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
17
from __future__ import absolute_import
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)
27
26
from bzrlib.lazy_import import lazy_import
28
27
lazy_import(globals(), """
29
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,
35
# We need to import both shutil and rmtree as we export the later on posix
36
# and need the former on windows
38
from shutil import rmtree
41
# We need to import both tempfile and mkdtemp as we export the later on posix
42
# and need the former on windows
44
from tempfile import mkdtemp
44
from tempfile import (
47
49
from bzrlib import (
54
from bzrlib.i18n import gettext
57
from bzrlib import symbol_versioning
57
58
from bzrlib.symbol_versioning import (
59
59
deprecated_function,
62
warn as warn_deprecated,
72
from bzrlib import symbol_versioning, _fs_enc
75
# Cross platform wall-clock time functionality with decent resolution.
76
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
77
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
78
# synchronized with ``time.time()``, this is only meant to be used to find
79
# delta times by subtracting from another call to this function.
80
timer_func = time.time
81
if sys.platform == 'win32':
82
timer_func = time.clock
61
from bzrlib.trace import mutter
84
64
# On win32, O_BINARY is used to indicate the file should
85
65
# be opened in binary mode, rather than text mode.
86
66
# On other platforms, O_BINARY doesn't exist, because
87
67
# they always open in binary mode, so it is okay to
88
# OR with 0 on those platforms.
89
# O_NOINHERIT and O_TEXT exists only on win32 too.
68
# OR with 0 on those platforms
90
69
O_BINARY = getattr(os, 'O_BINARY', 0)
91
O_TEXT = getattr(os, 'O_TEXT', 0)
92
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
95
def get_unicode_argv():
97
user_encoding = get_user_encoding()
98
return [a.decode(user_encoding) for a in sys.argv[1:]]
99
except UnicodeDecodeError:
100
raise errors.BzrError(gettext("Parameter {0!r} encoding is unsupported by {1} "
101
"application locale.").format(a, user_encoding))
104
72
def make_readonly(filename):
288
252
unlink_func(tmp_name)
290
254
rename_func(tmp_name, new)
291
if failure_exc is not None:
293
raise failure_exc[0], failure_exc[1], failure_exc[2]
298
257
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
299
258
# choke on a Unicode string containing a relative path if
300
259
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
261
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
302
262
def _posix_abspath(path):
303
263
# jam 20060426 rather than encoding to fsencoding
304
264
# copy posixpath.abspath, but use os.getcwdu instead
305
265
if not posixpath.isabs(path):
306
266
path = posixpath.join(getcwd(), path)
307
return _posix_normpath(path)
267
return posixpath.normpath(path)
310
270
def _posix_realpath(path):
311
271
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
314
def _posix_normpath(path):
315
path = posixpath.normpath(path)
316
# Bug 861008: posixpath.normpath() returns a path normalized according to
317
# the POSIX standard, which stipulates (for compatibility reasons) that two
318
# leading slashes must not be simplified to one, and only if there are 3 or
319
# more should they be simplified as one. So we treat the leading 2 slashes
320
# as a special case here by simply removing the first slash, as we consider
321
# that breaking POSIX compatibility for this obscure feature is acceptable.
322
# This is not a paranoid precaution, as we notably get paths like this when
323
# the repo is hosted at the root of the filesystem, i.e. in "/".
324
if path.startswith('//'):
329
def _posix_path_from_environ(key):
330
"""Get unicode path from `key` in environment or None if not present
332
Note that posix systems use arbitrary byte strings for filesystem objects,
333
so a path that raises BadFilenameEncoding here may still be accessible.
335
val = os.environ.get(key, None)
339
return val.decode(_fs_enc)
340
except UnicodeDecodeError:
341
# GZ 2011-12-12:Ideally want to include `key` in the exception message
342
raise errors.BadFilenameEncoding(val, _fs_enc)
345
def _posix_get_home_dir():
346
"""Get the home directory of the current user as a unicode path"""
347
path = posixpath.expanduser("~")
349
return path.decode(_fs_enc)
350
except UnicodeDecodeError:
351
raise errors.BadFilenameEncoding(path, _fs_enc)
354
def _posix_getuser_unicode():
355
"""Get username from environment or password database as unicode"""
356
name = getpass.getuser()
357
user_encoding = get_user_encoding()
359
return name.decode(user_encoding)
360
except UnicodeDecodeError:
361
raise errors.BzrError("Encoding of username %r is unsupported by %s "
362
"application locale." % (name, user_encoding))
365
274
def _win32_fixdrive(path):
366
275
"""Force drive letters to be consistent.
538
408
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
539
409
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
540
410
return shutil.rmtree(path, ignore_errors, onerror)
542
f = win32utils.get_unicode_argv # special function or None
545
path_from_environ = win32utils.get_environ_unicode
546
_get_home_dir = win32utils.get_home_location
547
getuser_unicode = win32utils.get_user_name
549
411
elif sys.platform == 'darwin':
550
412
getcwd = _mac_getcwd
553
def get_terminal_encoding(trace=False):
415
def get_terminal_encoding():
554
416
"""Find the best encoding for printing to the screen.
556
418
This attempts to check both sys.stdout and sys.stdin to see
557
419
what encoding they are in, and if that fails it falls back to
558
osutils.get_user_encoding().
420
bzrlib.user_encoding.
