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:
112
os.rename(path_from, path_to)
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)):
231
# This may throw an exception, in which case success will
233
rename_func(old, new)
235
except (IOError, OSError), e:
236
# source and target may be aliases of each other (e.g. on a
237
# case-insensitive filesystem), so we may have accidentally renamed
238
# source by when we tried to rename target
239
failure_exc = sys.exc_info()
240
if (file_existed and e.errno in (None, errno.ENOENT)
241
and old.lower() == new.lower()):
242
# source and target are the same file on a case-insensitive
243
# filesystem, so we don't generate an exception
247
# If the file used to exist, rename it back into place
248
# otherwise just delete it from the tmp location
250
unlink_func(tmp_name)
252
rename_func(tmp_name, new)
253
if failure_exc is not None:
254
raise failure_exc[0], failure_exc[1], failure_exc[2]
257
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
258
# choke on a Unicode string containing a relative path if
259
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
261
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
262
def _posix_abspath(path):
263
# jam 20060426 rather than encoding to fsencoding
264
# copy posixpath.abspath, but use os.getcwdu instead
265
if not posixpath.isabs(path):
266
path = posixpath.join(getcwd(), path)
267
return posixpath.normpath(path)
270
def _posix_realpath(path):
271
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
274
def _win32_fixdrive(path):
275
"""Force drive letters to be consistent.
277
win32 is inconsistent whether it returns lower or upper case
278
and even if it was consistent the user might type the other
279
so we force it to uppercase
280
running python.exe under cmd.exe return capital C:\\
281
running win32 python inside a cygwin shell returns lowercase c:\\
283
drive, path = _nt_splitdrive(path)
284
return drive.upper() + path
287
def _win32_abspath(path):
288
# Real _nt_abspath doesn't have a problem with a unicode cwd
289
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
292
def _win98_abspath(path):
293
"""Return the absolute version of a path.
294
Windows 98 safe implementation (python reimplementation
295
of Win32 API function GetFullPathNameW)
300
# \\HOST\path => //HOST/path
301
# //HOST/path => //HOST/path
302
# path => C:/cwd/path
305
# check for absolute path
306
drive = _nt_splitdrive(path)[0]
307
if drive == '' and path[:2] not in('//','\\\\'):
309
# we cannot simply os.path.join cwd and path
310
# because os.path.join('C:','/path') produce '/path'
311
# and this is incorrect
312
if path[:1] in ('/','\\'):
313
cwd = _nt_splitdrive(cwd)[0]
315
path = cwd + '\\' + path
316
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
319
def _win32_realpath(path):
320
# Real _nt_realpath doesn't have a problem with a unicode cwd
321
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
324
def _win32_pathjoin(*args):
325
return _nt_join(*args).replace('\\', '/')
328
def _win32_normpath(path):
329
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
333
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
336
def _win32_mkdtemp(*args, **kwargs):
337
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
340
def _win32_rename(old, new):
341
"""We expect to be able to atomically replace 'new' with old.
343
On win32, if new exists, it must be moved out of the way first,
347
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
349
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
350
# If we try to rename a non-existant file onto cwd, we get
351
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
352
# if the old path doesn't exist, sometimes we get EACCES
353
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
359
return unicodedata.normalize('NFC', os.getcwdu())
362
# Default is to just use the python builtins, but these can be rebound on
363
# particular platforms.
364
abspath = _posix_abspath
365
realpath = _posix_realpath
366
pathjoin = os.path.join
367
normpath = os.path.normpath
370
dirname = os.path.dirname
371
basename = os.path.basename
372
split = os.path.split
373
splitext = os.path.splitext
374
# These were already imported into local scope
375
# mkdtemp = tempfile.mkdtemp
376
# rmtree = shutil.rmtree
378
MIN_ABS_PATHLENGTH = 1
381
if sys.platform == 'win32':
382
if win32utils.winver == 'Windows 98':
383
abspath = _win98_abspath
385
abspath = _win32_abspath
386
realpath = _win32_realpath
387
pathjoin = _win32_pathjoin
388
normpath = _win32_normpath
389
getcwd = _win32_getcwd
390
mkdtemp = _win32_mkdtemp
391
rename = _win32_rename
393
MIN_ABS_PATHLENGTH = 3
395
def _win32_delete_readonly(function, path, excinfo):
396
"""Error handler for shutil.rmtree function [for win32]
397
Helps to remove files and dirs marked as read-only.
399
exception = excinfo[1]
400
if function in (os.remove, os.rmdir) \
401
and isinstance(exception, OSError) \
402
and exception.errno == errno.EACCES:
408
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
409
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
410
return shutil.rmtree(path, ignore_errors, onerror)
412
f = win32utils.get_unicode_argv # special function or None
416
elif sys.platform == 'darwin':
420
def get_terminal_encoding():
421
"""Find the best encoding for printing to the screen.
