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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 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)
26
25
from bzrlib.lazy_import import lazy_import
27
26
lazy_import(globals(), """
29
27
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,
32
# We need to import both shutil and rmtree as we export the later on posix
33
# and need the former on windows
35
from shutil import rmtree
38
# We need to import both tempfile and mkdtemp as we export the later on posix
39
# and need the former on windows
44
from tempfile import (
41
from tempfile import mkdtemp
49
44
from bzrlib import (
51
from bzrlib.i18n import gettext
57
from bzrlib import symbol_versioning
58
54
from bzrlib.symbol_versioning import (
59
55
deprecated_function,
62
from bzrlib.trace import mutter
66
from bzrlib import symbol_versioning, _fs_enc
69
# Cross platform wall-clock time functionality with decent resolution.
70
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
71
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
72
# synchronized with ``time.time()``, this is only meant to be used to find
73
# delta times by subtracting from another call to this function.
74
timer_func = time.time
75
if sys.platform == 'win32':
76
timer_func = time.clock
65
78
# On win32, O_BINARY is used to indicate the file should
66
79
# be opened in binary mode, rather than text mode.
67
80
# On other platforms, O_BINARY doesn't exist, because
68
81
# they always open in binary mode, so it is okay to
69
# OR with 0 on those platforms
82
# OR with 0 on those platforms.
83
# O_NOINHERIT and O_TEXT exists only on win32 too.
70
84
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)
85
O_TEXT = getattr(os, 'O_TEXT', 0)
86
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
89
def get_unicode_argv():
91
user_encoding = get_user_encoding()
92
return [a.decode(user_encoding) for a in sys.argv[1:]]
93
except UnicodeDecodeError:
94
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
95
"application locale.").format(a, user_encoding))
76
98
def make_readonly(filename):
77
99
"""Make a filename read-only."""
78
mod = lstat(filename).st_mode
100
mod = os.lstat(filename).st_mode
79
101
if not stat.S_ISLNK(mod):
80
102
mod = mod & 0777555
81
os.chmod(filename, mod)
103
chmod_if_possible(filename, mod)
84
106
def make_writable(filename):
85
mod = lstat(filename).st_mode
107
mod = os.lstat(filename).st_mode
86
108
if not stat.S_ISLNK(mod):
88
os.chmod(filename, mod)
110
chmod_if_possible(filename, mod)
113
def chmod_if_possible(filename, mode):
114
# Set file mode if that can be safely done.
115
# Sometimes even on unix the filesystem won't allow it - see
116
# https://bugs.launchpad.net/bzr/+bug/606537
118
# It is probably faster to just do the chmod, rather than
119
# doing a stat, and then trying to compare
120
os.chmod(filename, mode)
121
except (IOError, OSError),e:
122
# Permission/access denied seems to commonly happen on smbfs; there's
123
# probably no point warning about it.
124
# <https://bugs.launchpad.net/bzr/+bug/606537>
125
if getattr(e, 'errno') in (errno.EPERM, errno.EACCES):
126
trace.mutter("ignore error on chmod of %r: %r" % (
132
def minimum_path_selection(paths):
133
"""Return the smallset subset of paths which are outside paths.
135
:param paths: A container (and hence not None) of paths.
136
:return: A set of paths sufficient to include everything in paths via
137
is_inside, drawn from the paths parameter.
143
return path.split('/')
144
sorted_paths = sorted(list(paths), key=sort_key)
146
search_paths = [sorted_paths[0]]
147
for path in sorted_paths[1:]:
148
if not is_inside(search_paths[-1], path):
149
# This path is unique, add it
150
search_paths.append(path)
152
return set(search_paths)
388
471
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
389
472
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
390
473
return shutil.rmtree(path, ignore_errors, onerror)
475
f = win32utils.get_unicode_argv # special function or None
391
479
elif sys.platform == 'darwin':
392
480
getcwd = _mac_getcwd
395
def get_terminal_encoding():
483
def get_terminal_encoding(trace=False):
396
484
"""Find the best encoding for printing to the screen.
398
486
This attempts to check both sys.stdout and sys.stdin to see
399
487
what encoding they are in, and if that fails it falls back to
400
bzrlib.user_encoding.
488
osutils.get_user_encoding().
401
489
The problem is that on Windows, locale.getpreferredencoding()
402
490
is not the same encoding as that used by the console:
403
491
http://mail.python.org/pipermail/python-list/2003-May/162357.html
405
493
On my standard US Windows XP, the preferred encoding is
406
494
cp1252, but the console is cp437
496
:param trace: If True trace the selected encoding via mutter().