559
421
The problem is that on Windows, locale.getpreferredencoding()
560
422
is not the same encoding as that used by the console:
561
423
http://mail.python.org/pipermail/python-list/2003-May/162357.html
563
425
On my standard US Windows XP, the preferred encoding is
564
426
cp1252, but the console is cp437
566
:param trace: If True trace the selected encoding via mutter().
568
from bzrlib.trace import mutter
569
428
output_encoding = getattr(sys.stdout, 'encoding', None)
570
429
if not output_encoding:
571
430
input_encoding = getattr(sys.stdin, 'encoding', None)
572
431
if not input_encoding:
573
output_encoding = get_user_encoding()
575
mutter('encoding stdout as osutils.get_user_encoding() %r',
432
output_encoding = bzrlib.user_encoding
433
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
578
435
output_encoding = input_encoding
580
mutter('encoding stdout as sys.stdin encoding %r',
436
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
584
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
438
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
585
439
if output_encoding == 'cp0':
586
440
# invalid encoding (cp0 means 'no codepage' on Windows)
587
output_encoding = get_user_encoding()
589
mutter('cp0 is invalid encoding.'
590
' encoding stdout as osutils.get_user_encoding() %r',
441
output_encoding = bzrlib.user_encoding
442
mutter('cp0 is invalid encoding.'
443
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
594
446
codecs.lookup(output_encoding)
847
651
:param timezone: How to display the time: 'utc', 'original' for the
848
652
timezone specified by offset, or 'local' for the process's current
850
:param date_fmt: strftime format.
851
:param show_offset: Whether to append the timezone.
853
(date_fmt, tt, offset_str) = \
854
_format_date(t, offset, timezone, date_fmt, show_offset)
855
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
856
date_str = time.strftime(date_fmt, tt)
857
return date_str + offset_str
860
# Cache of formatted offset strings
864
def format_date_with_offset_in_original_timezone(t, offset=0,
865
_cache=_offset_cache):
866
"""Return a formatted date string in the original timezone.
868
This routine may be faster then format_date.
870
:param t: Seconds since the epoch.
871
:param offset: Timezone offset in seconds east of utc.
875
tt = time.gmtime(t + offset)
876
date_fmt = _default_format_by_weekday_num[tt[6]]
877
date_str = time.strftime(date_fmt, tt)
878
offset_str = _cache.get(offset, None)
879
if offset_str is None:
880
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
881
_cache[offset] = offset_str
882
return date_str + offset_str
885
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
887
"""Return an unicode date string formatted according to the current locale.
889
:param t: Seconds since the epoch.
890
:param offset: Timezone offset in seconds east of utc.
891
:param timezone: How to display the time: 'utc', 'original' for the
892
timezone specified by offset, or 'local' for the process's current
894
:param date_fmt: strftime format.
895
:param show_offset: Whether to append the timezone.
897
(date_fmt, tt, offset_str) = \
898
_format_date(t, offset, timezone, date_fmt, show_offset)
899
date_str = time.strftime(date_fmt, tt)
900
if not isinstance(date_str, unicode):
901
date_str = date_str.decode(get_user_encoding(), 'replace')
902
return date_str + offset_str
905
def _format_date(t, offset, timezone, date_fmt, show_offset):
654
:param show_offset: Whether to append the timezone.
655
:param date_fmt: strftime format.
906
657
if timezone == 'utc':
907
658
tt = time.gmtime(t)
1030
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
784
raise errors.BzrError("sorry, %r not allowed in path" % f)
1031
785
elif (f == '.') or (f == ''):
1038
791
def joinpath(p):
1040
793
if (f == '..') or (f is None) or (f == ''):
1041
raise errors.BzrError(gettext("sorry, %r not allowed in path") % f)
794
raise errors.BzrError("sorry, %r not allowed in path" % f)
1042
795
return pathjoin(*p)
1045
def parent_directories(filename):
1046
"""Return the list of parent directories, deepest first.
1048
For example, parent_directories("a/b/c") -> ["a/b", "a"].
1051
parts = splitpath(dirname(filename))
1053
parents.append(joinpath(parts))
1058
_extension_load_failures = []
1061
def failed_to_load_extension(exception):
1062
"""Handle failing to load a binary extension.
1064
This should be called from the ImportError block guarding the attempt to
1065
import the native extension. If this function returns, the pure-Python
1066
implementation should be loaded instead::
1069
>>> import bzrlib._fictional_extension_pyx
1070
>>> except ImportError, e:
1071
>>> bzrlib.osutils.failed_to_load_extension(e)
1072
>>> import bzrlib._fictional_extension_py
1074
# NB: This docstring is just an example, not a doctest, because doctest
1075
# currently can't cope with the use of lazy imports in this namespace --
1078
# This currently doesn't report the failure at the time it occurs, because
1079
# they tend to happen very early in startup when we can't check config
1080
# files etc, and also we want to report all failures but not spam the user
1082
exception_str = str(exception)
1083
if exception_str not in _extension_load_failures:
1084
trace.mutter("failed to load compiled extension: %s" % exception_str)
1085
_extension_load_failures.append(exception_str)
1088
def report_extension_load_failures():
1089
if not _extension_load_failures:
1091
if config.GlobalStack().get('ignore_missing_extensions'):
1093
# the warnings framework should by default show this only once
1094
from bzrlib.trace import warning
1096
"bzr: warning: some compiled extensions could not be loaded; "
1097
"see <https://answers.launchpad.net/bzr/+faq/703>")
1098
# we no longer show the specific missing extensions here, because it makes
1099
# the message too long and scary - see
1100
# https://bugs.launchpad.net/bzr/+bug/430529
1104
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1105
except ImportError, e:
1106
failed_to_load_extension(e)
1107
from bzrlib._chunks_to_lines_py import chunks_to_lines
1110
798
def split_lines(s):
1111
799
"""Split s into lines, but without removing the newline characters."""
