1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005-2011 Canonical Ltd
5
3
# This program is free software; you can redistribute it and/or modify
6
4
# it under the terms of the GNU General Public License as published by
7
5
# the Free Software Foundation; either version 2 of the License, or
8
6
# (at your option) any later version.
10
8
# This program is distributed in the hope that it will be useful,
11
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
11
# GNU General Public License for more details.
15
13
# You should have received a copy of the GNU General Public License
16
14
# along with this program; if not, write to the Free Software
17
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
import os, types, re, time, errno, sys
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
22
from errors import bailout, BzrError
23
from trace import mutter
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
27
from datetime import datetime
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
"""Return shell-quoted filename"""
43
## We could be a bit more terse by using double-quotes etc
44
f = _QUOTE_RE.sub(r'\\\1', f)
51
mode = os.lstat(f)[ST_MODE]
145
"""Return a quoted filename filename
147
This previously used backslash quoting, but that works poorly on
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.,:/\\_~-])')
154
if _QUOTE_RE.search(f):
59
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': '+',
62
180
def kind_marker(kind):
182
return _kind_marker_map[kind]
184
# Slightly faster than using .get(, '') when the common case is that
65
elif kind == 'directory':
67
elif kind == 'symlink':
70
raise BzrError('invalid file kind %r' % kind)
189
lexists = getattr(os.path, 'lexists', None)
193
stat = getattr(os, 'lstat', os.stat)
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
402
MIN_ABS_PATHLENGTH = 1
405
if sys.platform == 'win32':
406
if win32utils.winver == 'Windows 98':
407
abspath = _win98_abspath
409
abspath = _win32_abspath
410
realpath = _win32_realpath
411
pathjoin = _win32_pathjoin
412
normpath = _win32_normpath
413
getcwd = _win32_getcwd
414
mkdtemp = _win32_mkdtemp
415
rename = _win32_rename
417
from bzrlib import _walkdirs_win32
421
lstat = _walkdirs_win32.lstat
422
fstat = _walkdirs_win32.fstat
423
wrap_stat = _walkdirs_win32.wrap_stat
425
MIN_ABS_PATHLENGTH = 3
427
def _win32_delete_readonly(function, path, excinfo):
428
"""Error handler for shutil.rmtree function [for win32]
429
Helps to remove files and dirs marked as read-only.
431
exception = excinfo[1]
432
if function in (os.remove, os.rmdir) \
433
and isinstance(exception, OSError) \
434
and exception.errno == errno.EACCES:
440
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
441
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
442
return shutil.rmtree(path, ignore_errors, onerror)
444
f = win32utils.get_unicode_argv # special function or None
448
elif sys.platform == 'darwin':
452
def get_terminal_encoding(trace=False):
453
"""Find the best encoding for printing to the screen.
455
This attempts to check both sys.stdout and sys.stdin to see
456
what encoding they are in, and if that fails it falls back to
457
osutils.get_user_encoding().
458
The problem is that on Windows, locale.getpreferredencoding()
459
is not the same encoding as that used by the console:
460
http://mail.python.org/pipermail/python-list/2003-May/162357.html
462
On my standard US Windows XP, the preferred encoding is
463
cp1252, but the console is cp437
465
:param trace: If True trace the selected encoding via mutter().
467
from bzrlib.trace import mutter
468
output_encoding = getattr(sys.stdout, 'encoding', None)
469
if not output_encoding:
470
input_encoding = getattr(sys.stdin, 'encoding', None)
471
if not input_encoding:
472
output_encoding = get_user_encoding()
474
mutter('encoding stdout as osutils.get_user_encoding() %r',
477
output_encoding = input_encoding
479
mutter('encoding stdout as sys.stdin encoding %r',
483
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
484
if output_encoding == 'cp0':
485
# invalid encoding (cp0 means 'no codepage' on Windows)
486
output_encoding = get_user_encoding()
488
mutter('cp0 is invalid encoding.'
489
' encoding stdout as osutils.get_user_encoding() %r',
493
codecs.lookup(output_encoding)
495
sys.stderr.write('bzr: warning:'
496
' unknown terminal encoding %s.\n'
497
' Using encoding %s instead.\n'
498
% (output_encoding, get_user_encoding())
500
output_encoding = get_user_encoding()
502
return output_encoding
505
def normalizepath(f):
506
if getattr(os.path, 'realpath', None) is not None:
510
[p,e] = os.path.split(f)
511
if e == "" or e == "." or e == "..":
514
return pathjoin(F(p), e)
75
518
"""True if f is an accessible directory."""
77
return S_ISDIR(os.lstat(f)[ST_MODE])
520
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
84
526
"""True if f is a regular file."""
86
return S_ISREG(os.lstat(f)[ST_MODE])
528
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
533
"""True if f is a symlink."""
535
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
91
539
def is_inside(dir, fname):
92
540
"""True if fname is inside dir.
542
The parameters should typically be passed to osutils.normpath first, so
543
that . and .. and repeated slashes are eliminated, and the separators
544
are canonical for the platform.
546
The empty string as a dir name is taken as top-of-tree and matches
94
return os.path.commonprefix([dir, fname]) == dir
549
# XXX: Most callers of this can actually do something smarter by
550
# looking at the inventory
560
return fname.startswith(dir)
97
563
def is_inside_any(dir_list, fname):
98
564
"""True if fname is inside any of given dirs."""
99
# quick scan for perfect match
100
if fname in dir_list:
103
565
for dirname in dir_list:
104
566
if is_inside(dirname, fname):
571
def is_inside_or_parent_of_any(dir_list, fname):
572
"""True if fname is a child or a parent of any of the given files."""
573
for dirname in dir_list:
574
if is_inside(dirname, fname) or is_inside(fname, dirname):
579
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
580
report_activity=None, direction='read'):
581
"""Copy contents of one file to another.
583
The read_length can either be -1 to read to end-of-file (EOF) or
584
it can specify the maximum number of bytes to read.
586
The buff_size represents the maximum size for each read operation
587
performed on from_file.
589
:param report_activity: Call this as bytes are read, see
590
Transport._report_activity
591
:param direction: Will be passed to report_activity
593
:return: The number of bytes copied.
597
# read specified number of bytes
599
while read_length > 0:
600
num_bytes_to_read = min(read_length, buff_size)
602
block = from_file.read(num_bytes_to_read)
606
if report_activity is not None:
607
report_activity(len(block), direction)
610
actual_bytes_read = len(block)
611
read_length -= actual_bytes_read
612
length += actual_bytes_read
110
def pumpfile(fromfile, tofile):
111
"""Copy contents of one file to another."""
