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 (
63
from bzrlib import symbol_versioning
66
# Cross platform wall-clock time functionality with decent resolution.
67
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
68
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
69
# synchronized with ``time.time()``, this is only meant to be used to find
70
# delta times by subtracting from another call to this function.
71
timer_func = time.time
72
if sys.platform == 'win32':
73
timer_func = time.clock
75
# On win32, O_BINARY is used to indicate the file should
76
# be opened in binary mode, rather than text mode.
77
# On other platforms, O_BINARY doesn't exist, because
78
# they always open in binary mode, so it is okay to
79
# OR with 0 on those platforms.
80
# O_NOINHERIT and O_TEXT exists only on win32 too.
81
O_BINARY = getattr(os, 'O_BINARY', 0)
82
O_TEXT = getattr(os, 'O_TEXT', 0)
83
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
86
def get_unicode_argv():
88
user_encoding = get_user_encoding()
89
return [a.decode(user_encoding) for a in sys.argv[1:]]
90
except UnicodeDecodeError:
91
raise errors.BzrError("Parameter %r encoding is unsupported by %s "
92
"application locale." % (a, user_encoding))
26
95
def make_readonly(filename):
27
96
"""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)
97
mod = os.lstat(filename).st_mode
98
if not stat.S_ISLNK(mod):
100
os.chmod(filename, mod)
34
103
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.,:/_~-])')
104
mod = os.lstat(filename).st_mode
105
if not stat.S_ISLNK(mod):
107
os.chmod(filename, mod)
110
def minimum_path_selection(paths):
111
"""Return the smallset subset of paths which are outside paths.
113
:param paths: A container (and hence not None) of paths.
114
:return: A set of paths sufficient to include everything in paths via
115
is_inside, drawn from the paths parameter.
121
return path.split('/')
122
sorted_paths = sorted(list(paths), key=sort_key)
124
search_paths = [sorted_paths[0]]
125
for path in sorted_paths[1:]:
126
if not is_inside(search_paths[-1], path):
127
# This path is unique, add it
128
search_paths.append(path)
130
return set(search_paths)
42
137
"""Return a quoted filename filename
44
139
This previously used backslash quoting, but that works poorly on
46
141
# TODO: I'm not really sure this is the best format either.x
143
if _QUOTE_RE is None:
144
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
47
146
if _QUOTE_RE.search(f):
48
147
return '"' + f + '"'
54
mode = os.lstat(f)[ST_MODE]
62
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
152
_directory_kind = 'directory'
155
"""Return the current umask"""
156
# Assume that people aren't messing with the umask while running
157
# XXX: This is not thread safe, but there is no way to get the
158
# umask without setting it
166
_directory_kind: "/",
168
'tree-reference': '+',
65
172
def kind_marker(kind):
174
return _kind_marker_map[kind]
176
# 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':
181
lexists = getattr(os.path, 'lexists', None)
185
stat = getattr(os, 'lstat', os.stat)
106
188
except OSError, e:
107
if e.errno != e.ENOENT:
109
os.rename(path_from, path_to)
189
if e.errno == errno.ENOENT:
192
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
195
def fancy_rename(old, new, rename_func, unlink_func):
196
"""A fancy rename, when you don't have atomic rename.
198
:param old: The old path, to rename from
199
:param new: The new path, to rename to
200
:param rename_func: The potentially non-atomic rename function
201
:param unlink_func: A way to delete the target file if the full rename
204
# sftp rename doesn't allow overwriting, so play tricks:
205
base = os.path.basename(new)
206
dirname = os.path.dirname(new)
207
# callers use different encodings for the paths so the following MUST
208
# respect that. We rely on python upcasting to unicode if new is unicode
209
# and keeping a str if not.
210
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
211
os.getpid(), rand_chars(10))
212
tmp_name = pathjoin(dirname, tmp_name)
214
# Rename the file out of the way, but keep track if it didn't exist
215
# We don't want to grab just any exception
216
# something like EACCES should prevent us from continuing
217
# The downside is that the rename_func has to throw an exception
218
# with an errno = ENOENT, or NoSuchFile
221
rename_func(new, tmp_name)
222
except (errors.NoSuchFile,), e:
225
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
226
# function raises an IOError with errno is None when a rename fails.
227
# This then gets caught here.
228
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
231
if (getattr(e, 'errno', None) is None
232
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
241
# This may throw an exception, in which case success will
243
rename_func(old, new)
245
except (IOError, OSError), e:
246
# source and target may be aliases of each other (e.g. on a
247
# case-insensitive filesystem), so we may have accidentally renamed
248
# source by when we tried to rename target
249
failure_exc = sys.exc_info()
250
if (file_existed and e.errno in (None, errno.ENOENT)
251
and old.lower() == new.lower()):
252
# source and target are the same file on a case-insensitive
253
# filesystem, so we don't generate an exception
257
# If the file used to exist, rename it back into place
258
# otherwise just delete it from the tmp location
260
unlink_func(tmp_name)
262
rename_func(tmp_name, new)
263
if failure_exc is not None:
265
raise failure_exc[0], failure_exc[1], failure_exc[2]
270
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
271
# choke on a Unicode string containing a relative path if
272
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
274
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
275
def _posix_abspath(path):
276
# jam 20060426 rather than encoding to fsencoding
277
# copy posixpath.abspath, but use os.getcwdu instead
278
if not posixpath.isabs(path):
279
path = posixpath.join(getcwd(), path)
280
return posixpath.normpath(path)
283
def _posix_realpath(path):
284
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
287
def _win32_fixdrive(path):
288
"""Force drive letters to be consistent.
290
win32 is inconsistent whether it returns lower or upper case
291
and even if it was consistent the user might type the other
292
so we force it to uppercase
293
running python.exe under cmd.exe return capital C:\\
294
running win32 python inside a cygwin shell returns lowercase c:\\
296
drive, path = ntpath.splitdrive(path)
297
return drive.upper() + path
300
def _win32_abspath(path):
301
# Real ntpath.abspath doesn't have a problem with a unicode cwd
302
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
305
def _win98_abspath(path):
306
"""Return the absolute version of a path.
307
Windows 98 safe implementation (python reimplementation
308
of Win32 API function GetFullPathNameW)
313
# \\HOST\path => //HOST/path
314
# //HOST/path => //HOST/path
315
# path => C:/cwd/path
318
# check for absolute path
319
drive = ntpath.splitdrive(path)[0]
320
if drive == '' and path[:2] not in('//','\\\\'):
322
# we cannot simply os.path.join cwd and path
323
# because os.path.join('C:','/path') produce '/path'
324
# and this is incorrect
325
if path[:1] in ('/','\\'):
326
cwd = ntpath.splitdrive(cwd)[0]
328
path = cwd + '\\' + path
329
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
332
def _win32_realpath(path):
333
# Real ntpath.realpath doesn't have a problem with a unicode cwd
334
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
337
def _win32_pathjoin(*args):
338
return ntpath.join(*args).replace('\\', '/')
341
def _win32_normpath(path):
342
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
346
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
349
def _win32_mkdtemp(*args, **kwargs):
350
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
353
def _win32_rename(old, new):
354
"""We expect to be able to atomically replace 'new' with old.