498
from bzrlib.trace import mutter
408
499
output_encoding = getattr(sys.stdout, 'encoding', None)
409
500
if not output_encoding:
410
501
input_encoding = getattr(sys.stdin, 'encoding', None)
411
502
if not input_encoding:
412
output_encoding = bzrlib.user_encoding
413
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
503
output_encoding = get_user_encoding()
505
mutter('encoding stdout as osutils.get_user_encoding() %r',
415
508
output_encoding = input_encoding
416
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
510
mutter('encoding stdout as sys.stdin encoding %r',
418
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
514
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
419
515
if output_encoding == 'cp0':
420
516
# invalid encoding (cp0 means 'no codepage' on Windows)
421
output_encoding = bzrlib.user_encoding
422
mutter('cp0 is invalid encoding.'
423
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
517
output_encoding = get_user_encoding()
519
mutter('cp0 is invalid encoding.'
520
' encoding stdout as osutils.get_user_encoding() %r',
426
524
codecs.lookup(output_encoding)
620
777
:param timezone: How to display the time: 'utc', 'original' for the
621
778
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.
780
:param date_fmt: strftime format.
781
:param show_offset: Whether to append the timezone.
783
(date_fmt, tt, offset_str) = \
784
_format_date(t, offset, timezone, date_fmt, show_offset)
785
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
786
date_str = time.strftime(date_fmt, tt)
787
return date_str + offset_str
790
# Cache of formatted offset strings
794
def format_date_with_offset_in_original_timezone(t, offset=0,
795
_cache=_offset_cache):
796
"""Return a formatted date string in the original timezone.
798
This routine may be faster then format_date.
800
:param t: Seconds since the epoch.
801
:param offset: Timezone offset in seconds east of utc.
805
tt = time.gmtime(t + offset)
806
date_fmt = _default_format_by_weekday_num[tt[6]]
807
date_str = time.strftime(date_fmt, tt)
808
offset_str = _cache.get(offset, None)
809
if offset_str is None:
810
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
811
_cache[offset] = offset_str
812
return date_str + offset_str
815
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
817
"""Return an unicode date string formatted according to the current locale.
819
:param t: Seconds since the epoch.
820
:param offset: Timezone offset in seconds east of utc.
821
:param timezone: How to display the time: 'utc', 'original' for the
822
timezone specified by offset, or 'local' for the process's current
824
:param date_fmt: strftime format.
825
:param show_offset: Whether to append the timezone.
827
(date_fmt, tt, offset_str) = \
828
_format_date(t, offset, timezone, date_fmt, show_offset)
829
date_str = time.strftime(date_fmt, tt)
830
if not isinstance(date_str, unicode):
831
date_str = date_str.decode(get_user_encoding(), 'replace')
832
return date_str + offset_str
835
def _format_date(t, offset, timezone, date_fmt, show_offset):
626
836
if timezone == 'utc':
627
837
tt = time.gmtime(t)
756
raise errors.BzrError("sorry, %r not allowed in path" % f)
963
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
757
964
elif (f == '.') or (f == ''):
764
assert isinstance(p, (list, tuple))
766
973
if (f == '..') or (f is None) or (f == ''):
767
raise errors.BzrError("sorry, %r not allowed in path" % f)
974
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
768
975
return pathjoin(*p)
771
@deprecated_function(zero_nine)
772
def appendpath(p1, p2):
776
return pathjoin(p1, p2)
978
def parent_directories(filename):
979
"""Return the list of parent directories, deepest first.
981
For example, parent_directories("a/b/c") -> ["a/b", "a"].
984
parts = splitpath(dirname(filename))
986
parents.append(joinpath(parts))
991
_extension_load_failures = []
994
def failed_to_load_extension(exception):
995
"""Handle failing to load a binary extension.
997
This should be called from the ImportError block guarding the attempt to
998
import the native extension. If this function returns, the pure-Python
999
implementation should be loaded instead::
1002
>>> import bzrlib._fictional_extension_pyx
1003
>>> except ImportError, e:
1004
>>> bzrlib.osutils.failed_to_load_extension(e)
1005
>>> import bzrlib._fictional_extension_py
1007
# NB: This docstring is just an example, not a doctest, because doctest
1008
# currently can't cope with the use of lazy imports in this namespace --
1011
# This currently doesn't report the failure at the time it occurs, because
1012
# they tend to happen very early in startup when we can't check config
1013
# files etc, and also we want to report all failures but not spam the user
1015
exception_str = str(exception)
1016
if exception_str not in _extension_load_failures:
1017
trace.mutter("failed to load compiled extension: %s" % exception_str)
1018
_extension_load_failures.append(exception_str)
1021
def report_extension_load_failures():
1022
if not _extension_load_failures:
1024
if config.GlobalStack().get('ignore_missing_extensions'):
1026
# the warnings framework should by default show this only once
1027
from bzrlib.trace import warning
1029
"bzr: warning: some compiled extensions could not be loaded; "
1030
"see <https://answers.launchpad.net/bzr/+faq/703>")
1031
# we no longer show the specific missing extensions here, because it makes
1032
# the message too long and scary - see
1033
# https://bugs.launchpad.net/bzr/+bug/430529
1037
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1038
except ImportError, e:
1039
failed_to_load_extension(e)
1040
from bzrlib._chunks_to_lines_py import chunks_to_lines
779
1043
def split_lines(s):
780
1044
"""Split s into lines, but without removing the newline characters."""
