47
135
This previously used backslash quoting, but that works poorly on
49
137
# TODO: I'm not really sure this is the best format either.x
139
if _QUOTE_RE is None:
140
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
50
142
if _QUOTE_RE.search(f):
51
143
return '"' + f + '"'
57
mode = os.lstat(f)[ST_MODE]
65
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
148
_directory_kind = 'directory'
151
"""Return the current umask"""
152
# Assume that people aren't messing with the umask while running
153
# XXX: This is not thread safe, but there is no way to get the
154
# umask without setting it
162
_directory_kind: "/",
164
'tree-reference': '+',
68
168
def kind_marker(kind):
71
elif kind == 'directory':
73
elif kind == 'symlink':
76
raise BzrError('invalid file kind %r' % kind)
81
"""Copy a file to a backup.
83
Backups are named in GNU-style, with a ~ suffix.
85
If the file is already a backup, it's not copied.
98
outf = file(bfn, 'wb')
104
def rename(path_from, path_to):
105
"""Basically the same as os.rename() just special for win32"""
106
if sys.platform == 'win32':
170
return _kind_marker_map[kind]
172
raise errors.BzrError('invalid file kind %r' % kind)
175
lexists = getattr(os.path, 'lexists', None)
179
stat = getattr(os, 'lstat', os.stat)
109
182
except OSError, e:
110
if e.errno != e.ENOENT:
183
if e.errno == errno.ENOENT:
186
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
189
def fancy_rename(old, new, rename_func, unlink_func):
190
"""A fancy rename, when you don't have atomic rename.
192
:param old: The old path, to rename from
193
:param new: The new path, to rename to
194
:param rename_func: The potentially non-atomic rename function
195
:param unlink_func: A way to delete the target file if the full rename succeeds
198
# sftp rename doesn't allow overwriting, so play tricks:
199
base = os.path.basename(new)
200
dirname = os.path.dirname(new)
201
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
202
tmp_name = pathjoin(dirname, tmp_name)
204
# Rename the file out of the way, but keep track if it didn't exist
205
# We don't want to grab just any exception
206
# something like EACCES should prevent us from continuing
207
# The downside is that the rename_func has to throw an exception
208
# with an errno = ENOENT, or NoSuchFile
211
rename_func(new, tmp_name)
212
except (errors.NoSuchFile,), e:
215
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
216
# function raises an IOError with errno is None when a rename fails.
217
# This then gets caught here.
218
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
221
if (getattr(e, 'errno', None) is None
222
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
230
# This may throw an exception, in which case success will
232
rename_func(old, new)
234
except (IOError, OSError), e:
235
# source and target may be aliases of each other (e.g. on a
236
# case-insensitive filesystem), so we may have accidentally renamed
237
# source by when we tried to rename target
238
if not (file_existed and e.errno in (None, errno.ENOENT)):
112
os.rename(path_from, path_to)
242
# If the file used to exist, rename it back into place
243
# otherwise just delete it from the tmp location
245
unlink_func(tmp_name)
247
rename_func(tmp_name, new)
250
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
251
# choke on a Unicode string containing a relative path if
252
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
254
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
255
def _posix_abspath(path):
256
# jam 20060426 rather than encoding to fsencoding
257
# copy posixpath.abspath, but use os.getcwdu instead
258
if not posixpath.isabs(path):
259
path = posixpath.join(getcwd(), path)
260
return posixpath.normpath(path)
263
def _posix_realpath(path):
264
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
267
def _win32_fixdrive(path):
268
"""Force drive letters to be consistent.
270
win32 is inconsistent whether it returns lower or upper case
271
and even if it was consistent the user might type the other
272
so we force it to uppercase
273
running python.exe under cmd.exe return capital C:\\
274
running win32 python inside a cygwin shell returns lowercase c:\\
276
drive, path = _nt_splitdrive(path)
277
return drive.upper() + path
280
def _win32_abspath(path):
281
# Real _nt_abspath doesn't have a problem with a unicode cwd
282
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
285
def _win98_abspath(path):
286
"""Return the absolute version of a path.
287
Windows 98 safe implementation (python reimplementation
288
of Win32 API function GetFullPathNameW)
293
# \\HOST\path => //HOST/path
294
# //HOST/path => //HOST/path
295
# path => C:/cwd/path
298
# check for absolute path
299
drive = _nt_splitdrive(path)[0]
300
if drive == '' and path[:2] not in('//','\\\\'):
302
# we cannot simply os.path.join cwd and path
303
# because os.path.join('C:','/path') produce '/path'
304
# and this is incorrect
305
if path[:1] in ('/','\\'):
306
cwd = _nt_splitdrive(cwd)[0]
308
path = cwd + '\\' + path
309
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
312
def _win32_realpath(path):
313
# Real _nt_realpath doesn't have a problem with a unicode cwd
314
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
317
def _win32_pathjoin(*args):
318
return _nt_join(*args).replace('\\', '/')
321
def _win32_normpath(path):
322
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
326
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
329
def _win32_mkdtemp(*args, **kwargs):
330
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
333
def _win32_rename(old, new):
334
"""We expect to be able to atomically replace 'new' with old.