1112
# Trivially convert a fulltext into a 'chunked' representation, and let
1113
# chunks_to_lines do the heavy lifting.
1114
if isinstance(s, str):
1115
# chunks_to_lines only supports 8-bit strings
1116
return chunks_to_lines([s])
1118
return _split_lines(s)
1121
def _split_lines(s):
1122
"""Split s into lines, but without removing the newline characters.
1124
This supports Unicode or plain string objects.
1126
800
lines = s.split('\n')
1127
801
result = [line + '\n' for line in lines[:-1]]
1249
891
os.path.commonprefix (python2.4) has a bad bug that it works just
1250
892
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1251
893
avoids that problem.
1253
NOTE: `base` should not have a trailing slash otherwise you'll get
1254
PathNotChild exceptions regardless of `path`.
1257
896
if len(base) < MIN_ABS_PATHLENGTH:
1258
897
# must have space for e.g. a drive letter
1259
raise ValueError(gettext('%r is too short to calculate a relative path')
898
raise ValueError('%r is too short to calculate a relative path'
1262
901
rp = abspath(path)
1267
if len(head) <= len(base) and head != base:
1268
raise errors.PathNotChild(rp, base)
905
while len(head) >= len(base):
1269
906
if head == base:
1271
head, tail = split(head)
908
head, tail = os.path.split(head)
912
raise errors.PathNotChild(rp, base)
1276
return pathjoin(*reversed(s))
1281
def _cicp_canonical_relpath(base, path):
1282
"""Return the canonical path relative to base.
1284
Like relpath, but on case-insensitive-case-preserving file-systems, this
1285
will return the relpath as stored on the file-system rather than in the
1286
case specified in the input string, for all existing portions of the path.
1288
This will cause O(N) behaviour if called for every path in a tree; if you
1289
have a number of paths to convert, you should use canonical_relpaths().
1291
# TODO: it should be possible to optimize this for Windows by using the
1292
# win32 API FindFiles function to look for the specified name - but using
1293
# os.listdir() still gives us the correct, platform agnostic semantics in
1296
rel = relpath(base, path)
1297
# '.' will have been turned into ''
1301
abs_base = abspath(base)
1303
_listdir = os.listdir
1305
# use an explicit iterator so we can easily consume the rest on early exit.
1306
bit_iter = iter(rel.split('/'))
1307
for bit in bit_iter:
1310
next_entries = _listdir(current)
1311
except OSError: # enoent, eperm, etc
1312
# We can't find this in the filesystem, so just append the
1314
current = pathjoin(current, bit, *list(bit_iter))
1316
for look in next_entries:
1317
if lbit == look.lower():
1318
current = pathjoin(current, look)
1321
# got to the end, nothing matched, so we just return the
1322
# non-existing bits as they were specified (the filename may be
1323
# the target of a move, for example).
1324
current = pathjoin(current, bit, *list(bit_iter))
1326
return current[len(abs_base):].lstrip('/')
1328
# XXX - TODO - we need better detection/integration of case-insensitive
1329
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1330
# filesystems), for example, so could probably benefit from the same basic
1331
# support there. For now though, only Windows and OSX get that support, and
1332
# they get it for *all* file-systems!
1333
if sys.platform in ('win32', 'darwin'):
1334
canonical_relpath = _cicp_canonical_relpath
1336
canonical_relpath = relpath
1338
def canonical_relpaths(base, paths):
1339
"""Create an iterable to canonicalize a sequence of relative paths.
1341
The intent is for this implementation to use a cache, vastly speeding
1342
up multiple transformations in the same directory.
1344
# but for now, we haven't optimized...
1345
return [canonical_relpath(base, p) for p in paths]
1348
def decode_filename(filename):
1349
"""Decode the filename using the filesystem encoding
1351
If it is unicode, it is returned.
1352
Otherwise it is decoded from the the filesystem's encoding. If decoding
1353
fails, a errors.BadFilenameEncoding exception is raised.
1355
if type(filename) is unicode:
1358
return filename.decode(_fs_enc)
1359
except UnicodeDecodeError:
1360
raise errors.BadFilenameEncoding(filename, _fs_enc)
1363
920
def safe_unicode(unicode_or_utf8_string):
1364
921
"""Coerce unicode_or_utf8_string into unicode.
1366
923
If it is unicode, it is returned.
1367
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1368
wrapped in a BzrBadParameterNotUnicode exception.
924
Otherwise it is decoded from utf-8. If a decoding error
925
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
926
as a BzrBadParameter exception.
1370
928
if isinstance(unicode_or_utf8_string, unicode):
1371
929
return unicode_or_utf8_string
1484
1042
normalized_filename = _inaccessible_normalized_filename
1487
def set_signal_handler(signum, handler, restart_syscall=True):
1488
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1489
on platforms that support that.
1491
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1492
automatically restart (by calling `signal.siginterrupt(signum,
1493
False)`). May be ignored if the feature is not available on this
1494
platform or Python version.