1045
# Trivially convert a fulltext into a 'chunked' representation, and let
1046
# chunks_to_lines do the heavy lifting.
1047
if isinstance(s, str):
1048
# chunks_to_lines only supports 8-bit strings
1049
return chunks_to_lines([s])
1051
return _split_lines(s)
1054
def _split_lines(s):
1055
"""Split s into lines, but without removing the newline characters.
1057
This supports Unicode or plain string objects.
781
1059
lines = s.split('\n')
782
1060
result = [line + '\n' for line in lines[:-1]]
857
1182
os.path.commonprefix (python2.4) has a bad bug that it works just
858
1183
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
859
1184
avoids that problem.
1186
NOTE: `base` should not have a trailing slash otherwise you'll get
1187
PathNotChild exceptions regardless of `path`.
862
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
863
' exceed the platform minimum length (which is %d)' %
1190
if len(base) < MIN_ABS_PATHLENGTH:
1191
# must have space for e.g. a drive letter
1192
raise ValueError(gettext('%r is too short to calculate a relative path')
866
1195
rp = abspath(path)
870
while len(head) >= len(base):
1200
if len(head) <= len(base) and head != base:
1201
raise errors.PathNotChild(rp, base)
871
1202
if head == base:
873
head, tail = os.path.split(head)
1204
head, tail = split(head)
877
raise errors.PathNotChild(rp, base)
1209
return pathjoin(*reversed(s))
1214
def _cicp_canonical_relpath(base, path):
1215
"""Return the canonical path relative to base.
1217
Like relpath, but on case-insensitive-case-preserving file-systems, this
1218
will return the relpath as stored on the file-system rather than in the
1219
case specified in the input string, for all existing portions of the path.
1221
This will cause O(N) behaviour if called for every path in a tree; if you
1222
have a number of paths to convert, you should use canonical_relpaths().
1224
# TODO: it should be possible to optimize this for Windows by using the
1225
# win32 API FindFiles function to look for the specified name - but using
1226
# os.listdir() still gives us the correct, platform agnostic semantics in
1229
rel = relpath(base, path)
1230
# '.' will have been turned into ''
1234
abs_base = abspath(base)
1236
_listdir = os.listdir
1238
# use an explicit iterator so we can easily consume the rest on early exit.
1239
bit_iter = iter(rel.split('/'))
1240
for bit in bit_iter:
1243
next_entries = _listdir(current)
1244
except OSError: # enoent, eperm, etc
1245
# We can't find this in the filesystem, so just append the
1247
current = pathjoin(current, bit, *list(bit_iter))
1249
for look in next_entries:
1250
if lbit == look.lower():
1251
current = pathjoin(current, look)
1254
# got to the end, nothing matched, so we just return the
1255
# non-existing bits as they were specified (the filename may be
1256
# the target of a move, for example).
1257
current = pathjoin(current, bit, *list(bit_iter))
1259
return current[len(abs_base):].lstrip('/')
1261
# XXX - TODO - we need better detection/integration of case-insensitive
1262
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1263
# filesystems), for example, so could probably benefit from the same basic
1264
# support there. For now though, only Windows and OSX get that support, and
1265
# they get it for *all* file-systems!
1266
if sys.platform in ('win32', 'darwin'):
1267
canonical_relpath = _cicp_canonical_relpath
1269
canonical_relpath = relpath
1271
def canonical_relpaths(base, paths):
1272
"""Create an iterable to canonicalize a sequence of relative paths.
1274
The intent is for this implementation to use a cache, vastly speeding
1275
up multiple transformations in the same directory.
1277
# but for now, we haven't optimized...
1278
return [canonical_relpath(base, p) for p in paths]
1281
def decode_filename(filename):
1282
"""Decode the filename using the filesystem encoding
1284
If it is unicode, it is returned.
1285
Otherwise it is decoded from the the filesystem's encoding. If decoding
1286
fails, a errors.BadFilenameEncoding exception is raised.
1288
if type(filename) is unicode:
1291
return filename.decode(_fs_enc)
1292
except UnicodeDecodeError:
1293
raise errors.BadFilenameEncoding(filename, _fs_enc)
885
1296
def safe_unicode(unicode_or_utf8_string):
886
1297
"""Coerce unicode_or_utf8_string into unicode.
888
1299
If it is unicode, it is returned.