336
On win32, if new exists, it must be moved out of the way first,
340
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
342
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
343
# If we try to rename a non-existant file onto cwd, we get
344
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
345
# if the old path doesn't exist, sometimes we get EACCES
346
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
352
return unicodedata.normalize('NFC', os.getcwdu())
355
# Default is to just use the python builtins, but these can be rebound on
356
# particular platforms.
357
abspath = _posix_abspath
358
realpath = _posix_realpath
359
pathjoin = os.path.join
360
normpath = os.path.normpath
363
dirname = os.path.dirname
364
basename = os.path.basename
365
split = os.path.split
366
splitext = os.path.splitext
367
# These were already imported into local scope
368
# mkdtemp = tempfile.mkdtemp
369
# rmtree = shutil.rmtree
371
MIN_ABS_PATHLENGTH = 1
374
if sys.platform == 'win32':
375
if win32utils.winver == 'Windows 98':
376
abspath = _win98_abspath
378
abspath = _win32_abspath
379
realpath = _win32_realpath
380
pathjoin = _win32_pathjoin
381
normpath = _win32_normpath
382
getcwd = _win32_getcwd
383
mkdtemp = _win32_mkdtemp
384
rename = _win32_rename
386
MIN_ABS_PATHLENGTH = 3
388
def _win32_delete_readonly(function, path, excinfo):
389
"""Error handler for shutil.rmtree function [for win32]
390
Helps to remove files and dirs marked as read-only.
392
exception = excinfo[1]
393
if function in (os.remove, os.rmdir) \
394
and isinstance(exception, OSError) \
395
and exception.errno == errno.EACCES:
401
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
402
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
403
return shutil.rmtree(path, ignore_errors, onerror)
405
f = win32utils.get_unicode_argv # special function or None
409
elif sys.platform == 'darwin':
413
def get_terminal_encoding():
414
"""Find the best encoding for printing to the screen.
416
This attempts to check both sys.stdout and sys.stdin to see
417
what encoding they are in, and if that fails it falls back to
418
osutils.get_user_encoding().
419
The problem is that on Windows, locale.getpreferredencoding()
420
is not the same encoding as that used by the console:
421
http://mail.python.org/pipermail/python-list/2003-May/162357.html
423
On my standard US Windows XP, the preferred encoding is
424
cp1252, but the console is cp437
426
from bzrlib.trace import mutter
427
output_encoding = getattr(sys.stdout, 'encoding', None)
428
if not output_encoding:
429
input_encoding = getattr(sys.stdin, 'encoding', None)
430
if not input_encoding:
431
output_encoding = get_user_encoding()
432
mutter('encoding stdout as osutils.get_user_encoding() %r',
435
output_encoding = input_encoding
436
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
438
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
439
if output_encoding == 'cp0':
440
# invalid encoding (cp0 means 'no codepage' on Windows)
441
output_encoding = get_user_encoding()
442
mutter('cp0 is invalid encoding.'
443
' encoding stdout as osutils.get_user_encoding() %r',
447
codecs.lookup(output_encoding)
449
sys.stderr.write('bzr: warning:'
450
' unknown terminal encoding %s.\n'
451
' Using encoding %s instead.\n'
452
% (output_encoding, get_user_encoding())
454
output_encoding = get_user_encoding()
456
return output_encoding
459
def normalizepath(f):
460
if getattr(os.path, 'realpath', None) is not None:
464
[p,e] = os.path.split(f)
465
if e == "" or e == "." or e == "..":
468
return pathjoin(F(p), e)
430
raise BzrError("sorry, %r not allowed in path" % f)
857
raise errors.BzrError("sorry, %r not allowed in path" % f)
431
858
elif (f == '.') or (f == ''):
438
assert isinstance(p, list)
440
if (f == '..') or (f == None) or (f == ''):
441
raise BzrError("sorry, %r not allowed in path" % f)
442
return os.path.join(*p)
445
def appendpath(p1, p2):
449
return os.path.join(p1, p2)
452
def extern_command(cmd, ignore_errors = False):
453
mutter('external command: %s' % `cmd`)
455
if not ignore_errors:
456
raise BzrError('command failed')
459
def _read_config_value(name):
460
"""Read a config value from the file ~/.bzr.conf/<name>
461
Return None if the file does not exist"""
463
f = file(os.path.join(config_dir(), name), "r")
464
return f.read().decode(bzrlib.user_encoding).rstrip("\r\n")
466
if e.errno == errno.ENOENT:
867
if (f == '..') or (f is None) or (f == ''):
868
raise errors.BzrError("sorry, %r not allowed in path" % f)
872
def parent_directories(filename):
873
"""Return the list of parent directories, deepest first.
875
For example, parent_directories("a/b/c") -> ["a/b", "a"].
878
parts = splitpath(dirname(filename))
880
parents.append(joinpath(parts))
885
_extension_load_failures = []
888
def failed_to_load_extension(exception):
889
"""Handle failing to load a binary extension.