1498
siginterrupt = signal.siginterrupt
1500
# This python implementation doesn't provide signal support, hence no
1503
except AttributeError:
1504
# siginterrupt doesn't exist on this platform, or for this version
1506
siginterrupt = lambda signum, flag: None
1508
def sig_handler(*args):
1509
# Python resets the siginterrupt flag when a signal is
1510
# received. <http://bugs.python.org/issue8354>
1511
# As a workaround for some cases, set it back the way we want it.
1512
siginterrupt(signum, False)
1513
# Now run the handler function passed to set_signal_handler.
1516
sig_handler = handler
1517
old_handler = signal.signal(signum, sig_handler)
1519
siginterrupt(signum, False)
1523
default_terminal_width = 80
1524
"""The default terminal width for ttys.
1526
This is defined so that higher levels can share a common fallback value when
1527
terminal_width() returns None.
1530
# Keep some state so that terminal_width can detect if _terminal_size has
1531
# returned a different size since the process started. See docstring and
1532
# comments of terminal_width for details.
1533
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1534
_terminal_size_state = 'no_data'
1535
_first_terminal_size = None
1537
1045
def terminal_width():
1538
"""Return terminal width.
1540
None is returned if the width can't established precisely.
1543
- if BZR_COLUMNS is set, returns its value
1544
- if there is no controlling terminal, returns None
1545
- query the OS, if the queried size has changed since the last query,
1547
- if COLUMNS is set, returns its value,
1548
- if the OS has a value (even though it's never changed), return its value.
1550
From there, we need to query the OS to get the size of the controlling
1553
On Unices we query the OS by:
1554
- get termios.TIOCGWINSZ
1555
- if an error occurs or a negative value is obtained, returns None
1557
On Windows we query the OS by:
1558
- win32utils.get_console_size() decides,
1559
- returns None on error (provided default value)
1561
# Note to implementors: if changing the rules for determining the width,
1562
# make sure you've considered the behaviour in these cases:
1563
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1564
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1566
# - (add more interesting cases here, if you find any)
1567
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1568
# but we don't want to register a signal handler because it is impossible
1569
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1570
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1571
# time so we can notice if the reported size has changed, which should have
1574
# If BZR_COLUMNS is set, take it, user is always right
1575
# Except if they specified 0 in which case, impose no limit here
1577
width = int(os.environ['BZR_COLUMNS'])
1578
except (KeyError, ValueError):
1580
if width is not None:
1586
isatty = getattr(sys.stdout, 'isatty', None)
1587
if isatty is None or not isatty():
1588
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1592
width, height = os_size = _terminal_size(None, None)
1593
global _first_terminal_size, _terminal_size_state
1594
if _terminal_size_state == 'no_data':
1595
_first_terminal_size = os_size
1596
_terminal_size_state = 'unchanged'
1597
elif (_terminal_size_state == 'unchanged' and
1598
_first_terminal_size != os_size):
1599
_terminal_size_state = 'changed'
1601
# If the OS claims to know how wide the terminal is, and this value has
1602
# ever changed, use that.
1603
if _terminal_size_state == 'changed':
1604
if width is not None and width > 0:
1607
# If COLUMNS is set, use it.
1609
return int(os.environ['COLUMNS'])
1610
except (KeyError, ValueError):
1613
# Finally, use an unchanged size from the OS, if we have one.
1614
if _terminal_size_state == 'unchanged':
1615
if width is not None and width > 0:
1618
# The width could not be determined.
1622
def _win32_terminal_size(width, height):
1623
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1624
return width, height
1627
def _ioctl_terminal_size(width, height):
1046
"""Return estimated terminal width."""
1047
if sys.platform == 'win32':
1048
return win32utils.get_console_size()[0]
1629
1051
import struct, fcntl, termios
1630
1052
s = struct.pack('HHHH', 0, 0, 0, 0)
1631
1053
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1632
height, width = struct.unpack('HHHH', x)[0:2]
1633
except (IOError, AttributeError):
1054
width = struct.unpack('HHHH', x)[1]
1635
return width, height
1637
_terminal_size = None
1638
"""Returns the terminal size as (width, height).
1640
:param width: Default value for width.
1641
:param height: Default value for height.
1643
This is defined specifically for each OS and query the size of the controlling
1644
terminal. If any error occurs, the provided default values should be returned.
1646
if sys.platform == 'win32':
1647
_terminal_size = _win32_terminal_size
1649
_terminal_size = _ioctl_terminal_size
1059
width = int(os.environ['COLUMNS'])
1652
1068
def supports_executable():
1699
1115
raise errors.IllegalPath(path)
1702
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1704
def _is_error_enotdir(e):
1705
"""Check if this exception represents ENOTDIR.
1707
Unfortunately, python is very inconsistent about the exception
1708
here. The cases are:
1709
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1710
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1711
which is the windows error code.
1712
3) Windows, Python2.5 uses errno == EINVAL and
1713
winerror == ERROR_DIRECTORY
1715
:param e: An Exception object (expected to be OSError with an errno
1716
attribute, but we should be able to cope with anything)
1717
:return: True if this represents an ENOTDIR error. False otherwise.
1719
en = getattr(e, 'errno', None)
1720
if (en == errno.ENOTDIR
1721
or (sys.platform == 'win32'
1722
and (en == _WIN32_ERROR_DIRECTORY
1723
or (en == errno.EINVAL
1724
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1730
1118
def walkdirs(top, prefix=""):
1731
1119
"""Yield data about all the directories in a tree.
1733
1121
This yields all the data about the contents of a directory at a time.
