and use_cache:
2285
return _cached_local_concurrency
2287
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2288
if concurrency is None:
2290
import multiprocessing
2292
# multiprocessing is only available on Python >= 2.6
2294
concurrency = _local_concurrency()
2295
except (OSError, IOError):
2298
concurrency = multiprocessing.cpu_count()
2300
concurrency = int(concurrency)
2301
except (TypeError, ValueError):
2304
_cached_concurrency = concurrency
2308
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2309
"""A stream writer that doesn't decode str arguments."""
2311
def __init__(self, encode, stream, errors='strict'):
2312
codecs.StreamWriter.__init__(self, stream, errors)
2313
self.encode = encode
2315
def write(self, object):
2316
if type(object) is str:
2317
self.stream.write(object)
2319
data, _ = self.encode(object, self.errors)
2320
self.stream.write(data)
2322
if sys.platform == 'win32':
2323
def open_file(filename, mode='r', bufsize=-1):
2324
"""This function is used to override the ``open`` builtin.
2326
But it uses O_NOINHERIT flag so the file handle is not inherited by
2327
child processes. Deleting or renaming a closed file opened with this
2328
function is not blocking child processes.
2330
writing = 'w' in mode
2331
appending = 'a' in mode
2332
updating = '+' in mode
2333
binary = 'b' in mode
2336
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2337
# for flags for each modes.
2347
flags |= os.O_WRONLY
2348
flags |= os.O_CREAT | os.O_TRUNC
2353
flags |= os.O_WRONLY
2354
flags |= os.O_CREAT | os.O_APPEND
2359
flags |= os.O_RDONLY
2361
return os.fdopen(os.open(filename, flags), mode, bufsize)
2366
def getuser_unicode():
2367
"""Return the username as unicode.
2370
user_encoding = get_user_encoding()
2371
username = getpass.getuser().decode(user_encoding)
2372
except UnicodeDecodeError:
2373
raise errors.BzrError("Can't decode username as %s." % \
2378
def available_backup_name(base, exists):
2379
"""Find a non-existing backup file name.
2381
This will *not* create anything, this only return a 'free' entry. This
2382
should be used for checking names in a directory below a locked
2383
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2384
Leap) and generally discouraged.
2386
:param base: The base name.
2388
:param exists: A callable returning True if the path parameter exists.
2391
name = "%s.~%d~" % (base, counter)
2394
name = "%s.~%d~" % (base, counter)
2398
def set_fd_cloexec(fd):
2399
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2400
support for this is not available.
2404
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2405
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2406
except (ImportError, AttributeError):
2407
# Either the fcntl module or specific constants are not present
2411
def find_executable_on_path(name):
2412
"""Finds an executable on the PATH.
2414
On Windows, this will try to append each extension in the PATHEXT
2415
environment variable to the name, if it cannot be found with the name
2418
:param name: The base name of the executable.
2419
:return: The path to the executable found or None.
2421
path = os.environ.get('PATH')
2424
path = path.split(os.pathsep)
2425
if sys.platform == 'win32':
2426
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2427
exts = [ext.lower() for ext in exts]
2428
base, ext = os.path.splitext(name)
2430
if ext.lower() not in exts:
2438
f = os.path.join(d, name) + ext
2439
if os.access(f, os.X_OK):
added
removed
bzrlib/osutils.py