891
This should be called from the ImportError block guarding the attempt to
892
import the native extension. If this function returns, the pure-Python
893
implementation should be loaded instead::
896
>>> import bzrlib._fictional_extension_pyx
897
>>> except ImportError, e:
898
>>> bzrlib.osutils.failed_to_load_extension(e)
899
>>> import bzrlib._fictional_extension_py
901
# NB: This docstring is just an example, not a doctest, because doctest
902
# currently can't cope with the use of lazy imports in this namespace --
905
# This currently doesn't report the failure at the time it occurs, because
906
# they tend to happen very early in startup when we can't check config
907
# files etc, and also we want to report all failures but not spam the user
909
from bzrlib import trace
910
exception_str = str(exception)
911
if exception_str not in _extension_load_failures:
912
trace.mutter("failed to load compiled extension: %s" % exception_str)
913
_extension_load_failures.append(exception_str)
916
def report_extension_load_failures():
917
if not _extension_load_failures:
919
from bzrlib.config import GlobalConfig
920
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
922
# the warnings framework should by default show this only once
923
from bzrlib.trace import warning
925
"bzr: warning: some compiled extensions could not be loaded; "
926
"see <https://answers.launchpad.net/bzr/+faq/703>")
927
# we no longer show the specific missing extensions here, because it makes
928
# the message too long and scary - see
929
# https://bugs.launchpad.net/bzr/+bug/430529
933
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
934
except ImportError, e:
935
failed_to_load_extension(e)
936
from bzrlib._chunks_to_lines_py import chunks_to_lines
940
"""Split s into lines, but without removing the newline characters."""
941
# Trivially convert a fulltext into a 'chunked' representation, and let
942
# chunks_to_lines do the heavy lifting.
943
if isinstance(s, str):
944
# chunks_to_lines only supports 8-bit strings
945
return chunks_to_lines([s])
947
return _split_lines(s)
951
"""Split s into lines, but without removing the newline characters.
953
This supports Unicode or plain string objects.
955
lines = s.split('\n')
956
result = [line + '\n' for line in lines[:-1]]
958
result.append(lines[-1])
962
def hardlinks_good():
963
return sys.platform not in ('win32', 'cygwin', 'darwin')
966
def link_or_copy(src, dest):
967
"""Hardlink a file, or copy it if it can't be hardlinked."""
968
if not hardlinks_good():
969
shutil.copyfile(src, dest)
973
except (OSError, IOError), e:
974
if e.errno != errno.EXDEV:
976
shutil.copyfile(src, dest)
979
def delete_any(path):
980
"""Delete a file, symlink or directory.
982
Will delete even if readonly.
985
_delete_file_or_dir(path)
986
except (OSError, IOError), e:
987
if e.errno in (errno.EPERM, errno.EACCES):
988
# make writable and try again
991
except (OSError, IOError):
993
_delete_file_or_dir(path)
998
def _delete_file_or_dir(path):
999
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1000
# Forgiveness than Permission (EAFP) because:
1001
# - root can damage a solaris file system by using unlink,
1002
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1003
# EACCES, OSX: EPERM) when invoked on a directory.
1004
if isdir(path): # Takes care of symlinks
1011
if getattr(os, 'symlink', None) is not None:
1017
def has_hardlinks():
1018
if getattr(os, 'link', None) is not None:
1024
def host_os_dereferences_symlinks():
1025
return (has_symlinks()
1026
and sys.platform not in ('cygwin', 'win32'))
1029
def readlink(abspath):
1030
"""Return a string representing the path to which the symbolic link points.
1032
:param abspath: The link absolute unicode path.
1034
This his guaranteed to return the symbolic link in unicode in all python
1037
link = abspath.encode(_fs_enc)
1038
target = os.readlink(link)
1039
target = target.decode(_fs_enc)
1043
def contains_whitespace(s):
1044
"""True if there are any whitespace characters in s."""
1045
# string.whitespace can include '\xa0' in certain locales, because it is
1046
# considered "non-breaking-space" as part of ISO-8859-1. But it
1047
# 1) Isn't a breaking whitespace
1048
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1050
# 3) '\xa0' isn't unicode safe since it is >128.
1052
# This should *not* be a unicode set of characters in case the source
1053
# string is not a Unicode string. We can auto-up-cast the characters since
1054
# they are ascii, but we don't want to auto-up-cast the string in case it
1056
for ch in ' \t\n\r\v\f':
1063
def contains_linebreaks(s):
1064
"""True if there is any vertical whitespace in s."""
1072
def relpath(base, path):
1073
"""Return path relative to base, or raise exception.
1075
The path may be either an absolute path or a path relative to the
1076
current working directory.
1078
os.path.commonprefix (python2.4) has a bad bug that it works just
1079
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1080
avoids that problem.
1083
if len(base) < MIN_ABS_PATHLENGTH:
1084
# must have space for e.g. a drive letter
1085
raise ValueError('%r is too short to calculate a relative path'
1093
if len(head) <= len(base) and head != base:
1094
raise errors.PathNotChild(rp, base)
1097
head, tail = split(head)
1102
return pathjoin(*reversed(s))
1107
def _cicp_canonical_relpath(base, path):
1108
"""Return the canonical path relative to base.
1110
Like relpath, but on case-insensitive-case-preserving file-systems, this
1111
will return the relpath as stored on the file-system rather than in the
1112
case specified in the input string, for all existing portions of the path.
1114
This will cause O(N) behaviour if called for every path in a tree; if you
1115
have a number of paths to convert, you should use canonical_relpaths().