423
This attempts to check both sys.stdout and sys.stdin to see
424
what encoding they are in, and if that fails it falls back to
425
osutils.get_user_encoding().
426
The problem is that on Windows, locale.getpreferredencoding()
427
is not the same encoding as that used by the console:
428
http://mail.python.org/pipermail/python-list/2003-May/162357.html
430
On my standard US Windows XP, the preferred encoding is
431
cp1252, but the console is cp437
433
from bzrlib.trace import mutter
434
output_encoding = getattr(sys.stdout, 'encoding', None)
435
if not output_encoding:
436
input_encoding = getattr(sys.stdin, 'encoding', None)
437
if not input_encoding:
438
output_encoding = get_user_encoding()
439
mutter('encoding stdout as osutils.get_user_encoding() %r',
442
output_encoding = input_encoding
443
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
445
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
446
if output_encoding == 'cp0':
447
# invalid encoding (cp0 means 'no codepage' on Windows)
448
output_encoding = get_user_encoding()
449
mutter('cp0 is invalid encoding.'
450
' encoding stdout as osutils.get_user_encoding() %r',
454
codecs.lookup(output_encoding)
456
sys.stderr.write('bzr: warning:'
457
' unknown terminal encoding %s.\n'
458
' Using encoding %s instead.\n'
459
% (output_encoding, get_user_encoding())
461
output_encoding = get_user_encoding()
463
return output_encoding
466
def normalizepath(f):
467
if getattr(os.path, 'realpath', None) is not None:
471
[p,e] = os.path.split(f)
472
if e == "" or e == "." or e == "..":
475
return pathjoin(F(p), e)
430
raise BzrError("sorry, %r not allowed in path" % f)
864
raise errors.BzrError("sorry, %r not allowed in path" % f)
431
865
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:
874
if (f == '..') or (f is None) or (f == ''):
875
raise errors.BzrError("sorry, %r not allowed in path" % f)
879
def parent_directories(filename):
880
"""Return the list of parent directories, deepest first.
882
For example, parent_directories("a/b/c") -> ["a/b", "a"].
885
parts = splitpath(dirname(filename))
887
parents.append(joinpath(parts))
892
_extension_load_failures = []
895
def failed_to_load_extension(exception):
896
"""Handle failing to load a binary extension.
898
This should be called from the ImportError block guarding the attempt to
899
import the native extension. If this function returns, the pure-Python
900
implementation should be loaded instead::
903
>>> import bzrlib._fictional_extension_pyx
904
>>> except ImportError, e:
905
>>> bzrlib.osutils.failed_to_load_extension(e)
906
>>> import bzrlib._fictional_extension_py
908
# NB: This docstring is just an example, not a doctest, because doctest
909
# currently can't cope with the use of lazy imports in this namespace --
912
# This currently doesn't report the failure at the time it occurs, because
913
# they tend to happen very early in startup when we can't check config
914
# files etc, and also we want to report all failures but not spam the user
916
from bzrlib import trace
917
exception_str = str(exception)
918
if exception_str not in _extension_load_failures:
919
trace.mutter("failed to load compiled extension: %s" % exception_str)
920
_extension_load_failures.append(exception_str)
923
def report_extension_load_failures():
924
if not _extension_load_failures:
926
from bzrlib.config import GlobalConfig
927
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
929
# the warnings framework should by default show this only once
930
from bzrlib.trace import warning
932
"bzr: warning: some compiled extensions could not be loaded; "
933
"see <https://answers.launchpad.net/bzr/+faq/703>")
934
# we no longer show the specific missing extensions here, because it makes
935
# the message too long and scary - see
936
# https://bugs.launchpad.net/bzr/+bug/430529
940
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
941
except ImportError, e:
942
failed_to_load_extension(e)
943
from bzrlib._chunks_to_lines_py import chunks_to_lines
947
"""Split s into lines, but without removing the newline characters."""
948
# Trivially convert a fulltext into a 'chunked' representation, and let
949
# chunks_to_lines do the heavy lifting.
950
if isinstance(s, str):
951
# chunks_to_lines only supports 8-bit strings
952
return chunks_to_lines([s])
954
return _split_lines(s)
958
"""Split s into lines, but without removing the newline characters.
960
This supports Unicode or plain string objects.
962
lines = s.split('\n')
963
result = [line + '\n' for line in lines[:-1]]
965
result.append(lines[-1])
969
def hardlinks_good():
970
return sys.platform not in ('win32', 'cygwin', 'darwin')
973
def link_or_copy(src, dest):
974
"""Hardlink a file, or copy it if it can't be hardlinked."""