356
On win32, if new exists, it must be moved out of the way first,
360
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
362
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
363
# If we try to rename a non-existant file onto cwd, we get
364
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
365
# if the old path doesn't exist, sometimes we get EACCES
366
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
372
return unicodedata.normalize('NFC', os.getcwdu())
375
# Default is to just use the python builtins, but these can be rebound on
376
# particular platforms.
377
abspath = _posix_abspath
378
realpath = _posix_realpath
379
pathjoin = os.path.join
380
normpath = os.path.normpath
383
dirname = os.path.dirname
384
basename = os.path.basename
385
split = os.path.split
386
splitext = os.path.splitext
387
# These were already lazily imported into local scope
388
# mkdtemp = tempfile.mkdtemp
389
# rmtree = shutil.rmtree
397
MIN_ABS_PATHLENGTH = 1
400
if sys.platform == 'win32':
401
if win32utils.winver == 'Windows 98':
402
abspath = _win98_abspath
404
abspath = _win32_abspath
405
realpath = _win32_realpath
406
pathjoin = _win32_pathjoin
407
normpath = _win32_normpath
408
getcwd = _win32_getcwd
409
mkdtemp = _win32_mkdtemp
410
rename = _win32_rename
412
from bzrlib import _walkdirs_win32
416
lstat = _walkdirs_win32.lstat
417
fstat = _walkdirs_win32.fstat
418
wrap_stat = _walkdirs_win32.wrap_stat
420
MIN_ABS_PATHLENGTH = 3
422
def _win32_delete_readonly(function, path, excinfo):
423
"""Error handler for shutil.rmtree function [for win32]
424
Helps to remove files and dirs marked as read-only.
426
exception = excinfo[1]
427
if function in (os.remove, os.rmdir) \
428
and isinstance(exception, OSError) \
429
and exception.errno == errno.EACCES:
435
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
436
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
437
return shutil.rmtree(path, ignore_errors, onerror)
439
f = win32utils.get_unicode_argv # special function or None
443
elif sys.platform == 'darwin':
447
def get_terminal_encoding(trace=False):
448
"""Find the best encoding for printing to the screen.
450
This attempts to check both sys.stdout and sys.stdin to see
451
what encoding they are in, and if that fails it falls back to
452
osutils.get_user_encoding().
453
The problem is that on Windows, locale.getpreferredencoding()
454
is not the same encoding as that used by the console:
455
http://mail.python.org/pipermail/python-list/2003-May/162357.html
457
On my standard US Windows XP, the preferred encoding is
458
cp1252, but the console is cp437
460
:param trace: If True trace the selected encoding via mutter().
462
from bzrlib.trace import mutter
463
output_encoding = getattr(sys.stdout, 'encoding', None)
464
if not output_encoding:
465
input_encoding = getattr(sys.stdin, 'encoding', None)
466
if not input_encoding:
467
output_encoding = get_user_encoding()
469
mutter('encoding stdout as osutils.get_user_encoding() %r',
472
output_encoding = input_encoding
474
mutter('encoding stdout as sys.stdin encoding %r',
478
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
479
if output_encoding == 'cp0':
480
# invalid encoding (cp0 means 'no codepage' on Windows)
481
output_encoding = get_user_encoding()
483
mutter('cp0 is invalid encoding.'
484
' encoding stdout as osutils.get_user_encoding() %r',
488
codecs.lookup(output_encoding)
490
sys.stderr.write('bzr: warning:'
491
' unknown terminal encoding %s.\n'
492
' Using encoding %s instead.\n'
493
% (output_encoding, get_user_encoding())
495
output_encoding = get_user_encoding()
497
return output_encoding
500
def normalizepath(f):
501
if getattr(os.path, 'realpath', None) is not None:
505
[p,e] = os.path.split(f)
506
if e == "" or e == "." or e == "..":
509
return pathjoin(F(p), e)
116
513
"""True if f is an accessible directory."""
118
return S_ISDIR(os.lstat(f)[ST_MODE])
515
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
125
521
"""True if f is a regular file."""
127
return S_ISREG(os.lstat(f)[ST_MODE])
523
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
528
"""True if f is a symlink."""
530
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
132
534
def is_inside(dir, fname):
133
535
"""True if fname is inside dir.
537
The parameters should typically be passed to osutils.normpath first, so
538
that . and .. and repeated slashes are eliminated, and the separators
539
are canonical for the platform.
541
The empty string as a dir name is taken as top-of-tree and matches
135
return os.path.commonprefix([dir, fname]) == dir
544
# XXX: Most callers of this can actually do something smarter by
545
# looking at the inventory
555
return fname.startswith(dir)
138
558
def is_inside_any(dir_list, fname):
139
559
"""True if fname is inside any of given dirs."""
140
# quick scan for perfect match
141
if fname in dir_list:
144
560
for dirname in dir_list:
145
561
if is_inside(dirname, fname):
566
def is_inside_or_parent_of_any(dir_list, fname):
567
"""True if fname is a child or a parent of any of the given files."""
568
for dirname in dir_list:
569
if is_inside(dirname, fname) or is_inside(fname, dirname):
574
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
575
report_activity=None, direction='read'):
576
"""Copy contents of one file to another.
578
The read_length can either be -1 to read to end-of-file (EOF) or
579
it can specify the maximum number of bytes to read.
581
The buff_size represents the maximum size for each read operation
582
performed on from_file.
584
:param report_activity: Call this as bytes are read, see
585
Transport._report_activity
586
:param direction: Will be passed to report_activity
588
:return: The number of bytes copied.
592
# read specified number of bytes
594
while read_length > 0:
595
num_bytes_to_read = min(read_length, buff_size)
597
block = from_file.read(num_bytes_to_read)
601
if report_activity is not None:
602
report_activity(len(block), direction)
605
actual_bytes_read = len(block)
606
read_length -= actual_bytes_read
607
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())
611
block = from_file.read(buff_size)
615
if report_activity is not None:
616
report_activity(len(block), direction)
622
def pump_string_file(bytes, file_handle, segment_size=None):
623
"""Write bytes to file_handle in many smaller writes.
625
:param bytes: The string to write.
626
:param file_handle: The file to write to.
628
# Write data in chunks rather than all at once, because very large
629
# writes fail on some platforms (e.g. Windows with SMB mounted
632
segment_size = 5242880 # 5MB
633
segments = range(len(bytes) / segment_size + 1)
634
write = file_handle.write
635
for segment_index in segments:
636
segment = buffer(bytes, segment_index * segment_size, segment_size)
640
def file_iterator(input_file, readsize=32768):
642
b = input_file.read(readsize)
166
if hasattr(f, 'tell'):
649
"""Calculate the hexdigest of an open file.
651
The file cursor should be already at the start.
663
def size_sha_file(f):
664
"""Calculate the size and hexdigest of an open file.