1117
# TODO: it should be possible to optimize this for Windows by using the
1118
# win32 API FindFiles function to look for the specified name - but using
1119
# os.listdir() still gives us the correct, platform agnostic semantics in
1122
rel = relpath(base, path)
1123
# '.' will have been turned into ''
1127
abs_base = abspath(base)
1129
_listdir = os.listdir
1131
# use an explicit iterator so we can easily consume the rest on early exit.
1132
bit_iter = iter(rel.split('/'))
1133
for bit in bit_iter:
1135
for look in _listdir(current):
1136
if lbit == look.lower():
1137
current = pathjoin(current, look)
1140
# got to the end, nothing matched, so we just return the
1141
# non-existing bits as they were specified (the filename may be
1142
# the target of a move, for example).
1143
current = pathjoin(current, bit, *list(bit_iter))
1145
return current[len(abs_base)+1:]
1147
# XXX - TODO - we need better detection/integration of case-insensitive
1148
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1149
# filesystems), for example, so could probably benefit from the same basic
1150
# support there. For now though, only Windows and OSX get that support, and
1151
# they get it for *all* file-systems!
1152
if sys.platform in ('win32', 'darwin'):
1153
canonical_relpath = _cicp_canonical_relpath
1155
canonical_relpath = relpath
1157
def canonical_relpaths(base, paths):
1158
"""Create an iterable to canonicalize a sequence of relative paths.
1160
The intent is for this implementation to use a cache, vastly speeding
1161
up multiple transformations in the same directory.
1163
# but for now, we haven't optimized...
1164
return [canonical_relpath(base, p) for p in paths]
1166
def safe_unicode(unicode_or_utf8_string):
1167
"""Coerce unicode_or_utf8_string into unicode.
1169
If it is unicode, it is returned.
1170
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1171
wrapped in a BzrBadParameterNotUnicode exception.
1173
if isinstance(unicode_or_utf8_string, unicode):
1174
return unicode_or_utf8_string
1176
return unicode_or_utf8_string.decode('utf8')
1177
except UnicodeDecodeError:
1178
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1181
def safe_utf8(unicode_or_utf8_string):
1182
"""Coerce unicode_or_utf8_string to a utf8 string.
1184
If it is a str, it is returned.
1185
If it is Unicode, it is encoded into a utf-8 string.
1187
if isinstance(unicode_or_utf8_string, str):
1188
# TODO: jam 20070209 This is overkill, and probably has an impact on
1189
# performance if we are dealing with lots of apis that want a
1192
# Make sure it is a valid utf-8 string
1193
unicode_or_utf8_string.decode('utf-8')
1194
except UnicodeDecodeError:
1195
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1196
return unicode_or_utf8_string
1197
return unicode_or_utf8_string.encode('utf-8')
1200
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1201
' Revision id generators should be creating utf8'
1205
def safe_revision_id(unicode_or_utf8_string, warn=True):
1206
"""Revision ids should now be utf8, but at one point they were unicode.
1208
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1210
:param warn: Functions that are sanitizing user data can set warn=False
1211
:return: None or a utf8 revision id.
1213
if (unicode_or_utf8_string is None
1214
or unicode_or_utf8_string.__class__ == str):
1215
return unicode_or_utf8_string
1217
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1219
return cache_utf8.encode(unicode_or_utf8_string)
1222
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1223
' generators should be creating utf8 file ids.')
1226
def safe_file_id(unicode_or_utf8_string, warn=True):
1227
"""File ids should now be utf8, but at one point they were unicode.
1229
This is the same as safe_utf8, except it uses the cached encode functions
1230
to save a little bit of performance.
1232
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1234
:param warn: Functions that are sanitizing user data can set warn=False
1235
:return: None or a utf8 file id.
1237
if (unicode_or_utf8_string is None
1238
or unicode_or_utf8_string.__class__ == str):
1239
return unicode_or_utf8_string
1241
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1243
return cache_utf8.encode(unicode_or_utf8_string)
1246
_platform_normalizes_filenames = False
1247
if sys.platform == 'darwin':
1248
_platform_normalizes_filenames = True
1251
def normalizes_filenames():
1252
"""Return True if this platform normalizes unicode filenames.
1254
Mac OSX does, Windows/Linux do not.
1256
return _platform_normalizes_filenames
1259
def _accessible_normalized_filename(path):
1260
"""Get the unicode normalized path, and if you can access the file.
1262
On platforms where the system normalizes filenames (Mac OSX),
1263
you can access a file by any path which will normalize correctly.
1264
On platforms where the system does not normalize filenames
1265
(Windows, Linux), you have to access a file by its exact path.
1267
Internally, bzr only supports NFC normalization, since that is
1268
the standard for XML documents.
1270
So return the normalized path, and a flag indicating if the file
1271
can be accessed by that path.
1274
return unicodedata.normalize('NFC', unicode(path)), True
1277
def _inaccessible_normalized_filename(path):
1278
__doc__ = _accessible_normalized_filename.__doc__
1280
normalized = unicodedata.normalize('NFC', unicode(path))
1281
return normalized, normalized == path
1284
if _platform_normalizes_filenames:
1285
normalized_filename = _accessible_normalized_filename
1287
normalized_filename = _inaccessible_normalized_filename
1290
def terminal_width():
1291
"""Return estimated terminal width."""