975
if not hardlinks_good():
976
shutil.copyfile(src, dest)
980
except (OSError, IOError), e:
981
if e.errno != errno.EXDEV:
983
shutil.copyfile(src, dest)
986
def delete_any(path):
987
"""Delete a file, symlink or directory.
989
Will delete even if readonly.
992
_delete_file_or_dir(path)
993
except (OSError, IOError), e:
994
if e.errno in (errno.EPERM, errno.EACCES):
995
# make writable and try again
998
except (OSError, IOError):
1000
_delete_file_or_dir(path)
1005
def _delete_file_or_dir(path):
1006
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1007
# Forgiveness than Permission (EAFP) because:
1008
# - root can damage a solaris file system by using unlink,
1009
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1010
# EACCES, OSX: EPERM) when invoked on a directory.
1011
if isdir(path): # Takes care of symlinks
1018
if getattr(os, 'symlink', None) is not None:
1024
def has_hardlinks():
1025
if getattr(os, 'link', None) is not None:
1031
def host_os_dereferences_symlinks():
1032
return (has_symlinks()
1033
and sys.platform not in ('cygwin', 'win32'))
1036
def readlink(abspath):
1037
"""Return a string representing the path to which the symbolic link points.
1039
:param abspath: The link absolute unicode path.
1041
This his guaranteed to return the symbolic link in unicode in all python
1044
link = abspath.encode(_fs_enc)
1045
target = os.readlink(link)
1046
target = target.decode(_fs_enc)
1050
def contains_whitespace(s):
1051
"""True if there are any whitespace characters in s."""
1052
# string.whitespace can include '\xa0' in certain locales, because it is
1053
# considered "non-breaking-space" as part of ISO-8859-1. But it
1054
# 1) Isn't a breaking whitespace
1055
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1057
# 3) '\xa0' isn't unicode safe since it is >128.
1059
# This should *not* be a unicode set of characters in case the source
1060
# string is not a Unicode string. We can auto-up-cast the characters since
1061
# they are ascii, but we don't want to auto-up-cast the string in case it
1063
for ch in ' \t\n\r\v\f':
1070
def contains_linebreaks(s):
1071
"""True if there is any vertical whitespace in s."""
1079
def relpath(base, path):
1080
"""Return path relative to base, or raise exception.
1082
The path may be either an absolute path or a path relative to the
1083
current working directory.
1085
os.path.commonprefix (python2.4) has a bad bug that it works just
1086
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1087
avoids that problem.
1090
if len(base) < MIN_ABS_PATHLENGTH:
1091
# must have space for e.g. a drive letter
1092
raise ValueError('%r is too short to calculate a relative path'
1100
if len(head) <= len(base) and head != base:
1101
raise errors.PathNotChild(rp, base)
1104
head, tail = split(head)
1109
return pathjoin(*reversed(s))
1114
def _cicp_canonical_relpath(base, path):
1115
"""Return the canonical path relative to base.
1117
Like relpath, but on case-insensitive-case-preserving file-systems, this
1118
will return the relpath as stored on the file-system rather than in the
1119
case specified in the input string, for all existing portions of the path.
1121
This will cause O(N) behaviour if called for every path in a tree; if you
1122
have a number of paths to convert, you should use canonical_relpaths().
1124
# TODO: it should be possible to optimize this for Windows by using the
1125
# win32 API FindFiles function to look for the specified name - but using
1126
# os.listdir() still gives us the correct, platform agnostic semantics in
1129
rel = relpath(base, path)
1130
# '.' will have been turned into ''
1134
abs_base = abspath(base)
1136
_listdir = os.listdir
1138
# use an explicit iterator so we can easily consume the rest on early exit.
1139
bit_iter = iter(rel.split('/'))
1140
for bit in bit_iter:
1143
next_entries = _listdir(current)
1144
except OSError: # enoent, eperm, etc
1145
# We can't find this in the filesystem, so just append the
1147
current = pathjoin(current, bit, *list(bit_iter))
1149
for look in next_entries:
1150
if lbit == look.lower():
1151
current = pathjoin(current, look)
1154
# got to the end, nothing matched, so we just return the
1155
# non-existing bits as they were specified (the filename may be
1156
# the target of a move, for example).
1157
current = pathjoin(current, bit, *list(bit_iter))
1159
return current[len(abs_base):].lstrip('/')
1161
# XXX - TODO - we need better detection/integration of case-insensitive
1162
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1163
# filesystems), for example, so could probably benefit from the same basic
1164
# support there. For now though, only Windows and OSX get that support, and
1165
# they get it for *all* file-systems!
1166
if sys.platform in ('win32', 'darwin'):
1167
canonical_relpath = _cicp_canonical_relpath
1169
canonical_relpath = relpath
1171
def canonical_relpaths(base, paths):
1172
"""Create an iterable to canonicalize a sequence of relative paths.
