1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005, 2006 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
15
# 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
17
from cStringIO import StringIO
21
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
22
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
45
from tempfile import (
56
from bzrlib.symbol_versioning import (
23
60
from bzrlib.trace import mutter
63
# On win32, O_BINARY is used to indicate the file should
64
# be opened in binary mode, rather than text mode.
65
# On other platforms, O_BINARY doesn't exist, because
66
# they always open in binary mode, so it is okay to
67
# OR with 0 on those platforms
68
O_BINARY = getattr(os, 'O_BINARY', 0)
26
71
def make_readonly(filename):
27
72
"""Make a filename read-only."""
28
# TODO: probably needs to be fixed for windows
29
73
mod = os.stat(filename).st_mode
30
74
mod = mod & 0777555
31
75
os.chmod(filename, mod)
60
mode = os.lstat(f)[ST_MODE]
68
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
103
_directory_kind = 'directory'
106
stat.S_IFDIR:_directory_kind,
107
stat.S_IFCHR:'chardev',
108
stat.S_IFBLK:'block',
111
stat.S_IFLNK:'symlink',
112
stat.S_IFSOCK:'socket',
116
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
117
"""Generate a file kind from a stat mode. This is used in walkdirs.
119
Its performance is critical: Do not mutate without careful benchmarking.
122
return _formats[stat_mode & 0170000]
127
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
129
return _mapper(_lstat(f).st_mode)
131
if getattr(e, 'errno', None) == errno.ENOENT:
132
raise errors.NoSuchFile(f)
137
"""Return the current umask"""
138
# Assume that people aren't messing with the umask while running
139
# XXX: This is not thread safe, but there is no way to get the
140
# umask without setting it
71
146
def kind_marker(kind):
72
147
if kind == 'file':
74
elif kind == 'directory':
149
elif kind == _directory_kind:
76
151
elif kind == 'symlink':
79
raise BzrError('invalid file kind %r' % kind)
154
raise errors.BzrError('invalid file kind %r' % kind)
156
lexists = getattr(os.path, 'lexists', None)
160
if getattr(os, 'lstat') is not None:
166
if e.errno == errno.ENOENT:
169
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
172
def fancy_rename(old, new, rename_func, unlink_func):
173
"""A fancy rename, when you don't have atomic rename.
175
:param old: The old path, to rename from
176
:param new: The new path, to rename to
177
:param rename_func: The potentially non-atomic rename function
178
:param unlink_func: A way to delete the target file if the full rename succeeds
181
# sftp rename doesn't allow overwriting, so play tricks:
183
base = os.path.basename(new)
184
dirname = os.path.dirname(new)
185
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
186
tmp_name = pathjoin(dirname, tmp_name)
188
# Rename the file out of the way, but keep track if it didn't exist
189
# We don't want to grab just any exception
190
# something like EACCES should prevent us from continuing
191
# The downside is that the rename_func has to throw an exception
192
# with an errno = ENOENT, or NoSuchFile
195
rename_func(new, tmp_name)
196
except (errors.NoSuchFile,), e:
199
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
200
# function raises an IOError with errno is None when a rename fails.
201
# This then gets caught here.
202
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
205
if (getattr(e, 'errno', None) is None
206
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
213
# This may throw an exception, in which case success will
215
rename_func(old, new)
219
# If the file used to exist, rename it back into place
220
# otherwise just delete it from the tmp location
222
unlink_func(tmp_name)
224
rename_func(tmp_name, new)
227
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
228
# choke on a Unicode string containing a relative path if
229
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
231
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
232
def _posix_abspath(path):
233
# jam 20060426 rather than encoding to fsencoding
234
# copy posixpath.abspath, but use os.getcwdu instead
235
if not posixpath.isabs(path):
236
path = posixpath.join(getcwd(), path)
237
return posixpath.normpath(path)
240
def _posix_realpath(path):
241
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
244
def _win32_fixdrive(path):
245
"""Force drive letters to be consistent.