1734
1122
After each directory has been yielded, if the caller has mutated the list
1735
1123
to exclude some directories, they are then not descended into.
1737
1125
The data yielded is of the form:
1738
1126
((directory-relpath, directory-path-from-top),
1739
1127
[(relpath, basename, kind, lstat, path-from-top), ...]),
1740
1128
- directory-relpath is the relative path of the directory being returned
1741
1129
with respect to top. prefix is prepended to this.
1742
- directory-path-from-root is the path including top for this directory.
1130
- directory-path-from-root is the path including top for this directory.
1743
1131
It is suitable for use with os functions.
1744
1132
- relpath is the relative path within the subtree being walked.
1745
1133
- basename is the basename of the path
1835
1189
path-from-top might be unicode or utf8, but it is the correct path to
1836
1190
pass to os functions to affect the file in question. (such as os.lstat)
1838
global _selected_dir_reader
1839
if _selected_dir_reader is None:
1840
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1841
# Win98 doesn't have unicode apis like FindFirstFileW
1842
# TODO: We possibly could support Win98 by falling back to the
1843
# original FindFirstFile, and using TCHAR instead of WCHAR,
1844
# but that gets a bit tricky, and requires custom compiling
1847
from bzrlib._walkdirs_win32 import Win32ReadDir
1848
_selected_dir_reader = Win32ReadDir()
1851
elif _fs_enc in ('utf-8', 'ascii'):
1853
from bzrlib._readdir_pyx import UTF8DirReader
1854
_selected_dir_reader = UTF8DirReader()
1855
except ImportError, e:
1856
failed_to_load_extension(e)
1859
if _selected_dir_reader is None:
1860
# Fallback to the python version
1861
_selected_dir_reader = UnicodeDirReader()
1192
fs_encoding = _fs_enc.upper()
1193
if (sys.platform == 'win32' or
1194
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1195
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1197
return _walkdirs_fs_utf8(top, prefix=prefix)
1200
def _walkdirs_fs_utf8(top, prefix=""):
1201
"""See _walkdirs_utf8.
1203
This sub-function is called when we know the filesystem is already in utf8
1204
encoding. So we don't need to transcode filenames.
1207
_directory = _directory_kind
1208
_listdir = os.listdir
1209
_kind_from_mode = _formats.get
1863
1211
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1864
1212
# But we don't actually uses 1-3 in pending, so set them to None
1865
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1866
read_dir = _selected_dir_reader.read_dir
1867
_directory = _directory_kind
1213
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1869
relroot, _, _, _, top = pending[-1].pop()
1872
dirblock = sorted(read_dir(relroot, top))
1215
relroot, _, _, _, top = pending.pop()
1217
relprefix = relroot + '/'
1220
top_slash = top + '/'
1223
append = dirblock.append
1224
for name in sorted(_listdir(top)):
1225
abspath = top_slash + name
1226
statvalue = _lstat(abspath)
1227
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1228
append((relprefix + name, name, kind, statvalue, abspath))
1873
1229
yield (relroot, top), dirblock
1874
1231
# push the user specified dirs from dirblock
1875
next = [d for d in reversed(dirblock) if d[2] == _directory]
1877
pending.append(next)
1880
class UnicodeDirReader(DirReader):
1881
"""A dir reader for non-utf8 file systems, which transcodes."""
1883
__slots__ = ['_utf8_encode']
1886
self._utf8_encode = codecs.getencoder('utf8')
1888
def top_prefix_to_starting_dir(self, top, prefix=""):
1889
"""See DirReader.top_prefix_to_starting_dir."""
1890
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1892
def read_dir(self, prefix, top):
1893
"""Read a single directory from a non-utf8 file system.
1895
top, and the abspath element in the output are unicode, all other paths
1896
are utf8. Local disk IO is done via unicode calls to listdir etc.
1898
This is currently the fallback code path when the filesystem encoding is
1899
not UTF-8. It may be better to implement an alternative so that we can
1900
safely handle paths that are not properly decodable in the current
1903
See DirReader.read_dir for details.
1905
_utf8_encode = self._utf8_encode
1907
_listdir = os.listdir
1908
_kind_from_mode = file_kind_from_stat_mode
1911
relprefix = prefix + '/'
1232
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1235
def _walkdirs_unicode_to_utf8(top, prefix=""):
1236
"""See _walkdirs_utf8
1238
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1240
This is currently the fallback code path when the filesystem encoding is
1241
not UTF-8. It may be better to implement an alternative so that we can
1242
safely handle paths that are not properly decodable in the current
1245
_utf8_encode = codecs.getencoder('utf8')
1247
_directory = _directory_kind
1248
_listdir = os.listdir
1249
_kind_from_mode = _formats.get
1251
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1253
relroot, _, _, _, top = pending.pop()
1255
relprefix = relroot + '/'
1914
1258
top_slash = top + u'/'
2013
1331
_cached_user_encoding = None
2016
def get_user_encoding(use_cache=DEPRECATED_PARAMETER):
1334
def get_user_encoding(use_cache=True):
2017
1335
"""Find out what the preferred user encoding is.
2019
1337
This is generally the encoding that is used for command line parameters
2020
1338
and file contents. This may be different from the terminal encoding
2021
1339
or the filesystem encoding.
1341
:param use_cache: Enable cache for detected encoding.
