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 (
57
from bzrlib import symbol_versioning
51
58
from bzrlib.symbol_versioning import (
52
59
deprecated_function,
63
from bzrlib import symbol_versioning
66
# Cross platform wall-clock time functionality with decent resolution.
67
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
68
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
69
# synchronized with ``time.time()``, this is only meant to be used to find
70
# delta times by subtracting from another call to this function.
71
timer_func = time.time
72
if sys.platform == 'win32':
73
timer_func = time.clock
62
from bzrlib.trace import mutter
75
65
# On win32, O_BINARY is used to indicate the file should
76
66
# be opened in binary mode, rather than text mode.
77
67
# On other platforms, O_BINARY doesn't exist, because
78
68
# they always open in binary mode, so it is okay to
79
# OR with 0 on those platforms.
80
# O_NOINHERIT and O_TEXT exists only on win32 too.
69
# OR with 0 on those platforms
81
70
O_BINARY = getattr(os, 'O_BINARY', 0)
82
O_TEXT = getattr(os, 'O_TEXT', 0)
83
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
86
def get_unicode_argv():
88
user_encoding = get_user_encoding()
89
return [a.decode(user_encoding) for a in sys.argv[1:]]
90
except UnicodeDecodeError:
91
raise errors.BzrError("Parameter %r encoding is unsupported by %s "
92
"application locale." % (a, user_encoding))
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)
95
76
def make_readonly(filename):
96
77
"""Make a filename read-only."""
97
mod = os.lstat(filename).st_mode
78
mod = lstat(filename).st_mode
98
79
if not stat.S_ISLNK(mod):
99
80
mod = mod & 0777555
100
81
os.chmod(filename, mod)
103
84
def make_writable(filename):
104
mod = os.lstat(filename).st_mode
85
mod = lstat(filename).st_mode
105
86
if not stat.S_ISLNK(mod):
107
88
os.chmod(filename, mod)
110
def minimum_path_selection(paths):
111
"""Return the smallset subset of paths which are outside paths.
113
:param paths: A container (and hence not None) of paths.
114
:return: A set of paths sufficient to include everything in paths via
115
is_inside, drawn from the paths parameter.
121
return path.split('/')
122
sorted_paths = sorted(list(paths), key=sort_key)
124
search_paths = [sorted_paths[0]]
125
for path in sorted_paths[1:]:
126
if not is_inside(search_paths[-1], path):
127
# This path is unique, add it
128
search_paths.append(path)
130
return set(search_paths)
435
388
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
436
389
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
437
390
return shutil.rmtree(path, ignore_errors, onerror)
439
f = win32utils.get_unicode_argv # special function or None
443
391
elif sys.platform == 'darwin':
444
392
getcwd = _mac_getcwd
447
def get_terminal_encoding(trace=False):
395
def get_terminal_encoding():
448
396
"""Find the best encoding for printing to the screen.
450
398
This attempts to check both sys.stdout and sys.stdin to see
451
399
what encoding they are in, and if that fails it falls back to
452
osutils.get_user_encoding().
400
bzrlib.user_encoding.
453
401
The problem is that on Windows, locale.getpreferredencoding()
454
402
is not the same encoding as that used by the console:
455
403
http://mail.python.org/pipermail/python-list/2003-May/162357.html
457
405
On my standard US Windows XP, the preferred encoding is
458
406
cp1252, but the console is cp437
460
:param trace: If True trace the selected encoding via mutter().
462
from bzrlib.trace import mutter
463
408
output_encoding = getattr(sys.stdout, 'encoding', None)
464
409
if not output_encoding:
465
410
input_encoding = getattr(sys.stdin, 'encoding', None)
466
411
if not input_encoding:
467
output_encoding = get_user_encoding()
469
mutter('encoding stdout as osutils.get_user_encoding() %r',
412
output_encoding = bzrlib.user_encoding
413
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
472
415
output_encoding = input_encoding
474
mutter('encoding stdout as sys.stdin encoding %r',
416
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
478
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
418
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
479
419
if output_encoding == 'cp0':
480
420
# invalid encoding (cp0 means 'no codepage' on Windows)
481
output_encoding = get_user_encoding()
483
mutter('cp0 is invalid encoding.'
484
' encoding stdout as osutils.get_user_encoding() %r',
421
output_encoding = bzrlib.user_encoding
422
mutter('cp0 is invalid encoding.'
423
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
488
426
codecs.lookup(output_encoding)
574
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
575
report_activity=None, direction='read'):
576
"""Copy contents of one file to another.
578
The read_length can either be -1 to read to end-of-file (EOF) or
579
it can specify the maximum number of bytes to read.