247
win32 is inconsistent whether it returns lower or upper case
248
and even if it was consistent the user might type the other
249
so we force it to uppercase
250
running python.exe under cmd.exe return capital C:\\
251
running win32 python inside a cygwin shell returns lowercase c:\\
253
drive, path = _nt_splitdrive(path)
254
return drive.upper() + path
257
def _win32_abspath(path):
258
# Real _nt_abspath doesn't have a problem with a unicode cwd
259
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
262
def _win32_realpath(path):
263
# Real _nt_realpath doesn't have a problem with a unicode cwd
264
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
267
def _win32_pathjoin(*args):
268
return _nt_join(*args).replace('\\', '/')
271
def _win32_normpath(path):
272
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
276
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
279
def _win32_mkdtemp(*args, **kwargs):
280
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
283
def _win32_rename(old, new):
284
"""We expect to be able to atomically replace 'new' with old.
286
On win32, if new exists, it must be moved out of the way first,
290
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
292
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
293
# If we try to rename a non-existant file onto cwd, we get
294
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
295
# if the old path doesn't exist, sometimes we get EACCES
296
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
302
return unicodedata.normalize('NFKC', os.getcwdu())
305
# Default is to just use the python builtins, but these can be rebound on
306
# particular platforms.
307
abspath = _posix_abspath
308
realpath = _posix_realpath
309
pathjoin = os.path.join
310
normpath = os.path.normpath
313
dirname = os.path.dirname
314
basename = os.path.basename
315
split = os.path.split
316
splitext = os.path.splitext
317
# These were already imported into local scope
318
# mkdtemp = tempfile.mkdtemp
319
# rmtree = shutil.rmtree
321
MIN_ABS_PATHLENGTH = 1
324
if sys.platform == 'win32':
325
abspath = _win32_abspath
326
realpath = _win32_realpath
327
pathjoin = _win32_pathjoin
328
normpath = _win32_normpath
329
getcwd = _win32_getcwd
330
mkdtemp = _win32_mkdtemp
331
rename = _win32_rename
333
MIN_ABS_PATHLENGTH = 3
335
def _win32_delete_readonly(function, path, excinfo):
336
"""Error handler for shutil.rmtree function [for win32]
337
Helps to remove files and dirs marked as read-only.
339
exception = excinfo[1]
340
if function in (os.remove, os.rmdir) \
341
and isinstance(exception, OSError) \
342
and exception.errno == errno.EACCES:
348
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
349
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
350
return shutil.rmtree(path, ignore_errors, onerror)
351
elif sys.platform == 'darwin':
355
def get_terminal_encoding():
356
"""Find the best encoding for printing to the screen.
358
This attempts to check both sys.stdout and sys.stdin to see
359
what encoding they are in, and if that fails it falls back to
360
bzrlib.user_encoding.
361
The problem is that on Windows, locale.getpreferredencoding()
362
is not the same encoding as that used by the console:
363
http://mail.python.org/pipermail/python-list/2003-May/162357.html
365
On my standard US Windows XP, the preferred encoding is
366
cp1252, but the console is cp437
368
output_encoding = getattr(sys.stdout, 'encoding', None)
369
if not output_encoding:
370
input_encoding = getattr(sys.stdin, 'encoding', None)
371
if not input_encoding:
372
output_encoding = bzrlib.user_encoding
373
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
375
output_encoding = input_encoding
376
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
378
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
379
if output_encoding == 'cp0':
380
# invalid encoding (cp0 means 'no codepage' on Windows)
381
output_encoding = bzrlib.user_encoding
382
mutter('cp0 is invalid encoding.'
