13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from cStringIO import StringIO
21
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
22
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
25
26
from bzrlib.lazy_import import lazy_import
26
27
lazy_import(globals(), """
27
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
31
# We need to import both shutil and rmtree as we export the later on posix
32
# and need the former on windows
34
from shutil import rmtree
37
# We need to import both tempfile and mkdtemp as we export the later on posix
38
# and need the former on windows
40
from tempfile import mkdtemp
44
from tempfile import (
43
49
from bzrlib import (
58
from bzrlib import symbol_versioning
51
59
from bzrlib.symbol_versioning import (
52
60
deprecated_function,
56
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
58
if sys.version_info < (2, 5):
59
import md5 as _mod_md5
61
import sha as _mod_sha
71
from bzrlib import symbol_versioning
74
# Cross platform wall-clock time functionality with decent resolution.
75
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
76
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
77
# synchronized with ``time.time()``, this is only meant to be used to find
78
# delta times by subtracting from another call to this function.
79
timer_func = time.time
80
if sys.platform == 'win32':
81
timer_func = time.clock
62
from bzrlib.trace import mutter
83
65
# On win32, O_BINARY is used to indicate the file should
84
66
# be opened in binary mode, rather than text mode.
85
67
# On other platforms, O_BINARY doesn't exist, because
86
68
# they always open in binary mode, so it is okay to
87
# OR with 0 on those platforms.
88
# O_NOINHERIT and O_TEXT exists only on win32 too.
69
# OR with 0 on those platforms
89
70
O_BINARY = getattr(os, 'O_BINARY', 0)
90
O_TEXT = getattr(os, 'O_TEXT', 0)
91
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
94
def get_unicode_argv():
96
user_encoding = get_user_encoding()
97
return [a.decode(user_encoding) for a in sys.argv[1:]]
98
except UnicodeDecodeError:
99
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
103
73
def make_readonly(filename):
203
196
def fancy_rename(old, new, rename_func, unlink_func):
204
197
"""A fancy rename, when you don't have atomic rename.
206
199
:param old: The old path, to rename from
207
200
:param new: The new path, to rename to
208
201
:param rename_func: The potentially non-atomic rename function
209
:param unlink_func: A way to delete the target file if the full rename
202
:param unlink_func: A way to delete the target file if the full rename succeeds
212
205
# sftp rename doesn't allow overwriting, so play tricks:
213
207
base = os.path.basename(new)
214
208
dirname = os.path.dirname(new)
215
# callers use different encodings for the paths so the following MUST
216
# respect that. We rely on python upcasting to unicode if new is unicode
217
# and keeping a str if not.
218
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
219
os.getpid(), rand_chars(10))
209
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
220
210
tmp_name = pathjoin(dirname, tmp_name)
222
212
# Rename the file out of the way, but keep track if it didn't exist
321
304
# /path => C:/path
322
305
path = unicode(path)
323
306
# check for absolute path
324
drive = ntpath.splitdrive(path)[0]
307
drive = _nt_splitdrive(path)[0]
325
308
if drive == '' and path[:2] not in('//','\\\\'):
326
309
cwd = os.getcwdu()
327
310
# we cannot simply os.path.join cwd and path
328
311
# because os.path.join('C:','/path') produce '/path'
329
312
# and this is incorrect
330
313
if path[:1] in ('/','\\'):
331
cwd = ntpath.splitdrive(cwd)[0]
314
cwd = _nt_splitdrive(cwd)[0]
333
316
path = cwd + '\\' + path
334
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
317
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
319
if win32utils.winver == 'Windows 98':
320
_win32_abspath = _win98_abspath
337
323
def _win32_realpath(path):
338
# Real ntpath.realpath doesn't have a problem with a unicode cwd
339
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
324
# Real _nt_realpath doesn't have a problem with a unicode cwd
325
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
342
328
def _win32_pathjoin(*args):
343
return ntpath.join(*args).replace('\\', '/')
329
return _nt_join(*args).replace('\\', '/')
346
332
def _win32_normpath(path):
347
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
333
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
350
336
def _win32_getcwd():
426
409
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
427
410
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
428
411
return shutil.rmtree(path, ignore_errors, onerror)
430
f = win32utils.get_unicode_argv # special function or None
434
412
elif sys.platform == 'darwin':
435
413
getcwd = _mac_getcwd
438
def get_terminal_encoding(trace=False):
416
def get_terminal_encoding():
439
417
"""Find the best encoding for printing to the screen.