889
Otherwise it is decoded from utf-8. If a decoding error
890
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
891
as a BzrBadParameter exception.
1300
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1301
wrapped in a BzrBadParameterNotUnicode exception.
893
1303
if isinstance(unicode_or_utf8_string, unicode):
894
1304
return unicode_or_utf8_string
1007
1417
normalized_filename = _inaccessible_normalized_filename
1420
def set_signal_handler(signum, handler, restart_syscall=True):
1421
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1422
on platforms that support that.
1424
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1425
automatically restart (by calling `signal.siginterrupt(signum,
1426
False)`). May be ignored if the feature is not available on this
1427
platform or Python version.
1431
siginterrupt = signal.siginterrupt
1433
# This python implementation doesn't provide signal support, hence no
1436
except AttributeError:
1437
# siginterrupt doesn't exist on this platform, or for this version
1439
siginterrupt = lambda signum, flag: None
1441
def sig_handler(*args):
1442
# Python resets the siginterrupt flag when a signal is
1443
# received. <http://bugs.python.org/issue8354>
1444
# As a workaround for some cases, set it back the way we want it.
1445
siginterrupt(signum, False)
1446
# Now run the handler function passed to set_signal_handler.
1449
sig_handler = handler
1450
old_handler = signal.signal(signum, sig_handler)
1452
siginterrupt(signum, False)
1456
default_terminal_width = 80
1457
"""The default terminal width for ttys.
1459
This is defined so that higher levels can share a common fallback value when
1460
terminal_width() returns None.
1463
# Keep some state so that terminal_width can detect if _terminal_size has
1464
# returned a different size since the process started. See docstring and
1465
# comments of terminal_width for details.
1466
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1467
_terminal_size_state = 'no_data'
1468
_first_terminal_size = None
1010
1470
def terminal_width():
1011
"""Return estimated terminal width."""
1012
if sys.platform == 'win32':
1013
return win32utils.get_console_size()[0]
1471
"""Return terminal width.
1473
None is returned if the width can't established precisely.
1476
- if BZR_COLUMNS is set, returns its value
1477
- if there is no controlling terminal, returns None
1478
- query the OS, if the queried size has changed since the last query,
1480
- if COLUMNS is set, returns its value,
1481
- if the OS has a value (even though it's never changed), return its value.
1483
From there, we need to query the OS to get the size of the controlling
1486
On Unices we query the OS by:
1487
- get termios.TIOCGWINSZ
1488
- if an error occurs or a negative value is obtained, returns None
1490
On Windows we query the OS by:
1491
- win32utils.get_console_size() decides,
1492
- returns None on error (provided default value)
1494
# Note to implementors: if changing the rules for determining the width,
1495
# make sure you've considered the behaviour in these cases:
1496
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1497
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1499
# - (add more interesting cases here, if you find any)
1500
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1501
# but we don't want to register a signal handler because it is impossible
1502
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1503
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1504
# time so we can notice if the reported size has changed, which should have
1507
# If BZR_COLUMNS is set, take it, user is always right
1508
# Except if they specified 0 in which case, impose no limit here
1510
width = int(os.environ['BZR_COLUMNS'])
1511
except (KeyError, ValueError):
1513
if width is not None:
1519
isatty = getattr(sys.stdout, 'isatty', None)
1520
if isatty is None or not isatty():
1521
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1525
width, height = os_size = _terminal_size(None, None)
1526
global _first_terminal_size, _terminal_size_state
1527
if _terminal_size_state == 'no_data':
1528
_first_terminal_size = os_size
1529
_terminal_size_state = 'unchanged'
1530
elif (_terminal_size_state == 'unchanged' and
1531
_first_terminal_size != os_size):
1532
_terminal_size_state = 'changed'
1534
# If the OS claims to know how wide the terminal is, and this value has
1535
# ever changed, use that.
1536
if _terminal_size_state == 'changed':
1537
if width is not None and width > 0:
1540
# If COLUMNS is set, use it.
1542
return int(os.environ['COLUMNS'])
1543
except (KeyError, ValueError):
1546
# Finally, use an unchanged size from the OS, if we have one.
1547
if _terminal_size_state == 'unchanged':
1548
if width is not None and width > 0:
1551
# The width could not be determined.
1555
def _win32_terminal_size(width, height):
1556
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1557
return width, height
1560
def _ioctl_terminal_size(width, height):
1016
1562
import struct, fcntl, termios
1017
1563
s = struct.pack('HHHH', 0, 0, 0, 0)
1018
1564
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1019
width = struct.unpack('HHHH', x)[1]
1565
height, width = struct.unpack('HHHH', x)[0:2]
1566
except (IOError, AttributeError):
1024
width = int(os.environ['COLUMNS'])
1568
return width, height
1570
_terminal_size = None
1571
"""Returns the terminal size as (width, height).