1292
if sys.platform == 'win32':
1293
return win32utils.get_console_size()[0]
1296
import struct, fcntl, termios
1297
s = struct.pack('HHHH', 0, 0, 0, 0)
1298
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1299
width = struct.unpack('HHHH', x)[1]
1304
width = int(os.environ['COLUMNS'])
1313
def supports_executable():
1314
return sys.platform != "win32"
1317
def supports_posix_readonly():
1318
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1320
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1321
directory controls creation/deletion, etc.
1323
And under win32, readonly means that the directory itself cannot be
1324
deleted. The contents of a readonly directory can be changed, unlike POSIX
1325
where files in readonly directories cannot be added, deleted or renamed.
1327
return sys.platform != "win32"
1330
def set_or_unset_env(env_variable, value):
1331
"""Modify the environment, setting or removing the env_variable.
1333
:param env_variable: The environment variable in question
1334
:param value: The value to set the environment to. If None, then
1335
the variable will be removed.
1336
:return: The original value of the environment variable.
1338
orig_val = os.environ.get(env_variable)
1340
if orig_val is not None:
1341
del os.environ[env_variable]
1343
if isinstance(value, unicode):
1344
value = value.encode(get_user_encoding())
1345
os.environ[env_variable] = value
1349
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1352
def check_legal_path(path):
1353
"""Check whether the supplied path is legal.
1354
This is only required on Windows, so we don't test on other platforms
1357
if sys.platform != "win32":
1359
if _validWin32PathRE.match(path) is None:
1360
raise errors.IllegalPath(path)
1363
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1365
def _is_error_enotdir(e):
1366
"""Check if this exception represents ENOTDIR.
1368
Unfortunately, python is very inconsistent about the exception
1369
here. The cases are:
1370
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1371
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1372
which is the windows error code.
1373
3) Windows, Python2.5 uses errno == EINVAL and
1374
winerror == ERROR_DIRECTORY
1376
:param e: An Exception object (expected to be OSError with an errno
1377
attribute, but we should be able to cope with anything)
1378
:return: True if this represents an ENOTDIR error. False otherwise.
1380
en = getattr(e, 'errno', None)
1381
if (en == errno.ENOTDIR
1382
or (sys.platform == 'win32'
1383
and (en == _WIN32_ERROR_DIRECTORY
1384
or (en == errno.EINVAL
1385
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1391
def walkdirs(top, prefix=""):
1392
"""Yield data about all the directories in a tree.
1394
This yields all the data about the contents of a directory at a time.
1395
After each directory has been yielded, if the caller has mutated the list
1396
to exclude some directories, they are then not descended into.
1398
The data yielded is of the form:
1399
((directory-relpath, directory-path-from-top),
1400
[(relpath, basename, kind, lstat, path-from-top), ...]),
1401
- directory-relpath is the relative path of the directory being returned
1402
with respect to top. prefix is prepended to this.
1403
- directory-path-from-root is the path including top for this directory.
1404
It is suitable for use with os functions.
1405
- relpath is the relative path within the subtree being walked.
1406
- basename is the basename of the path
1407
- kind is the kind of the file now. If unknown then the file is not
1408
present within the tree - but it may be recorded as versioned. See
1410
- lstat is the stat data *if* the file was statted.
1411
- planned, not implemented:
1412
path_from_tree_root is the path from the root of the tree.
1414
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1415
allows one to walk a subtree but get paths that are relative to a tree
1417
:return: an iterator over the dirs.
1419
#TODO there is a bit of a smell where the results of the directory-
1420
# summary in this, and the path from the root, may not agree
1421
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1422
# potentially confusing output. We should make this more robust - but
1423
# not at a speed cost. RBC 20060731
1425
_directory = _directory_kind
1426
_listdir = os.listdir
1427
_kind_from_mode = file_kind_from_stat_mode
1428
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1430
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1431
relroot, _, _, _, top = pending.pop()
1433
relprefix = relroot + u'/'
1436
top_slash = top + u'/'
1439
append = dirblock.append
1441
names = sorted(_listdir(top))
1443
if not _is_error_enotdir(e):
1447
abspath = top_slash + name
1448
statvalue = _lstat(abspath)
1449
kind = _kind_from_mode(statvalue.st_mode)
1450
append((relprefix + name, name, kind, statvalue, abspath))
1451
yield (relroot, top), dirblock
1453
# push the user specified dirs from dirblock
1454
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1457
class DirReader(object):
1458
"""An interface for reading directories."""
1460
def top_prefix_to_starting_dir(self, top, prefix=""):
1461
"""Converts top and prefix to a starting dir entry
1463
:param top: A utf8 path
1464
:param prefix: An optional utf8 path to prefix output relative paths
1466
:return: A tuple starting with prefix, and ending with the native
1469
raise NotImplementedError(self.top_prefix_to_starting_dir)
1471
def read_dir(self, prefix, top):
1472
"""Read a specific dir.
1474
:param prefix: A utf8 prefix to be preprended to the path basenames.
1475
:param top: A natively encoded path to read.