666
The file cursor should be already at the start and
667
the caller is responsible for closing the file afterwards.
678
return size, s.hexdigest()
681
def sha_file_by_name(fname):
682
"""Calculate the SHA1 of a file by reading the full text"""
684
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
687
b = os.read(f, 1<<16)
695
def sha_strings(strings, _factory=sha):
696
"""Return the sha-1 of concatenation of strings"""
698
map(s.update, strings)
702
def sha_string(f, _factory=sha):
703
return _factory(f).hexdigest()
186
706
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]
708
return {'size': len(b),
709
'sha1': sha(b).hexdigest()}
299
712
def compare_files(a, b):
400
raise BzrError("sorry, %r not allowed in path" % f)
927
raise errors.BzrError("sorry, %r not allowed in path" % f)
401
928
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:
937
if (f == '..') or (f is None) or (f == ''):
938
raise errors.BzrError("sorry, %r not allowed in path" % f)
942
def parent_directories(filename):
943
"""Return the list of parent directories, deepest first.
945
For example, parent_directories("a/b/c") -> ["a/b", "a"].
948
parts = splitpath(dirname(filename))
950
parents.append(joinpath(parts))
955
_extension_load_failures = []
958
def failed_to_load_extension(exception):
959
"""Handle failing to load a binary extension.
961
This should be called from the ImportError block guarding the attempt to
962
import the native extension. If this function returns, the pure-Python
963
implementation should be loaded instead::
966
>>> import bzrlib._fictional_extension_pyx
967
>>> except ImportError, e:
968
>>> bzrlib.osutils.failed_to_load_extension(e)
969
>>> import bzrlib._fictional_extension_py
971
# NB: This docstring is just an example, not a doctest, because doctest
972
# currently can't cope with the use of lazy imports in this namespace --
975
# This currently doesn't report the failure at the time it occurs, because
976
# they tend to happen very early in startup when we can't check config
977
# files etc, and also we want to report all failures but not spam the user
979
exception_str = str(exception)
980
if exception_str not in _extension_load_failures:
981
trace.mutter("failed to load compiled extension: %s" % exception_str)
982
_extension_load_failures.append(exception_str)
985
def report_extension_load_failures():
986
if not _extension_load_failures:
988
from bzrlib.config import GlobalConfig
989
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
991
# the warnings framework should by default show this only once
992
from bzrlib.trace import warning
994
"bzr: warning: some compiled extensions could not be loaded; "
995
"see <https://answers.launchpad.net/bzr/+faq/703>")
996
# we no longer show the specific missing extensions here, because it makes
997
# the message too long and scary - see
998
# https://bugs.launchpad.net/bzr/+bug/430529
1002
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1003
except ImportError, e:
1004
failed_to_load_extension(e)
1005
from bzrlib._chunks_to_lines_py import chunks_to_lines
1009
"""Split s into lines, but without removing the newline characters."""
1010
# Trivially convert a fulltext into a 'chunked' representation, and let
1011
# chunks_to_lines do the heavy lifting.
1012
if isinstance(s, str):
1013
# chunks_to_lines only supports 8-bit strings
1014
return chunks_to_lines([s])
1016
return _split_lines(s)
1019
def _split_lines(s):
1020
"""Split s into lines, but without removing the newline characters.
1022
This supports Unicode or plain string objects.
1024
lines = s.split('\n')
1025
result = [line + '\n' for line in lines[:-1]]
1027
result.append(lines[-1])
1031
def hardlinks_good():
1032
return sys.platform not in ('win32', 'cygwin', 'darwin')
1035
def link_or_copy(src, dest):
1036
"""Hardlink a file, or copy it if it can't be hardlinked."""
1037
if not hardlinks_good():
1038
shutil.copyfile(src, dest)
1042
except (OSError, IOError), e:
1043
if e.errno != errno.EXDEV:
1045
shutil.copyfile(src, dest)
1048
def delete_any(path):
1049
"""Delete a file, symlink or directory.
1051
Will delete even if readonly.
1054
_delete_file_or_dir(path)
1055
except (OSError, IOError), e:
1056
if e.errno in (errno.EPERM, errno.EACCES):
1057
# make writable and try again
1060
except (OSError, IOError):
1062
_delete_file_or_dir(path)
1067
def _delete_file_or_dir(path):
1068
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1069
# Forgiveness than Permission (EAFP) because:
1070
# - root can damage a solaris file system by using unlink,
1071
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1072
# EACCES, OSX: EPERM) when invoked on a directory.
1073
if isdir(path): # Takes care of symlinks
1080
if getattr(os, 'symlink', None) is not None:
1086
def has_hardlinks():
1087
if getattr(os, 'link', None) is not None:
1093
def host_os_dereferences_symlinks():
1094
return (has_symlinks()
1095
and sys.platform not in ('cygwin', 'win32'))
1098
def readlink(abspath):
1099
"""Return a string representing the path to which the symbolic link points.
1101
:param abspath: The link absolute unicode path.
1103
This his guaranteed to return the symbolic link in unicode in all python
1106
link = abspath.encode(_fs_enc)
1107
target = os.readlink(link)
1108
target = target.decode(_fs_enc)
1112
def contains_whitespace(s):
1113
"""True if there are any whitespace characters in s."""
1114
# string.whitespace can include '\xa0' in certain locales, because it is
1115
# considered "non-breaking-space" as part of ISO-8859-1. But it
1116
# 1) Isn't a breaking whitespace
1117
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1119
# 3) '\xa0' isn't unicode safe since it is >128.
1121
# This should *not* be a unicode set of characters in case the source
1122
# string is not a Unicode string. We can auto-up-cast the characters since
1123
# they are ascii, but we don't want to auto-up-cast the string in case it
1125
for ch in ' \t\n\r\v\f':
1132
def contains_linebreaks(s):
1133
"""True if there is any vertical whitespace in s."""
1141
def relpath(base, path):
1142
"""Return path relative to base, or raise PathNotChild exception.
1144
The path may be either an absolute path or a path relative to the
1145
current working directory.
1147
os.path.commonprefix (python2.4) has a bad bug that it works just
1148
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1149
avoids that problem.
1151
NOTE: `base` should not have a trailing slash otherwise you'll get
1152
PathNotChild exceptions regardless of `path`.
1155
if len(base) < MIN_ABS_PATHLENGTH:
1156
# must have space for e.g. a drive letter
1157
raise ValueError('%r is too short to calculate a relative path'
1165
if len(head) <= len(base) and head != base:
1166
raise errors.PathNotChild(rp, base)
1169
head, tail = split(head)
1174
return pathjoin(*reversed(s))
1179
def _cicp_canonical_relpath(base, path):
1180
"""Return the canonical path relative to base.
1182
Like relpath, but on case-insensitive-case-preserving file-systems, this
1183
will return the relpath as stored on the file-system rather than in the
1184
case specified in the input string, for all existing portions of the path.