441
419
This attempts to check both sys.stdout and sys.stdin to see
442
420
what encoding they are in, and if that fails it falls back to
443
osutils.get_user_encoding().
421
bzrlib.user_encoding.
444
422
The problem is that on Windows, locale.getpreferredencoding()
445
423
is not the same encoding as that used by the console:
446
424
http://mail.python.org/pipermail/python-list/2003-May/162357.html
448
426
On my standard US Windows XP, the preferred encoding is
449
427
cp1252, but the console is cp437
451
:param trace: If True trace the selected encoding via mutter().
453
from bzrlib.trace import mutter
454
429
output_encoding = getattr(sys.stdout, 'encoding', None)
455
430
if not output_encoding:
456
431
input_encoding = getattr(sys.stdin, 'encoding', None)
457
432
if not input_encoding:
458
output_encoding = get_user_encoding()
460
mutter('encoding stdout as osutils.get_user_encoding() %r',
433
output_encoding = bzrlib.user_encoding
434
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
463
436
output_encoding = input_encoding
465
mutter('encoding stdout as sys.stdin encoding %r',
437
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
469
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
439
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
470
440
if output_encoding == 'cp0':
471
441
# invalid encoding (cp0 means 'no codepage' on Windows)
472
output_encoding = get_user_encoding()
474
mutter('cp0 is invalid encoding.'
475
' encoding stdout as osutils.get_user_encoding() %r',
442
output_encoding = bzrlib.user_encoding
443
mutter('cp0 is invalid encoding.'
444
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
479
447
codecs.lookup(output_encoding)
732
653
:param timezone: How to display the time: 'utc', 'original' for the
733
654
timezone specified by offset, or 'local' for the process's current
735
:param date_fmt: strftime format.
736
:param show_offset: Whether to append the timezone.
738
(date_fmt, tt, offset_str) = \
739
_format_date(t, offset, timezone, date_fmt, show_offset)
740
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
741
date_str = time.strftime(date_fmt, tt)
742
return date_str + offset_str
745
# Cache of formatted offset strings
749
def format_date_with_offset_in_original_timezone(t, offset=0,
750
_cache=_offset_cache):
751
"""Return a formatted date string in the original timezone.
753
This routine may be faster then format_date.
755
:param t: Seconds since the epoch.
756
:param offset: Timezone offset in seconds east of utc.
760
tt = time.gmtime(t + offset)
761
date_fmt = _default_format_by_weekday_num[tt[6]]
762
date_str = time.strftime(date_fmt, tt)
763
offset_str = _cache.get(offset, None)
764
if offset_str is None:
765
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
766
_cache[offset] = offset_str
767
return date_str + offset_str
770
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
772
"""Return an unicode date string formatted according to the current locale.
774
:param t: Seconds since the epoch.
775
:param offset: Timezone offset in seconds east of utc.
776
:param timezone: How to display the time: 'utc', 'original' for the
777
timezone specified by offset, or 'local' for the process's current
779
:param date_fmt: strftime format.
780
:param show_offset: Whether to append the timezone.
782
(date_fmt, tt, offset_str) = \
783
_format_date(t, offset, timezone, date_fmt, show_offset)
784
date_str = time.strftime(date_fmt, tt)
785
if not isinstance(date_str, unicode):
786
date_str = date_str.decode(get_user_encoding(), 'replace')
787
return date_str + offset_str
790
def _format_date(t, offset, timezone, date_fmt, show_offset):
656
:param show_offset: Whether to append the timezone.
657
:param date_fmt: strftime format.
791
659
if timezone == 'utc':
792
660
tt = time.gmtime(t)
930
799
return pathjoin(*p)
933
def parent_directories(filename):
934
"""Return the list of parent directories, deepest first.
936
For example, parent_directories("a/b/c") -> ["a/b", "a"].
939
parts = splitpath(dirname(filename))
941
parents.append(joinpath(parts))
946
_extension_load_failures = []
949
def failed_to_load_extension(exception):
950
"""Handle failing to load a binary extension.