1342
(This parameter is turned on by default,
1343
and required only for selftesting)
2023
1345
:return: A string defining the preferred user encoding
2025
1347
global _cached_user_encoding
2026
if deprecated_passed(use_cache):
2027
warn_deprecated("use_cache should only have been used for tests",
2028
DeprecationWarning, stacklevel=2)
2029
if _cached_user_encoding is not None:
1348
if _cached_user_encoding is not None and use_cache:
2030
1349
return _cached_user_encoding
2032
if os.name == 'posix' and getattr(locale, 'CODESET', None) is not None:
2033
# Use the existing locale settings and call nl_langinfo directly
2034
# rather than going through getpreferredencoding. This avoids
2035
# <http://bugs.python.org/issue6202> on OSX Python 2.6 and the
2036
# possibility of the setlocale call throwing an error.
2037
user_encoding = locale.nl_langinfo(locale.CODESET)
1351
if sys.platform == 'darwin':
1352
# work around egregious python 2.4 bug
1353
sys.platform = 'posix'
1357
sys.platform = 'darwin'
2039
# GZ 2011-12-19: On windows could call GetACP directly instead.
2040
user_encoding = locale.getpreferredencoding(False)
2043
user_encoding = codecs.lookup(user_encoding).name
2045
if user_encoding not in ("", "cp0"):
1362
user_encoding = locale.getpreferredencoding()
1363
except locale.Error, e:
1364
sys.stderr.write('bzr: warning: %s\n'
1365
' Could not determine what text encoding to use.\n'
1366
' This error usually means your Python interpreter\n'
1367
' doesn\'t support the locale set by $LANG (%s)\n'
1368
" Continuing with ascii encoding.\n"
1369
% (e, os.environ.get('LANG')))
1370
user_encoding = 'ascii'
1372
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1373
# treat that as ASCII, and not support printing unicode characters to the
1376
# For python scripts run under vim, we get '', so also treat that as ASCII
1377
if user_encoding in (None, 'cp0', ''):
1378
user_encoding = 'ascii'
1382
codecs.lookup(user_encoding)
2046
1384
sys.stderr.write('bzr: warning:'
2047
1385
' unknown encoding %s.'
2048
1386
' Continuing with ascii encoding.\n'
2049
1387
% user_encoding
2051
user_encoding = 'ascii'
2053
# Get 'ascii' when setlocale has not been called or LANG=C or unset.
2054
if user_encoding == 'ascii':
2055
if sys.platform == 'darwin':
2056
# OSX is special-cased in Python to have a UTF-8 filesystem
2057
# encoding and previously had LANG set here if not present.
2058
user_encoding = 'utf-8'
2059
# GZ 2011-12-19: Maybe UTF-8 should be the default in this case
2060
# for some other posix platforms as well.
2062
_cached_user_encoding = user_encoding
1389
user_encoding = 'ascii'
1392
_cached_user_encoding = user_encoding
2063
1394
return user_encoding
2066
def get_diff_header_encoding():
2067
return get_terminal_encoding()
2070
def get_host_name():
2071
"""Return the current unicode host name.
2073
This is meant to be used in place of socket.gethostname() because that
2074
behaves inconsistently on different platforms.
2076
if sys.platform == "win32":
2077
return win32utils.get_host_name()
2080
return socket.gethostname().decode(get_user_encoding())
2083
# We must not read/write any more than 64k at a time from/to a socket so we
2084
# don't risk "no buffer space available" errors on some platforms. Windows in
2085
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2087
MAX_SOCKET_CHUNK = 64 * 1024
2089
_end_of_stream_errors = [errno.ECONNRESET, errno.EPIPE, errno.EINVAL]
2090
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2091
_eno = getattr(errno, _eno, None)
2092
if _eno is not None:
2093
_end_of_stream_errors.append(_eno)
2097
def read_bytes_from_socket(sock, report_activity=None,
2098
max_read_size=MAX_SOCKET_CHUNK):
2099
"""Read up to max_read_size of bytes from sock and notify of progress.
2101
Translates "Connection reset by peer" into file-like EOF (return an
2102
empty string rather than raise an error), and repeats the recv if
2103
interrupted by a signal.
2107
bytes = sock.recv(max_read_size)
2108
except socket.error, e:
2110
if eno in _end_of_stream_errors:
2111
# The connection was closed by the other side. Callers expect
2112
# an empty string to signal end-of-stream.
2114
elif eno == errno.EINTR:
2115
# Retry the interrupted recv.
2119
if report_activity is not None:
2120
report_activity(len(bytes), 'read')
2124
def recv_all(socket, count):
1397
def recv_all(socket, bytes):
2125
1398
"""Receive an exact number of bytes.
2127
1400
Regular Socket.recv() may return less than the requested number of bytes,
2128
depending on what's in the OS buffer. MSG_WAITALL is not available
1401
dependning on what's in the OS buffer. MSG_WAITALL is not available
2129
1402
on all platforms, but this should work everywhere. This will return
2130
1403
less than the requested amount if the remote end closes.
2132
1405
This isn't optimized and is intended mostly for use in testing.
2135
while len(b) < count:
2136
new = read_bytes_from_socket(socket, None, count - len(b))
1408
while len(b) < bytes:
1409
new = socket.recv(bytes - len(b))
2143
def send_all(sock, bytes, report_activity=None):
1416
def send_all(socket, bytes):
2144
1417
"""Send all bytes on a socket.
2146
Breaks large blocks in smaller chunks to avoid buffering limitations on
2147
some platforms, and catches EINTR which may be thrown if the send is
2148
interrupted by a signal.