1573
:param width: Default value for width.
1574
:param height: Default value for height.
1576
This is defined specifically for each OS and query the size of the controlling
1577
terminal. If any error occurs, the provided default values should be returned.
1579
if sys.platform == 'win32':
1580
_terminal_size = _win32_terminal_size
1582
_terminal_size = _ioctl_terminal_size
1033
1585
def supports_executable():
1080
1632
raise errors.IllegalPath(path)
1635
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1637
def _is_error_enotdir(e):
1638
"""Check if this exception represents ENOTDIR.
1640
Unfortunately, python is very inconsistent about the exception
1641
here. The cases are:
1642
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1643
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1644
which is the windows error code.
1645
3) Windows, Python2.5 uses errno == EINVAL and
1646
winerror == ERROR_DIRECTORY
1648
:param e: An Exception object (expected to be OSError with an errno
1649
attribute, but we should be able to cope with anything)
1650
:return: True if this represents an ENOTDIR error. False otherwise.
1652
en = getattr(e, 'errno', None)
1653
if (en == errno.ENOTDIR
1654
or (sys.platform == 'win32'
1655
and (en == _WIN32_ERROR_DIRECTORY
1656
or (en == errno.EINVAL
1657
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1083
1663
def walkdirs(top, prefix=""):
1084
1664
"""Yield data about all the directories in a tree.
1086
1666
This yields all the data about the contents of a directory at a time.
1087
1667
After each directory has been yielded, if the caller has mutated the list
1088
1668
to exclude some directories, they are then not descended into.
1090
1670
The data yielded is of the form:
1091
1671
((directory-relpath, directory-path-from-top),
1092
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1672
[(relpath, basename, kind, lstat, path-from-top), ...]),
1093
1673
- directory-relpath is the relative path of the directory being returned
1094
1674
with respect to top. prefix is prepended to this.
1095
- directory-path-from-root is the path including top for this directory.
1675
- directory-path-from-root is the path including top for this directory.
1096
1676
It is suitable for use with os functions.
1097
1677
- relpath is the relative path within the subtree being walked.
1098
1678
- basename is the basename of the path
1154
1768
path-from-top might be unicode or utf8, but it is the correct path to
1155
1769
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
1771
global _selected_dir_reader
1772
if _selected_dir_reader is None:
1773
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1774
# Win98 doesn't have unicode apis like FindFirstFileW
1775
# TODO: We possibly could support Win98 by falling back to the
1776
# original FindFirstFile, and using TCHAR instead of WCHAR,
1777
# but that gets a bit tricky, and requires custom compiling
1780
from bzrlib._walkdirs_win32 import Win32ReadDir
1781
_selected_dir_reader = Win32ReadDir()
1784
elif _fs_enc in ('utf-8', 'ascii'):
1786
from bzrlib._readdir_pyx import UTF8DirReader
1787
_selected_dir_reader = UTF8DirReader()
1788
except ImportError, e:
1789
failed_to_load_extension(e)
1792
if _selected_dir_reader is None:
1793
# Fallback to the python version
1794
_selected_dir_reader = UnicodeDirReader()
1176
1796
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1177
1797
# 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))]
1798
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1799
read_dir = _selected_dir_reader.read_dir
1800
_directory = _directory_kind
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))
1802
relroot, _, _, _, top = pending[-1].pop()
1805
dirblock = sorted(read_dir(relroot, top))
1194
1806
yield (relroot, top), dirblock
1196
1807
# 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 + '/'
1808
next = [d for d in reversed(dirblock) if d[2] == _directory]
1810
pending.append(next)
1813
class UnicodeDirReader(DirReader):
1814
"""A dir reader for non-utf8 file systems, which transcodes."""
1816
__slots__ = ['_utf8_encode']
1819
self._utf8_encode = codecs.getencoder('utf8')
1821
def top_prefix_to_starting_dir(self, top, prefix=""):
1822
"""See DirReader.top_prefix_to_starting_dir."""
1823
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1825
def read_dir(self, prefix, top):
1826
"""Read a single directory from a non-utf8 file system.
1828
top, and the abspath element in the output are unicode, all other paths
1829
are utf8. Local disk IO is done via unicode calls to listdir etc.
1831
This is currently the fallback code path when the filesystem encoding is
1832
not UTF-8. It may be better to implement an alternative so that we can
1833
safely handle paths that are not properly decodable in the current
1836
See DirReader.read_dir for details.
1838
_utf8_encode = self._utf8_encode
1840
_listdir = os.listdir
1841
_kind_from_mode = file_kind_from_stat_mode
1844
relprefix = prefix + '/'
1223
1847
top_slash = top + u'/'
1357
2019
return user_encoding
1360
def recv_all(socket, bytes):
2022
def get_diff_header_encoding():
2023
return get_terminal_encoding()
2026
_message_encoding = None
2029
def get_message_encoding():
2030
"""Return the encoding used for messages
2032
While the message encoding is a general setting it should usually only be
2033
needed for decoding system error strings such as from OSError instances.