952
This should be called from the ImportError block guarding the attempt to
953
import the native extension. If this function returns, the pure-Python
954
implementation should be loaded instead::
957
>>> import bzrlib._fictional_extension_pyx
958
>>> except ImportError, e:
959
>>> bzrlib.osutils.failed_to_load_extension(e)
960
>>> import bzrlib._fictional_extension_py
962
# NB: This docstring is just an example, not a doctest, because doctest
963
# currently can't cope with the use of lazy imports in this namespace --
966
# This currently doesn't report the failure at the time it occurs, because
967
# they tend to happen very early in startup when we can't check config
968
# files etc, and also we want to report all failures but not spam the user
970
from bzrlib import trace
971
exception_str = str(exception)
972
if exception_str not in _extension_load_failures:
973
trace.mutter("failed to load compiled extension: %s" % exception_str)
974
_extension_load_failures.append(exception_str)
977
def report_extension_load_failures():
978
if not _extension_load_failures:
980
from bzrlib.config import GlobalConfig
981
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
983
# the warnings framework should by default show this only once
984
from bzrlib.trace import warning
986
"bzr: warning: some compiled extensions could not be loaded; "
987
"see <https://answers.launchpad.net/bzr/+faq/703>")
988
# we no longer show the specific missing extensions here, because it makes
989
# the message too long and scary - see
990
# https://bugs.launchpad.net/bzr/+bug/430529
994
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
995
except ImportError, e:
996
failed_to_load_extension(e)
997
from bzrlib._chunks_to_lines_py import chunks_to_lines
1000
802
def split_lines(s):
1001
803
"""Split s into lines, but without removing the newline characters."""
1002
# Trivially convert a fulltext into a 'chunked' representation, and let
1003
# chunks_to_lines do the heavy lifting.
1004
if isinstance(s, str):
1005
# chunks_to_lines only supports 8-bit strings
1006
return chunks_to_lines([s])
1008
return _split_lines(s)
1011
def _split_lines(s):
1012
"""Split s into lines, but without removing the newline characters.
1014
This supports Unicode or plain string objects.
1016
804
lines = s.split('\n')
1017
805
result = [line + '\n' for line in lines[:-1]]
1037
825
shutil.copyfile(src, dest)
828
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
829
# Forgiveness than Permission (EAFP) because:
830
# - root can damage a solaris file system by using unlink,
831
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
832
# EACCES, OSX: EPERM) when invoked on a directory.
1040
833
def delete_any(path):
1041
"""Delete a file, symlink or directory.
1043
Will delete even if readonly.
1046
_delete_file_or_dir(path)
1047
except (OSError, IOError), e:
1048
if e.errno in (errno.EPERM, errno.EACCES):
1049
# make writable and try again
1052
except (OSError, IOError):
1054
_delete_file_or_dir(path)
1059
def _delete_file_or_dir(path):
1060
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1061
# Forgiveness than Permission (EAFP) because:
1062
# - root can damage a solaris file system by using unlink,
1063
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1064
# EACCES, OSX: EPERM) when invoked on a directory.
834
"""Delete a file or directory."""
1065
835
if isdir(path): # Takes care of symlinks
1157
if len(head) <= len(base) and head != base:
1158
raise errors.PathNotChild(rp, base)
909
while len(head) >= len(base):
1159
910
if head == base:
1161
head, tail = split(head)
912
head, tail = os.path.split(head)
916
raise errors.PathNotChild(rp, base)
1166
return pathjoin(*reversed(s))
1171
def _cicp_canonical_relpath(base, path):
1172
"""Return the canonical path relative to base.
1174
Like relpath, but on case-insensitive-case-preserving file-systems, this
1175
will return the relpath as stored on the file-system rather than in the
1176
case specified in the input string, for all existing portions of the path.
1178
This will cause O(N) behaviour if called for every path in a tree; if you
1179
have a number of paths to convert, you should use canonical_relpaths().
1181
# TODO: it should be possible to optimize this for Windows by using the
1182
# win32 API FindFiles function to look for the specified name - but using
1183
# os.listdir() still gives us the correct, platform agnostic semantics in
1186
rel = relpath(base, path)
1187
# '.' will have been turned into ''
1191
abs_base = abspath(base)
1193
_listdir = os.listdir
1195
# use an explicit iterator so we can easily consume the rest on early exit.
1196
bit_iter = iter(rel.split('/'))
1197
for bit in bit_iter:
1200
next_entries = _listdir(current)
1201
except OSError: # enoent, eperm, etc
1202
# We can't find this in the filesystem, so just append the
1204
current = pathjoin(current, bit, *list(bit_iter))
1206
for look in next_entries:
1207
if lbit == look.lower():
1208
current = pathjoin(current, look)
1211
# got to the end, nothing matched, so we just return the
1212
# non-existing bits as they were specified (the filename may be
1213
# the target of a move, for example).
1214
current = pathjoin(current, bit, *list(bit_iter))
1216
return current[len(abs_base):].lstrip('/')
1218
# XXX - TODO - we need better detection/integration of case-insensitive
1219
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1220
# filesystems), for example, so could probably benefit from the same basic
1221
# support there. For now though, only Windows and OSX get that support, and
1222
# they get it for *all* file-systems!
1223
if sys.platform in ('win32', 'darwin'):
1224
canonical_relpath = _cicp_canonical_relpath
1226
canonical_relpath = relpath
1228
def canonical_relpaths(base, paths):
1229
"""Create an iterable to canonicalize a sequence of relative paths.