383
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
386
codecs.lookup(output_encoding)
388
sys.stderr.write('bzr: warning:'
389
' unknown terminal encoding %s.\n'
390
' Using encoding %s instead.\n'
391
% (output_encoding, bzrlib.user_encoding)
393
output_encoding = bzrlib.user_encoding
395
return output_encoding
398
def normalizepath(f):
399
if getattr(os.path, 'realpath', None) is not None:
403
[p,e] = os.path.split(f)
404
if e == "" or e == "." or e == "..":
407
return pathjoin(F(p), e)
83
410
def backup_file(fn):
468
720
assert isinstance(p, list)
470
if (f == '..') or (f == None) or (f == ''):
471
raise BzrError("sorry, %r not allowed in path" % f)
472
return os.path.join(*p)
722
if (f == '..') or (f is None) or (f == ''):
723
raise errors.BzrError("sorry, %r not allowed in path" % f)
727
@deprecated_function(zero_nine)
475
728
def appendpath(p1, p2):
479
return os.path.join(p1, p2)
482
def extern_command(cmd, ignore_errors = False):
483
mutter('external command: %s' % `cmd`)
485
if not ignore_errors:
486
raise BzrError('command failed')
489
def _read_config_value(name):
490
"""Read a config value from the file ~/.bzr.conf/<name>
491
Return None if the file does not exist"""
493
f = file(os.path.join(config_dir(), name), "r")
494
return f.read().decode(bzrlib.user_encoding).rstrip("\r\n")
496
if e.errno == errno.ENOENT:
732
return pathjoin(p1, p2)
736
"""Split s into lines, but without removing the newline characters."""
737
lines = s.split('\n')
738
result = [line + '\n' for line in lines[:-1]]
740
result.append(lines[-1])
744
def hardlinks_good():
745
return sys.platform not in ('win32', 'cygwin', 'darwin')
748
def link_or_copy(src, dest):
749
"""Hardlink a file, or copy it if it can't be hardlinked."""
750
if not hardlinks_good():
751
shutil.copyfile(src, dest)
755
except (OSError, IOError), e:
756
if e.errno != errno.EXDEV:
758
shutil.copyfile(src, dest)
760
def delete_any(full_path):
761
"""Delete a file or directory."""
765
# We may be renaming a dangling inventory id
766
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
772
if getattr(os, 'symlink', None) is not None:
778
def contains_whitespace(s):
779
"""True if there are any whitespace characters in s."""
780
for ch in string.whitespace:
787
def contains_linebreaks(s):
788
"""True if there is any vertical whitespace in s."""
796
def relpath(base, path):
797
"""Return path relative to base, or raise exception.
799
The path may be either an absolute path or a path relative to the
800
current working directory.
802
os.path.commonprefix (python2.4) has a bad bug that it works just
803
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
807
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
808
' exceed the platform minimum length (which is %d)' %
815
while len(head) >= len(base):
818
head, tail = os.path.split(head)
822
raise errors.PathNotChild(rp, base)
830
def safe_unicode(unicode_or_utf8_string):
831
"""Coerce unicode_or_utf8_string into unicode.
833
If it is unicode, it is returned.
834
Otherwise it is decoded from utf-8. If a decoding error
835
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
836
as a BzrBadParameter exception.
838
if isinstance(unicode_or_utf8_string, unicode):
839
return unicode_or_utf8_string
841
return unicode_or_utf8_string.decode('utf8')
842
except UnicodeDecodeError:
843
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
846
_platform_normalizes_filenames = False
847
if sys.platform == 'darwin':
848
_platform_normalizes_filenames = True
851
def normalizes_filenames():
852
"""Return True if this platform normalizes unicode filenames.
854
Mac OSX does, Windows/Linux do not.
856
return _platform_normalizes_filenames
859
def _accessible_normalized_filename(path):
860
"""Get the unicode normalized path, and if you can access the file.
862
On platforms where the system normalizes filenames (Mac OSX),
863
you can access a file by any path which will normalize correctly.
864
On platforms where the system does not normalize filenames
865
(Windows, Linux), you have to access a file by its exact path.
867
Internally, bzr only supports NFC/NFKC normalization, since that is
868
the standard for XML documents.
870
So return the normalized path, and a flag indicating if the file
871
can be accessed by that path.