1186
This will cause O(N) behaviour if called for every path in a tree; if you
1187
have a number of paths to convert, you should use canonical_relpaths().
1189
# TODO: it should be possible to optimize this for Windows by using the
1190
# win32 API FindFiles function to look for the specified name - but using
1191
# os.listdir() still gives us the correct, platform agnostic semantics in
1194
rel = relpath(base, path)
1195
# '.' will have been turned into ''
1199
abs_base = abspath(base)
1201
_listdir = os.listdir
1203
# use an explicit iterator so we can easily consume the rest on early exit.
1204
bit_iter = iter(rel.split('/'))
1205
for bit in bit_iter:
1208
next_entries = _listdir(current)
1209
except OSError: # enoent, eperm, etc
1210
# We can't find this in the filesystem, so just append the
1212
current = pathjoin(current, bit, *list(bit_iter))
1214
for look in next_entries:
1215
if lbit == look.lower():
1216
current = pathjoin(current, look)
1219
# got to the end, nothing matched, so we just return the
1220
# non-existing bits as they were specified (the filename may be
1221
# the target of a move, for example).
1222
current = pathjoin(current, bit, *list(bit_iter))
1224
return current[len(abs_base):].lstrip('/')
1226
# XXX - TODO - we need better detection/integration of case-insensitive
1227
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1228
# filesystems), for example, so could probably benefit from the same basic
1229
# support there. For now though, only Windows and OSX get that support, and
1230
# they get it for *all* file-systems!
1231
if sys.platform in ('win32', 'darwin'):
1232
canonical_relpath = _cicp_canonical_relpath
1234
canonical_relpath = relpath
1236
def canonical_relpaths(base, paths):
1237
"""Create an iterable to canonicalize a sequence of relative paths.
1239
The intent is for this implementation to use a cache, vastly speeding
1240
up multiple transformations in the same directory.
1242
# but for now, we haven't optimized...
1243
return [canonical_relpath(base, p) for p in paths]
1246
def decode_filename(filename):
1247
"""Decode the filename using the filesystem encoding
1249
If it is unicode, it is returned.
1250
Otherwise it is decoded from the the filesystem's encoding. If decoding
1251
fails, a errors.BadFilenameEncoding exception is raised.
1253
if type(filename) is unicode:
1256
return filename.decode(_fs_enc)
1257
except UnicodeDecodeError:
1258
raise errors.BadFilenameEncoding(filename, _fs_enc)
1261
def safe_unicode(unicode_or_utf8_string):
1262
"""Coerce unicode_or_utf8_string into unicode.
1264
If it is unicode, it is returned.
1265
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1266
wrapped in a BzrBadParameterNotUnicode exception.
1268
if isinstance(unicode_or_utf8_string, unicode):
1269
return unicode_or_utf8_string
1271
return unicode_or_utf8_string.decode('utf8')
1272
except UnicodeDecodeError:
1273
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1276
def safe_utf8(unicode_or_utf8_string):
1277
"""Coerce unicode_or_utf8_string to a utf8 string.
1279
If it is a str, it is returned.
1280
If it is Unicode, it is encoded into a utf-8 string.
1282
if isinstance(unicode_or_utf8_string, str):
1283
# TODO: jam 20070209 This is overkill, and probably has an impact on
1284
# performance if we are dealing with lots of apis that want a
1287
# Make sure it is a valid utf-8 string
1288
unicode_or_utf8_string.decode('utf-8')
1289
except UnicodeDecodeError:
1290
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1291
return unicode_or_utf8_string
1292
return unicode_or_utf8_string.encode('utf-8')
1295
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1296
' Revision id generators should be creating utf8'
1300
def safe_revision_id(unicode_or_utf8_string, warn=True):
1301
"""Revision ids should now be utf8, but at one point they were unicode.
1303
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1305
:param warn: Functions that are sanitizing user data can set warn=False
1306
:return: None or a utf8 revision id.
1308
if (unicode_or_utf8_string is None
1309
or unicode_or_utf8_string.__class__ == str):
1310
return unicode_or_utf8_string
1312
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1314
return cache_utf8.encode(unicode_or_utf8_string)
1317
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1318
' generators should be creating utf8 file ids.')
1321
def safe_file_id(unicode_or_utf8_string, warn=True):
1322
"""File ids should now be utf8, but at one point they were unicode.
1324
This is the same as safe_utf8, except it uses the cached encode functions
1325
to save a little bit of performance.
1327
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1329
:param warn: Functions that are sanitizing user data can set warn=False
1330
:return: None or a utf8 file id.
1332
if (unicode_or_utf8_string is None
1333
or unicode_or_utf8_string.__class__ == str):
1334
return unicode_or_utf8_string
1336
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1338
return cache_utf8.encode(unicode_or_utf8_string)
1341
_platform_normalizes_filenames = False
1342
if sys.platform == 'darwin':
1343
_platform_normalizes_filenames = True
1346
def normalizes_filenames():
1347
"""Return True if this platform normalizes unicode filenames.
1351
return _platform_normalizes_filenames
1354
def _accessible_normalized_filename(path):
1355
"""Get the unicode normalized path, and if you can access the file.
1357
On platforms where the system normalizes filenames (Mac OSX),
1358
you can access a file by any path which will normalize correctly.
1359
On platforms where the system does not normalize filenames
1360
(everything else), you have to access a file by its exact path.
1362
Internally, bzr only supports NFC normalization, since that is
1363
the standard for XML documents.
1365
So return the normalized path, and a flag indicating if the file
1366
can be accessed by that path.
1369
return unicodedata.normalize('NFC', unicode(path)), True
1372
def _inaccessible_normalized_filename(path):
1373
__doc__ = _accessible_normalized_filename.__doc__
1375
normalized = unicodedata.normalize('NFC', unicode(path))
1376
return normalized, normalized == path
1379
if _platform_normalizes_filenames:
1380
normalized_filename = _accessible_normalized_filename
1382
normalized_filename = _inaccessible_normalized_filename
1385
def set_signal_handler(signum, handler, restart_syscall=True):
1386
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1387
on platforms that support that.
1389
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1390
automatically restart (by calling `signal.siginterrupt(signum,
1391
False)`). May be ignored if the feature is not available on this
1392
platform or Python version.
1396
siginterrupt = signal.siginterrupt
1398
# This python implementation doesn't provide signal support, hence no
1401
except AttributeError:
1402
# siginterrupt doesn't exist on this platform, or for this version
1404
siginterrupt = lambda signum, flag: None
1406
def sig_handler(*args):
1407
# Python resets the siginterrupt flag when a signal is
1408
# received. <http://bugs.python.org/issue8354>
1409
# As a workaround for some cases, set it back the way we want it.
1410
siginterrupt(signum, False)
1411
# Now run the handler function passed to set_signal_handler.
1414
sig_handler = handler
1415
old_handler = signal.signal(signum, sig_handler)
1417
siginterrupt(signum, False)
1421
default_terminal_width = 80
1422
"""The default terminal width for ttys.
1424
This is defined so that higher levels can share a common fallback value when
1425
terminal_width() returns None.