1174
The intent is for this implementation to use a cache, vastly speeding
1175
up multiple transformations in the same directory.
1177
# but for now, we haven't optimized...
1178
return [canonical_relpath(base, p) for p in paths]
1180
def safe_unicode(unicode_or_utf8_string):
1181
"""Coerce unicode_or_utf8_string into unicode.
1183
If it is unicode, it is returned.
1184
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1185
wrapped in a BzrBadParameterNotUnicode exception.
1187
if isinstance(unicode_or_utf8_string, unicode):
1188
return unicode_or_utf8_string
1190
return unicode_or_utf8_string.decode('utf8')
1191
except UnicodeDecodeError:
1192
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1195
def safe_utf8(unicode_or_utf8_string):
1196
"""Coerce unicode_or_utf8_string to a utf8 string.
1198
If it is a str, it is returned.
1199
If it is Unicode, it is encoded into a utf-8 string.
1201
if isinstance(unicode_or_utf8_string, str):
1202
# TODO: jam 20070209 This is overkill, and probably has an impact on
1203
# performance if we are dealing with lots of apis that want a
1206
# Make sure it is a valid utf-8 string
1207
unicode_or_utf8_string.decode('utf-8')
1208
except UnicodeDecodeError:
1209
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1210
return unicode_or_utf8_string
1211
return unicode_or_utf8_string.encode('utf-8')
1214
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1215
' Revision id generators should be creating utf8'
1219
def safe_revision_id(unicode_or_utf8_string, warn=True):
1220
"""Revision ids should now be utf8, but at one point they were unicode.
1222
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1224
:param warn: Functions that are sanitizing user data can set warn=False
1225
:return: None or a utf8 revision id.
1227
if (unicode_or_utf8_string is None
1228
or unicode_or_utf8_string.__class__ == str):
1229
return unicode_or_utf8_string
1231
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1233
return cache_utf8.encode(unicode_or_utf8_string)
1236
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1237
' generators should be creating utf8 file ids.')
1240
def safe_file_id(unicode_or_utf8_string, warn=True):
1241
"""File ids should now be utf8, but at one point they were unicode.
1243
This is the same as safe_utf8, except it uses the cached encode functions
1244
to save a little bit of performance.
1246
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1248
:param warn: Functions that are sanitizing user data can set warn=False
1249
:return: None or a utf8 file id.
1251
if (unicode_or_utf8_string is None
1252
or unicode_or_utf8_string.__class__ == str):
1253
return unicode_or_utf8_string
1255
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1257
return cache_utf8.encode(unicode_or_utf8_string)
1260
_platform_normalizes_filenames = False
1261
if sys.platform == 'darwin':
1262
_platform_normalizes_filenames = True
1265
def normalizes_filenames():
1266
"""Return True if this platform normalizes unicode filenames.
1268
Mac OSX does, Windows/Linux do not.
1270
return _platform_normalizes_filenames
1273
def _accessible_normalized_filename(path):
1274
"""Get the unicode normalized path, and if you can access the file.
1276
On platforms where the system normalizes filenames (Mac OSX),
1277
you can access a file by any path which will normalize correctly.
1278
On platforms where the system does not normalize filenames
1279
(Windows, Linux), you have to access a file by its exact path.
1281
Internally, bzr only supports NFC normalization, since that is
1282
the standard for XML documents.
1284
So return the normalized path, and a flag indicating if the file
1285
can be accessed by that path.
1288
return unicodedata.normalize('NFC', unicode(path)), True
1291
def _inaccessible_normalized_filename(path):
1292
__doc__ = _accessible_normalized_filename.__doc__
1294
normalized = unicodedata.normalize('NFC', unicode(path))
1295
return normalized, normalized == path
1298
if _platform_normalizes_filenames:
1299
normalized_filename = _accessible_normalized_filename
1301
normalized_filename = _inaccessible_normalized_filename
1304
def terminal_width():
1305
"""Return estimated terminal width."""
1306
if sys.platform == 'win32':
1307
return win32utils.get_console_size()[0]
1310
import struct, fcntl, termios
1311
s = struct.pack('HHHH', 0, 0, 0, 0)
1312
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1313
width = struct.unpack('HHHH', x)[1]
1318
width = int(os.environ['COLUMNS'])
1327
def supports_executable():
1328
return sys.platform != "win32"
1331
def supports_posix_readonly():
1332
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1334
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1335
directory controls creation/deletion, etc.