112
tofile.write(fromfile.read())
116
"""Return a new UUID"""
118
return file('/proc/sys/kernel/random/uuid').readline().rstrip('\n')
120
return chomp(os.popen('uuidgen').readline())
616
block = from_file.read(buff_size)
620
if report_activity is not None:
621
report_activity(len(block), direction)
627
def pump_string_file(bytes, file_handle, segment_size=None):
628
"""Write bytes to file_handle in many smaller writes.
630
:param bytes: The string to write.
631
:param file_handle: The file to write to.
633
# Write data in chunks rather than all at once, because very large
634
# writes fail on some platforms (e.g. Windows with SMB mounted
637
segment_size = 5242880 # 5MB
638
segments = range(len(bytes) / segment_size + 1)
639
write = file_handle.write
640
for segment_index in segments:
641
segment = buffer(bytes, segment_index * segment_size, segment_size)
645
def file_iterator(input_file, readsize=32768):
647
b = input_file.read(readsize)
125
if hasattr(f, 'tell'):
654
"""Calculate the hexdigest of an open file.
656
The file cursor should be already at the start.
668
def size_sha_file(f):
669
"""Calculate the size and hexdigest of an open file.
671
The file cursor should be already at the start and
672
the caller is responsible for closing the file afterwards.
683
return size, s.hexdigest()
686
def sha_file_by_name(fname):
687
"""Calculate the SHA1 of a file by reading the full text"""
689
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
692
b = os.read(f, 1<<16)
700
def sha_strings(strings, _factory=sha):
701
"""Return the sha-1 of concatenation of strings"""
703
map(s.update, strings)
707
def sha_string(f, _factory=sha):
708
return _factory(f).hexdigest()
145
711
def fingerprint_file(f):
151
return {'size': size,
152
'sha1': s.hexdigest()}
156
"""Return per-user configuration directory.
158
By default this is ~/.bzr.conf/
160
TODO: Global option --config-dir to override this.
162
return os.path.expanduser("~/.bzr.conf")
166
"""Calculate automatic user identification.
168
Returns (realname, email).
170
Only used when none is set in the environment or the id file.
172
This previously used the FQDN as the default domain, but that can
173
be very slow on machines where DNS is broken. So now we simply
178
# XXX: Any good way to get real user name on win32?
183
w = pwd.getpwuid(uid)
184
gecos = w.pw_gecos.decode(bzrlib.user_encoding)
185
username = w.pw_name.decode(bzrlib.user_encoding)
186
comma = gecos.find(',')
190
realname = gecos[:comma]
196
realname = username = getpass.getuser().decode(bzrlib.user_encoding)
198
return realname, (username + '@' + socket.gethostname())
202
"""Return the full user id from a file or environment variable.
204
TODO: Allow taking this from a file in the branch directory too
205
for per-branch ids."""
206
v = os.environ.get('BZREMAIL')
208
return v.decode(bzrlib.user_encoding)
211
return (open(os.path.join(config_dir(), "email"))
213
.decode(bzrlib.user_encoding)
216
if e.errno != errno.ENOENT:
219
v = os.environ.get('EMAIL')
221
return v.decode(bzrlib.user_encoding)
227
"""Return email-style username.
229
Something similar to 'Martin Pool <mbp@sourcefrog.net>'
231
TODO: Check it's reasonably well-formed.
237
name, email = _auto_user_id()
239
return '%s <%s>' % (name, email)
244
_EMAIL_RE = re.compile(r'[\w+.-]+@[\w+.-]+')
246
"""Return just the email component of a username."""
249
m = _EMAIL_RE.search(e)
251
bailout("%r doesn't seem to contain a reasonable email address" % e)
254
return _auto_user_id()[1]
713
return {'size': len(b),
714
'sha1': sha(b).hexdigest()}
258
717
def compare_files(a, b):
359
bailout("sorry, %r not allowed in path" % f)
932
raise errors.BzrError("sorry, %r not allowed in path" % f)
360
933
elif (f == '.') or (f == ''):
367
assert isinstance(p, list)
369
if (f == '..') or (f == None) or (f == ''):
370
bailout("sorry, %r not allowed in path" % f)
371
return os.path.join(*p)
374
def appendpath(p1, p2):
378
return os.path.join(p1, p2)
381
def extern_command(cmd, ignore_errors = False):
382
mutter('external command: %s' % `cmd`)
384
if not ignore_errors:
385
bailout('command failed')
942
if (f == '..') or (f is None) or (f == ''):
943
raise errors.BzrError("sorry, %r not allowed in path" % f)
947
def parent_directories(filename):
948
"""Return the list of parent directories, deepest first.
950
For example, parent_directories("a/b/c") -> ["a/b", "a"].
953
parts = splitpath(dirname(filename))
955
parents.append(joinpath(parts))
960
_extension_load_failures = []
963
def failed_to_load_extension(exception):
964
"""Handle failing to load a binary extension.
966
This should be called from the ImportError block guarding the attempt to
967
import the native extension. If this function returns, the pure-Python
968
implementation should be loaded instead::
971
>>> import bzrlib._fictional_extension_pyx
972
>>> except ImportError, e:
973
>>> bzrlib.osutils.failed_to_load_extension(e)
974
>>> import bzrlib._fictional_extension_py
976
# NB: This docstring is just an example, not a doctest, because doctest
977
# currently can't cope with the use of lazy imports in this namespace --
980
# This currently doesn't report the failure at the time it occurs, because
981
# they tend to happen very early in startup when we can't check config
982
# files etc, and also we want to report all failures but not spam the user
984
exception_str = str(exception)
985
if exception_str not in _extension_load_failures:
986
trace.mutter("failed to load compiled extension: %s" % exception_str)
987
_extension_load_failures.append(exception_str)
990
def report_extension_load_failures():
991
if not _extension_load_failures:
993
from bzrlib.config import GlobalConfig
994
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
996
# the warnings framework should by default show this only once
997
from bzrlib.trace import warning
999
"bzr: warning: some compiled extensions could not be loaded; "
1000
"see <https://answers.launchpad.net/bzr/+faq/703>")
1001
# we no longer show the specific missing extensions here, because it makes
1002
# the message too long and scary - see
1003
# https://bugs.launchpad.net/bzr/+bug/430529
1007
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1008
except ImportError, e:
1009
failed_to_load_extension(e)
1010
from bzrlib._chunks_to_lines_py import chunks_to_lines
1014
"""Split s into lines, but without removing the newline characters."""