581
The buff_size represents the maximum size for each read operation
582
performed on from_file.
584
:param report_activity: Call this as bytes are read, see
585
Transport._report_activity
586
:param direction: Will be passed to report_activity
588
:return: The number of bytes copied.
592
# read specified number of bytes
594
while read_length > 0:
595
num_bytes_to_read = min(read_length, buff_size)
597
block = from_file.read(num_bytes_to_read)
601
if report_activity is not None:
602
report_activity(len(block), direction)
605
actual_bytes_read = len(block)
606
read_length -= actual_bytes_read
607
length += actual_bytes_read
611
block = from_file.read(buff_size)
615
if report_activity is not None:
616
report_activity(len(block), direction)
622
def pump_string_file(bytes, file_handle, segment_size=None):
623
"""Write bytes to file_handle in many smaller writes.
625
:param bytes: The string to write.
626
:param file_handle: The file to write to.
628
# Write data in chunks rather than all at once, because very large
629
# writes fail on some platforms (e.g. Windows with SMB mounted
632
segment_size = 5242880 # 5MB
633
segments = range(len(bytes) / segment_size + 1)
634
write = file_handle.write
635
for segment_index in segments:
636
segment = buffer(bytes, segment_index * segment_size, segment_size)
540
def pumpfile(fromfile, tofile):
541
"""Copy contents of one file to another."""
544
b = fromfile.read(BUFSIZE)
640
550
def file_iterator(input_file, readsize=32768):
741
620
:param timezone: How to display the time: 'utc', 'original' for the
742
621
timezone specified by offset, or 'local' for the process's current
744
:param date_fmt: strftime format.
745
:param show_offset: Whether to append the timezone.
747
(date_fmt, tt, offset_str) = \
748
_format_date(t, offset, timezone, date_fmt, show_offset)
749
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
750
date_str = time.strftime(date_fmt, tt)
751
return date_str + offset_str
754
# Cache of formatted offset strings
758
def format_date_with_offset_in_original_timezone(t, offset=0,
759
_cache=_offset_cache):
760
"""Return a formatted date string in the original timezone.
762
This routine may be faster then format_date.
764
:param t: Seconds since the epoch.
765
:param offset: Timezone offset in seconds east of utc.
769
tt = time.gmtime(t + offset)
770
date_fmt = _default_format_by_weekday_num[tt[6]]
771
date_str = time.strftime(date_fmt, tt)
772
offset_str = _cache.get(offset, None)
773
if offset_str is None:
774
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
775
_cache[offset] = offset_str
776
return date_str + offset_str
779
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
781
"""Return an unicode date string formatted according to the current locale.
783
:param t: Seconds since the epoch.
784
:param offset: Timezone offset in seconds east of utc.
785
:param timezone: How to display the time: 'utc', 'original' for the
786
timezone specified by offset, or 'local' for the process's current
788
:param date_fmt: strftime format.
789
:param show_offset: Whether to append the timezone.
791
(date_fmt, tt, offset_str) = \
792
_format_date(t, offset, timezone, date_fmt, show_offset)
793
date_str = time.strftime(date_fmt, tt)
794
if not isinstance(date_str, unicode):
795
date_str = date_str.decode(get_user_encoding(), 'replace')
796
return date_str + offset_str
799
def _format_date(t, offset, timezone, date_fmt, show_offset):
623
:param show_offset: Whether to append the timezone.
624
:param date_fmt: strftime format.
800
626
if timezone == 'utc':
801
627
tt = time.gmtime(t)
764
assert isinstance(p, (list, tuple))
937
766
if (f == '..') or (f is None) or (f == ''):
938
767
raise errors.BzrError("sorry, %r not allowed in path" % f)
939
768
return pathjoin(*p)
942
def parent_directories(filename):
943
"""Return the list of parent directories, deepest first.
945
For example, parent_directories("a/b/c") -> ["a/b", "a"].
948
parts = splitpath(dirname(filename))
950
parents.append(joinpath(parts))
955
_extension_load_failures = []
958
def failed_to_load_extension(exception):
959
"""Handle failing to load a binary extension.