1337
And under win32, readonly means that the directory itself cannot be
1338
deleted. The contents of a readonly directory can be changed, unlike POSIX
1339
where files in readonly directories cannot be added, deleted or renamed.
1341
return sys.platform != "win32"
1344
def set_or_unset_env(env_variable, value):
1345
"""Modify the environment, setting or removing the env_variable.
1347
:param env_variable: The environment variable in question
1348
:param value: The value to set the environment to. If None, then
1349
the variable will be removed.
1350
:return: The original value of the environment variable.
1352
orig_val = os.environ.get(env_variable)
1354
if orig_val is not None:
1355
del os.environ[env_variable]
1357
if isinstance(value, unicode):
1358
value = value.encode(get_user_encoding())
1359
os.environ[env_variable] = value
1363
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1366
def check_legal_path(path):
1367
"""Check whether the supplied path is legal.
1368
This is only required on Windows, so we don't test on other platforms
1371
if sys.platform != "win32":
1373
if _validWin32PathRE.match(path) is None:
1374
raise errors.IllegalPath(path)
1377
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1379
def _is_error_enotdir(e):
1380
"""Check if this exception represents ENOTDIR.
1382
Unfortunately, python is very inconsistent about the exception
1383
here. The cases are:
1384
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1385
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1386
which is the windows error code.
1387
3) Windows, Python2.5 uses errno == EINVAL and
1388
winerror == ERROR_DIRECTORY
1390
:param e: An Exception object (expected to be OSError with an errno
1391
attribute, but we should be able to cope with anything)
1392
:return: True if this represents an ENOTDIR error. False otherwise.
1394
en = getattr(e, 'errno', None)
1395
if (en == errno.ENOTDIR
1396
or (sys.platform == 'win32'
1397
and (en == _WIN32_ERROR_DIRECTORY
1398
or (en == errno.EINVAL
1399
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1405
def walkdirs(top, prefix=""):
1406
"""Yield data about all the directories in a tree.
1408
This yields all the data about the contents of a directory at a time.
1409
After each directory has been yielded, if the caller has mutated the list
1410
to exclude some directories, they are then not descended into.
1412
The data yielded is of the form:
1413
((directory-relpath, directory-path-from-top),
1414
[(relpath, basename, kind, lstat, path-from-top), ...]),
1415
- directory-relpath is the relative path of the directory being returned
1416
with respect to top. prefix is prepended to this.
1417
- directory-path-from-root is the path including top for this directory.
1418
It is suitable for use with os functions.
1419
- relpath is the relative path within the subtree being walked.
1420
- basename is the basename of the path
1421
- kind is the kind of the file now. If unknown then the file is not
1422
present within the tree - but it may be recorded as versioned. See
1424
- lstat is the stat data *if* the file was statted.
1425
- planned, not implemented:
1426
path_from_tree_root is the path from the root of the tree.
1428
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1429
allows one to walk a subtree but get paths that are relative to a tree
1431
:return: an iterator over the dirs.
1433
#TODO there is a bit of a smell where the results of the directory-
1434
# summary in this, and the path from the root, may not agree
1435
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1436
# potentially confusing output. We should make this more robust - but
1437
# not at a speed cost. RBC 20060731
1439
_directory = _directory_kind
1440
_listdir = os.listdir
1441
_kind_from_mode = file_kind_from_stat_mode
1442
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1444
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1445
relroot, _, _, _, top = pending.pop()
1447
relprefix = relroot + u'/'
1450
top_slash = top + u'/'
1453
append = dirblock.append
1455
names = sorted(_listdir(top))
1457
if not _is_error_enotdir(e):
1461
abspath = top_slash + name
1462
statvalue = _lstat(abspath)
1463
kind = _kind_from_mode(statvalue.st_mode)
1464
append((relprefix + name, name, kind, statvalue, abspath))
1465
yield (relroot, top), dirblock
1467
# push the user specified dirs from dirblock
1468
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1471
class DirReader(object):
1472
"""An interface for reading directories."""
1474
def top_prefix_to_starting_dir(self, top, prefix=""):
1475
"""Converts top and prefix to a starting dir entry
1477
:param top: A utf8 path
1478
:param prefix: An optional utf8 path to prefix output relative paths
1480
:return: A tuple starting with prefix, and ending with the native
1483
raise NotImplementedError(self.top_prefix_to_starting_dir)
1485
def read_dir(self, prefix, top):
1486
"""Read a specific dir.
1488
:param prefix: A utf8 prefix to be preprended to the path basenames.
1489
:param top: A natively encoded path to read.