1015
# Trivially convert a fulltext into a 'chunked' representation, and let
1016
# chunks_to_lines do the heavy lifting.
1017
if isinstance(s, str):
1018
# chunks_to_lines only supports 8-bit strings
1019
return chunks_to_lines([s])
1021
return _split_lines(s)
1024
def _split_lines(s):
1025
"""Split s into lines, but without removing the newline characters.
1027
This supports Unicode or plain string objects.
1029
lines = s.split('\n')
1030
result = [line + '\n' for line in lines[:-1]]
1032
result.append(lines[-1])
1036
def hardlinks_good():
1037
return sys.platform not in ('win32', 'cygwin', 'darwin')
1040
def link_or_copy(src, dest):
1041
"""Hardlink a file, or copy it if it can't be hardlinked."""
1042
if not hardlinks_good():
1043
shutil.copyfile(src, dest)
1047
except (OSError, IOError), e:
1048
if e.errno != errno.EXDEV:
1050
shutil.copyfile(src, dest)
1053
def delete_any(path):
1054
"""Delete a file, symlink or directory.
1056
Will delete even if readonly.
1059
_delete_file_or_dir(path)
1060
except (OSError, IOError), e:
1061
if e.errno in (errno.EPERM, errno.EACCES):
1062
# make writable and try again
1065
except (OSError, IOError):
1067
_delete_file_or_dir(path)
1072
def _delete_file_or_dir(path):
1073
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1074
# Forgiveness than Permission (EAFP) because:
1075
# - root can damage a solaris file system by using unlink,
1076
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1077
# EACCES, OSX: EPERM) when invoked on a directory.
1078
if isdir(path): # Takes care of symlinks
1085
if getattr(os, 'symlink', None) is not None:
1091
def has_hardlinks():
1092
if getattr(os, 'link', None) is not None:
1098
def host_os_dereferences_symlinks():
1099
return (has_symlinks()
1100
and sys.platform not in ('cygwin', 'win32'))
1103
def readlink(abspath):
1104
"""Return a string representing the path to which the symbolic link points.
1106
:param abspath: The link absolute unicode path.
1108
This his guaranteed to return the symbolic link in unicode in all python
1111
link = abspath.encode(_fs_enc)
1112
target = os.readlink(link)
1113
target = target.decode(_fs_enc)
1117
def contains_whitespace(s):
1118
"""True if there are any whitespace characters in s."""
1119
# string.whitespace can include '\xa0' in certain locales, because it is
1120
# considered "non-breaking-space" as part of ISO-8859-1. But it
1121
# 1) Isn't a breaking whitespace
1122
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1124
# 3) '\xa0' isn't unicode safe since it is >128.
1126
# This should *not* be a unicode set of characters in case the source
1127
# string is not a Unicode string. We can auto-up-cast the characters since
1128
# they are ascii, but we don't want to auto-up-cast the string in case it
1130
for ch in ' \t\n\r\v\f':
1137
def contains_linebreaks(s):
1138
"""True if there is any vertical whitespace in s."""
1146
def relpath(base, path):
1147
"""Return path relative to base, or raise PathNotChild exception.
1149
The path may be either an absolute path or a path relative to the
1150
current working directory.
1152
os.path.commonprefix (python2.4) has a bad bug that it works just
1153
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1154
avoids that problem.
1156
NOTE: `base` should not have a trailing slash otherwise you'll get
1157
PathNotChild exceptions regardless of `path`.
1160
if len(base) < MIN_ABS_PATHLENGTH:
1161
# must have space for e.g. a drive letter
1162
raise ValueError('%r is too short to calculate a relative path'
1170
if len(head) <= len(base) and head != base:
1171
raise errors.PathNotChild(rp, base)
1174
head, tail = split(head)
1179
return pathjoin(*reversed(s))
1184
def _cicp_canonical_relpath(base, path):
1185
"""Return the canonical path relative to base.
1187
Like relpath, but on case-insensitive-case-preserving file-systems, this
1188
will return the relpath as stored on the file-system rather than in the
1189
case specified in the input string, for all existing portions of the path.
1191
This will cause O(N) behaviour if called for every path in a tree; if you
1192
have a number of paths to convert, you should use canonical_relpaths().
1194
# TODO: it should be possible to optimize this for Windows by using the
1195
# win32 API FindFiles function to look for the specified name - but using
1196
# os.listdir() still gives us the correct, platform agnostic semantics in
1199
rel = relpath(base, path)
1200
# '.' will have been turned into ''
1204
abs_base = abspath(base)
1206
_listdir = os.listdir
1208
# use an explicit iterator so we can easily consume the rest on early exit.
1209
bit_iter = iter(rel.split('/'))
1210
for bit in bit_iter:
1213
next_entries = _listdir(current)
1214
except OSError: # enoent, eperm, etc
1215
# We can't find this in the filesystem, so just append the
1217
current = pathjoin(current, bit, *list(bit_iter))
1219
for look in next_entries:
1220
if lbit == look.lower():
1221
current = pathjoin(current, look)
1224
# got to the end, nothing matched, so we just return the
1225
# non-existing bits as they were specified (the filename may be
1226
# the target of a move, for example).
1227
current = pathjoin(current, bit, *list(bit_iter))
1229
return current[len(abs_base):].lstrip('/')
1231
# XXX - TODO - we need better detection/integration of case-insensitive
1232
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1233
# filesystems), for example, so could probably benefit from the same basic
1234
# support there. For now though, only Windows and OSX get that support, and
1235
# they get it for *all* file-systems!
1236
if sys.platform in ('win32', 'darwin'):
1237
canonical_relpath = _cicp_canonical_relpath
1239
canonical_relpath = relpath
1241
def canonical_relpaths(base, paths):
1242
"""Create an iterable to canonicalize a sequence of relative paths.
1244
The intent is for this implementation to use a cache, vastly speeding
1245
up multiple transformations in the same directory.
1247
# but for now, we haven't optimized...
1248
return [canonical_relpath(base, p) for p in paths]
1251
def decode_filename(filename):
1252
"""Decode the filename using the filesystem encoding
1254
If it is unicode, it is returned.
1255
Otherwise it is decoded from the the filesystem's encoding. If decoding
1256
fails, a errors.BadFilenameEncoding exception is raised.
1258
if type(filename) is unicode:
1261
return filename.decode(_fs_enc)
1262
except UnicodeDecodeError:
1263
raise errors.BadFilenameEncoding(filename, _fs_enc)
1266
def safe_unicode(unicode_or_utf8_string):
1267
"""Coerce unicode_or_utf8_string into unicode.