961
This should be called from the ImportError block guarding the attempt to
962
import the native extension. If this function returns, the pure-Python
963
implementation should be loaded instead::
966
>>> import bzrlib._fictional_extension_pyx
967
>>> except ImportError, e:
968
>>> bzrlib.osutils.failed_to_load_extension(e)
969
>>> import bzrlib._fictional_extension_py
971
# NB: This docstring is just an example, not a doctest, because doctest
972
# currently can't cope with the use of lazy imports in this namespace --
975
# This currently doesn't report the failure at the time it occurs, because
976
# they tend to happen very early in startup when we can't check config
977
# files etc, and also we want to report all failures but not spam the user
979
exception_str = str(exception)
980
if exception_str not in _extension_load_failures:
981
trace.mutter("failed to load compiled extension: %s" % exception_str)
982
_extension_load_failures.append(exception_str)
985
def report_extension_load_failures():
986
if not _extension_load_failures:
988
from bzrlib.config import GlobalConfig
989
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
991
# the warnings framework should by default show this only once
992
from bzrlib.trace import warning
994
"bzr: warning: some compiled extensions could not be loaded; "
995
"see <https://answers.launchpad.net/bzr/+faq/703>")
996
# we no longer show the specific missing extensions here, because it makes
997
# the message too long and scary - see
998
# https://bugs.launchpad.net/bzr/+bug/430529
1002
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1003
except ImportError, e:
1004
failed_to_load_extension(e)
1005
from bzrlib._chunks_to_lines_py import chunks_to_lines
771
@deprecated_function(zero_nine)
772
def appendpath(p1, p2):
776
return pathjoin(p1, p2)
1008
779
def split_lines(s):
1009
780
"""Split s into lines, but without removing the newline characters."""
1010
# Trivially convert a fulltext into a 'chunked' representation, and let
1011
# chunks_to_lines do the heavy lifting.
1012
if isinstance(s, str):
1013
# chunks_to_lines only supports 8-bit strings
1014
return chunks_to_lines([s])
1016
return _split_lines(s)
1019
def _split_lines(s):
1020
"""Split s into lines, but without removing the newline characters.
1022
This supports Unicode or plain string objects.
1024
781
lines = s.split('\n')
1025
782
result = [line + '\n' for line in lines[:-1]]
1147
857
os.path.commonprefix (python2.4) has a bad bug that it works just
1148
858
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1149
859
avoids that problem.
1151
NOTE: `base` should not have a trailing slash otherwise you'll get
1152
PathNotChild exceptions regardless of `path`.
1155
if len(base) < MIN_ABS_PATHLENGTH:
1156
# must have space for e.g. a drive letter
1157
raise ValueError('%r is too short to calculate a relative path'
862
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
863
' exceed the platform minimum length (which is %d)' %
1160
866
rp = abspath(path)
1165
if len(head) <= len(base) and head != base:
1166
raise errors.PathNotChild(rp, base)
870
while len(head) >= len(base):
1167
871
if head == base:
1169
head, tail = split(head)
873
head, tail = os.path.split(head)
877
raise errors.PathNotChild(rp, base)
1174
return pathjoin(*reversed(s))
1179
def _cicp_canonical_relpath(base, path):
1180
"""Return the canonical path relative to base.
1182
Like relpath, but on case-insensitive-case-preserving file-systems, this
1183
will return the relpath as stored on the file-system rather than in the
1184
case specified in the input string, for all existing portions of the path.
1186
This will cause O(N) behaviour if called for every path in a tree; if you
1187
have a number of paths to convert, you should use canonical_relpaths().
1189
# TODO: it should be possible to optimize this for Windows by using the
1190
# win32 API FindFiles function to look for the specified name - but using
1191
# os.listdir() still gives us the correct, platform agnostic semantics in
1194
rel = relpath(base, path)
1195
# '.' will have been turned into ''
1199
abs_base = abspath(base)
1201
_listdir = os.listdir
1203
# use an explicit iterator so we can easily consume the rest on early exit.
1204
bit_iter = iter(rel.split('/'))
1205
for bit in bit_iter:
1208
next_entries = _listdir(current)
1209
except OSError: # enoent, eperm, etc
1210
# We can't find this in the filesystem, so just append the
1212
current = pathjoin(current, bit, *list(bit_iter))
1214
for look in next_entries:
1215
if lbit == look.lower():
1216
current = pathjoin(current, look)
1219
# got to the end, nothing matched, so we just return the
1220
# non-existing bits as they were specified (the filename may be
1221
# the target of a move, for example).
1222
current = pathjoin(current, bit, *list(bit_iter))
1224
return current[len(abs_base):].lstrip('/')
1226
# XXX - TODO - we need better detection/integration of case-insensitive
1227
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1228
# filesystems), for example, so could probably benefit from the same basic
1229
# support there. For now though, only Windows and OSX get that support, and
1230
# they get it for *all* file-systems!
1231
if sys.platform in ('win32', 'darwin'):
1232
canonical_relpath = _cicp_canonical_relpath
1234
canonical_relpath = relpath
1236
def canonical_relpaths(base, paths):
1237
"""Create an iterable to canonicalize a sequence of relative paths.
