34
24
realpath as _nt_realpath,
35
25
splitdrive as _nt_splitdrive,
28
from os import listdir
33
from shutil import copyfile
35
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
36
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
44
from tempfile import (
57
from bzrlib import symbol_versioning
58
from bzrlib.symbol_versioning import (
45
from bzrlib.errors import (BzrError,
46
BzrBadParameterNotUnicode,
51
from bzrlib.symbol_versioning import (deprecated_function,
62
53
from bzrlib.trace import mutter
65
# On win32, O_BINARY is used to indicate the file should
66
# be opened in binary mode, rather than text mode.
67
# On other platforms, O_BINARY doesn't exist, because
68
# they always open in binary mode, so it is okay to
69
# OR with 0 on those platforms
70
O_BINARY = getattr(os, 'O_BINARY', 0)
72
# On posix, use lstat instead of stat so that we can
73
# operate on broken symlinks. On Windows revert to stat.
74
lstat = getattr(os, 'lstat', os.stat)
76
56
def make_readonly(filename):
77
57
"""Make a filename read-only."""
78
mod = lstat(filename).st_mode
79
if not stat.S_ISLNK(mod):
81
os.chmod(filename, mod)
58
mod = os.stat(filename).st_mode
60
os.chmod(filename, mod)
84
63
def make_writable(filename):
85
mod = lstat(filename).st_mode
86
if not stat.S_ISLNK(mod):
88
os.chmod(filename, mod)
64
mod = os.stat(filename).st_mode
66
os.chmod(filename, mod)
606
530
def local_time_offset(t=None):
607
531
"""Return offset of local zone from GMT, either at present or at time t."""
532
# python2.3 localtime() can't take None
610
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
611
return offset.days * 86400 + offset.seconds
536
if time.localtime(t).tm_isdst and time.daylight:
539
return -time.timezone
614
def format_date(t, offset=0, timezone='original', date_fmt=None,
542
def format_date(t, offset=0, timezone='original', date_fmt=None,
615
543
show_offset=True):
616
"""Return a formatted date string.
618
:param t: Seconds since the epoch.
619
:param offset: Timezone offset in seconds east of utc.
620
:param timezone: How to display the time: 'utc', 'original' for the
621
timezone specified by offset, or 'local' for the process's current
623
:param show_offset: Whether to append the timezone.
624
:param date_fmt: strftime format.
544
## TODO: Perhaps a global option to use either universal or local time?
545
## Or perhaps just let people set $TZ?
546
assert isinstance(t, float)
626
548
if timezone == 'utc':
627
549
tt = time.gmtime(t)
629
551
elif timezone == 'original':
632
554
tt = time.gmtime(t + offset)
633
555
elif timezone == 'local':
634
556
tt = time.localtime(t)
635
557
offset = local_time_offset(t)
637
raise errors.BzrError("unsupported timezone format %r" % timezone,
638
['options are "utc", "original", "local"'])
559
raise BzrError("unsupported timezone format %r" % timezone,
560
['options are "utc", "original", "local"'])
639
561
if date_fmt is None:
640
562
date_fmt = "%a %Y-%m-%d %H:%M:%S"
896
773
return unicode_or_utf8_string.decode('utf8')
897
774
except UnicodeDecodeError:
898
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
901
def safe_utf8(unicode_or_utf8_string):
902
"""Coerce unicode_or_utf8_string to a utf8 string.
904
If it is a str, it is returned.
905
If it is Unicode, it is encoded into a utf-8 string.
907
if isinstance(unicode_or_utf8_string, str):
908
# TODO: jam 20070209 This is overkill, and probably has an impact on
909
# performance if we are dealing with lots of apis that want a
912
# Make sure it is a valid utf-8 string
913
unicode_or_utf8_string.decode('utf-8')
914
except UnicodeDecodeError:
915
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
916
return unicode_or_utf8_string
917
return unicode_or_utf8_string.encode('utf-8')
920
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
921
' Revision id generators should be creating utf8'
925
def safe_revision_id(unicode_or_utf8_string, warn=True):
926
"""Revision ids should now be utf8, but at one point they were unicode.