1231
The intent is for this implementation to use a cache, vastly speeding
1232
up multiple transformations in the same directory.
1234
# but for now, we haven't optimized...
1235
return [canonical_relpath(base, p) for p in paths]
1238
def decode_filename(filename):
1239
"""Decode the filename using the filesystem encoding
1241
If it is unicode, it is returned.
1242
Otherwise it is decoded from the the filesystem's encoding. If decoding
1243
fails, a errors.BadFilenameEncoding exception is raised.
1245
if type(filename) is unicode:
1248
return filename.decode(_fs_enc)
1249
except UnicodeDecodeError:
1250
raise errors.BadFilenameEncoding(filename, _fs_enc)
1253
924
def safe_unicode(unicode_or_utf8_string):
1254
925
"""Coerce unicode_or_utf8_string into unicode.
1256
927
If it is unicode, it is returned.
1257
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1258
wrapped in a BzrBadParameterNotUnicode exception.
928
Otherwise it is decoded from utf-8. If a decoding error
929
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
930
as a BzrBadParameter exception.
1260
932
if isinstance(unicode_or_utf8_string, unicode):
1261
933
return unicode_or_utf8_string
1374
1046
normalized_filename = _inaccessible_normalized_filename
1377
def set_signal_handler(signum, handler, restart_syscall=True):
1378
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1379
on platforms that support that.
1381
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1382
automatically restart (by calling `signal.siginterrupt(signum,
1383
False)`). May be ignored if the feature is not available on this
1384
platform or Python version.
1388
siginterrupt = signal.siginterrupt
1390
# This python implementation doesn't provide signal support, hence no
1393
except AttributeError:
1394
# siginterrupt doesn't exist on this platform, or for this version
1396
siginterrupt = lambda signum, flag: None
1398
def sig_handler(*args):
1399
# Python resets the siginterrupt flag when a signal is
1400
# received. <http://bugs.python.org/issue8354>
1401
# As a workaround for some cases, set it back the way we want it.
1402
siginterrupt(signum, False)
1403
# Now run the handler function passed to set_signal_handler.
1406
sig_handler = handler
1407
old_handler = signal.signal(signum, sig_handler)
1409
siginterrupt(signum, False)
1413
default_terminal_width = 80
1414
"""The default terminal width for ttys.
1416
This is defined so that higher levels can share a common fallback value when
1417
terminal_width() returns None.
1420
# Keep some state so that terminal_width can detect if _terminal_size has
1421
# returned a different size since the process started. See docstring and
1422
# comments of terminal_width for details.
1423
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1424
_terminal_size_state = 'no_data'
1425
_first_terminal_size = None
1427
1049
def terminal_width():
1428
"""Return terminal width.
1430
None is returned if the width can't established precisely.
1433
- if BZR_COLUMNS is set, returns its value
1434
- if there is no controlling terminal, returns None
1435
- query the OS, if the queried size has changed since the last query,
1437
- if COLUMNS is set, returns its value,
1438
- if the OS has a value (even though it's never changed), return its value.
1440
From there, we need to query the OS to get the size of the controlling
1443
On Unices we query the OS by:
1444
- get termios.TIOCGWINSZ
1445
- if an error occurs or a negative value is obtained, returns None
1447
On Windows we query the OS by:
1448
- win32utils.get_console_size() decides,
1449
- returns None on error (provided default value)
1451
# Note to implementors: if changing the rules for determining the width,
1452
# make sure you've considered the behaviour in these cases:
1453
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1454
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1456
# - (add more interesting cases here, if you find any)
1457
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1458
# but we don't want to register a signal handler because it is impossible
1459
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1460
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1461
# time so we can notice if the reported size has changed, which should have
1464
# If BZR_COLUMNS is set, take it, user is always right
1466
return int(os.environ['BZR_COLUMNS'])
1467
except (KeyError, ValueError):
1470
isatty = getattr(sys.stdout, 'isatty', None)
1471
if isatty is None or not isatty():
1472
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1476
width, height = os_size = _terminal_size(None, None)
1477
global _first_terminal_size, _terminal_size_state
1478
if _terminal_size_state == 'no_data':
1479
_first_terminal_size = os_size
1480
_terminal_size_state = 'unchanged'
1481
elif (_terminal_size_state == 'unchanged' and
1482
_first_terminal_size != os_size):
1483
_terminal_size_state = 'changed'