1239
The intent is for this implementation to use a cache, vastly speeding
1240
up multiple transformations in the same directory.
1242
# but for now, we haven't optimized...
1243
return [canonical_relpath(base, p) for p in paths]
1246
def decode_filename(filename):
1247
"""Decode the filename using the filesystem encoding
1249
If it is unicode, it is returned.
1250
Otherwise it is decoded from the the filesystem's encoding. If decoding
1251
fails, a errors.BadFilenameEncoding exception is raised.
1253
if type(filename) is unicode:
1256
return filename.decode(_fs_enc)
1257
except UnicodeDecodeError:
1258
raise errors.BadFilenameEncoding(filename, _fs_enc)
1261
885
def safe_unicode(unicode_or_utf8_string):
1262
886
"""Coerce unicode_or_utf8_string into unicode.
1264
888
If it is unicode, it is returned.
1265
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1266
wrapped in a BzrBadParameterNotUnicode exception.
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.
1268
893
if isinstance(unicode_or_utf8_string, unicode):
1269
894
return unicode_or_utf8_string
1382
1007
normalized_filename = _inaccessible_normalized_filename
1385
def set_signal_handler(signum, handler, restart_syscall=True):
1386
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1387
on platforms that support that.
1389
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1390
automatically restart (by calling `signal.siginterrupt(signum,
1391
False)`). May be ignored if the feature is not available on this
1392
platform or Python version.
1396
siginterrupt = signal.siginterrupt
1398
# This python implementation doesn't provide signal support, hence no
1401
except AttributeError:
1402
# siginterrupt doesn't exist on this platform, or for this version
1404
siginterrupt = lambda signum, flag: None
1406
def sig_handler(*args):
1407
# Python resets the siginterrupt flag when a signal is
1408
# received. <http://bugs.python.org/issue8354>
1409
# As a workaround for some cases, set it back the way we want it.
1410
siginterrupt(signum, False)
1411
# Now run the handler function passed to set_signal_handler.
1414
sig_handler = handler
1415
old_handler = signal.signal(signum, sig_handler)
1417
siginterrupt(signum, False)
1421
default_terminal_width = 80
1422
"""The default terminal width for ttys.
1424
This is defined so that higher levels can share a common fallback value when
1425
terminal_width() returns None.
1428
# Keep some state so that terminal_width can detect if _terminal_size has
1429
# returned a different size since the process started. See docstring and
1430
# comments of terminal_width for details.
1431
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1432
_terminal_size_state = 'no_data'
1433
_first_terminal_size = None
1435
1010
def terminal_width():
1436
"""Return terminal width.
1438
None is returned if the width can't established precisely.
1441
- if BZR_COLUMNS is set, returns its value
1442
- if there is no controlling terminal, returns None
1443
- query the OS, if the queried size has changed since the last query,
1445
- if COLUMNS is set, returns its value,
1446
- if the OS has a value (even though it's never changed), return its value.
1448
From there, we need to query the OS to get the size of the controlling
1451
On Unices we query the OS by:
1452
- get termios.TIOCGWINSZ
1453
- if an error occurs or a negative value is obtained, returns None
1455
On Windows we query the OS by:
1456
- win32utils.get_console_size() decides,
1457
- returns None on error (provided default value)
1459
# Note to implementors: if changing the rules for determining the width,
1460
# make sure you've considered the behaviour in these cases:
1461
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1462
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1464
# - (add more interesting cases here, if you find any)
1465
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1466
# but we don't want to register a signal handler because it is impossible
1467
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1468
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1469
# time so we can notice if the reported size has changed, which should have
1472
# If BZR_COLUMNS is set, take it, user is always right
1473
# Except if they specified 0 in which case, impose no limit here
1475
width = int(os.environ['BZR_COLUMNS'])
1476
except (KeyError, ValueError):
1478
if width is not None:
1484
isatty = getattr(sys.stdout, 'isatty', None)
1485
if isatty is None or not isatty():
1486
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1490
width, height = os_size = _terminal_size(None, None)
1491
global _first_terminal_size, _terminal_size_state
1492
if _terminal_size_state == 'no_data':
1493
_first_terminal_size = os_size
1494
_terminal_size_state = 'unchanged'
1495
elif (_terminal_size_state == 'unchanged' and
1496
_first_terminal_size != os_size):
1497
_terminal_size_state = 'changed'
1499
# If the OS claims to know how wide the terminal is, and this value has
1500
# ever changed, use that.
1501
if _terminal_size_state == 'changed':
1502
if width is not None and width > 0:
1505
# If COLUMNS is set, use it.
1507
return int(os.environ['COLUMNS'])
1508
except (KeyError, ValueError):
1511
# Finally, use an unchanged size from the OS, if we have one.
1512
if _terminal_size_state == 'unchanged':
1513
if width is not None and width > 0:
1516
# The width could not be determined.
1520
def _win32_terminal_size(width, height):
1521
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1522
return width, height
1525
def _ioctl_terminal_size(width, height):
1011
"""Return estimated terminal width."""