928
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
930
:param warn: Functions that are sanitizing user data can set warn=False
931
:return: None or a utf8 revision id.
933
if (unicode_or_utf8_string is None
934
or unicode_or_utf8_string.__class__ == str):
935
return unicode_or_utf8_string
937
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
939
return cache_utf8.encode(unicode_or_utf8_string)
942
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
943
' generators should be creating utf8 file ids.')
946
def safe_file_id(unicode_or_utf8_string, warn=True):
947
"""File ids should now be utf8, but at one point they were unicode.
949
This is the same as safe_utf8, except it uses the cached encode functions
950
to save a little bit of performance.
952
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
954
:param warn: Functions that are sanitizing user data can set warn=False
955
:return: None or a utf8 file id.
957
if (unicode_or_utf8_string is None
958
or unicode_or_utf8_string.__class__ == str):
959
return unicode_or_utf8_string
961
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
963
return cache_utf8.encode(unicode_or_utf8_string)
775
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
966
778
_platform_normalizes_filenames = False
976
788
return _platform_normalizes_filenames
979
def _accessible_normalized_filename(path):
980
"""Get the unicode normalized path, and if you can access the file.
982
On platforms where the system normalizes filenames (Mac OSX),
983
you can access a file by any path which will normalize correctly.
984
On platforms where the system does not normalize filenames
985
(Windows, Linux), you have to access a file by its exact path.
987
Internally, bzr only supports NFC/NFKC normalization, since that is
988
the standard for XML documents.
990
So return the normalized path, and a flag indicating if the file
991
can be accessed by that path.
994
return unicodedata.normalize('NFKC', unicode(path)), True
997
def _inaccessible_normalized_filename(path):
998
__doc__ = _accessible_normalized_filename.__doc__
1000
normalized = unicodedata.normalize('NFKC', unicode(path))
1001
return normalized, normalized == path
1004
791
if _platform_normalizes_filenames:
1005
normalized_filename = _accessible_normalized_filename
792
def unicode_filename(path):
793
"""Make sure 'path' is a properly normalized filename.
795
On platforms where the system normalizes filenames (Mac OSX),
796
you can access a file by any path which will normalize
798
Internally, bzr only supports NFC/NFKC normalization, since
799
that is the standard for XML documents.
800
So we return an normalized path, and indicate this has been
803
:return: (path, is_normalized) Return a path which can
804
access the file, and whether or not this path is
807
return unicodedata.normalize('NFKC', path), True
1007
normalized_filename = _inaccessible_normalized_filename
809
def unicode_filename(path):
810
"""Make sure 'path' is a properly normalized filename.
812
On platforms where the system does not normalize filenames
813
(Windows, Linux), you have to access a file by its exact path.
814
Internally, bzr only supports NFC/NFKC normalization, since
815
that is the standard for XML documents.
816
So we return the original path, and indicate if this is
819
:return: (path, is_normalized) Return a path which can
820
access the file, and whether or not this path is
823
return path, unicodedata.normalize('NFKC', path) == path
1010
826
def terminal_width():
1011
827
"""Return estimated terminal width."""
1012
828
if sys.platform == 'win32':
1013
return win32utils.get_console_size()[0]
829
import bzrlib.win32console
830
return bzrlib.win32console.get_console_size()[0]
1016
833
import struct, fcntl, termios
1088
872
to exclude some directories, they are then not descended into.
1090
874
The data yielded is of the form:
1091
((directory-relpath, directory-path-from-top),
1092
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1093
- directory-relpath is the relative path of the directory being returned
1094
with respect to top. prefix is prepended to this.
1095
- directory-path-from-root is the path including top for this directory.
1096
It is suitable for use with os functions.
1097
- relpath is the relative path within the subtree being walked.
1098
- basename is the basename of the path
1099
- kind is the kind of the file now. If unknown then the file is not
1100
present within the tree - but it may be recorded as versioned. See
1102
- lstat is the stat data *if* the file was statted.
1103
- planned, not implemented:
1104
path_from_tree_root is the path from the root of the tree.
875
[(relpath, basename, kind, lstat, path_from_top), ...]