1269
If it is unicode, it is returned.
1270
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1271
wrapped in a BzrBadParameterNotUnicode exception.
1273
if isinstance(unicode_or_utf8_string, unicode):
1274
return unicode_or_utf8_string
1276
return unicode_or_utf8_string.decode('utf8')
1277
except UnicodeDecodeError:
1278
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1281
def safe_utf8(unicode_or_utf8_string):
1282
"""Coerce unicode_or_utf8_string to a utf8 string.
1284
If it is a str, it is returned.
1285
If it is Unicode, it is encoded into a utf-8 string.
1287
if isinstance(unicode_or_utf8_string, str):
1288
# TODO: jam 20070209 This is overkill, and probably has an impact on
1289
# performance if we are dealing with lots of apis that want a
1292
# Make sure it is a valid utf-8 string
1293
unicode_or_utf8_string.decode('utf-8')
1294
except UnicodeDecodeError:
1295
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1296
return unicode_or_utf8_string
1297
return unicode_or_utf8_string.encode('utf-8')
1300
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1301
' Revision id generators should be creating utf8'
1305
def safe_revision_id(unicode_or_utf8_string, warn=True):
1306
"""Revision ids should now be utf8, but at one point they were unicode.
1308
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1310
:param warn: Functions that are sanitizing user data can set warn=False
1311
:return: None or a utf8 revision id.
1313
if (unicode_or_utf8_string is None
1314
or unicode_or_utf8_string.__class__ == str):
1315
return unicode_or_utf8_string
1317
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1319
return cache_utf8.encode(unicode_or_utf8_string)
1322
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1323
' generators should be creating utf8 file ids.')
1326
def safe_file_id(unicode_or_utf8_string, warn=True):
1327
"""File ids should now be utf8, but at one point they were unicode.
1329
This is the same as safe_utf8, except it uses the cached encode functions
1330
to save a little bit of performance.
1332
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1334
:param warn: Functions that are sanitizing user data can set warn=False
1335
:return: None or a utf8 file id.
1337
if (unicode_or_utf8_string is None
1338
or unicode_or_utf8_string.__class__ == str):
1339
return unicode_or_utf8_string
1341
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1343
return cache_utf8.encode(unicode_or_utf8_string)
1346
_platform_normalizes_filenames = False
1347
if sys.platform == 'darwin':
1348
_platform_normalizes_filenames = True
1351
def normalizes_filenames():
1352
"""Return True if this platform normalizes unicode filenames.
1356
return _platform_normalizes_filenames
1359
def _accessible_normalized_filename(path):
1360
"""Get the unicode normalized path, and if you can access the file.
1362
On platforms where the system normalizes filenames (Mac OSX),
1363
you can access a file by any path which will normalize correctly.
1364
On platforms where the system does not normalize filenames
1365
(everything else), you have to access a file by its exact path.
1367
Internally, bzr only supports NFC normalization, since that is
1368
the standard for XML documents.
1370
So return the normalized path, and a flag indicating if the file
1371
can be accessed by that path.
1374
return unicodedata.normalize('NFC', unicode(path)), True
1377
def _inaccessible_normalized_filename(path):
1378
__doc__ = _accessible_normalized_filename.__doc__
1380
normalized = unicodedata.normalize('NFC', unicode(path))
1381
return normalized, normalized == path
1384
if _platform_normalizes_filenames:
1385
normalized_filename = _accessible_normalized_filename
1387
normalized_filename = _inaccessible_normalized_filename
1390
def set_signal_handler(signum, handler, restart_syscall=True):
1391
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1392
on platforms that support that.
1394
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1395
automatically restart (by calling `signal.siginterrupt(signum,
1396
False)`). May be ignored if the feature is not available on this
1397
platform or Python version.
1401
siginterrupt = signal.siginterrupt
1403
# This python implementation doesn't provide signal support, hence no
1406
except AttributeError:
1407
# siginterrupt doesn't exist on this platform, or for this version
1409
siginterrupt = lambda signum, flag: None
1411
def sig_handler(*args):
1412
# Python resets the siginterrupt flag when a signal is
1413
# received. <http://bugs.python.org/issue8354>
1414
# As a workaround for some cases, set it back the way we want it.
1415
siginterrupt(signum, False)
1416
# Now run the handler function passed to set_signal_handler.
1419
sig_handler = handler
1420
old_handler = signal.signal(signum, sig_handler)
1422
siginterrupt(signum, False)
1426
default_terminal_width = 80
1427
"""The default terminal width for ttys.
1429
This is defined so that higher levels can share a common fallback value when
1430
terminal_width() returns None.
1433
# Keep some state so that terminal_width can detect if _terminal_size has
1434
# returned a different size since the process started. See docstring and
1435
# comments of terminal_width for details.
1436
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1437
_terminal_size_state = 'no_data'
1438
_first_terminal_size = None
1440
def terminal_width():
1441
"""Return terminal width.
1443
None is returned if the width can't established precisely.
1446
- if BZR_COLUMNS is set, returns its value
1447
- if there is no controlling terminal, returns None
1448
- query the OS, if the queried size has changed since the last query,
1450
- if COLUMNS is set, returns its value,
1451
- if the OS has a value (even though it's never changed), return its value.
1453
From there, we need to query the OS to get the size of the controlling
1456
On Unices we query the OS by:
1457
- get termios.TIOCGWINSZ
1458
- if an error occurs or a negative value is obtained, returns None
1460
On Windows we query the OS by:
1461
- win32utils.get_console_size() decides,
1462
- returns None on error (provided default value)
1464
# Note to implementors: if changing the rules for determining the width,
1465
# make sure you've considered the behaviour in these cases:
1466
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1467
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1469
# - (add more interesting cases here, if you find any)
1470
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1471
# but we don't want to register a signal handler because it is impossible
1472
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1473
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1474
# time so we can notice if the reported size has changed, which should have
1477
# If BZR_COLUMNS is set, take it, user is always right
1478
# Except if they specified 0 in which case, impose no limit here
1480
width = int(os.environ['BZR_COLUMNS'])
1481
except (KeyError, ValueError):
1483
if width is not None:
1489
isatty = getattr(sys.stdout, 'isatty', None)
1490
if isatty is None or not isatty():
1491
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1495
width, height = os_size = _terminal_size(None, None)
1496
global _first_terminal_size, _terminal_size_state
1497
if _terminal_size_state == 'no_data':
1498
_first_terminal_size = os_size
1499
_terminal_size_state = 'unchanged'
1500
elif (_terminal_size_state == 'unchanged' and
1501
_first_terminal_size != os_size):
1502
_terminal_size_state = 'changed'