1012
if sys.platform == 'win32':
1013
return win32utils.get_console_size()[0]
1527
1016
import struct, fcntl, termios
1528
1017
s = struct.pack('HHHH', 0, 0, 0, 0)
1529
1018
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1530
height, width = struct.unpack('HHHH', x)[0:2]
1531
except (IOError, AttributeError):
1019
width = struct.unpack('HHHH', x)[1]
1533
return width, height
1535
_terminal_size = None
1536
"""Returns the terminal size as (width, height).
1538
:param width: Default value for width.
1539
:param height: Default value for height.
1541
This is defined specifically for each OS and query the size of the controlling
1542
terminal. If any error occurs, the provided default values should be returned.
1544
if sys.platform == 'win32':
1545
_terminal_size = _win32_terminal_size
1547
_terminal_size = _ioctl_terminal_size
1024
width = int(os.environ['COLUMNS'])
1550
1033
def supports_executable():
1597
1080
raise errors.IllegalPath(path)
1600
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1602
def _is_error_enotdir(e):
1603
"""Check if this exception represents ENOTDIR.
1605
Unfortunately, python is very inconsistent about the exception
1606
here. The cases are:
1607
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1608
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1609
which is the windows error code.
1610
3) Windows, Python2.5 uses errno == EINVAL and
1611
winerror == ERROR_DIRECTORY
1613
:param e: An Exception object (expected to be OSError with an errno
1614
attribute, but we should be able to cope with anything)
1615
:return: True if this represents an ENOTDIR error. False otherwise.
1617
en = getattr(e, 'errno', None)
1618
if (en == errno.ENOTDIR
1619
or (sys.platform == 'win32'
1620
and (en == _WIN32_ERROR_DIRECTORY
1621
or (en == errno.EINVAL
1622
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1628
1083
def walkdirs(top, prefix=""):
1629
1084
"""Yield data about all the directories in a tree.
1631
1086
This yields all the data about the contents of a directory at a time.
1632
1087
After each directory has been yielded, if the caller has mutated the list
1633
1088
to exclude some directories, they are then not descended into.
1635
1090
The data yielded is of the form:
1636
1091
((directory-relpath, directory-path-from-top),
1637
[(relpath, basename, kind, lstat, path-from-top), ...]),
1092
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1638
1093
- directory-relpath is the relative path of the directory being returned
1639
1094
with respect to top. prefix is prepended to this.
1640
- directory-path-from-root is the path including top for this directory.
1095
- directory-path-from-root is the path including top for this directory.
1641
1096
It is suitable for use with os functions.
1642
1097
- relpath is the relative path within the subtree being walked.
1643
1098
- basename is the basename of the path
1733
1154
path-from-top might be unicode or utf8, but it is the correct path to
1734
1155
pass to os functions to affect the file in question. (such as os.lstat)
1736
global _selected_dir_reader
1737
if _selected_dir_reader is None:
1738
fs_encoding = _fs_enc.upper()
1739
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1740
# Win98 doesn't have unicode apis like FindFirstFileW
1741
# TODO: We possibly could support Win98 by falling back to the
1742
# original FindFirstFile, and using TCHAR instead of WCHAR,
1743
# but that gets a bit tricky, and requires custom compiling
1746
from bzrlib._walkdirs_win32 import Win32ReadDir
1747
_selected_dir_reader = Win32ReadDir()
1750
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1751
# ANSI_X3.4-1968 is a form of ASCII
1753
from bzrlib._readdir_pyx import UTF8DirReader
1754
_selected_dir_reader = UTF8DirReader()
1755
except ImportError, e:
1756
failed_to_load_extension(e)
1759
if _selected_dir_reader is None:
1760
# Fallback to the python version
1761
_selected_dir_reader = UnicodeDirReader()
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
1763
1176
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1764
1177
# But we don't actually uses 1-3 in pending, so set them to None
1765
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1766
read_dir = _selected_dir_reader.read_dir
1767
_directory = _directory_kind
1178
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1769
relroot, _, _, _, top = pending[-1].pop()
1772
dirblock = sorted(read_dir(relroot, top))
1180
relroot, _, _, _, top = pending.pop()
1182
relprefix = relroot + '/'
1185
top_slash = top + '/'
1188
append = dirblock.append
1189
for name in sorted(_listdir(top)):
1190
abspath = top_slash + name
1191
statvalue = _lstat(abspath)
1192
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1193
append((relprefix + name, name, kind, statvalue, abspath))
1773
1194
yield (relroot, top), dirblock
1774
1196
# push the user specified dirs from dirblock
1775
next = [d for d in reversed(dirblock) if d[2] == _directory]
1777
pending.append(next)
1780
class UnicodeDirReader(DirReader):
1781
"""A dir reader for non-utf8 file systems, which transcodes."""