1106
877
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1107
878
allows one to walk a subtree but get paths that are relative to a tree
1108
879
rooted higher up.
1109
880
:return: an iterator over the dirs.
1111
#TODO there is a bit of a smell where the results of the directory-
1112
# summary in this, and the path from the root, may not agree
1113
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1114
# potentially confusing output. We should make this more robust - but
1115
# not at a speed cost. RBC 20060731
1117
884
_directory = _directory_kind
1118
_listdir = os.listdir
1119
_kind_from_mode = _formats.get
1120
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
886
pending = [(prefix, "", _directory, None, top)]
889
currentdir = pending.pop()
1122
890
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1123
relroot, _, _, _, top = pending.pop()
1125
relprefix = relroot + u'/'
1128
top_slash = top + u'/'
1131
append = dirblock.append
1132
for name in sorted(_listdir(top)):
1133
abspath = top_slash + name
1134
statvalue = _lstat(abspath)
1135
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1136
append((relprefix + name, name, kind, statvalue, abspath))
1137
yield (relroot, top), dirblock
1139
# push the user specified dirs from dirblock
1140
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1143
def _walkdirs_utf8(top, prefix=""):
1144
"""Yield data about all the directories in a tree.
1146
This yields the same information as walkdirs() only each entry is yielded
1147
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1148
are returned as exact byte-strings.
1150
:return: yields a tuple of (dir_info, [file_info])
1151
dir_info is (utf8_relpath, path-from-top)
1152
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1153
if top is an absolute path, path-from-top is also an absolute path.
1154
path-from-top might be unicode or utf8, but it is the correct path to
1155
pass to os functions to affect the file in question. (such as os.lstat)
1157
fs_encoding = _fs_enc.upper()
1158
if (sys.platform == 'win32' or
1159
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1160
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1162
return _walkdirs_fs_utf8(top, prefix=prefix)
1165
def _walkdirs_fs_utf8(top, prefix=""):
1166
"""See _walkdirs_utf8.
1168
This sub-function is called when we know the filesystem is already in utf8
1169
encoding. So we don't need to transcode filenames.
1172
_directory = _directory_kind
1173
_listdir = os.listdir
1174
_kind_from_mode = _formats.get
1176
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1177
# But we don't actually uses 1-3 in pending, so set them to None
1178
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1180
relroot, _, _, _, top = pending.pop()
1182
relprefix = relroot + '/'
1185
top_slash = top + '/'
1188
append = dirblock.append
1189
for name in sorted(_listdir(top)):
1190
abspath = top_slash + name
1191
statvalue = _lstat(abspath)
1192
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1193
append((relprefix + name, name, kind, statvalue, abspath))
1194
yield (relroot, top), dirblock
1196
# push the user specified dirs from dirblock
1197
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1200
def _walkdirs_unicode_to_utf8(top, prefix=""):
1201
"""See _walkdirs_utf8
1203
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1205
This is currently the fallback code path when the filesystem encoding is
1206
not UTF-8. It may be better to implement an alternative so that we can
1207
safely handle paths that are not properly decodable in the current
1210
_utf8_encode = codecs.getencoder('utf8')
1212
_directory = _directory_kind
1213
_listdir = os.listdir
1214
_kind_from_mode = _formats.get
1216
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1218
relroot, _, _, _, top = pending.pop()
1220
relprefix = relroot + '/'
1223
top_slash = top + u'/'
1226
append = dirblock.append
1227
for name in sorted(_listdir(top)):
1228
name_utf8 = _utf8_encode(name)[0]
1229
abspath = top_slash + name
1230
statvalue = _lstat(abspath)
1231
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1232
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1233
yield (relroot, top), dirblock
1235
# push the user specified dirs from dirblock
1236
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1239
def copy_tree(from_path, to_path, handlers={}):
1240
"""Copy all of the entries in from_path into to_path.
1242
:param from_path: The base directory to copy.
1243
:param to_path: The target directory. If it does not exist, it will
1245
:param handlers: A dictionary of functions, which takes a source and
1246
destinations for files, directories, etc.
1247
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1248
'file', 'directory', and 'symlink' should always exist.