1428
# Keep some state so that terminal_width can detect if _terminal_size has
1429
# returned a different size since the process started. See docstring and
1430
# comments of terminal_width for details.
1431
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1432
_terminal_size_state = 'no_data'
1433
_first_terminal_size = None
1435
def terminal_width():
1436
"""Return terminal width.
1438
None is returned if the width can't established precisely.
1441
- if BZR_COLUMNS is set, returns its value
1442
- if there is no controlling terminal, returns None
1443
- query the OS, if the queried size has changed since the last query,
1445
- if COLUMNS is set, returns its value,
1446
- if the OS has a value (even though it's never changed), return its value.
1448
From there, we need to query the OS to get the size of the controlling
1451
On Unices we query the OS by:
1452
- get termios.TIOCGWINSZ
1453
- if an error occurs or a negative value is obtained, returns None
1455
On Windows we query the OS by:
1456
- win32utils.get_console_size() decides,
1457
- returns None on error (provided default value)
1459
# Note to implementors: if changing the rules for determining the width,
1460
# make sure you've considered the behaviour in these cases:
1461
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1462
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1464
# - (add more interesting cases here, if you find any)
1465
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1466
# but we don't want to register a signal handler because it is impossible
1467
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1468
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1469
# time so we can notice if the reported size has changed, which should have
1472
# If BZR_COLUMNS is set, take it, user is always right
1473
# Except if they specified 0 in which case, impose no limit here
1475
width = int(os.environ['BZR_COLUMNS'])
1476
except (KeyError, ValueError):
1478
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")
1484
isatty = getattr(sys.stdout, 'isatty', None)
1485
if isatty is None or not isatty():
1486
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1490
width, height = os_size = _terminal_size(None, None)
1491
global _first_terminal_size, _terminal_size_state
1492
if _terminal_size_state == 'no_data':
1493
_first_terminal_size = os_size
1494
_terminal_size_state = 'unchanged'
1495
elif (_terminal_size_state == 'unchanged' and
1496
_first_terminal_size != os_size):
1497
_terminal_size_state = 'changed'
1499
# If the OS claims to know how wide the terminal is, and this value has
1500
# ever changed, use that.
1501
if _terminal_size_state == 'changed':
1502
if width is not None and width > 0:
1505
# If COLUMNS is set, use it.
1507
return int(os.environ['COLUMNS'])
1508
except (KeyError, ValueError):
1511
# Finally, use an unchanged size from the OS, if we have one.
1512
if _terminal_size_state == 'unchanged':
1513
if width is not None and width > 0:
1516
# The width could not be determined.
1520
def _win32_terminal_size(width, height):
1521
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1522
return width, height
1525
def _ioctl_terminal_size(width, height):
1527
import struct, fcntl, termios
1528
s = struct.pack('HHHH', 0, 0, 0, 0)
1529
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1530
height, width = struct.unpack('HHHH', x)[0:2]
1531
except (IOError, AttributeError):
1533
return width, height
1535
_terminal_size = None
1536
"""Returns the terminal size as (width, height).
1538
:param width: Default value for width.
1539
:param height: Default value for height.
1541
This is defined specifically for each OS and query the size of the controlling
1542
terminal. If any error occurs, the provided default values should be returned.
1544
if sys.platform == 'win32':
1545
_terminal_size = _win32_terminal_size
1547
_terminal_size = _ioctl_terminal_size
1550
def supports_executable():
1551
return sys.platform != "win32"
1554
def supports_posix_readonly():
1555
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1557
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1558
directory controls creation/deletion, etc.
1560
And under win32, readonly means that the directory itself cannot be
1561
deleted. The contents of a readonly directory can be changed, unlike POSIX
1562
where files in readonly directories cannot be added, deleted or renamed.
1564
return sys.platform != "win32"
1567
def set_or_unset_env(env_variable, value):
1568
"""Modify the environment, setting or removing the env_variable.
1570
:param env_variable: The environment variable in question
1571
:param value: The value to set the environment to. If None, then
1572
the variable will be removed.
1573
:return: The original value of the environment variable.
1575
orig_val = os.environ.get(env_variable)
1577
if orig_val is not None:
1578
del os.environ[env_variable]
1580
if isinstance(value, unicode):
1581
value = value.encode(get_user_encoding())
1582
os.environ[env_variable] = value
1586
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1589
def check_legal_path(path):
1590
"""Check whether the supplied path is legal.
1591
This is only required on Windows, so we don't test on other platforms
1594
if sys.platform != "win32":
1596
if _validWin32PathRE.match(path) is None:
1597
raise errors.IllegalPath(path)
1600
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1602
def _is_error_enotdir(e):
1603
"""Check if this exception represents ENOTDIR.
1605
Unfortunately, python is very inconsistent about the exception
1606
here. The cases are:
1607
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1608
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1609
which is the windows error code.
1610
3) Windows, Python2.5 uses errno == EINVAL and
1611
winerror == ERROR_DIRECTORY
1613
:param e: An Exception object (expected to be OSError with an errno
1614
attribute, but we should be able to cope with anything)
1615
:return: True if this represents an ENOTDIR error. False otherwise.
1617
en = getattr(e, 'errno', None)
1618
if (en == errno.ENOTDIR
1619
or (sys.platform == 'win32'
1620
and (en == _WIN32_ERROR_DIRECTORY
1621
or (en == errno.EINVAL
1622
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")
1628
def walkdirs(top, prefix=""):
1629
"""Yield data about all the directories in a tree.
1631
This yields all the data about the contents of a directory at a time.
1632
After each directory has been yielded, if the caller has mutated the list
1633
to exclude some directories, they are then not descended into.
1635
The data yielded is of the form:
1636
((directory-relpath, directory-path-from-top),
1637
[(relpath, basename, kind, lstat, path-from-top), ...]),
1638
- directory-relpath is the relative path of the directory being returned
1639
with respect to top. prefix is prepended to this.
1640
- directory-path-from-root is the path including top for this directory.
1641
It is suitable for use with os functions.
1642
- relpath is the relative path within the subtree being walked.
1643
- basename is the basename of the path
1644
- kind is the kind of the file now. If unknown then the file is not
1645
present within the tree - but it may be recorded as versioned. See
1647
- lstat is the stat data *if* the file was statted.
1648
- planned, not implemented:
1649
path_from_tree_root is the path from the root of the tree.
1651
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1652
allows one to walk a subtree but get paths that are relative to a tree
1654
:return: an iterator over the dirs.