2035
global _message_encoding
2036
if _message_encoding is None:
2037
if os.name == "posix":
2039
# This is a process-global setting that can change, but should in
2040
# general just get set once at process startup then be constant.
2041
_message_encoding = locale.getlocale(locale.LC_MESSAGES)[1]
2043
# On windows want the result of GetACP() which this boils down to.
2044
_message_encoding = get_user_encoding()
2045
return _message_encoding or "ascii"
2048
def get_host_name():
2049
"""Return the current unicode host name.
2051
This is meant to be used in place of socket.gethostname() because that
2052
behaves inconsistently on different platforms.
2054
if sys.platform == "win32":
2056
return win32utils.get_host_name()
2059
return socket.gethostname().decode(get_user_encoding())
2062
# We must not read/write any more than 64k at a time from/to a socket so we
2063
# don't risk "no buffer space available" errors on some platforms. Windows in
2064
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2066
MAX_SOCKET_CHUNK = 64 * 1024
2068
_end_of_stream_errors = [errno.ECONNRESET]
2069
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2070
_eno = getattr(errno, _eno, None)
2071
if _eno is not None:
2072
_end_of_stream_errors.append(_eno)
2076
def read_bytes_from_socket(sock, report_activity=None,
2077
max_read_size=MAX_SOCKET_CHUNK):
2078
"""Read up to max_read_size of bytes from sock and notify of progress.
2080
Translates "Connection reset by peer" into file-like EOF (return an
2081
empty string rather than raise an error), and repeats the recv if
2082
interrupted by a signal.
2086
bytes = sock.recv(max_read_size)
2087
except socket.error, e:
2089
if eno in _end_of_stream_errors:
2090
# The connection was closed by the other side. Callers expect
2091
# an empty string to signal end-of-stream.
2093
elif eno == errno.EINTR:
2094
# Retry the interrupted recv.
2098
if report_activity is not None:
2099
report_activity(len(bytes), 'read')
2103
def recv_all(socket, count):
1361
2104
"""Receive an exact number of bytes.
1363
2106
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
2107
depending on what's in the OS buffer. MSG_WAITALL is not available
1365
2108
on all platforms, but this should work everywhere. This will return
1366
2109
less than the requested amount if the remote end closes.
1368
2111
This isn't optimized and is intended mostly for use in testing.
1371
while len(b) < bytes:
1372
new = socket.recv(bytes - len(b))
2114
while len(b) < count:
2115
new = read_bytes_from_socket(socket, None, count - len(b))
2122
def send_all(sock, bytes, report_activity=None):
2123
"""Send all bytes on a socket.
2125
Breaks large blocks in smaller chunks to avoid buffering limitations on
2126
some platforms, and catches EINTR which may be thrown if the send is
2127
interrupted by a signal.
2129
This is preferred to socket.sendall(), because it avoids portability bugs
2130
and provides activity reporting.
2132
:param report_activity: Call this as bytes are read, see
2133
Transport._report_activity
2136
byte_count = len(bytes)
2137
while sent_total < byte_count:
2139
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2140
except socket.error, e:
2141
if e.args[0] != errno.EINTR:
2145
report_activity(sent, 'write')
2148
def connect_socket(address):
2149
# Slight variation of the socket.create_connection() function (provided by
2150
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2151
# provide it for previous python versions. Also, we don't use the timeout
2152
# parameter (provided by the python implementation) so we don't implement
2154
err = socket.error('getaddrinfo returns an empty list')
2155
host, port = address
2156
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2157
af, socktype, proto, canonname, sa = res
2160
sock = socket.socket(af, socktype, proto)
2164
except socket.error, err:
2165
# 'err' is now the most recent error
2166
if sock is not None:
1378
2171
def dereference_path(path):
1379
2172
"""Determine the real path to a file.
1387
2180
# The pathjoin for '.' is a workaround for Python bug #1213894.
1388
2181
# (initial path components aren't dereferenced)
1389
2182
return pathjoin(realpath(pathjoin('.', parent)), base)
2185
def supports_mapi():
2186
"""Return True if we can use MAPI to launch a mail client."""
2187
return sys.platform == "win32"
2190
def resource_string(package, resource_name):
2191
"""Load a resource from a package and return it as a string.
2193
Note: Only packages that start with bzrlib are currently supported.