1249
If they are missing, they will be replaced with 'os.mkdir()',
1250
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1252
# Now, just copy the existing cached tree to the new location
1253
# We use a cheap trick here.
1254
# Absolute paths are prefixed with the first parameter
1255
# relative paths are prefixed with the second.
1256
# So we can get both the source and target returned
1257
# without any extra work.
1259
def copy_dir(source, dest):
1262
def copy_link(source, dest):
1263
"""Copy the contents of a symlink"""
1264
link_to = os.readlink(source)
1265
os.symlink(link_to, dest)
1267
real_handlers = {'file':shutil.copy2,
1268
'symlink':copy_link,
1269
'directory':copy_dir,
1271
real_handlers.update(handlers)
1273
if not os.path.exists(to_path):
1274
real_handlers['directory'](from_path, to_path)
1276
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1277
for relpath, name, kind, st, abspath in entries:
1278
real_handlers[kind](abspath, relpath)
893
relroot = currentdir[0] + '/'
896
for name in sorted(_listdir(top)):
897
abspath = top + '/' + name
898
statvalue = lstat(abspath)
899
dirblock.append ((relroot + name, name, file_kind_from_stat_mode(statvalue.st_mode), statvalue, abspath))
901
# push the user specified dirs from dirblock
902
for dir in reversed(dirblock):
903
if dir[2] == _directory:
1281
907
def path_prefix_key(path):
1291
917
key_a = path_prefix_key(path_a)
1292
918
key_b = path_prefix_key(path_b)
1293
919
return cmp(key_a, key_b)
1296
_cached_user_encoding = None
1299
def get_user_encoding(use_cache=True):
1300
"""Find out what the preferred user encoding is.
1302
This is generally the encoding that is used for command line parameters
1303
and file contents. This may be different from the terminal encoding
1304
or the filesystem encoding.
1306
:param use_cache: Enable cache for detected encoding.
1307
(This parameter is turned on by default,
1308
and required only for selftesting)
1310
:return: A string defining the preferred user encoding
1312
global _cached_user_encoding
1313
if _cached_user_encoding is not None and use_cache:
1314
return _cached_user_encoding
1316
if sys.platform == 'darwin':
1317
# work around egregious python 2.4 bug
1318
sys.platform = 'posix'
1322
sys.platform = 'darwin'
1327
user_encoding = locale.getpreferredencoding()
1328
except locale.Error, e:
1329
sys.stderr.write('bzr: warning: %s\n'
1330
' Could not determine what text encoding to use.\n'
1331
' This error usually means your Python interpreter\n'
1332
' doesn\'t support the locale set by $LANG (%s)\n'
1333
" Continuing with ascii encoding.\n"
1334
% (e, os.environ.get('LANG')))
1335
user_encoding = 'ascii'
1337
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1338
# treat that as ASCII, and not support printing unicode characters to the
1340
if user_encoding in (None, 'cp0'):
1341
user_encoding = 'ascii'
1345
codecs.lookup(user_encoding)
1347
sys.stderr.write('bzr: warning:'
1348
' unknown encoding %s.'
1349
' Continuing with ascii encoding.\n'
1352
user_encoding = 'ascii'
1355
_cached_user_encoding = user_encoding
1357
return user_encoding
1360
def recv_all(socket, bytes):
1361
"""Receive an exact number of bytes.
1363
Regular Socket.recv() may return less than the requested number of bytes,
1364
dependning on what's in the OS buffer. MSG_WAITALL is not available
1365
on all platforms, but this should work everywhere. This will return
1366
less than the requested amount if the remote end closes.
1368
This isn't optimized and is intended mostly for use in testing.
1371
while len(b) < bytes:
1372
new = socket.recv(bytes - len(b))
1378
def dereference_path(path):
1379
"""Determine the real path to a file.
1381
All parent elements are dereferenced. But the file itself is not
1383
:param path: The original path. May be absolute or relative.
1384
:return: the real path *to* the file
1386
parent, base = os.path.split(path)
1387
# The pathjoin for '.' is a workaround for Python bug #1213894.
1388
# (initial path components aren't dereferenced)
1389
return pathjoin(realpath(pathjoin('.', parent)), base)