1656
#TODO there is a bit of a smell where the results of the directory-
1657
# summary in this, and the path from the root, may not agree
1658
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1659
# potentially confusing output. We should make this more robust - but
1660
# not at a speed cost. RBC 20060731
1662
_directory = _directory_kind
1663
_listdir = os.listdir
1664
_kind_from_mode = file_kind_from_stat_mode
1665
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1667
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1668
relroot, _, _, _, top = pending.pop()
1670
relprefix = relroot + u'/'
1673
top_slash = top + u'/'
1676
append = dirblock.append
1678
names = sorted(map(decode_filename, _listdir(top)))
1680
if not _is_error_enotdir(e):
1684
abspath = top_slash + name
1685
statvalue = _lstat(abspath)
1686
kind = _kind_from_mode(statvalue.st_mode)
1687
append((relprefix + name, name, kind, statvalue, abspath))
1688
yield (relroot, top), dirblock
1690
# push the user specified dirs from dirblock
1691
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1694
class DirReader(object):
1695
"""An interface for reading directories."""
1697
def top_prefix_to_starting_dir(self, top, prefix=""):
1698
"""Converts top and prefix to a starting dir entry
1700
:param top: A utf8 path
1701
:param prefix: An optional utf8 path to prefix output relative paths
1703
:return: A tuple starting with prefix, and ending with the native
1706
raise NotImplementedError(self.top_prefix_to_starting_dir)
1708
def read_dir(self, prefix, top):
1709
"""Read a specific dir.
1711
:param prefix: A utf8 prefix to be preprended to the path basenames.
1712
:param top: A natively encoded path to read.
1713
:return: A list of the directories contents. Each item contains:
1714
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1716
raise NotImplementedError(self.read_dir)
1719
_selected_dir_reader = None
1722
def _walkdirs_utf8(top, prefix=""):
1723
"""Yield data about all the directories in a tree.
1725
This yields the same information as walkdirs() only each entry is yielded
1726
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1727
are returned as exact byte-strings.
1729
:return: yields a tuple of (dir_info, [file_info])
1730
dir_info is (utf8_relpath, path-from-top)
1731
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1732
if top is an absolute path, path-from-top is also an absolute path.
1733
path-from-top might be unicode or utf8, but it is the correct path to
1734
pass to os functions to affect the file in question. (such as os.lstat)
1736
global _selected_dir_reader
1737
if _selected_dir_reader is None:
1738
fs_encoding = _fs_enc.upper()
1739
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1740
# Win98 doesn't have unicode apis like FindFirstFileW
1741
# TODO: We possibly could support Win98 by falling back to the
1742
# original FindFirstFile, and using TCHAR instead of WCHAR,
1743
# but that gets a bit tricky, and requires custom compiling
1746
from bzrlib._walkdirs_win32 import Win32ReadDir
1747
_selected_dir_reader = Win32ReadDir()
1750
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1751
# ANSI_X3.4-1968 is a form of ASCII
1753
from bzrlib._readdir_pyx import UTF8DirReader
1754
_selected_dir_reader = UTF8DirReader()
1755
except ImportError, e:
1756
failed_to_load_extension(e)
1759
if _selected_dir_reader is None:
1760
# Fallback to the python version
1761
_selected_dir_reader = UnicodeDirReader()
1763
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1764
# But we don't actually uses 1-3 in pending, so set them to None
1765
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1766
read_dir = _selected_dir_reader.read_dir
1767
_directory = _directory_kind
1769
relroot, _, _, _, top = pending[-1].pop()
1772
dirblock = sorted(read_dir(relroot, top))
1773
yield (relroot, top), dirblock
1774
# push the user specified dirs from dirblock
1775
next = [d for d in reversed(dirblock) if d[2] == _directory]
1777
pending.append(next)
1780
class UnicodeDirReader(DirReader):
1781
"""A dir reader for non-utf8 file systems, which transcodes."""
1783
__slots__ = ['_utf8_encode']
1786
self._utf8_encode = codecs.getencoder('utf8')
1788
def top_prefix_to_starting_dir(self, top, prefix=""):
1789
"""See DirReader.top_prefix_to_starting_dir."""
1790
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1792
def read_dir(self, prefix, top):
1793
"""Read a single directory from a non-utf8 file system.
1795
top, and the abspath element in the output are unicode, all other paths
1796
are utf8. Local disk IO is done via unicode calls to listdir etc.
1798
This is currently the fallback code path when the filesystem encoding is
1799
not UTF-8. It may be better to implement an alternative so that we can
1800
safely handle paths that are not properly decodable in the current
1803
See DirReader.read_dir for details.
1805
_utf8_encode = self._utf8_encode
1807
_listdir = os.listdir
1808
_kind_from_mode = file_kind_from_stat_mode
1811
relprefix = prefix + '/'
1814
top_slash = top + u'/'
1817
append = dirblock.append
1818
for name in sorted(_listdir(top)):
1820
name_utf8 = _utf8_encode(name)[0]
1821
except UnicodeDecodeError:
1822
raise errors.BadFilenameEncoding(
1823
_utf8_encode(relprefix)[0] + name, _fs_enc)
1824
abspath = top_slash + name
1825
statvalue = _lstat(abspath)
1826
kind = _kind_from_mode(statvalue.st_mode)
1827
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1831
def copy_tree(from_path, to_path, handlers={}):
1832
"""Copy all of the entries in from_path into to_path.
1834
:param from_path: The base directory to copy.
1835
:param to_path: The target directory. If it does not exist, it will
1837
:param handlers: A dictionary of functions, which takes a source and
1838
destinations for files, directories, etc.
1839
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1840
'file', 'directory', and 'symlink' should always exist.
1841
If they are missing, they will be replaced with 'os.mkdir()',
1842
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1844
# Now, just copy the existing cached tree to the new location
1845
# We use a cheap trick here.
1846
# Absolute paths are prefixed with the first parameter
1847
# relative paths are prefixed with the second.
1848
# So we can get both the source and target returned
1849
# without any extra work.
1851
def copy_dir(source, dest):
1854
def copy_link(source, dest):
1855
"""Copy the contents of a symlink"""
1856
link_to = os.readlink(source)
1857
os.symlink(link_to, dest)
1859
real_handlers = {'file':shutil.copy2,
1860
'symlink':copy_link,
1861
'directory':copy_dir,
1863
real_handlers.update(handlers)
1865
if not os.path.exists(to_path):
1866
real_handlers['directory'](from_path, to_path)
1868
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1869
for relpath, name, kind, st, abspath in entries:
1870
real_handlers[kind](abspath, relpath)
1873
def copy_ownership_from_path(dst, src=None):
1874
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1876
If src is None, the containing directory is used as source. If chown
1877
fails, the error is ignored and a warning is printed.
1879
chown = getattr(os, 'chown', None)
1884
src = os.path.dirname(dst)
1890
chown(dst, s.st_uid, s.st_gid)
1893
'Unable to copy ownership from "%s" to "%s". '
1894
'You may want to set it manually.', src, dst)
1895
trace.log_exception_quietly()
1898
def path_prefix_key(path):
1899
"""Generate a prefix-order path key for path.
1901
This can be used to sort paths in the same way that walkdirs does.
1903
return (dirname(path) , path)
1906
def compare_paths_prefix_order(path_a, path_b):
1907
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1908
key_a = path_prefix_key(path_a)
1909
key_b = path_prefix_key(path_b)
1910
return cmp(key_a, key_b)
1913
_cached_user_encoding = None
1916
def get_user_encoding(use_cache=True):
1917
"""Find out what the preferred user encoding is.