2150
This is preferred to socket.sendall(), because it avoids portability bugs
2151
and provides activity reporting.
2153
:param report_activity: Call this as bytes are read, see
2154
Transport._report_activity
1419
Regular socket.sendall() can give socket error 10053 on Windows. This
1420
implementation sends no more than 64k at a time, which avoids this problem.
2157
byte_count = len(bytes)
2158
while sent_total < byte_count:
2160
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2161
except (socket.error, IOError), e:
2162
if e.args[0] in _end_of_stream_errors:
2163
raise errors.ConnectionReset(
2164
"Error trying to write to socket", e)
2165
if e.args[0] != errno.EINTR:
2169
raise errors.ConnectionReset('Sending to %s returned 0 bytes'
2172
if report_activity is not None:
2173
report_activity(sent, 'write')
2176
def connect_socket(address):
2177
# Slight variation of the socket.create_connection() function (provided by
2178
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2179
# provide it for previous python versions. Also, we don't use the timeout
2180
# parameter (provided by the python implementation) so we don't implement
2182
err = socket.error('getaddrinfo returns an empty list')
2183
host, port = address
2184
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2185
af, socktype, proto, canonname, sa = res
2188
sock = socket.socket(af, socktype, proto)
2192
except socket.error, err:
2193
# 'err' is now the most recent error
2194
if sock is not None:
1423
for pos in xrange(0, len(bytes), chunk_size):
1424
socket.sendall(bytes[pos:pos+chunk_size])
2199
1427
def dereference_path(path):
2240
1468
base = dirname(bzrlib.__file__)
2241
1469
if getattr(sys, 'frozen', None): # bzr.exe
2242
1470
base = abspath(pathjoin(base, '..', '..'))
2243
f = file(pathjoin(base, resource_relpath), "rU")
2249
def file_kind_from_stat_mode_thunk(mode):
2250
global file_kind_from_stat_mode
2251
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2253
from bzrlib._readdir_pyx import UTF8DirReader
2254
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2255
except ImportError, e:
2256
# This is one time where we won't warn that an extension failed to
2257
# load. The extension is never available on Windows anyway.
2258
from bzrlib._readdir_py import (
2259
_kind_from_mode as file_kind_from_stat_mode
2261
return file_kind_from_stat_mode(mode)
2262
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2264
def file_stat(f, _lstat=os.lstat):
2269
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2270
raise errors.NoSuchFile(f)
2273
def file_kind(f, _lstat=os.lstat):
2274
stat_value = file_stat(f, _lstat)
2275
return file_kind_from_stat_mode(stat_value.st_mode)
2277
def until_no_eintr(f, *a, **kw):
2278
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2280
WARNING: you must be certain that it is safe to retry the call repeatedly
2281
if EINTR does occur. This is typically only true for low-level operations
2282
like os.read. If in any doubt, don't use this.
2284
Keep in mind that this is not a complete solution to EINTR. There is
2285
probably code in the Python standard library and other dependencies that
2286
may encounter EINTR if a signal arrives (and there is signal handler for
2287
that signal). So this function can reduce the impact for IO that bzrlib
2288
directly controls, but it is not a complete solution.
2290
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2294
except (IOError, OSError), e:
2295
if e.errno == errno.EINTR:
2300
@deprecated_function(deprecated_in((2, 2, 0)))
2301
def re_compile_checked(re_string, flags=0, where=""):
2302
"""Return a compiled re, or raise a sensible error.
2304
This should only be used when compiling user-supplied REs.
2306
:param re_string: Text form of regular expression.
2307
:param flags: eg re.IGNORECASE
2308
:param where: Message explaining to the user the context where
2309
it occurred, eg 'log search filter'.
2311
# from https://bugs.launchpad.net/bzr/+bug/251352
2313
re_obj = re.compile(re_string, flags)
2316
except errors.InvalidPattern, e:
2318
where = ' in ' + where
2319
# despite the name 'error' is a type
2320
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2324
if sys.platform == "win32":
2327
return msvcrt.getch()
2332
fd = sys.stdin.fileno()
2333
settings = termios.tcgetattr(fd)
2336
ch = sys.stdin.read(1)
2338
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2341
if sys.platform.startswith('linux'):
2342
def _local_concurrency():
2344
return os.sysconf('SC_NPROCESSORS_ONLN')
2345
except (ValueError, OSError, AttributeError):
2347
elif sys.platform == 'darwin':
2348
def _local_concurrency():
2349
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2350
stdout=subprocess.PIPE).communicate()[0]
2351
elif "bsd" in sys.platform:
2352
def _local_concurrency():
2353
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2354
stdout=subprocess.PIPE).communicate()[0]
2355
elif sys.platform == 'sunos5':
2356
def _local_concurrency():
2357
return subprocess.Popen(['psrinfo', '-p',],
2358
stdout=subprocess.PIPE).communicate()[0]
2359
elif sys.platform == "win32":
2360
def _local_concurrency():
2361
# This appears to return the number of cores.
2362
return os.environ.get('NUMBER_OF_PROCESSORS')
2364
def _local_concurrency():
2369
_cached_local_concurrency = None
2371
def local_concurrency(use_cache=True):
2372
"""Return how many processes can be run concurrently.
2374
Rely on platform specific implementations and default to 1 (one) if
2375
anything goes wrong.