2195
This is designed to be a lightweight implementation of resource
2196
loading in a way which is API compatible with the same API from
2198
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2199
If and when pkg_resources becomes a standard library, this routine
2202
# Check package name is within bzrlib
2203
if package == "bzrlib":
2204
resource_relpath = resource_name
2205
elif package.startswith("bzrlib."):
2206
package = package[len("bzrlib."):].replace('.', os.sep)
2207
resource_relpath = pathjoin(package, resource_name)
2209
raise errors.BzrError('resource package %s not in bzrlib' % package)
2211
# Map the resource to a file and read its contents
2212
base = dirname(bzrlib.__file__)
2213
if getattr(sys, 'frozen', None): # bzr.exe
2214
base = abspath(pathjoin(base, '..', '..'))
2215
f = file(pathjoin(base, resource_relpath), "rU")
2221
def file_kind_from_stat_mode_thunk(mode):
2222
global file_kind_from_stat_mode
2223
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2225
from bzrlib._readdir_pyx import UTF8DirReader
2226
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2227
except ImportError, e:
2228
# This is one time where we won't warn that an extension failed to
2229
# load. The extension is never available on Windows anyway.
2230
from bzrlib._readdir_py import (
2231
_kind_from_mode as file_kind_from_stat_mode
2233
return file_kind_from_stat_mode(mode)
2234
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2236
def file_stat(f, _lstat=os.lstat):
2241
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2242
raise errors.NoSuchFile(f)
2245
def file_kind(f, _lstat=os.lstat):
2246
stat_value = file_stat(f, _lstat)
2247
return file_kind_from_stat_mode(stat_value.st_mode)
2249
def until_no_eintr(f, *a, **kw):
2250
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2252
WARNING: you must be certain that it is safe to retry the call repeatedly
2253
if EINTR does occur. This is typically only true for low-level operations
2254
like os.read. If in any doubt, don't use this.
2256
Keep in mind that this is not a complete solution to EINTR. There is
2257
probably code in the Python standard library and other dependencies that
2258
may encounter EINTR if a signal arrives (and there is signal handler for
2259
that signal). So this function can reduce the impact for IO that bzrlib
2260
directly controls, but it is not a complete solution.
2262
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2266
except (IOError, OSError), e:
2267
if e.errno == errno.EINTR:
2272
@deprecated_function(deprecated_in((2, 2, 0)))
2273
def re_compile_checked(re_string, flags=0, where=""):
2274
"""Return a compiled re, or raise a sensible error.
2276
This should only be used when compiling user-supplied REs.
2278
:param re_string: Text form of regular expression.
2279
:param flags: eg re.IGNORECASE
2280
:param where: Message explaining to the user the context where
2281
it occurred, eg 'log search filter'.
2283
# from https://bugs.launchpad.net/bzr/+bug/251352
2285
re_obj = re.compile(re_string, flags)
2288
except errors.InvalidPattern, e:
2290
where = ' in ' + where
2291
# despite the name 'error' is a type
2292
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2296
if sys.platform == "win32":
2299
return msvcrt.getch()
2304
fd = sys.stdin.fileno()
2305
settings = termios.tcgetattr(fd)
2308
ch = sys.stdin.read(1)
2310
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2313
if sys.platform.startswith('linux'):
2314
def _local_concurrency():
2316
return os.sysconf('SC_NPROCESSORS_ONLN')
2317
except (ValueError, OSError, AttributeError):
2319
elif sys.platform == 'darwin':
2320
def _local_concurrency():
2321
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2322
stdout=subprocess.PIPE).communicate()[0]
2323
elif "bsd" in sys.platform:
2324
def _local_concurrency():
2325
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2326
stdout=subprocess.PIPE).communicate()[0]
2327
elif sys.platform == 'sunos5':
2328
def _local_concurrency():
2329
return subprocess.Popen(['psrinfo', '-p',],
2330
stdout=subprocess.PIPE).communicate()[0]
2331
elif sys.platform == "win32":
2332
def _local_concurrency():
2333
# This appears to return the number of cores.
2334
return os.environ.get('NUMBER_OF_PROCESSORS')
2336
def _local_concurrency():
2341
_cached_local_concurrency = None
2343
def local_concurrency(use_cache=True):
2344
"""Return how many processes can be run concurrently.
2346
Rely on platform specific implementations and default to 1 (one) if
2347
anything goes wrong.
2349
global _cached_local_concurrency
2351
if _cached_local_concurrency is not None and use_cache:
2352
return _cached_local_concurrency
2354
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2355
if concurrency is None:
2357
import multiprocessing
2358
concurrency = multiprocessing.cpu_count()
2359
except (ImportError, NotImplementedError):
2360
# multiprocessing is only available on Python >= 2.6
2361
# and multiprocessing.cpu_count() isn't implemented on all
2364
concurrency = _local_concurrency()
2365
except (OSError, IOError):
2368
concurrency = int(concurrency)
2369
except (TypeError, ValueError):
2372
_cached_concurrency = concurrency
2376
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2377
"""A stream writer that doesn't decode str arguments."""