1476
:return: A list of the directories contents. Each item contains:
1477
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1479
raise NotImplementedError(self.read_dir)
1482
_selected_dir_reader = None
1485
def _walkdirs_utf8(top, prefix=""):
1486
"""Yield data about all the directories in a tree.
1488
This yields the same information as walkdirs() only each entry is yielded
1489
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1490
are returned as exact byte-strings.
1492
:return: yields a tuple of (dir_info, [file_info])
1493
dir_info is (utf8_relpath, path-from-top)
1494
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1495
if top is an absolute path, path-from-top is also an absolute path.
1496
path-from-top might be unicode or utf8, but it is the correct path to
1497
pass to os functions to affect the file in question. (such as os.lstat)
1499
global _selected_dir_reader
1500
if _selected_dir_reader is None:
1501
fs_encoding = _fs_enc.upper()
1502
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1503
# Win98 doesn't have unicode apis like FindFirstFileW
1504
# TODO: We possibly could support Win98 by falling back to the
1505
# original FindFirstFile, and using TCHAR instead of WCHAR,
1506
# but that gets a bit tricky, and requires custom compiling
1509
from bzrlib._walkdirs_win32 import Win32ReadDir
1510
_selected_dir_reader = Win32ReadDir()
1513
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1514
# ANSI_X3.4-1968 is a form of ASCII
1516
from bzrlib._readdir_pyx import UTF8DirReader
1517
_selected_dir_reader = UTF8DirReader()
1518
except ImportError, e:
1519
failed_to_load_extension(e)
1522
if _selected_dir_reader is None:
1523
# Fallback to the python version
1524
_selected_dir_reader = UnicodeDirReader()
1526
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1527
# But we don't actually uses 1-3 in pending, so set them to None
1528
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1529
read_dir = _selected_dir_reader.read_dir
1530
_directory = _directory_kind
1532
relroot, _, _, _, top = pending[-1].pop()
1535
dirblock = sorted(read_dir(relroot, top))
1536
yield (relroot, top), dirblock
1537
# push the user specified dirs from dirblock
1538
next = [d for d in reversed(dirblock) if d[2] == _directory]
1540
pending.append(next)
1543
class UnicodeDirReader(DirReader):
1544
"""A dir reader for non-utf8 file systems, which transcodes."""
1546
__slots__ = ['_utf8_encode']
1549
self._utf8_encode = codecs.getencoder('utf8')
1551
def top_prefix_to_starting_dir(self, top, prefix=""):
1552
"""See DirReader.top_prefix_to_starting_dir."""
1553
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1555
def read_dir(self, prefix, top):
1556
"""Read a single directory from a non-utf8 file system.
1558
top, and the abspath element in the output are unicode, all other paths
1559
are utf8. Local disk IO is done via unicode calls to listdir etc.
1561
This is currently the fallback code path when the filesystem encoding is
1562
not UTF-8. It may be better to implement an alternative so that we can
1563
safely handle paths that are not properly decodable in the current
1566
See DirReader.read_dir for details.
1568
_utf8_encode = self._utf8_encode
1570
_listdir = os.listdir
1571
_kind_from_mode = file_kind_from_stat_mode
1574
relprefix = prefix + '/'
1577
top_slash = top + u'/'
1580
append = dirblock.append
1581
for name in sorted(_listdir(top)):
1583
name_utf8 = _utf8_encode(name)[0]
1584
except UnicodeDecodeError:
1585
raise errors.BadFilenameEncoding(
1586
_utf8_encode(relprefix)[0] + name, _fs_enc)
1587
abspath = top_slash + name
1588
statvalue = _lstat(abspath)
1589
kind = _kind_from_mode(statvalue.st_mode)
1590
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1594
def copy_tree(from_path, to_path, handlers={}):
1595
"""Copy all of the entries in from_path into to_path.
1597
:param from_path: The base directory to copy.
1598
:param to_path: The target directory. If it does not exist, it will
1600
:param handlers: A dictionary of functions, which takes a source and
1601
destinations for files, directories, etc.
1602
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1603
'file', 'directory', and 'symlink' should always exist.
1604
If they are missing, they will be replaced with 'os.mkdir()',
1605
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1607
# Now, just copy the existing cached tree to the new location
1608
# We use a cheap trick here.
1609
# Absolute paths are prefixed with the first parameter
1610
# relative paths are prefixed with the second.
1611
# So we can get both the source and target returned
1612
# without any extra work.
1614
def copy_dir(source, dest):
1617
def copy_link(source, dest):
1618
"""Copy the contents of a symlink"""
1619
link_to = os.readlink(source)
1620
os.symlink(link_to, dest)
1622
real_handlers = {'file':shutil.copy2,
1623
'symlink':copy_link,
1624
'directory':copy_dir,
1626
real_handlers.update(handlers)
1628
if not os.path.exists(to_path):
1629
real_handlers['directory'](from_path, to_path)
1631
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1632
for relpath, name, kind, st, abspath in entries:
1633
real_handlers[kind](abspath, relpath)
1636
def path_prefix_key(path):
1637
"""Generate a prefix-order path key for path.
1639
This can be used to sort paths in the same way that walkdirs does.