1919
This is generally the encoding that is used for command line parameters
1920
and file contents. This may be different from the terminal encoding
1921
or the filesystem encoding.
1923
:param use_cache: Enable cache for detected encoding.
1924
(This parameter is turned on by default,
1925
and required only for selftesting)
1927
:return: A string defining the preferred user encoding
1929
global _cached_user_encoding
1930
if _cached_user_encoding is not None and use_cache:
1931
return _cached_user_encoding
1933
if sys.platform == 'darwin':
1934
# python locale.getpreferredencoding() always return
1935
# 'mac-roman' on darwin. That's a lie.
1936
sys.platform = 'posix'
1938
if os.environ.get('LANG', None) is None:
1939
# If LANG is not set, we end up with 'ascii', which is bad
1940
# ('mac-roman' is more than ascii), so we set a default which
1941
# will give us UTF-8 (which appears to work in all cases on
1942
# OSX). Users are still free to override LANG of course, as
1943
# long as it give us something meaningful. This work-around
1944
# *may* not be needed with python 3k and/or OSX 10.5, but will
1945
# work with them too -- vila 20080908
1946
os.environ['LANG'] = 'en_US.UTF-8'
1949
sys.platform = 'darwin'
1954
user_encoding = locale.getpreferredencoding()
1955
except locale.Error, e:
1956
sys.stderr.write('bzr: warning: %s\n'
1957
' Could not determine what text encoding to use.\n'
1958
' This error usually means your Python interpreter\n'
1959
' doesn\'t support the locale set by $LANG (%s)\n'
1960
" Continuing with ascii encoding.\n"
1961
% (e, os.environ.get('LANG')))
1962
user_encoding = 'ascii'
1964
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1965
# treat that as ASCII, and not support printing unicode characters to the
1968
# For python scripts run under vim, we get '', so also treat that as ASCII
1969
if user_encoding in (None, 'cp0', ''):
1970
user_encoding = 'ascii'
1974
codecs.lookup(user_encoding)
1976
sys.stderr.write('bzr: warning:'
1977
' unknown encoding %s.'
1978
' Continuing with ascii encoding.\n'
1981
user_encoding = 'ascii'
1984
_cached_user_encoding = user_encoding
1986
return user_encoding
1989
def get_diff_header_encoding():
1990
return get_terminal_encoding()
1993
def get_host_name():
1994
"""Return the current unicode host name.
1996
This is meant to be used in place of socket.gethostname() because that
1997
behaves inconsistently on different platforms.
1999
if sys.platform == "win32":
2001
return win32utils.get_host_name()
2004
return socket.gethostname().decode(get_user_encoding())
2007
# We must not read/write any more than 64k at a time from/to a socket so we
2008
# don't risk "no buffer space available" errors on some platforms. Windows in
2009
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2011
MAX_SOCKET_CHUNK = 64 * 1024
2013
_end_of_stream_errors = [errno.ECONNRESET]
2014
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2015
_eno = getattr(errno, _eno, None)
2016
if _eno is not None:
2017
_end_of_stream_errors.append(_eno)
2021
def read_bytes_from_socket(sock, report_activity=None,
2022
max_read_size=MAX_SOCKET_CHUNK):
2023
"""Read up to max_read_size of bytes from sock and notify of progress.
2025
Translates "Connection reset by peer" into file-like EOF (return an
2026
empty string rather than raise an error), and repeats the recv if
2027
interrupted by a signal.
2031
bytes = sock.recv(max_read_size)
2032
except socket.error, e:
2034
if eno in _end_of_stream_errors:
2035
# The connection was closed by the other side. Callers expect
2036
# an empty string to signal end-of-stream.
2038
elif eno == errno.EINTR:
2039
# Retry the interrupted recv.
2043
if report_activity is not None:
2044
report_activity(len(bytes), 'read')
2048
def recv_all(socket, count):
2049
"""Receive an exact number of bytes.
2051
Regular Socket.recv() may return less than the requested number of bytes,
2052
depending on what's in the OS buffer. MSG_WAITALL is not available
2053
on all platforms, but this should work everywhere. This will return
2054
less than the requested amount if the remote end closes.
2056
This isn't optimized and is intended mostly for use in testing.
2059
while len(b) < count:
2060
new = read_bytes_from_socket(socket, None, count - len(b))
2067
def send_all(sock, bytes, report_activity=None):
2068
"""Send all bytes on a socket.
2070
Breaks large blocks in smaller chunks to avoid buffering limitations on
2071
some platforms, and catches EINTR which may be thrown if the send is
2072
interrupted by a signal.
2074
This is preferred to socket.sendall(), because it avoids portability bugs
2075
and provides activity reporting.
2077
:param report_activity: Call this as bytes are read, see
2078
Transport._report_activity
2081
byte_count = len(bytes)
2082
while sent_total < byte_count:
2084
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2085
except socket.error, e:
2086
if e.args[0] != errno.EINTR:
2090
report_activity(sent, 'write')
2093
def connect_socket(address):
2094
# Slight variation of the socket.create_connection() function (provided by
2095
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2096
# provide it for previous python versions. Also, we don't use the timeout
2097
# parameter (provided by the python implementation) so we don't implement
2099
err = socket.error('getaddrinfo returns an empty list')
2100
host, port = address
2101
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2102
af, socktype, proto, canonname, sa = res
2105
sock = socket.socket(af, socktype, proto)
2109
except socket.error, err:
2110
# 'err' is now the most recent error
2111
if sock is not None:
2116
def dereference_path(path):
2117
"""Determine the real path to a file.
2119
All parent elements are dereferenced. But the file itself is not
2121
:param path: The original path. May be absolute or relative.
2122
:return: the real path *to* the file
2124
parent, base = os.path.split(path)
2125
# The pathjoin for '.' is a workaround for Python bug #1213894.
2126
# (initial path components aren't dereferenced)
2127
return pathjoin(realpath(pathjoin('.', parent)), base)
2130
def supports_mapi():
2131
"""Return True if we can use MAPI to launch a mail client."""
2132
return sys.platform == "win32"
2135
def resource_string(package, resource_name):
2136
"""Load a resource from a package and return it as a string.
2138
Note: Only packages that start with bzrlib are currently supported.
2140
This is designed to be a lightweight implementation of resource
2141
loading in a way which is API compatible with the same API from
2143
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2144
If and when pkg_resources becomes a standard library, this routine
2147
# Check package name is within bzrlib
2148
if package == "bzrlib":
2149
resource_relpath = resource_name
2150
elif package.startswith("bzrlib."):
2151
package = package[len("bzrlib."):].replace('.', os.sep)
2152
resource_relpath = pathjoin(package, resource_name)
2154
raise errors.BzrError('resource package %s not in bzrlib' % package)
2156
# Map the resource to a file and read its contents
2157
base = dirname(bzrlib.__file__)
2158
if getattr(sys, 'frozen', None): # bzr.exe
2159
base = abspath(pathjoin(base, '..', '..'))