1490
:return: A list of the directories contents. Each item contains:
1491
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1493
raise NotImplementedError(self.read_dir)
1496
_selected_dir_reader = None
1499
def _walkdirs_utf8(top, prefix=""):
1500
"""Yield data about all the directories in a tree.
1502
This yields the same information as walkdirs() only each entry is yielded
1503
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1504
are returned as exact byte-strings.
1506
:return: yields a tuple of (dir_info, [file_info])
1507
dir_info is (utf8_relpath, path-from-top)
1508
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1509
if top is an absolute path, path-from-top is also an absolute path.
1510
path-from-top might be unicode or utf8, but it is the correct path to
1511
pass to os functions to affect the file in question. (such as os.lstat)
1513
global _selected_dir_reader
1514
if _selected_dir_reader is None:
1515
fs_encoding = _fs_enc.upper()
1516
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1517
# Win98 doesn't have unicode apis like FindFirstFileW
1518
# TODO: We possibly could support Win98 by falling back to the
1519
# original FindFirstFile, and using TCHAR instead of WCHAR,
1520
# but that gets a bit tricky, and requires custom compiling
1523
from bzrlib._walkdirs_win32 import Win32ReadDir
1524
_selected_dir_reader = Win32ReadDir()
1527
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1528
# ANSI_X3.4-1968 is a form of ASCII
1530
from bzrlib._readdir_pyx import UTF8DirReader
1531
_selected_dir_reader = UTF8DirReader()
1532
except ImportError, e:
1533
failed_to_load_extension(e)
1536
if _selected_dir_reader is None:
1537
# Fallback to the python version
1538
_selected_dir_reader = UnicodeDirReader()
1540
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1541
# But we don't actually uses 1-3 in pending, so set them to None
1542
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1543
read_dir = _selected_dir_reader.read_dir
1544
_directory = _directory_kind
1546
relroot, _, _, _, top = pending[-1].pop()
1549
dirblock = sorted(read_dir(relroot, top))
1550
yield (relroot, top), dirblock
1551
# push the user specified dirs from dirblock
1552
next = [d for d in reversed(dirblock) if d[2] == _directory]
1554
pending.append(next)
1557
class UnicodeDirReader(DirReader):
1558
"""A dir reader for non-utf8 file systems, which transcodes."""
1560
__slots__ = ['_utf8_encode']
1563
self._utf8_encode = codecs.getencoder('utf8')
1565
def top_prefix_to_starting_dir(self, top, prefix=""):
1566
"""See DirReader.top_prefix_to_starting_dir."""
1567
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1569
def read_dir(self, prefix, top):
1570
"""Read a single directory from a non-utf8 file system.
1572
top, and the abspath element in the output are unicode, all other paths
1573
are utf8. Local disk IO is done via unicode calls to listdir etc.
1575
This is currently the fallback code path when the filesystem encoding is
1576
not UTF-8. It may be better to implement an alternative so that we can
1577
safely handle paths that are not properly decodable in the current
1580
See DirReader.read_dir for details.
1582
_utf8_encode = self._utf8_encode
1584
_listdir = os.listdir
1585
_kind_from_mode = file_kind_from_stat_mode
1588
relprefix = prefix + '/'
1591
top_slash = top + u'/'
1594
append = dirblock.append
1595
for name in sorted(_listdir(top)):
1597
name_utf8 = _utf8_encode(name)[0]
1598
except UnicodeDecodeError:
1599
raise errors.BadFilenameEncoding(
1600
_utf8_encode(relprefix)[0] + name, _fs_enc)
1601
abspath = top_slash + name
1602
statvalue = _lstat(abspath)
1603
kind = _kind_from_mode(statvalue.st_mode)
1604
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1608
def copy_tree(from_path, to_path, handlers={}):
1609
"""Copy all of the entries in from_path into to_path.
1611
:param from_path: The base directory to copy.
1612
:param to_path: The target directory. If it does not exist, it will
1614
:param handlers: A dictionary of functions, which takes a source and
1615
destinations for files, directories, etc.
1616
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1617
'file', 'directory', and 'symlink' should always exist.
1618
If they are missing, they will be replaced with 'os.mkdir()',
1619
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1621
# Now, just copy the existing cached tree to the new location
1622
# We use a cheap trick here.
1623
# Absolute paths are prefixed with the first parameter
1624
# relative paths are prefixed with the second.
1625
# So we can get both the source and target returned
1626
# without any extra work.
1628
def copy_dir(source, dest):
1631
def copy_link(source, dest):
1632
"""Copy the contents of a symlink"""
1633
link_to = os.readlink(source)
1634
os.symlink(link_to, dest)
1636
real_handlers = {'file':shutil.copy2,
1637
'symlink':copy_link,
1638
'directory':copy_dir,
1640
real_handlers.update(handlers)
1642
if not os.path.exists(to_path):
1643
real_handlers['directory'](from_path, to_path)
1645
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1646
for relpath, name, kind, st, abspath in entries:
1647
real_handlers[kind](abspath, relpath)
1650
def path_prefix_key(path):
1651
"""Generate a prefix-order path key for path.