1783
__slots__ = ['_utf8_encode']
1786
self._utf8_encode = codecs.getencoder('utf8')
1788
def top_prefix_to_starting_dir(self, top, prefix=""):
1789
"""See DirReader.top_prefix_to_starting_dir."""
1790
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1792
def read_dir(self, prefix, top):
1793
"""Read a single directory from a non-utf8 file system.
1795
top, and the abspath element in the output are unicode, all other paths
1796
are utf8. Local disk IO is done via unicode calls to listdir etc.
1798
This is currently the fallback code path when the filesystem encoding is
1799
not UTF-8. It may be better to implement an alternative so that we can
1800
safely handle paths that are not properly decodable in the current
1803
See DirReader.read_dir for details.
1805
_utf8_encode = self._utf8_encode
1807
_listdir = os.listdir
1808
_kind_from_mode = file_kind_from_stat_mode
1811
relprefix = prefix + '/'
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 + '/'
1814
1223
top_slash = top + u'/'
1986
1357
return user_encoding
1989
def get_diff_header_encoding():
1990
return get_terminal_encoding()
1993
def get_host_name():
1994
"""Return the current unicode host name.
1996
This is meant to be used in place of socket.gethostname() because that
1997
behaves inconsistently on different platforms.
1999
if sys.platform == "win32":
2001
return win32utils.get_host_name()
2004
return socket.gethostname().decode(get_user_encoding())
2007
# We must not read/write any more than 64k at a time from/to a socket so we
2008
# don't risk "no buffer space available" errors on some platforms. Windows in
2009
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2011
MAX_SOCKET_CHUNK = 64 * 1024
2013
_end_of_stream_errors = [errno.ECONNRESET]
2014
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2015
_eno = getattr(errno, _eno, None)
2016
if _eno is not None:
2017
_end_of_stream_errors.append(_eno)
2021
def read_bytes_from_socket(sock, report_activity=None,
2022
max_read_size=MAX_SOCKET_CHUNK):
2023
"""Read up to max_read_size of bytes from sock and notify of progress.
2025
Translates "Connection reset by peer" into file-like EOF (return an
2026
empty string rather than raise an error), and repeats the recv if
2027
interrupted by a signal.
2031
bytes = sock.recv(max_read_size)
2032
except socket.error, e:
2034
if eno in _end_of_stream_errors:
2035
# The connection was closed by the other side. Callers expect
2036
# an empty string to signal end-of-stream.
2038
elif eno == errno.EINTR:
2039
# Retry the interrupted recv.
2043
if report_activity is not None:
2044
report_activity(len(bytes), 'read')
2048
def recv_all(socket, count):
1360
def recv_all(socket, bytes):
2049
1361
"""Receive an exact number of bytes.
2051
1363
Regular Socket.recv() may return less than the requested number of bytes,
2052
depending on what's in the OS buffer. MSG_WAITALL is not available
1364
dependning on what's in the OS buffer. MSG_WAITALL is not available
2053
1365
on all platforms, but this should work everywhere. This will return
2054
1366
less than the requested amount if the remote end closes.
2056
1368
This isn't optimized and is intended mostly for use in testing.
2059
while len(b) < count:
2060
new = read_bytes_from_socket(socket, None, count - len(b))
1371
while len(b) < bytes:
1372
new = socket.recv(bytes - len(b))
2067
def send_all(sock, bytes, report_activity=None):
2068
"""Send all bytes on a socket.
2070
Breaks large blocks in smaller chunks to avoid buffering limitations on
2071
some platforms, and catches EINTR which may be thrown if the send is
2072
interrupted by a signal.
2074
This is preferred to socket.sendall(), because it avoids portability bugs
2075
and provides activity reporting.