1485
# If the OS claims to know how wide the terminal is, and this value has
1486
# ever changed, use that.
1487
if _terminal_size_state == 'changed':
1488
if width is not None and width > 0:
1491
# If COLUMNS is set, use it.
1493
return int(os.environ['COLUMNS'])
1494
except (KeyError, ValueError):
1497
# Finally, use an unchanged size from the OS, if we have one.
1498
if _terminal_size_state == 'unchanged':
1499
if width is not None and width > 0:
1502
# The width could not be determined.
1506
def _win32_terminal_size(width, height):
1507
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1508
return width, height
1511
def _ioctl_terminal_size(width, height):
1050
"""Return estimated terminal width."""
1051
if sys.platform == 'win32':
1052
return win32utils.get_console_size()[0]
1513
1055
import struct, fcntl, termios
1514
1056
s = struct.pack('HHHH', 0, 0, 0, 0)
1515
1057
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1516
height, width = struct.unpack('HHHH', x)[0:2]
1517
except (IOError, AttributeError):
1058
width = struct.unpack('HHHH', x)[1]
1519
return width, height
1521
_terminal_size = None
1522
"""Returns the terminal size as (width, height).
1524
:param width: Default value for width.
1525
:param height: Default value for height.
1527
This is defined specifically for each OS and query the size of the controlling
1528
terminal. If any error occurs, the provided default values should be returned.
1530
if sys.platform == 'win32':
1531
_terminal_size = _win32_terminal_size
1533
_terminal_size = _ioctl_terminal_size
1063
width = int(os.environ['COLUMNS'])
1536
1072
def supports_executable():
1631
1167
present within the tree - but it may be recorded as versioned. See
1632
1168
versioned_kind.
1633
1169
- lstat is the stat data *if* the file was statted.
1634
- planned, not implemented:
1170
- planned, not implemented:
1635
1171
path_from_tree_root is the path from the root of the tree.
1637
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1173
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1638
1174
allows one to walk a subtree but get paths that are relative to a tree
1639
1175
rooted higher up.
1640
1176
:return: an iterator over the dirs.
1642
1178
#TODO there is a bit of a smell where the results of the directory-
1643
# summary in this, and the path from the root, may not agree
1179
# summary in this, and the path from the root, may not agree
1644
1180
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1645
1181
# potentially confusing output. We should make this more robust - but
1646
1182
# not at a speed cost. RBC 20060731
1647
1183
_lstat = os.lstat
1648
1184
_directory = _directory_kind
1649
1185
_listdir = os.listdir
1650
_kind_from_mode = file_kind_from_stat_mode
1186
_kind_from_mode = _formats.get
1651
1187
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1653
1189
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1719
1229
path-from-top might be unicode or utf8, but it is the correct path to
1720
1230
pass to os functions to affect the file in question. (such as os.lstat)
1722
global _selected_dir_reader
1723
if _selected_dir_reader is None:
1232
global _real_walkdirs_utf8
1233
if _real_walkdirs_utf8 is None:
1724
1234
fs_encoding = _fs_enc.upper()
1725
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1235
if win32utils.winver == 'Windows NT':
1726
1236
# Win98 doesn't have unicode apis like FindFirstFileW
1727
1237
# TODO: We possibly could support Win98 by falling back to the
1728
1238
# original FindFirstFile, and using TCHAR instead of WCHAR,
1729
1239
# but that gets a bit tricky, and requires custom compiling
1730
1240
# for win98 anyway.
1732
from bzrlib._walkdirs_win32 import Win32ReadDir
1733
_selected_dir_reader = Win32ReadDir()
1242
from bzrlib._walkdirs_win32 import _walkdirs_utf8_win32_find_file
1734
1243
except ImportError:
1736
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1244
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1246
_real_walkdirs_utf8 = _walkdirs_utf8_win32_find_file
1247
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1737
1248
# ANSI_X3.4-1968 is a form of ASCII
1739
from bzrlib._readdir_pyx import UTF8DirReader
1740
_selected_dir_reader = UTF8DirReader()
1741
except ImportError, e:
1742
failed_to_load_extension(e)
1745
if _selected_dir_reader is None:
1746
# Fallback to the python version
1747
_selected_dir_reader = UnicodeDirReader()
1249
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1251
_real_walkdirs_utf8 = _walkdirs_fs_utf8
1252
return _real_walkdirs_utf8(top, prefix=prefix)
1255
def _walkdirs_fs_utf8(top, prefix=""):
1256
"""See _walkdirs_utf8.