874
return unicodedata.normalize('NFKC', unicode(path)), True
877
def _inaccessible_normalized_filename(path):
878
__doc__ = _accessible_normalized_filename.__doc__
880
normalized = unicodedata.normalize('NFKC', unicode(path))
881
return normalized, normalized == path
884
if _platform_normalizes_filenames:
885
normalized_filename = _accessible_normalized_filename
887
normalized_filename = _inaccessible_normalized_filename
890
def terminal_width():
891
"""Return estimated terminal width."""
892
if sys.platform == 'win32':
893
import bzrlib.win32console
894
return bzrlib.win32console.get_console_size()[0]
897
import struct, fcntl, termios
898
s = struct.pack('HHHH', 0, 0, 0, 0)
899
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
900
width = struct.unpack('HHHH', x)[1]
905
width = int(os.environ['COLUMNS'])
914
def supports_executable():
915
return sys.platform != "win32"
918
def set_or_unset_env(env_variable, value):
919
"""Modify the environment, setting or removing the env_variable.
921
:param env_variable: The environment variable in question
922
:param value: The value to set the environment to. If None, then
923
the variable will be removed.
924
:return: The original value of the environment variable.
926
orig_val = os.environ.get(env_variable)
928
if orig_val is not None:
929
del os.environ[env_variable]
931
if isinstance(value, unicode):
932
value = value.encode(bzrlib.user_encoding)
933
os.environ[env_variable] = value
937
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
940
def check_legal_path(path):
941
"""Check whether the supplied path is legal.
942
This is only required on Windows, so we don't test on other platforms
945
if sys.platform != "win32":
947
if _validWin32PathRE.match(path) is None:
948
raise errors.IllegalPath(path)
951
def walkdirs(top, prefix=""):
952
"""Yield data about all the directories in a tree.
954
This yields all the data about the contents of a directory at a time.
955
After each directory has been yielded, if the caller has mutated the list
956
to exclude some directories, they are then not descended into.
958
The data yielded is of the form:
959
((directory-relpath, directory-path-from-top),
960
[(relpath, basename, kind, lstat), ...]),
961
- directory-relpath is the relative path of the directory being returned
962
with respect to top. prefix is prepended to this.
963
- directory-path-from-root is the path including top for this directory.
964
It is suitable for use with os functions.
965
- relpath is the relative path within the subtree being walked.
966
- basename is the basename of the path
967
- kind is the kind of the file now. If unknown then the file is not
968
present within the tree - but it may be recorded as versioned. See
970
- lstat is the stat data *if* the file was statted.
971
- planned, not implemented:
972
path_from_tree_root is the path from the root of the tree.
974
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
975
allows one to walk a subtree but get paths that are relative to a tree
977
:return: an iterator over the dirs.
979
#TODO there is a bit of a smell where the results of the directory-
980
# summary in this, and the path from the root, may not agree
981
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
982
# potentially confusing output. We should make this more robust - but
983
# not at a speed cost. RBC 20060731
986
_directory = _directory_kind
987
_listdir = os.listdir
988
pending = [(prefix, "", _directory, None, top)]
991
currentdir = pending.pop()
992
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
995
relroot = currentdir[0] + '/'
998
for name in sorted(_listdir(top)):
999
abspath = top + '/' + name
1000
statvalue = lstat(abspath)
1001
dirblock.append((relroot + name, name,
1002
file_kind_from_stat_mode(statvalue.st_mode),
1003
statvalue, abspath))
1004
yield (currentdir[0], top), dirblock
1005
# push the user specified dirs from dirblock
1006
for dir in reversed(dirblock):
1007
if dir[2] == _directory:
1011
def copy_tree(from_path, to_path, handlers={}):
1012
"""Copy all of the entries in from_path into to_path.
1014
:param from_path: The base directory to copy.
1015
:param to_path: The target directory. If it does not exist, it will
1017
:param handlers: A dictionary of functions, which takes a source and
1018
destinations for files, directories, etc.
1019
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1020
'file', 'directory', and 'symlink' should always exist.