1653
This can be used to sort paths in the same way that walkdirs does.
1655
return (dirname(path) , path)
1658
def compare_paths_prefix_order(path_a, path_b):
1659
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1660
key_a = path_prefix_key(path_a)
1661
key_b = path_prefix_key(path_b)
1662
return cmp(key_a, key_b)
1665
_cached_user_encoding = None
1668
def get_user_encoding(use_cache=True):
1669
"""Find out what the preferred user encoding is.
1671
This is generally the encoding that is used for command line parameters
1672
and file contents. This may be different from the terminal encoding
1673
or the filesystem encoding.
1675
:param use_cache: Enable cache for detected encoding.
1676
(This parameter is turned on by default,
1677
and required only for selftesting)
1679
:return: A string defining the preferred user encoding
1681
global _cached_user_encoding
1682
if _cached_user_encoding is not None and use_cache:
1683
return _cached_user_encoding
1685
if sys.platform == 'darwin':
1686
# python locale.getpreferredencoding() always return
1687
# 'mac-roman' on darwin. That's a lie.
1688
sys.platform = 'posix'
1690
if os.environ.get('LANG', None) is None:
1691
# If LANG is not set, we end up with 'ascii', which is bad
1692
# ('mac-roman' is more than ascii), so we set a default which
1693
# will give us UTF-8 (which appears to work in all cases on
1694
# OSX). Users are still free to override LANG of course, as
1695
# long as it give us something meaningful. This work-around
1696
# *may* not be needed with python 3k and/or OSX 10.5, but will
1697
# work with them too -- vila 20080908
1698
os.environ['LANG'] = 'en_US.UTF-8'
1701
sys.platform = 'darwin'
1706
user_encoding = locale.getpreferredencoding()
1707
except locale.Error, e:
1708
sys.stderr.write('bzr: warning: %s\n'
1709
' Could not determine what text encoding to use.\n'
1710
' This error usually means your Python interpreter\n'
1711
' doesn\'t support the locale set by $LANG (%s)\n'
1712
" Continuing with ascii encoding.\n"
1713
% (e, os.environ.get('LANG')))
1714
user_encoding = 'ascii'
1716
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1717
# treat that as ASCII, and not support printing unicode characters to the
1720
# For python scripts run under vim, we get '', so also treat that as ASCII
1721
if user_encoding in (None, 'cp0', ''):
1722
user_encoding = 'ascii'
1726
codecs.lookup(user_encoding)
1728
sys.stderr.write('bzr: warning:'
1729
' unknown encoding %s.'
1730
' Continuing with ascii encoding.\n'
1733
user_encoding = 'ascii'
1736
_cached_user_encoding = user_encoding
1738
return user_encoding
1741
def get_host_name():
1742
"""Return the current unicode host name.
1744
This is meant to be used in place of socket.gethostname() because that
1745
behaves inconsistently on different platforms.
1747
if sys.platform == "win32":
1749
return win32utils.get_host_name()
1752
return socket.gethostname().decode(get_user_encoding())
1755
def recv_all(socket, bytes):
1756
"""Receive an exact number of bytes.
1758
Regular Socket.recv() may return less than the requested number of bytes,
1759
dependning on what's in the OS buffer. MSG_WAITALL is not available
1760
on all platforms, but this should work everywhere. This will return
1761
less than the requested amount if the remote end closes.
1763
This isn't optimized and is intended mostly for use in testing.
1766
while len(b) < bytes:
1767
new = until_no_eintr(socket.recv, bytes - len(b))
1774
def send_all(socket, bytes, report_activity=None):
1775
"""Send all bytes on a socket.
1777
Regular socket.sendall() can give socket error 10053 on Windows. This
1778
implementation sends no more than 64k at a time, which avoids this problem.
1780
:param report_activity: Call this as bytes are read, see
1781
Transport._report_activity
1784
for pos in xrange(0, len(bytes), chunk_size):
1785
block = bytes[pos:pos+chunk_size]
1786
if report_activity is not None:
1787
report_activity(len(block), 'write')
1788
until_no_eintr(socket.sendall, block)
1791
def dereference_path(path):
1792
"""Determine the real path to a file.
1794
All parent elements are dereferenced. But the file itself is not
1796
:param path: The original path. May be absolute or relative.