2377
global _cached_local_concurrency
2379
if _cached_local_concurrency is not None and use_cache:
2380
return _cached_local_concurrency
2382
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2383
if concurrency is None:
2385
import multiprocessing
2386
concurrency = multiprocessing.cpu_count()
2387
except (ImportError, NotImplementedError):
2388
# multiprocessing is only available on Python >= 2.6
2389
# and multiprocessing.cpu_count() isn't implemented on all
2392
concurrency = _local_concurrency()
2393
except (OSError, IOError):
2396
concurrency = int(concurrency)
2397
except (TypeError, ValueError):
2400
_cached_concurrency = concurrency
2404
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2405
"""A stream writer that doesn't decode str arguments."""
2407
def __init__(self, encode, stream, errors='strict'):
2408
codecs.StreamWriter.__init__(self, stream, errors)
2409
self.encode = encode
2411
def write(self, object):
2412
if type(object) is str:
2413
self.stream.write(object)
2415
data, _ = self.encode(object, self.errors)
2416
self.stream.write(data)
2418
if sys.platform == 'win32':
2419
def open_file(filename, mode='r', bufsize=-1):
2420
"""This function is used to override the ``open`` builtin.
2422
But it uses O_NOINHERIT flag so the file handle is not inherited by
2423
child processes. Deleting or renaming a closed file opened with this
2424
function is not blocking child processes.
2426
writing = 'w' in mode
2427
appending = 'a' in mode
2428
updating = '+' in mode
2429
binary = 'b' in mode
2432
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2433
# for flags for each modes.
2443
flags |= os.O_WRONLY
2444
flags |= os.O_CREAT | os.O_TRUNC
2449
flags |= os.O_WRONLY
2450
flags |= os.O_CREAT | os.O_APPEND
2455
flags |= os.O_RDONLY
2457
return os.fdopen(os.open(filename, flags), mode, bufsize)
2462
def available_backup_name(base, exists):
2463
"""Find a non-existing backup file name.
2465
This will *not* create anything, this only return a 'free' entry. This
2466
should be used for checking names in a directory below a locked
2467
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2468
Leap) and generally discouraged.
2470
:param base: The base name.
2472
:param exists: A callable returning True if the path parameter exists.
2475
name = "%s.~%d~" % (base, counter)
2478
name = "%s.~%d~" % (base, counter)
2482
def set_fd_cloexec(fd):
2483
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2484
support for this is not available.
2488
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2489
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2490
except (ImportError, AttributeError):
2491
# Either the fcntl module or specific constants are not present
2495
def find_executable_on_path(name):
2496
"""Finds an executable on the PATH.
2498
On Windows, this will try to append each extension in the PATHEXT
2499
environment variable to the name, if it cannot be found with the name
2502
:param name: The base name of the executable.
2503
:return: The path to the executable found or None.
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:
2516
path = os.environ.get('PATH')
2517
if path is not None:
2518
path = path.split(os.pathsep)
2521
f = os.path.join(d, name) + ext
2522
if os.access(f, os.X_OK):
2524
if sys.platform == 'win32':
2525
app_path = win32utils.get_app_path(name)
2526
if app_path != name:
2531
def _posix_is_local_pid_dead(pid):
2532
"""True if pid doesn't correspond to live process on this machine"""
2534
# Special meaning of unix kill: just check if it's there.
2537
if e.errno == errno.ESRCH:
2538
# On this machine, and really not found: as sure as we can be
2541
elif e.errno == errno.EPERM:
2542
# exists, though not ours
2545
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2546
# Don't really know.
2549
# Exists and our process: not dead.
2552
if sys.platform == "win32":
2553
is_local_pid_dead = win32utils.is_local_pid_dead
2555
is_local_pid_dead = _posix_is_local_pid_dead
2557
_maybe_ignored = ['EAGAIN', 'EINTR', 'ENOTSUP', 'EOPNOTSUPP', 'EACCES']
2558
_fdatasync_ignored = [getattr(errno, name) for name in _maybe_ignored
2559
if getattr(errno, name, None) is not None]
2562
def fdatasync(fileno):
2563
"""Flush file contents to disk if possible.
2565
:param fileno: Integer OS file handle.
2566
:raises TransportNotPossible: If flushing to disk is not possible.
2568
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
2573
# See bug #1075108, on some platforms fdatasync exists, but can
2574
# raise ENOTSUP. However, we are calling fdatasync to be helpful
2575
# and reduce the chance of corruption-on-powerloss situations. It
2576
# is not a mandatory call, so it is ok to suppress failures.
2577
trace.mutter("ignoring error calling fdatasync: %s" % (e,))
2578
if getattr(e, 'errno', None) not in _fdatasync_ignored:
2582
def ensure_empty_directory_exists(path, exception_class):
2583
"""Make sure a local directory exists and is empty.
2585
If it does not exist, it is created. If it exists and is not empty, an
2586
instance of exception_class is raised.
2591
if e.errno != errno.EEXIST:
2593
if os.listdir(path) != []:
2594
raise exception_class(path)
2597
def is_environment_error(evalue):
2598
"""True if exception instance is due to a process environment issue
2600
This includes OSError and IOError, but also other errors that come from
2601
the operating system or core libraries but are not subclasses of those.
2603
if isinstance(evalue, (EnvironmentError, select.error)):
2605
if sys.platform == "win32" and win32utils._is_pywintypes_error(evalue):
1471
filename = pathjoin(base, resource_relpath)
1472
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