1504
# If the OS claims to know how wide the terminal is, and this value has
1505
# ever changed, use that.
1506
if _terminal_size_state == 'changed':
1507
if width is not None and width > 0:
1510
# If COLUMNS is set, use it.
1512
return int(os.environ['COLUMNS'])
1513
except (KeyError, ValueError):
1516
# Finally, use an unchanged size from the OS, if we have one.
1517
if _terminal_size_state == 'unchanged':
1518
if width is not None and width > 0:
1521
# The width could not be determined.
1525
def _win32_terminal_size(width, height):
1526
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1527
return width, height
1530
def _ioctl_terminal_size(width, height):
1532
import struct, fcntl, termios
1533
s = struct.pack('HHHH', 0, 0, 0, 0)
1534
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1535
height, width = struct.unpack('HHHH', x)[0:2]
1536
except (IOError, AttributeError):
1538
return width, height
1540
_terminal_size = None
1541
"""Returns the terminal size as (width, height).
1543
:param width: Default value for width.
1544
:param height: Default value for height.
1546
This is defined specifically for each OS and query the size of the controlling
1547
terminal. If any error occurs, the provided default values should be returned.
1549
if sys.platform == 'win32':
1550
_terminal_size = _win32_terminal_size
1552
_terminal_size = _ioctl_terminal_size
1555
def supports_executable():
1556
return sys.platform != "win32"
1559
def supports_posix_readonly():
1560
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1562
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1563
directory controls creation/deletion, etc.
1565
And under win32, readonly means that the directory itself cannot be
1566
deleted. The contents of a readonly directory can be changed, unlike POSIX
1567
where files in readonly directories cannot be added, deleted or renamed.
1569
return sys.platform != "win32"
1572
def set_or_unset_env(env_variable, value):
1573
"""Modify the environment, setting or removing the env_variable.
1575
:param env_variable: The environment variable in question
1576
:param value: The value to set the environment to. If None, then
1577
the variable will be removed.
1578
:return: The original value of the environment variable.
1580
orig_val = os.environ.get(env_variable)
1582
if orig_val is not None:
1583
del os.environ[env_variable]
1585
if isinstance(value, unicode):
1586
value = value.encode(get_user_encoding())
1587
os.environ[env_variable] = value
1591
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1594
def check_legal_path(path):
1595
"""Check whether the supplied path is legal.
1596
This is only required on Windows, so we don't test on other platforms
1599
if sys.platform != "win32":
1601
if _validWin32PathRE.match(path) is None:
1602
raise errors.IllegalPath(path)
1605
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1607
def _is_error_enotdir(e):
1608
"""Check if this exception represents ENOTDIR.
1610
Unfortunately, python is very inconsistent about the exception
1611
here. The cases are:
1612
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1613
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1614
which is the windows error code.
1615
3) Windows, Python2.5 uses errno == EINVAL and
1616
winerror == ERROR_DIRECTORY
1618
:param e: An Exception object (expected to be OSError with an errno
1619
attribute, but we should be able to cope with anything)
1620
:return: True if this represents an ENOTDIR error. False otherwise.
1622
en = getattr(e, 'errno', None)
1623
if (en == errno.ENOTDIR
1624
or (sys.platform == 'win32'
1625
and (en == _WIN32_ERROR_DIRECTORY
1626
or (en == errno.EINVAL
1627
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1633
def walkdirs(top, prefix=""):
1634
"""Yield data about all the directories in a tree.
1636
This yields all the data about the contents of a directory at a time.
1637
After each directory has been yielded, if the caller has mutated the list
1638
to exclude some directories, they are then not descended into.
1640
The data yielded is of the form:
1641
((directory-relpath, directory-path-from-top),
1642
[(relpath, basename, kind, lstat, path-from-top), ...]),
1643
- directory-relpath is the relative path of the directory being returned
1644
with respect to top. prefix is prepended to this.
1645
- directory-path-from-root is the path including top for this directory.
1646
It is suitable for use with os functions.
1647
- relpath is the relative path within the subtree being walked.
1648
- basename is the basename of the path
1649
- kind is the kind of the file now. If unknown then the file is not
1650
present within the tree - but it may be recorded as versioned. See
1652
- lstat is the stat data *if* the file was statted.
1653
- planned, not implemented:
1654
path_from_tree_root is the path from the root of the tree.
1656
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1657
allows one to walk a subtree but get paths that are relative to a tree
1659
:return: an iterator over the dirs.
1661
#TODO there is a bit of a smell where the results of the directory-
1662
# summary in this, and the path from the root, may not agree
1663
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1664
# potentially confusing output. We should make this more robust - but
1665
# not at a speed cost. RBC 20060731
1667
_directory = _directory_kind
1668
_listdir = os.listdir
1669
_kind_from_mode = file_kind_from_stat_mode
1670
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1672
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1673
relroot, _, _, _, top = pending.pop()
1675
relprefix = relroot + u'/'
1678
top_slash = top + u'/'
1681
append = dirblock.append
1683
names = sorted(map(decode_filename, _listdir(top)))
1685
if not _is_error_enotdir(e):
1689
abspath = top_slash + name
1690
statvalue = _lstat(abspath)
1691
kind = _kind_from_mode(statvalue.st_mode)
1692
append((relprefix + name, name, kind, statvalue, abspath))
1693
yield (relroot, top), dirblock
1695
# push the user specified dirs from dirblock
1696
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1699
class DirReader(object):
1700
"""An interface for reading directories."""
1702
def top_prefix_to_starting_dir(self, top, prefix=""):
1703
"""Converts top and prefix to a starting dir entry
1705
:param top: A utf8 path
1706
:param prefix: An optional utf8 path to prefix output relative paths
1708
:return: A tuple starting with prefix, and ending with the native
1711
raise NotImplementedError(self.top_prefix_to_starting_dir)
1713
def read_dir(self, prefix, top):
1714
"""Read a specific dir.
1716
:param prefix: A utf8 prefix to be preprended to the path basenames.
1717
:param top: A natively encoded path to read.
1718
:return: A list of the directories contents. Each item contains:
1719
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1721
raise NotImplementedError(self.read_dir)
1724
_selected_dir_reader = None
1727
def _walkdirs_utf8(top, prefix=""):
1728
"""Yield data about all the directories in a tree.