1797
:return: the real path *to* the file
1799
parent, base = os.path.split(path)
1800
# The pathjoin for '.' is a workaround for Python bug #1213894.
1801
# (initial path components aren't dereferenced)
1802
return pathjoin(realpath(pathjoin('.', parent)), base)
1805
def supports_mapi():
1806
"""Return True if we can use MAPI to launch a mail client."""
1807
return sys.platform == "win32"
1810
def resource_string(package, resource_name):
1811
"""Load a resource from a package and return it as a string.
1813
Note: Only packages that start with bzrlib are currently supported.
1815
This is designed to be a lightweight implementation of resource
1816
loading in a way which is API compatible with the same API from
1818
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1819
If and when pkg_resources becomes a standard library, this routine
1822
# Check package name is within bzrlib
1823
if package == "bzrlib":
1824
resource_relpath = resource_name
1825
elif package.startswith("bzrlib."):
1826
package = package[len("bzrlib."):].replace('.', os.sep)
1827
resource_relpath = pathjoin(package, resource_name)
1829
raise errors.BzrError('resource package %s not in bzrlib' % package)
1831
# Map the resource to a file and read its contents
1832
base = dirname(bzrlib.__file__)
1833
if getattr(sys, 'frozen', None): # bzr.exe
1834
base = abspath(pathjoin(base, '..', '..'))
1835
filename = pathjoin(base, resource_relpath)
1836
return open(filename, 'rU').read()
1839
def file_kind_from_stat_mode_thunk(mode):
1840
global file_kind_from_stat_mode
1841
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1843
from bzrlib._readdir_pyx import UTF8DirReader
1844
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1845
except ImportError, e:
1846
# This is one time where we won't warn that an extension failed to
1847
# load. The extension is never available on Windows anyway.
1848
from bzrlib._readdir_py import (
1849
_kind_from_mode as file_kind_from_stat_mode
1851
return file_kind_from_stat_mode(mode)
1852
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1855
def file_kind(f, _lstat=os.lstat):
1857
return file_kind_from_stat_mode(_lstat(f).st_mode)
1859
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1860
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)
1864
def until_no_eintr(f, *a, **kw):
1865
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1866
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1870
except (IOError, OSError), e:
1871
if e.errno == errno.EINTR:
1875
def re_compile_checked(re_string, flags=0, where=""):
1876
"""Return a compiled re, or raise a sensible error.
1878
This should only be used when compiling user-supplied REs.
1880
:param re_string: Text form of regular expression.
1881
:param flags: eg re.IGNORECASE
1882
:param where: Message explaining to the user the context where
1883
it occurred, eg 'log search filter'.
1885
# from https://bugs.launchpad.net/bzr/+bug/251352
1887
re_obj = re.compile(re_string, flags)
1892
where = ' in ' + where
1893
# despite the name 'error' is a type
1894
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
1895
% (where, re_string, e))
1898
if sys.platform == "win32":
1901
return msvcrt.getch()
1906
fd = sys.stdin.fileno()
1907
settings = termios.tcgetattr(fd)
1910
ch = sys.stdin.read(1)
1912
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
1916
if sys.platform == 'linux2':
1917
def _local_concurrency():
1919
prefix = 'processor'
1920
for line in file('/proc/cpuinfo', 'rb'):
1921
if line.startswith(prefix):
1922
concurrency = int(line[line.find(':')+1:]) + 1
1924
elif sys.platform == 'darwin':
1925
def _local_concurrency():
1926
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
1927
stdout=subprocess.PIPE).communicate()[0]
1928
elif sys.platform[0:7] == 'freebsd':
1929
def _local_concurrency():
1930
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
1931
stdout=subprocess.PIPE).communicate()[0]
1932
elif sys.platform == 'sunos5':
1933
def _local_concurrency():
1934
return subprocess.Popen(['psrinfo', '-p',],
1935
stdout=subprocess.PIPE).communicate()[0]
1936
elif sys.platform == "win32":
1937
def _local_concurrency():
1938
# This appears to return the number of cores.
1939
return os.environ.get('NUMBER_OF_PROCESSORS')
1941
def _local_concurrency():
1946
_cached_local_concurrency = None
1948
def local_concurrency(use_cache=True):
1949
"""Return how many processes can be run concurrently.
1951
Rely on platform specific implementations and default to 1 (one) if
1952
anything goes wrong.
1954
global _cached_local_concurrency
1955
if _cached_local_concurrency is not None and use_cache:
1956
return _cached_local_concurrency
1959
concurrency = _local_concurrency()
1960
except (OSError, IOError):
1963
concurrency = int(concurrency)
1964
except (TypeError, ValueError):
1967
_cached_concurrency = concurrency
added
removed
bzrlib/osutils.py