1641
return (dirname(path) , path)
1644
def compare_paths_prefix_order(path_a, path_b):
1645
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1646
key_a = path_prefix_key(path_a)
1647
key_b = path_prefix_key(path_b)
1648
return cmp(key_a, key_b)
1651
_cached_user_encoding = None
1654
def get_user_encoding(use_cache=True):
1655
"""Find out what the preferred user encoding is.
1657
This is generally the encoding that is used for command line parameters
1658
and file contents. This may be different from the terminal encoding
1659
or the filesystem encoding.
1661
:param use_cache: Enable cache for detected encoding.
1662
(This parameter is turned on by default,
1663
and required only for selftesting)
1665
:return: A string defining the preferred user encoding
1667
global _cached_user_encoding
1668
if _cached_user_encoding is not None and use_cache:
1669
return _cached_user_encoding
1671
if sys.platform == 'darwin':
1672
# python locale.getpreferredencoding() always return
1673
# 'mac-roman' on darwin. That's a lie.
1674
sys.platform = 'posix'
1676
if os.environ.get('LANG', None) is None:
1677
# If LANG is not set, we end up with 'ascii', which is bad
1678
# ('mac-roman' is more than ascii), so we set a default which
1679
# will give us UTF-8 (which appears to work in all cases on
1680
# OSX). Users are still free to override LANG of course, as
1681
# long as it give us something meaningful. This work-around
1682
# *may* not be needed with python 3k and/or OSX 10.5, but will
1683
# work with them too -- vila 20080908
1684
os.environ['LANG'] = 'en_US.UTF-8'
1687
sys.platform = 'darwin'
1692
user_encoding = locale.getpreferredencoding()
1693
except locale.Error, e:
1694
sys.stderr.write('bzr: warning: %s\n'
1695
' Could not determine what text encoding to use.\n'
1696
' This error usually means your Python interpreter\n'
1697
' doesn\'t support the locale set by $LANG (%s)\n'
1698
" Continuing with ascii encoding.\n"
1699
% (e, os.environ.get('LANG')))
1700
user_encoding = 'ascii'
1702
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1703
# treat that as ASCII, and not support printing unicode characters to the
1706
# For python scripts run under vim, we get '', so also treat that as ASCII
1707
if user_encoding in (None, 'cp0', ''):
1708
user_encoding = 'ascii'
1712
codecs.lookup(user_encoding)
1714
sys.stderr.write('bzr: warning:'
1715
' unknown encoding %s.'
1716
' Continuing with ascii encoding.\n'
1719
user_encoding = 'ascii'
1722
_cached_user_encoding = user_encoding
1724
return user_encoding
1727
def get_host_name():
1728
"""Return the current unicode host name.
1730
This is meant to be used in place of socket.gethostname() because that
1731
behaves inconsistently on different platforms.
1733
if sys.platform == "win32":
1735
return win32utils.get_host_name()
1738
return socket.gethostname().decode(get_user_encoding())
1741
def recv_all(socket, bytes):
1742
"""Receive an exact number of bytes.
1744
Regular Socket.recv() may return less than the requested number of bytes,
1745
dependning on what's in the OS buffer. MSG_WAITALL is not available
1746
on all platforms, but this should work everywhere. This will return
1747
less than the requested amount if the remote end closes.
1749
This isn't optimized and is intended mostly for use in testing.
1752
while len(b) < bytes:
1753
new = until_no_eintr(socket.recv, bytes - len(b))
1760
def send_all(socket, bytes, report_activity=None):
1761
"""Send all bytes on a socket.
1763
Regular socket.sendall() can give socket error 10053 on Windows. This
1764
implementation sends no more than 64k at a time, which avoids this problem.
1766
:param report_activity: Call this as bytes are read, see
1767
Transport._report_activity
1770
for pos in xrange(0, len(bytes), chunk_size):
1771
block = bytes[pos:pos+chunk_size]
1772
if report_activity is not None:
1773
report_activity(len(block), 'write')
1774
until_no_eintr(socket.sendall, block)
1777
def dereference_path(path):
1778
"""Determine the real path to a file.
1780
All parent elements are dereferenced. But the file itself is not
1782
:param path: The original path. May be absolute or relative.
1783
:return: the real path *to* the file
1785
parent, base = os.path.split(path)
1786
# The pathjoin for '.' is a workaround for Python bug #1213894.
1787
# (initial path components aren't dereferenced)
1788
return pathjoin(realpath(pathjoin('.', parent)), base)
1791
def supports_mapi():
1792
"""Return True if we can use MAPI to launch a mail client."""
1793
return sys.platform == "win32"
1796
def resource_string(package, resource_name):
1797
"""Load a resource from a package and return it as a string.
1799
Note: Only packages that start with bzrlib are currently supported.
1801
This is designed to be a lightweight implementation of resource
1802
loading in a way which is API compatible with the same API from
1804
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1805
If and when pkg_resources becomes a standard library, this routine
1808
# Check package name is within bzrlib
1809
if package == "bzrlib":
1810
resource_relpath = resource_name
1811
elif package.startswith("bzrlib."):
1812
package = package[len("bzrlib."):].replace('.', os.sep)
1813
resource_relpath = pathjoin(package, resource_name)
1815
raise errors.BzrError('resource package %s not in bzrlib' % package)
1817
# Map the resource to a file and read its contents
1818
base = dirname(bzrlib.__file__)
1819
if getattr(sys, 'frozen', None): # bzr.exe
1820
base = abspath(pathjoin(base, '..', '..'))