2160
f = file(pathjoin(base, resource_relpath), "rU")
522
# delete the msg file in any case
523
try: os.unlink(msgfilename)
2166
def file_kind_from_stat_mode_thunk(mode):
2167
global file_kind_from_stat_mode
2168
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2170
from bzrlib._readdir_pyx import UTF8DirReader
2171
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2172
except ImportError, e:
2173
# This is one time where we won't warn that an extension failed to
2174
# load. The extension is never available on Windows anyway.
2175
from bzrlib._readdir_py import (
2176
_kind_from_mode as file_kind_from_stat_mode
2178
return file_kind_from_stat_mode(mode)
2179
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2182
def file_kind(f, _lstat=os.lstat):
2184
return file_kind_from_stat_mode(_lstat(f).st_mode)
2186
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2187
raise errors.NoSuchFile(f)
2191
def until_no_eintr(f, *a, **kw):
2192
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2194
WARNING: you must be certain that it is safe to retry the call repeatedly
2195
if EINTR does occur. This is typically only true for low-level operations
2196
like os.read. If in any doubt, don't use this.
2198
Keep in mind that this is not a complete solution to EINTR. There is
2199
probably code in the Python standard library and other dependencies that
2200
may encounter EINTR if a signal arrives (and there is signal handler for
2201
that signal). So this function can reduce the impact for IO that bzrlib
2202
directly controls, but it is not a complete solution.
2204
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2208
except (IOError, OSError), e:
2209
if e.errno == errno.EINTR:
2214
@deprecated_function(deprecated_in((2, 2, 0)))
2215
def re_compile_checked(re_string, flags=0, where=""):
2216
"""Return a compiled re, or raise a sensible error.
2218
This should only be used when compiling user-supplied REs.
2220
:param re_string: Text form of regular expression.
2221
:param flags: eg re.IGNORECASE
2222
:param where: Message explaining to the user the context where
2223
it occurred, eg 'log search filter'.
2225
# from https://bugs.launchpad.net/bzr/+bug/251352
2227
re_obj = re.compile(re_string, flags)
2230
except errors.InvalidPattern, e:
2232
where = ' in ' + where
2233
# despite the name 'error' is a type
2234
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2238
if sys.platform == "win32":
2241
return msvcrt.getch()
2246
fd = sys.stdin.fileno()
2247
settings = termios.tcgetattr(fd)
2250
ch = sys.stdin.read(1)
2252
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2255
if sys.platform == 'linux2':
2256
def _local_concurrency():
2258
return os.sysconf('SC_NPROCESSORS_ONLN')
2259
except (ValueError, OSError, AttributeError):
2261
elif sys.platform == 'darwin':
2262
def _local_concurrency():
2263
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2264
stdout=subprocess.PIPE).communicate()[0]
2265
elif "bsd" in sys.platform:
2266
def _local_concurrency():
2267
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2268
stdout=subprocess.PIPE).communicate()[0]
2269
elif sys.platform == 'sunos5':
2270
def _local_concurrency():
2271
return subprocess.Popen(['psrinfo', '-p',],
2272
stdout=subprocess.PIPE).communicate()[0]
2273
elif sys.platform == "win32":
2274
def _local_concurrency():
2275
# This appears to return the number of cores.
2276
return os.environ.get('NUMBER_OF_PROCESSORS')
2278
def _local_concurrency():
2283
_cached_local_concurrency = None
2285
def local_concurrency(use_cache=True):
2286
"""Return how many processes can be run concurrently.
2288
Rely on platform specific implementations and default to 1 (one) if
2289
anything goes wrong.
2291
global _cached_local_concurrency
2293
if _cached_local_concurrency is not None and use_cache:
2294
return _cached_local_concurrency
2296
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2297
if concurrency is None:
2299
import multiprocessing
2301
# multiprocessing is only available on Python >= 2.6
2303
concurrency = _local_concurrency()
2304
except (OSError, IOError):
2307
concurrency = multiprocessing.cpu_count()
2309
concurrency = int(concurrency)
2310
except (TypeError, ValueError):
2313
_cached_concurrency = concurrency
2317
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2318
"""A stream writer that doesn't decode str arguments."""
2320
def __init__(self, encode, stream, errors='strict'):
2321
codecs.StreamWriter.__init__(self, stream, errors)
2322
self.encode = encode
2324
def write(self, object):
2325
if type(object) is str:
2326
self.stream.write(object)
2328
data, _ = self.encode(object, self.errors)
2329
self.stream.write(data)
2331
if sys.platform == 'win32':
2332
def open_file(filename, mode='r', bufsize=-1):
2333
"""This function is used to override the ``open`` builtin.
2335
But it uses O_NOINHERIT flag so the file handle is not inherited by
2336
child processes. Deleting or renaming a closed file opened with this
2337
function is not blocking child processes.
2339
writing = 'w' in mode
2340
appending = 'a' in mode
2341
updating = '+' in mode
2342
binary = 'b' in mode
2345
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2346
# for flags for each modes.
2356
flags |= os.O_WRONLY
2357
flags |= os.O_CREAT | os.O_TRUNC
2362
flags |= os.O_WRONLY
2363
flags |= os.O_CREAT | os.O_APPEND
2368
flags |= os.O_RDONLY
2370
return os.fdopen(os.open(filename, flags), mode, bufsize)
2375
def getuser_unicode():
2376
"""Return the username as unicode.
2379
user_encoding = get_user_encoding()
2380
username = getpass.getuser().decode(user_encoding)
2381
except UnicodeDecodeError:
2382
raise errors.BzrError("Can't decode username as %s." % \
2387
def available_backup_name(base, exists):
2388
"""Find a non-existing backup file name.
2390
This will *not* create anything, this only return a 'free' entry. This
2391
should be used for checking names in a directory below a locked
2392
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2393
Leap) and generally discouraged.
2395
:param base: The base name.
2397
:param exists: A callable returning True if the path parameter exists.
2400
name = "%s.~%d~" % (base, counter)
2403
name = "%s.~%d~" % (base, counter)
2407
def set_fd_cloexec(fd):
2408
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2409
support for this is not available.
2413
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2414
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2415
except (ImportError, AttributeError):
2416
# Either the fcntl module or specific constants are not present
2420
def find_executable_on_path(name):
2421
"""Finds an executable on the PATH.
2423
On Windows, this will try to append each extension in the PATHEXT
2424
environment variable to the name, if it cannot be found with the name
2427
:param name: The base name of the executable.
2428
:return: The path to the executable found or None.
2430
path = os.environ.get('PATH')
2433
path = path.split(os.pathsep)
2434
if sys.platform == 'win32':
2435
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2436
exts = [ext.lower() for ext in exts]
2437
base, ext = os.path.splitext(name)
2439
if ext.lower() not in exts:
2447
f = os.path.join(d, name) + ext
2448
if os.access(f, os.X_OK):
added
removed
bzrlib/osutils.py