1258
This sub-function is called when we know the filesystem is already in utf8
1259
encoding. So we don't need to transcode filenames.
1262
_directory = _directory_kind
1263
_listdir = os.listdir
1264
_kind_from_mode = _formats.get
1749
1266
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1750
1267
# But we don't actually uses 1-3 in pending, so set them to None
1751
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1752
read_dir = _selected_dir_reader.read_dir
1753
_directory = _directory_kind
1268
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1755
relroot, _, _, _, top = pending[-1].pop()
1758
dirblock = sorted(read_dir(relroot, top))
1270
relroot, _, _, _, top = pending.pop()
1272
relprefix = relroot + '/'
1275
top_slash = top + '/'
1278
append = dirblock.append
1279
for name in sorted(_listdir(top)):
1280
abspath = top_slash + name
1281
statvalue = _lstat(abspath)
1282
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1283
append((relprefix + name, name, kind, statvalue, abspath))
1759
1284
yield (relroot, top), dirblock
1760
1286
# push the user specified dirs from dirblock
1761
next = [d for d in reversed(dirblock) if d[2] == _directory]
1763
pending.append(next)
1766
class UnicodeDirReader(DirReader):
1767
"""A dir reader for non-utf8 file systems, which transcodes."""
1769
__slots__ = ['_utf8_encode']
1772
self._utf8_encode = codecs.getencoder('utf8')
1774
def top_prefix_to_starting_dir(self, top, prefix=""):
1775
"""See DirReader.top_prefix_to_starting_dir."""
1776
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1778
def read_dir(self, prefix, top):
1779
"""Read a single directory from a non-utf8 file system.
1781
top, and the abspath element in the output are unicode, all other paths
1782
are utf8. Local disk IO is done via unicode calls to listdir etc.
1784
This is currently the fallback code path when the filesystem encoding is
1785
not UTF-8. It may be better to implement an alternative so that we can
1786
safely handle paths that are not properly decodable in the current
1789
See DirReader.read_dir for details.
1791
_utf8_encode = self._utf8_encode
1793
_listdir = os.listdir
1794
_kind_from_mode = file_kind_from_stat_mode
1797
relprefix = prefix + '/'
1287
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1290
def _walkdirs_unicode_to_utf8(top, prefix=""):
1291
"""See _walkdirs_utf8
1293
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1295
This is currently the fallback code path when the filesystem encoding is
1296
not UTF-8. It may be better to implement an alternative so that we can
1297
safely handle paths that are not properly decodable in the current
1300
_utf8_encode = codecs.getencoder('utf8')
1302
_directory = _directory_kind
1303
_listdir = os.listdir
1304
_kind_from_mode = _formats.get
1306
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1308
relroot, _, _, _, top = pending.pop()
1310
relprefix = relroot + '/'
1800
1313
top_slash = top + u'/'
1803
1316
append = dirblock.append
1804
1317
for name in sorted(_listdir(top)):
1806
name_utf8 = _utf8_encode(name)[0]
1807
except UnicodeDecodeError:
1808
raise errors.BadFilenameEncoding(
1809
_utf8_encode(relprefix)[0] + name, _fs_enc)
1318
name_utf8 = _utf8_encode(name)[0]
1810
1319
abspath = top_slash + name
1811
1320
statvalue = _lstat(abspath)
1812
kind = _kind_from_mode(statvalue.st_mode)
1321
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1813
1322
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1323
yield (relroot, top), dirblock
1325
# push the user specified dirs from dirblock
1326
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1817
1329
def copy_tree(from_path, to_path, handlers={}):
1818
1330
"""Copy all of the entries in from_path into to_path.
1820
:param from_path: The base directory to copy.
1332
:param from_path: The base directory to copy.