1730
This yields the same information as walkdirs() only each entry is yielded
1731
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1732
are returned as exact byte-strings.
1734
:return: yields a tuple of (dir_info, [file_info])
1735
dir_info is (utf8_relpath, path-from-top)
1736
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1737
if top is an absolute path, path-from-top is also an absolute path.
1738
path-from-top might be unicode or utf8, but it is the correct path to
1739
pass to os functions to affect the file in question. (such as os.lstat)
1741
global _selected_dir_reader
1742
if _selected_dir_reader is None:
1743
fs_encoding = _fs_enc.upper()
1744
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1745
# Win98 doesn't have unicode apis like FindFirstFileW
1746
# TODO: We possibly could support Win98 by falling back to the
1747
# original FindFirstFile, and using TCHAR instead of WCHAR,
1748
# but that gets a bit tricky, and requires custom compiling
1751
from bzrlib._walkdirs_win32 import Win32ReadDir
1752
_selected_dir_reader = Win32ReadDir()
1755
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1756
# ANSI_X3.4-1968 is a form of ASCII
1758
from bzrlib._readdir_pyx import UTF8DirReader
1759
_selected_dir_reader = UTF8DirReader()
1760
except ImportError, e:
1761
failed_to_load_extension(e)
1764
if _selected_dir_reader is None:
1765
# Fallback to the python version
1766
_selected_dir_reader = UnicodeDirReader()
1768
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1769
# But we don't actually uses 1-3 in pending, so set them to None
1770
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1771
read_dir = _selected_dir_reader.read_dir
1772
_directory = _directory_kind
1774
relroot, _, _, _, top = pending[-1].pop()
1777
dirblock = sorted(read_dir(relroot, top))
1778
yield (relroot, top), dirblock
1779
# push the user specified dirs from dirblock
1780
next = [d for d in reversed(dirblock) if d[2] == _directory]
1782
pending.append(next)
1785
class UnicodeDirReader(DirReader):
1786
"""A dir reader for non-utf8 file systems, which transcodes."""
1788
__slots__ = ['_utf8_encode']
1791
self._utf8_encode = codecs.getencoder('utf8')
1793
def top_prefix_to_starting_dir(self, top, prefix=""):
1794
"""See DirReader.top_prefix_to_starting_dir."""
1795
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1797
def read_dir(self, prefix, top):
1798
"""Read a single directory from a non-utf8 file system.
1800
top, and the abspath element in the output are unicode, all other paths
1801
are utf8. Local disk IO is done via unicode calls to listdir etc.
1803
This is currently the fallback code path when the filesystem encoding is
1804
not UTF-8. It may be better to implement an alternative so that we can
1805
safely handle paths that are not properly decodable in the current
1808
See DirReader.read_dir for details.
1810
_utf8_encode = self._utf8_encode
1812
_listdir = os.listdir
1813
_kind_from_mode = file_kind_from_stat_mode
1816
relprefix = prefix + '/'
1819
top_slash = top + u'/'
1822
append = dirblock.append
1823
for name in sorted(_listdir(top)):
1825
name_utf8 = _utf8_encode(name)[0]
1826
except UnicodeDecodeError:
1827
raise errors.BadFilenameEncoding(
1828
_utf8_encode(relprefix)[0] + name, _fs_enc)
1829
abspath = top_slash + name
1830
statvalue = _lstat(abspath)
1831
kind = _kind_from_mode(statvalue.st_mode)
1832
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1836
def copy_tree(from_path, to_path, handlers={}):
1837
"""Copy all of the entries in from_path into to_path.
1839
:param from_path: The base directory to copy.
1840
:param to_path: The target directory. If it does not exist, it will
1842
:param handlers: A dictionary of functions, which takes a source and
1843
destinations for files, directories, etc.
1844
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1845
'file', 'directory', and 'symlink' should always exist.
1846
If they are missing, they will be replaced with 'os.mkdir()',
1847
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1849
# Now, just copy the existing cached tree to the new location
1850
# We use a cheap trick here.
1851
# Absolute paths are prefixed with the first parameter
1852
# relative paths are prefixed with the second.
1853
# So we can get both the source and target returned
1854
# without any extra work.
1856
def copy_dir(source, dest):
1859
def copy_link(source, dest):
1860
"""Copy the contents of a symlink"""
1861
link_to = os.readlink(source)
1862
os.symlink(link_to, dest)
1864
real_handlers = {'file':shutil.copy2,
1865
'symlink':copy_link,
1866
'directory':copy_dir,
1868
real_handlers.update(handlers)
1870
if not os.path.exists(to_path):
1871
real_handlers['directory'](from_path, to_path)
1873
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1874
for relpath, name, kind, st, abspath in entries:
1875
real_handlers[kind](abspath, relpath)
1878
def copy_ownership_from_path(dst, src=None):
1879
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1881
If src is None, the containing directory is used as source. If chown
1882
fails, the error is ignored and a warning is printed.
1884
chown = getattr(os, 'chown', None)
1889
src = os.path.dirname(dst)
1895
chown(dst, s.st_uid, s.st_gid)
1898
'Unable to copy ownership from "%s" to "%s". '
1899
'You may want to set it manually.', src, dst)
1900
trace.log_exception_quietly()
1903
def path_prefix_key(path):
1904
"""Generate a prefix-order path key for path.
1906
This can be used to sort paths in the same way that walkdirs does.
1908
return (dirname(path) , path)
1911
def compare_paths_prefix_order(path_a, path_b):
1912
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1913
key_a = path_prefix_key(path_a)
1914
key_b = path_prefix_key(path_b)
1915
return cmp(key_a, key_b)
1918
_cached_user_encoding = None
1921
def get_user_encoding(use_cache=True):
1922
"""Find out what the preferred user encoding is.
1924
This is generally the encoding that is used for command line parameters
1925
and file contents. This may be different from the terminal encoding
1926
or the filesystem encoding.
1928
:param use_cache: Enable cache for detected encoding.