2077
:param report_activity: Call this as bytes are read, see
2078
Transport._report_activity
2081
byte_count = len(bytes)
2082
while sent_total < byte_count:
2084
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2085
except socket.error, e:
2086
if e.args[0] != errno.EINTR:
2090
report_activity(sent, 'write')
2093
def connect_socket(address):
2094
# Slight variation of the socket.create_connection() function (provided by
2095
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2096
# provide it for previous python versions. Also, we don't use the timeout
2097
# parameter (provided by the python implementation) so we don't implement
2099
err = socket.error('getaddrinfo returns an empty list')
2100
host, port = address
2101
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2102
af, socktype, proto, canonname, sa = res
2105
sock = socket.socket(af, socktype, proto)
2109
except socket.error, err:
2110
# 'err' is now the most recent error
2111
if sock is not None:
2116
1378
def dereference_path(path):
2117
1379
"""Determine the real path to a file.
2125
1387
# The pathjoin for '.' is a workaround for Python bug #1213894.
2126
1388
# (initial path components aren't dereferenced)
2127
1389
return pathjoin(realpath(pathjoin('.', parent)), base)
2130
def supports_mapi():
2131
"""Return True if we can use MAPI to launch a mail client."""
2132
return sys.platform == "win32"
2135
def resource_string(package, resource_name):
2136
"""Load a resource from a package and return it as a string.
2138
Note: Only packages that start with bzrlib are currently supported.
2140
This is designed to be a lightweight implementation of resource
2141
loading in a way which is API compatible with the same API from
2143
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2144
If and when pkg_resources becomes a standard library, this routine
2147
# Check package name is within bzrlib
2148
if package == "bzrlib":
2149
resource_relpath = resource_name
2150
elif package.startswith("bzrlib."):
2151
package = package[len("bzrlib."):].replace('.', os.sep)
2152
resource_relpath = pathjoin(package, resource_name)
2154
raise errors.BzrError('resource package %s not in bzrlib' % package)
2156
# Map the resource to a file and read its contents
2157
base = dirname(bzrlib.__file__)
2158
if getattr(sys, 'frozen', None): # bzr.exe
2159
base = abspath(pathjoin(base, '..', '..'))
2160
f = file(pathjoin(base, resource_relpath), "rU")
2166
def file_kind_from_stat_mode_thunk(mode):
2167
global file_kind_from_stat_mode
2168
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2170
from bzrlib._readdir_pyx import UTF8DirReader
2171
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2172
except ImportError, e:
2173
# This is one time where we won't warn that an extension failed to
2174
# load. The extension is never available on Windows anyway.
2175
from bzrlib._readdir_py import (
2176
_kind_from_mode as file_kind_from_stat_mode
2178
return file_kind_from_stat_mode(mode)
2179
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2182
def file_kind(f, _lstat=os.lstat):
2184
return file_kind_from_stat_mode(_lstat(f).st_mode)
2186
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2187
raise errors.NoSuchFile(f)
2191
def until_no_eintr(f, *a, **kw):
2192
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2194
WARNING: you must be certain that it is safe to retry the call repeatedly
2195
if EINTR does occur. This is typically only true for low-level operations
2196
like os.read. If in any doubt, don't use this.
2198
Keep in mind that this is not a complete solution to EINTR. There is
2199
probably code in the Python standard library and other dependencies that
2200
may encounter EINTR if a signal arrives (and there is signal handler for
2201
that signal). So this function can reduce the impact for IO that bzrlib
2202
directly controls, but it is not a complete solution.
2204
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2208
except (IOError, OSError), e:
2209
if e.errno == errno.EINTR:
2214
@deprecated_function(deprecated_in((2, 2, 0)))
2215
def re_compile_checked(re_string, flags=0, where=""):
2216
"""Return a compiled re, or raise a sensible error.
2218
This should only be used when compiling user-supplied REs.
2220
:param re_string: Text form of regular expression.
2221
:param flags: eg re.IGNORECASE
2222
:param where: Message explaining to the user the context where
2223
it occurred, eg 'log search filter'.