1821
1333
:param to_path: The target directory. If it does not exist, it will
1823
1335
:param handlers: A dictionary of functions, which takes a source and
1969
1449
return user_encoding
1972
def get_diff_header_encoding():
1973
return get_terminal_encoding()
1976
def get_host_name():
1977
"""Return the current unicode host name.
1979
This is meant to be used in place of socket.gethostname() because that
1980
behaves inconsistently on different platforms.
1982
if sys.platform == "win32":
1984
return win32utils.get_host_name()
1987
return socket.gethostname().decode(get_user_encoding())
1990
# We must not read/write any more than 64k at a time from/to a socket so we
1991
# don't risk "no buffer space available" errors on some platforms. Windows in
1992
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1994
MAX_SOCKET_CHUNK = 64 * 1024
1996
def read_bytes_from_socket(sock, report_activity=None,
1997
max_read_size=MAX_SOCKET_CHUNK):
1998
"""Read up to max_read_size of bytes from sock and notify of progress.
2000
Translates "Connection reset by peer" into file-like EOF (return an
2001
empty string rather than raise an error), and repeats the recv if
2002
interrupted by a signal.
2006
bytes = sock.recv(max_read_size)
2007
except socket.error, e:
2009
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2010
# The connection was closed by the other side. Callers expect
2011
# an empty string to signal end-of-stream.
2013
elif eno == errno.EINTR:
2014
# Retry the interrupted recv.
2018
if report_activity is not None:
2019
report_activity(len(bytes), 'read')
2023
def recv_all(socket, count):
1452
def recv_all(socket, bytes):
2024
1453
"""Receive an exact number of bytes.
2026
1455
Regular Socket.recv() may return less than the requested number of bytes,
2027
depending on what's in the OS buffer. MSG_WAITALL is not available
1456
dependning on what's in the OS buffer. MSG_WAITALL is not available
2028
1457
on all platforms, but this should work everywhere. This will return
2029
1458
less than the requested amount if the remote end closes.
2031
1460
This isn't optimized and is intended mostly for use in testing.
2034
while len(b) < count:
2035
new = read_bytes_from_socket(socket, None, count - len(b))
1463
while len(b) < bytes:
1464
new = socket.recv(bytes - len(b))
2042
def send_all(sock, bytes, report_activity=None):
1471
def send_all(socket, bytes):
2043
1472
"""Send all bytes on a socket.
2045
Breaks large blocks in smaller chunks to avoid buffering limitations on
2046
some platforms, and catches EINTR which may be thrown if the send is
2047
interrupted by a signal.
2049
This is preferred to socket.sendall(), because it avoids portability bugs
2050
and provides activity reporting.
2052
:param report_activity: Call this as bytes are read, see
2053
Transport._report_activity
1474
Regular socket.sendall() can give socket error 10053 on Windows. This
1475
implementation sends no more than 64k at a time, which avoids this problem.
2056
byte_count = len(bytes)
2057
while sent_total < byte_count:
2059
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2060
except socket.error, e:
2061
if e.args[0] != errno.EINTR:
2065
report_activity(sent, 'write')
1478
for pos in xrange(0, len(bytes), chunk_size):
1479
socket.sendall(bytes[pos:pos+chunk_size])
2068
1482
def dereference_path(path):
2109
1523
base = dirname(bzrlib.__file__)
2110
1524
if getattr(sys, 'frozen', None): # bzr.exe
2111
1525
base = abspath(pathjoin(base, '..', '..'))
2112
f = file(pathjoin(base, resource_relpath), "rU")
2118
def file_kind_from_stat_mode_thunk(mode):
2119
global file_kind_from_stat_mode
2120
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2122
from bzrlib._readdir_pyx import UTF8DirReader
2123
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2124
except ImportError, e:
2125
# This is one time where we won't warn that an extension failed to
2126
# load. The extension is never available on Windows anyway.
2127
from bzrlib._readdir_py import (
2128
_kind_from_mode as file_kind_from_stat_mode
2130
return file_kind_from_stat_mode(mode)
2131
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2134
def file_kind(f, _lstat=os.lstat):
2136
return file_kind_from_stat_mode(_lstat(f).st_mode)
2138
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2139
raise errors.NoSuchFile(f)
2143
def until_no_eintr(f, *a, **kw):
2144
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2146
WARNING: you must be certain that it is safe to retry the call repeatedly
2147
if EINTR does occur. This is typically only true for low-level operations
2148
like os.read. If in any doubt, don't use this.