1929
(This parameter is turned on by default,
1930
and required only for selftesting)
1932
:return: A string defining the preferred user encoding
1934
global _cached_user_encoding
1935
if _cached_user_encoding is not None and use_cache:
1936
return _cached_user_encoding
1938
if sys.platform == 'darwin':
1939
# python locale.getpreferredencoding() always return
1940
# 'mac-roman' on darwin. That's a lie.
1941
sys.platform = 'posix'
1943
if os.environ.get('LANG', None) is None:
1944
# If LANG is not set, we end up with 'ascii', which is bad
1945
# ('mac-roman' is more than ascii), so we set a default which
1946
# will give us UTF-8 (which appears to work in all cases on
1947
# OSX). Users are still free to override LANG of course, as
1948
# long as it give us something meaningful. This work-around
1949
# *may* not be needed with python 3k and/or OSX 10.5, but will
1950
# work with them too -- vila 20080908
1951
os.environ['LANG'] = 'en_US.UTF-8'
1954
sys.platform = 'darwin'
1959
user_encoding = locale.getpreferredencoding()
1960
except locale.Error, e:
1961
sys.stderr.write('bzr: warning: %s\n'
1962
' Could not determine what text encoding to use.\n'
1963
' This error usually means your Python interpreter\n'
1964
' doesn\'t support the locale set by $LANG (%s)\n'
1965
" Continuing with ascii encoding.\n"
1966
% (e, os.environ.get('LANG')))
1967
user_encoding = 'ascii'
1969
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1970
# treat that as ASCII, and not support printing unicode characters to the
1973
# For python scripts run under vim, we get '', so also treat that as ASCII
1974
if user_encoding in (None, 'cp0', ''):
1975
user_encoding = 'ascii'
1979
codecs.lookup(user_encoding)
1981
sys.stderr.write('bzr: warning:'
1982
' unknown encoding %s.'
1983
' Continuing with ascii encoding.\n'
1986
user_encoding = 'ascii'
1989
_cached_user_encoding = user_encoding
1991
return user_encoding
1994
def get_diff_header_encoding():
1995
return get_terminal_encoding()
1998
def get_host_name():
1999
"""Return the current unicode host name.
2001
This is meant to be used in place of socket.gethostname() because that
2002
behaves inconsistently on different platforms.
2004
if sys.platform == "win32":
2006
return win32utils.get_host_name()
2009
return socket.gethostname().decode(get_user_encoding())
2012
# We must not read/write any more than 64k at a time from/to a socket so we
2013
# don't risk "no buffer space available" errors on some platforms. Windows in
2014
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2016
MAX_SOCKET_CHUNK = 64 * 1024
2018
_end_of_stream_errors = [errno.ECONNRESET]
2019
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2020
_eno = getattr(errno, _eno, None)
2021
if _eno is not None:
2022
_end_of_stream_errors.append(_eno)
2026
def read_bytes_from_socket(sock, report_activity=None,
2027
max_read_size=MAX_SOCKET_CHUNK):
2028
"""Read up to max_read_size of bytes from sock and notify of progress.
2030
Translates "Connection reset by peer" into file-like EOF (return an
2031
empty string rather than raise an error), and repeats the recv if
2032
interrupted by a signal.
2036
bytes = sock.recv(max_read_size)
2037
except socket.error, e:
2039
if eno in _end_of_stream_errors:
2040
# The connection was closed by the other side. Callers expect
2041
# an empty string to signal end-of-stream.
2043
elif eno == errno.EINTR:
2044
# Retry the interrupted recv.
2048
if report_activity is not None:
2049
report_activity(len(bytes), 'read')
2053
def recv_all(socket, count):
2054
"""Receive an exact number of bytes.
2056
Regular Socket.recv() may return less than the requested number of bytes,
2057
depending on what's in the OS buffer. MSG_WAITALL is not available
2058
on all platforms, but this should work everywhere. This will return
2059
less than the requested amount if the remote end closes.
2061
This isn't optimized and is intended mostly for use in testing.
2064
while len(b) < count:
2065
new = read_bytes_from_socket(socket, None, count - len(b))
2072
def send_all(sock, bytes, report_activity=None):
2073
"""Send all bytes on a socket.
2075
Breaks large blocks in smaller chunks to avoid buffering limitations on
2076
some platforms, and catches EINTR which may be thrown if the send is
2077
interrupted by a signal.
2079
This is preferred to socket.sendall(), because it avoids portability bugs
2080
and provides activity reporting.
2082
:param report_activity: Call this as bytes are read, see
2083
Transport._report_activity
2086
byte_count = len(bytes)
2087
while sent_total < byte_count:
2089
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2090
except socket.error, e:
2091
if e.args[0] != errno.EINTR:
2095
report_activity(sent, 'write')
2098
def connect_socket(address):
2099
# Slight variation of the socket.create_connection() function (provided by
2100
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2101
# provide it for previous python versions. Also, we don't use the timeout
2102
# parameter (provided by the python implementation) so we don't implement
2104
err = socket.error('getaddrinfo returns an empty list')
2105
host, port = address
2106
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2107
af, socktype, proto, canonname, sa = res
2110
sock = socket.socket(af, socktype, proto)
2114
except socket.error, err:
2115
# 'err' is now the most recent error
2116
if sock is not None:
2121
def dereference_path(path):
2122
"""Determine the real path to a file.
2124
All parent elements are dereferenced. But the file itself is not
2126
:param path: The original path. May be absolute or relative.
2127
:return: the real path *to* the file
2129
parent, base = os.path.split(path)
2130
# The pathjoin for '.' is a workaround for Python bug #1213894.
2131
# (initial path components aren't dereferenced)
2132
return pathjoin(realpath(pathjoin('.', parent)), base)
2135
def supports_mapi():
2136
"""Return True if we can use MAPI to launch a mail client."""
2137
return sys.platform == "win32"
2140
def resource_string(package, resource_name):
2141
"""Load a resource from a package and return it as a string.
2143
Note: Only packages that start with bzrlib are currently supported.
2145
This is designed to be a lightweight implementation of resource
2146
loading in a way which is API compatible with the same API from
2148
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2149
If and when pkg_resources becomes a standard library, this routine
2152
# Check package name is within bzrlib
2153
if package == "bzrlib":
2154
resource_relpath = resource_name
2155
elif package.startswith("bzrlib."):
2156
package = package[len("bzrlib."):].replace('.', os.sep)
2157
resource_relpath = pathjoin(package, resource_name)
2159
raise errors.BzrError('resource package %s not in bzrlib' % package)
2161
# Map the resource to a file and read its contents
2162
base = dirname(bzrlib.__file__)
2163
if getattr(sys, 'frozen', None): # bzr.exe
2164
base = abspath(pathjoin(base, '..', '..'))
