171
224
rename_func(tmp_name, new)
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
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
176
309
pathjoin = os.path.join
177
310
normpath = os.path.normpath
178
311
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
180
312
rename = os.rename
181
313
dirname = os.path.dirname
182
314
basename = os.path.basename
184
if os.name == "posix":
185
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
186
# choke on a Unicode string containing a relative path if
187
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
189
_fs_enc = sys.getfilesystemencoding()
191
return os.path.abspath(path.encode(_fs_enc)).decode(_fs_enc)
194
return os.path.realpath(path.encode(_fs_enc)).decode(_fs_enc)
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
196
324
if sys.platform == 'win32':
197
# We need to use the Unicode-aware os.path.abspath and
198
# os.path.realpath on Windows systems.
200
return os.path.abspath(path).replace('\\', '/')
203
return os.path.realpath(path).replace('\\', '/')
206
return os.path.join(*args).replace('\\', '/')
209
return os.path.normpath(path).replace('\\', '/')
212
return os.getcwdu().replace('\\', '/')
214
def mkdtemp(*args, **kwargs):
215
return tempfile.mkdtemp(*args, **kwargs).replace('\\', '/')
217
def rename(old, new):
218
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
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
221
398
def normalizepath(f):
222
if hasattr(os.path, 'realpath'):
399
if getattr(os.path, 'realpath', None) is not None:
636
841
return unicode_or_utf8_string.decode('utf8')
637
842
except UnicodeDecodeError:
638
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
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
641
890
def terminal_width():
642
891
"""Return estimated terminal width."""
644
# TODO: Do something smart on Windows?
646
# TODO: Is there anything that gets a better update when the window
647
# is resized while the program is running? We could use the Python termcap
892
if sys.platform == 'win32':
893
import bzrlib.win32console
894
return bzrlib.win32console.get_console_size()[0]
650
return int(os.environ['COLUMNS'])
651
except (IndexError, KeyError, ValueError):
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'])
654
914
def supports_executable():
655
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)