1021
If they are missing, they will be replaced with 'os.mkdir()',
1022
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1024
# Now, just copy the existing cached tree to the new location
1025
# We use a cheap trick here.
1026
# Absolute paths are prefixed with the first parameter
1027
# relative paths are prefixed with the second.
1028
# So we can get both the source and target returned
1029
# without any extra work.
1031
def copy_dir(source, dest):
1034
def copy_link(source, dest):
1035
"""Copy the contents of a symlink"""
1036
link_to = os.readlink(source)
1037
os.symlink(link_to, dest)
1039
real_handlers = {'file':shutil.copy2,
1040
'symlink':copy_link,
1041
'directory':copy_dir,
1043
real_handlers.update(handlers)
1045
if not os.path.exists(to_path):
1046
real_handlers['directory'](from_path, to_path)
1048
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1049
for relpath, name, kind, st, abspath in entries:
1050
real_handlers[kind](abspath, relpath)
1053
def path_prefix_key(path):
1054
"""Generate a prefix-order path key for path.
1056
This can be used to sort paths in the same way that walkdirs does.
1058
return (dirname(path) , path)
1061
def compare_paths_prefix_order(path_a, path_b):
1062
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1063
key_a = path_prefix_key(path_a)
1064
key_b = path_prefix_key(path_b)
1065
return cmp(key_a, key_b)
1068
_cached_user_encoding = None
1071
def get_user_encoding(use_cache=True):
1072
"""Find out what the preferred user encoding is.
1074
This is generally the encoding that is used for command line parameters
1075
and file contents. This may be different from the terminal encoding
1076
or the filesystem encoding.
1078
:param use_cache: Enable cache for detected encoding.
1079
(This parameter is turned on by default,
1080
and required only for selftesting)
1082
:return: A string defining the preferred user encoding
1084
global _cached_user_encoding
1085
if _cached_user_encoding is not None and use_cache:
1086
return _cached_user_encoding
1088
if sys.platform == 'darwin':
1089
# work around egregious python 2.4 bug
1090
sys.platform = 'posix'
1094
sys.platform = 'darwin'
1099
user_encoding = locale.getpreferredencoding()
1100
except locale.Error, e:
1101
sys.stderr.write('bzr: warning: %s\n'
1102
' Could not determine what text encoding to use.\n'
1103
' This error usually means your Python interpreter\n'
1104
' doesn\'t support the locale set by $LANG (%s)\n'
1105
" Continuing with ascii encoding.\n"
1106
% (e, os.environ.get('LANG')))
1107
user_encoding = 'ascii'
1109
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1110
# treat that as ASCII, and not support printing unicode characters to the
1112
if user_encoding in (None, 'cp0'):
1113
user_encoding = 'ascii'
1117
codecs.lookup(user_encoding)
1119
sys.stderr.write('bzr: warning:'
1120
' unknown encoding %s.'
1121
' Continuing with ascii encoding.\n'
1124
user_encoding = 'ascii'
1127
_cached_user_encoding = user_encoding
1129
return user_encoding
1132
def recv_all(socket, bytes):
1133
"""Receive an exact number of bytes.
1135
Regular Socket.recv() may return less than the requested number of bytes,
1136
dependning on what's in the OS buffer. MSG_WAITALL is not available
1137
on all platforms, but this should work everywhere. This will return
1138
less than the requested amount if the remote end closes.
1140
This isn't optimized and is intended mostly for use in testing.
1143
while len(b) < bytes:
1144
new = socket.recv(bytes - len(b))
1150
def dereference_path(path):
1151
"""Determine the real path to a file.
1153
All parent elements are dereferenced. But the file itself is not
1155
:param path: The original path. May be absolute or relative.
1156
:return: the real path *to* the file
1158
parent, base = os.path.split(path)
1159
# The pathjoin for '.' is a workaround for Python bug #1213894.
1160
# (initial path components aren't dereferenced)
1161
return pathjoin(realpath(pathjoin('.', parent)), base)
added
removed
bzrlib/osutils.py