2379
def __init__(self, encode, stream, errors='strict'):
2380
codecs.StreamWriter.__init__(self, stream, errors)
2381
self.encode = encode
2383
def write(self, object):
2384
if type(object) is str:
2385
self.stream.write(object)
2387
data, _ = self.encode(object, self.errors)
2388
self.stream.write(data)
2390
if sys.platform == 'win32':
2391
def open_file(filename, mode='r', bufsize=-1):
2392
"""This function is used to override the ``open`` builtin.
2394
But it uses O_NOINHERIT flag so the file handle is not inherited by
2395
child processes. Deleting or renaming a closed file opened with this
2396
function is not blocking child processes.
2398
writing = 'w' in mode
2399
appending = 'a' in mode
2400
updating = '+' in mode
2401
binary = 'b' in mode
2404
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2405
# for flags for each modes.
2415
flags |= os.O_WRONLY
2416
flags |= os.O_CREAT | os.O_TRUNC
2421
flags |= os.O_WRONLY
2422
flags |= os.O_CREAT | os.O_APPEND
2427
flags |= os.O_RDONLY
2429
return os.fdopen(os.open(filename, flags), mode, bufsize)
2434
def getuser_unicode():
2435
"""Return the username as unicode.
2438
user_encoding = get_user_encoding()
2439
username = getpass.getuser().decode(user_encoding)
2440
except UnicodeDecodeError:
2441
raise errors.BzrError("Can't decode username as %s." % \
2443
except ImportError, e:
2444
if sys.platform != 'win32':
2446
if str(e) != 'No module named pwd':
2448
# https://bugs.launchpad.net/bzr/+bug/660174
2449
# getpass.getuser() is unable to return username on Windows
2450
# if there is no USERNAME environment variable set.
2451
# That could be true if bzr is running as a service,
2452
# e.g. running `bzr serve` as a service on Windows.
2453
# We should not fail with traceback in this case.
2454
username = u'UNKNOWN'
2458
def available_backup_name(base, exists):
2459
"""Find a non-existing backup file name.
2461
This will *not* create anything, this only return a 'free' entry. This
2462
should be used for checking names in a directory below a locked
2463
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2464
Leap) and generally discouraged.
2466
:param base: The base name.
2468
:param exists: A callable returning True if the path parameter exists.
2471
name = "%s.~%d~" % (base, counter)
2474
name = "%s.~%d~" % (base, counter)
2478
def set_fd_cloexec(fd):
2479
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2480
support for this is not available.
2484
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2485
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2486
except (ImportError, AttributeError):
2487
# Either the fcntl module or specific constants are not present
2491
def find_executable_on_path(name):
2492
"""Finds an executable on the PATH.
2494
On Windows, this will try to append each extension in the PATHEXT
2495
environment variable to the name, if it cannot be found with the name
2498
:param name: The base name of the executable.
2499
:return: The path to the executable found or None.
2501
path = os.environ.get('PATH')
2504
path = path.split(os.pathsep)
2505
if sys.platform == 'win32':
2506
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2507
exts = [ext.lower() for ext in exts]
2508
base, ext = os.path.splitext(name)
2510
if ext.lower() not in exts:
2518
f = os.path.join(d, name) + ext
2519
if os.access(f, os.X_OK):
2524
def _posix_is_local_pid_dead(pid):
2525
"""True if pid doesn't correspond to live process on this machine"""
2527
# Special meaning of unix kill: just check if it's there.
2530
if e.errno == errno.ESRCH:
2531
# On this machine, and really not found: as sure as we can be
2534
elif e.errno == errno.EPERM:
2535
# exists, though not ours
2538
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2539
# Don't really know.
2542
# Exists and our process: not dead.
2545
if sys.platform == "win32":
2546
is_local_pid_dead = win32utils.is_local_pid_dead
2548
is_local_pid_dead = _posix_is_local_pid_dead
2551
def fdatasync(fileno):
2552
"""Flush file contents to disk if possible.
2554
:param fileno: Integer OS file handle.
2555
:raises TransportNotPossible: If flushing to disk is not possible.
2557
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2562
def ensure_empty_directory_exists(path, exception_class):
2563
"""Make sure a local directory exists and is empty.
2565
If it does not exist, it is created. If it exists and is not empty, an
2566
instance of exception_class is raised.
2571
if e.errno != errno.EEXIST:
2573
if os.listdir(path) != []:
2574
raise exception_class(path)
2577
def is_environment_error(evalue):
2578
"""True if exception instance is due to a process environment issue
2580
This includes OSError and IOError, but also other errors that come from
2581
the operating system or core libraries but are not subclasses of those.
2583
if isinstance(evalue, (EnvironmentError, select.error)):
2585
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
added
removed
bzrlib/osutils.py