2150
Keep in mind that this is not a complete solution to EINTR. There is
2151
probably code in the Python standard library and other dependencies that
2152
may encounter EINTR if a signal arrives (and there is signal handler for
2153
that signal). So this function can reduce the impact for IO that bzrlib
2154
directly controls, but it is not a complete solution.
2156
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2160
except (IOError, OSError), e:
2161
if e.errno == errno.EINTR:
2166
@deprecated_function(deprecated_in((2, 2, 0)))
2167
def re_compile_checked(re_string, flags=0, where=""):
2168
"""Return a compiled re, or raise a sensible error.
2170
This should only be used when compiling user-supplied REs.
2172
:param re_string: Text form of regular expression.
2173
:param flags: eg re.IGNORECASE
2174
:param where: Message explaining to the user the context where
2175
it occurred, eg 'log search filter'.
2177
# from https://bugs.launchpad.net/bzr/+bug/251352
2179
re_obj = re.compile(re_string, flags)
2182
except errors.InvalidPattern, e:
2184
where = ' in ' + where
2185
# despite the name 'error' is a type
2186
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2190
if sys.platform == "win32":
2193
return msvcrt.getch()
2198
fd = sys.stdin.fileno()
2199
settings = termios.tcgetattr(fd)
2202
ch = sys.stdin.read(1)
2204
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2208
if sys.platform == 'linux2':
2209
def _local_concurrency():
2211
prefix = 'processor'
2212
for line in file('/proc/cpuinfo', 'rb'):
2213
if line.startswith(prefix):
2214
concurrency = int(line[line.find(':')+1:]) + 1
2216
elif sys.platform == 'darwin':
2217
def _local_concurrency():
2218
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2219
stdout=subprocess.PIPE).communicate()[0]
2220
elif sys.platform[0:7] == 'freebsd':
2221
def _local_concurrency():
2222
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2223
stdout=subprocess.PIPE).communicate()[0]
2224
elif sys.platform == 'sunos5':
2225
def _local_concurrency():
2226
return subprocess.Popen(['psrinfo', '-p',],
2227
stdout=subprocess.PIPE).communicate()[0]
2228
elif sys.platform == "win32":
2229
def _local_concurrency():
2230
# This appears to return the number of cores.
2231
return os.environ.get('NUMBER_OF_PROCESSORS')
2233
def _local_concurrency():
2238
_cached_local_concurrency = None
2240
def local_concurrency(use_cache=True):
2241
"""Return how many processes can be run concurrently.
2243
Rely on platform specific implementations and default to 1 (one) if
2244
anything goes wrong.
2246
global _cached_local_concurrency
2248
if _cached_local_concurrency is not None and use_cache:
2249
return _cached_local_concurrency
2251
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2252
if concurrency is None:
2254
concurrency = _local_concurrency()
2255
except (OSError, IOError):
2258
concurrency = int(concurrency)
2259
except (TypeError, ValueError):
2262
_cached_concurrency = concurrency
2266
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2267
"""A stream writer that doesn't decode str arguments."""
2269
def __init__(self, encode, stream, errors='strict'):
2270
codecs.StreamWriter.__init__(self, stream, errors)
2271
self.encode = encode
2273
def write(self, object):
2274
if type(object) is str:
2275
self.stream.write(object)
2277
data, _ = self.encode(object, self.errors)
2278
self.stream.write(data)
2280
if sys.platform == 'win32':
2281
def open_file(filename, mode='r', bufsize=-1):
2282
"""This function is used to override the ``open`` builtin.
2284
But it uses O_NOINHERIT flag so the file handle is not inherited by
2285
child processes. Deleting or renaming a closed file opened with this
2286
function is not blocking child processes.
2288
writing = 'w' in mode
2289
appending = 'a' in mode
2290
updating = '+' in mode
2291
binary = 'b' in mode
2294
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2295
# for flags for each modes.
2305
flags |= os.O_WRONLY
2306
flags |= os.O_CREAT | os.O_TRUNC
2311
flags |= os.O_WRONLY
2312
flags |= os.O_CREAT | os.O_APPEND
2317
flags |= os.O_RDONLY
2319
return os.fdopen(os.open(filename, flags), mode, bufsize)
2324
def getuser_unicode():
2325
"""Return the username as unicode.
2328
user_encoding = get_user_encoding()
2329
username = getpass.getuser().decode(user_encoding)
2330
except UnicodeDecodeError:
2331
raise errors.BzrError("Can't decode username as %s." % \
1526
filename = pathjoin(base, resource_relpath)
1527
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