1821
filename = pathjoin(base, resource_relpath)
1822
return open(filename, 'rU').read()
1825
def file_kind_from_stat_mode_thunk(mode):
1826
global file_kind_from_stat_mode
1827
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1829
from bzrlib._readdir_pyx import UTF8DirReader
1830
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1831
except ImportError, e:
1832
# This is one time where we won't warn that an extension failed to
1833
# load. The extension is never available on Windows anyway.
1834
from bzrlib._readdir_py import (
1835
_kind_from_mode as file_kind_from_stat_mode
1837
return file_kind_from_stat_mode(mode)
1838
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1841
def file_kind(f, _lstat=os.lstat):
1843
return file_kind_from_stat_mode(_lstat(f).st_mode)
1845
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1846
raise errors.NoSuchFile(f)
472
"""Return a sequence of possible editor binaries for the current platform"""
473
e = _read_config_value("editor")
477
if os.name == "windows":
479
elif os.name == "posix":
481
yield os.environ["EDITOR"]
486
def _run_editor(filename):
487
"""Try to execute an editor to edit the commit message. Returns True on success,
489
for e in _get_editor():
490
x = os.spawnvp(os.P_WAIT, e, (e, filename))
497
raise BzrError("Could not start any editor. Please specify $EDITOR or use ~/.bzr.conf/editor")
501
def get_text_message(infotext, ignoreline = "default"):
504
if ignoreline == "default":
505
ignoreline = "-- This line and the following will be ignored --"
508
tmp_fileno, msgfilename = tempfile.mkstemp()
509
msgfile = os.close(tmp_fileno)
510
if infotext is not None and infotext != "":
512
msgfile = file(msgfilename, "w")
513
msgfile.write("\n\n%s\n\n%s" % (ignoreline, infotext))
518
if not _run_editor(msgfilename):
523
lastline, nlines = 0, 0
524
for line in file(msgfilename, "r"):
525
stripped_line = line.strip()
526
# strip empty line before the log message starts
528
if stripped_line != "":
532
# check for the ignore line only if there
533
# is additional information at the end
534
if hasinfo and stripped_line == ignoreline:
537
# keep track of the last line that had some content
538
if stripped_line != "":
544
# delete empty lines at the end
546
# add a newline at the end, if needed
547
if not msg[-1].endswith("\n"):
548
return "%s%s" % ("".join(msg), "\n")
552
# delete the msg file in any case
553
try: os.unlink(msgfilename)
1850
def until_no_eintr(f, *a, **kw):
1851
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1852
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1856
except (IOError, OSError), e:
1857
if e.errno == errno.EINTR:
1861
def re_compile_checked(re_string, flags=0, where=""):
1862
"""Return a compiled re, or raise a sensible error.
1864
This should only be used when compiling user-supplied REs.
1866
:param re_string: Text form of regular expression.
1867
:param flags: eg re.IGNORECASE
1868
:param where: Message explaining to the user the context where
1869
it occurred, eg 'log search filter'.
1871
# from https://bugs.launchpad.net/bzr/+bug/251352
1873
re_obj = re.compile(re_string, flags)
1878
where = ' in ' + where
1879
# despite the name 'error' is a type
1880
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
1881
% (where, re_string, e))
1884
if sys.platform == "win32":
1887
return msvcrt.getch()
1892
fd = sys.stdin.fileno()
1893
settings = termios.tcgetattr(fd)
1896
ch = sys.stdin.read(1)
1898
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
1902
if sys.platform == 'linux2':
1903
def _local_concurrency():
1905
prefix = 'processor'
1906
for line in file('/proc/cpuinfo', 'rb'):
1907
if line.startswith(prefix):
1908
concurrency = int(line[line.find(':')+1:]) + 1
1910
elif sys.platform == 'darwin':
1911
def _local_concurrency():
1912
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
1913
stdout=subprocess.PIPE).communicate()[0]
1914
elif sys.platform[0:7] == 'freebsd':
1915
def _local_concurrency():
1916
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
1917
stdout=subprocess.PIPE).communicate()[0]
1918
elif sys.platform == 'sunos5':
1919
def _local_concurrency():
1920
return subprocess.Popen(['psrinfo', '-p',],
1921
stdout=subprocess.PIPE).communicate()[0]
1922
elif sys.platform == "win32":
1923
def _local_concurrency():
1924
# This appears to return the number of cores.
1925
return os.environ.get('NUMBER_OF_PROCESSORS')
1927
def _local_concurrency():
1932
_cached_local_concurrency = None
1934
def local_concurrency(use_cache=True):
1935
"""Return how many processes can be run concurrently.
1937
Rely on platform specific implementations and default to 1 (one) if
1938
anything goes wrong.
1940
global _cached_local_concurrency
1941
if _cached_local_concurrency is not None and use_cache:
1942
return _cached_local_concurrency
1945
concurrency = _local_concurrency()
1946
except (OSError, IOError):
1949
concurrency = int(concurrency)
1950
except (TypeError, ValueError):
1953
_cached_concurrency = concurrency
added
removed
bzrlib/osutils.py