2165
f = file(pathjoin(base, resource_relpath), "rU")
2171
def file_kind_from_stat_mode_thunk(mode):
2172
global file_kind_from_stat_mode
2173
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2175
from bzrlib._readdir_pyx import UTF8DirReader
2176
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2177
except ImportError, e:
2178
# This is one time where we won't warn that an extension failed to
2179
# load. The extension is never available on Windows anyway.
2180
from bzrlib._readdir_py import (
2181
_kind_from_mode as file_kind_from_stat_mode
2183
return file_kind_from_stat_mode(mode)
2184
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2187
def file_kind(f, _lstat=os.lstat):
2189
return file_kind_from_stat_mode(_lstat(f).st_mode)
2191
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2192
raise errors.NoSuchFile(f)
2196
def until_no_eintr(f, *a, **kw):
2197
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2199
WARNING: you must be certain that it is safe to retry the call repeatedly
2200
if EINTR does occur. This is typically only true for low-level operations
2201
like os.read. If in any doubt, don't use this.
2203
Keep in mind that this is not a complete solution to EINTR. There is
2204
probably code in the Python standard library and other dependencies that
2205
may encounter EINTR if a signal arrives (and there is signal handler for
2206
that signal). So this function can reduce the impact for IO that bzrlib
2207
directly controls, but it is not a complete solution.
2209
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2213
except (IOError, OSError), e:
2214
if e.errno == errno.EINTR:
2219
@deprecated_function(deprecated_in((2, 2, 0)))
2220
def re_compile_checked(re_string, flags=0, where=""):
2221
"""Return a compiled re, or raise a sensible error.
2223
This should only be used when compiling user-supplied REs.
2225
:param re_string: Text form of regular expression.
2226
:param flags: eg re.IGNORECASE
2227
:param where: Message explaining to the user the context where
2228
it occurred, eg 'log search filter'.
2230
# from https://bugs.launchpad.net/bzr/+bug/251352
2232
re_obj = re.compile(re_string, flags)
2235
except errors.InvalidPattern, e:
2237
where = ' in ' + where
2238
# despite the name 'error' is a type
2239
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2243
if sys.platform == "win32":
2246
return msvcrt.getch()
2251
fd = sys.stdin.fileno()
2252
settings = termios.tcgetattr(fd)
2255
ch = sys.stdin.read(1)
2257
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2260
if sys.platform == 'linux2':
2261
def _local_concurrency():
2263
return os.sysconf('SC_NPROCESSORS_ONLN')
2264
except (ValueError, OSError, AttributeError):
2266
elif sys.platform == 'darwin':
2267
def _local_concurrency():
2268
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2269
stdout=subprocess.PIPE).communicate()[0]
2270
elif "bsd" in sys.platform:
2271
def _local_concurrency():
2272
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2273
stdout=subprocess.PIPE).communicate()[0]
2274
elif sys.platform == 'sunos5':
2275
def _local_concurrency():
2276
return subprocess.Popen(['psrinfo', '-p',],
2277
stdout=subprocess.PIPE).communicate()[0]
2278
elif sys.platform == "win32":
2279
def _local_concurrency():
2280
# This appears to return the number of cores.
2281
return os.environ.get('NUMBER_OF_PROCESSORS')
2283
def _local_concurrency():
2288
_cached_local_concurrency = None
2290
def local_concurrency(use_cache=True):
2291
"""Return how many processes can be run concurrently.
2293
Rely on platform specific implementations and default to 1 (one) if
2294
anything goes wrong.
2296
global _cached_local_concurrency
2298
if _cached_local_concurrency is not None and use_cache:
2299
return _cached_local_concurrency
2301
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2302
if concurrency is None:
2304
import multiprocessing
2306
# multiprocessing is only available on Python >= 2.6
2308
concurrency = _local_concurrency()
2309
except (OSError, IOError):
2312
concurrency = multiprocessing.cpu_count()
2314
concurrency = int(concurrency)
2315
except (TypeError, ValueError):
2318
_cached_concurrency = concurrency
2322
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2323
"""A stream writer that doesn't decode str arguments."""
2325
def __init__(self, encode, stream, errors='strict'):
2326
codecs.StreamWriter.__init__(self, stream, errors)
2327
self.encode = encode
2329
def write(self, object):
2330
if type(object) is str:
2331
self.stream.write(object)
2333
data, _ = self.encode(object, self.errors)
2334
self.stream.write(data)
2336
if sys.platform == 'win32':
2337
def open_file(filename, mode='r', bufsize=-1):
2338
"""This function is used to override the ``open`` builtin.
2340
But it uses O_NOINHERIT flag so the file handle is not inherited by
2341
child processes. Deleting or renaming a closed file opened with this
2342
function is not blocking child processes.
2344
writing = 'w' in mode
2345
appending = 'a' in mode
2346
updating = '+' in mode
2347
binary = 'b' in mode
2350
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2351
# for flags for each modes.
2361
flags |= os.O_WRONLY
2362
flags |= os.O_CREAT | os.O_TRUNC
2367
flags |= os.O_WRONLY
2368
flags |= os.O_CREAT | os.O_APPEND
2373
flags |= os.O_RDONLY
2375
return os.fdopen(os.open(filename, flags), mode, bufsize)
2380
def getuser_unicode():
2381
"""Return the username as unicode.
2384
user_encoding = get_user_encoding()
2385
username = getpass.getuser().decode(user_encoding)
2386
except UnicodeDecodeError:
2387
raise errors.BzrError("Can't decode username as %s." % \
2392
def available_backup_name(base, exists):
2393
"""Find a non-existing backup file name.
2395
This will *not* create anything, this only return a 'free' entry. This
2396
should be used for checking names in a directory below a locked
2397
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2398
Leap) and generally discouraged.
2400
:param base: The base name.
2402
:param exists: A callable returning True if the path parameter exists.
2405
name = "%s.~%d~" % (base, counter)
2408
name = "%s.~%d~" % (base, counter)
2412
def set_fd_cloexec(fd):
2413
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2414
support for this is not available.
2418
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2419
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2420
except (ImportError, AttributeError):
2421
# Either the fcntl module or specific constants are not present
2425
def find_executable_on_path(name):
2426
"""Finds an executable on the PATH.
2428
On Windows, this will try to append each extension in the PATHEXT
2429
environment variable to the name, if it cannot be found with the name
2432
:param name: The base name of the executable.
2433
:return: The path to the executable found or None.
2435
path = os.environ.get('PATH')
2438
path = path.split(os.pathsep)
2439
if sys.platform == 'win32':
2440
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2441
exts = [ext.lower() for ext in exts]
2442
base, ext = os.path.splitext(name)
2444
if ext.lower() not in exts:
2452
f = os.path.join(d, name) + ext
2453
if os.access(f, os.X_OK):
added
removed
bzrlib/osutils.py