2225
# from https://bugs.launchpad.net/bzr/+bug/251352
2227
re_obj = re.compile(re_string, flags)
2230
except errors.InvalidPattern, e:
2232
where = ' in ' + where
2233
# despite the name 'error' is a type
2234
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2238
if sys.platform == "win32":
2241
return msvcrt.getch()
2246
fd = sys.stdin.fileno()
2247
settings = termios.tcgetattr(fd)
2250
ch = sys.stdin.read(1)
2252
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2255
if sys.platform == 'linux2':
2256
def _local_concurrency():
2258
return os.sysconf('SC_NPROCESSORS_ONLN')
2259
except (ValueError, OSError, AttributeError):
2261
elif sys.platform == 'darwin':
2262
def _local_concurrency():
2263
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2264
stdout=subprocess.PIPE).communicate()[0]
2265
elif "bsd" in sys.platform:
2266
def _local_concurrency():
2267
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2268
stdout=subprocess.PIPE).communicate()[0]
2269
elif sys.platform == 'sunos5':
2270
def _local_concurrency():
2271
return subprocess.Popen(['psrinfo', '-p',],
2272
stdout=subprocess.PIPE).communicate()[0]
2273
elif sys.platform == "win32":
2274
def _local_concurrency():
2275
# This appears to return the number of cores.
2276
return os.environ.get('NUMBER_OF_PROCESSORS')
2278
def _local_concurrency():
2283
_cached_local_concurrency = None
2285
def local_concurrency(use_cache=True):
2286
"""Return how many processes can be run concurrently.
2288
Rely on platform specific implementations and default to 1 (one) if
2289
anything goes wrong.
2291
global _cached_local_concurrency
2293
if _cached_local_concurrency is not None and use_cache:
2294
return _cached_local_concurrency
2296
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2297
if concurrency is None:
2299
import multiprocessing
2301
# multiprocessing is only available on Python >= 2.6
2303
concurrency = _local_concurrency()
2304
except (OSError, IOError):
2307
concurrency = multiprocessing.cpu_count()
2309
concurrency = int(concurrency)
2310
except (TypeError, ValueError):
2313
_cached_concurrency = concurrency
2317
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2318
"""A stream writer that doesn't decode str arguments."""
2320
def __init__(self, encode, stream, errors='strict'):
2321
codecs.StreamWriter.__init__(self, stream, errors)
2322
self.encode = encode
2324
def write(self, object):
2325
if type(object) is str:
2326
self.stream.write(object)
2328
data, _ = self.encode(object, self.errors)
2329
self.stream.write(data)
2331
if sys.platform == 'win32':
2332
def open_file(filename, mode='r', bufsize=-1):
2333
"""This function is used to override the ``open`` builtin.
2335
But it uses O_NOINHERIT flag so the file handle is not inherited by
2336
child processes. Deleting or renaming a closed file opened with this
2337
function is not blocking child processes.
2339
writing = 'w' in mode
2340
appending = 'a' in mode
2341
updating = '+' in mode
2342
binary = 'b' in mode
2345
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2346
# for flags for each modes.
2356
flags |= os.O_WRONLY
2357
flags |= os.O_CREAT | os.O_TRUNC
2362
flags |= os.O_WRONLY
2363
flags |= os.O_CREAT | os.O_APPEND
2368
flags |= os.O_RDONLY
2370
return os.fdopen(os.open(filename, flags), mode, bufsize)
2375
def getuser_unicode():
2376
"""Return the username as unicode.
2379
user_encoding = get_user_encoding()
2380
username = getpass.getuser().decode(user_encoding)
2381
except UnicodeDecodeError:
2382
raise errors.BzrError("Can't decode username as %s." % \
2384
except ImportError, e:
2385
if sys.platform != 'win32':
2387
if str(e) != 'No module named pwd':
2389
# https://bugs.launchpad.net/bzr/+bug/660174
2390
# getpass.getuser() is unable to return username on Windows
2391
# if there is no USERNAME environment variable set.
2392
# That could be true if bzr is running as a service,
2393
# e.g. running `bzr serve` as a service on Windows.
2394
# We should not fail with traceback in this case.
2395
username = u'UNKNOWN'
2399
def available_backup_name(base, exists):
2400
"""Find a non-existing backup file name.
2402
This will *not* create anything, this only return a 'free' entry. This
2403
should be used for checking names in a directory below a locked
2404
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2405
Leap) and generally discouraged.
2407
:param base: The base name.
2409
:param exists: A callable returning True if the path parameter exists.
2412
name = "%s.~%d~" % (base, counter)
2415
name = "%s.~%d~" % (base, counter)
2419
def set_fd_cloexec(fd):
2420
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2421
support for this is not available.
2425
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2426
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2427
except (ImportError, AttributeError):
2428
# Either the fcntl module or specific constants are not present
2432
def find_executable_on_path(name):
2433
"""Finds an executable on the PATH.
2435
On Windows, this will try to append each extension in the PATHEXT
2436
environment variable to the name, if it cannot be found with the name
2439
:param name: The base name of the executable.
2440
:return: The path to the executable found or None.
2442
path = os.environ.get('PATH')
2445
path = path.split(os.pathsep)
2446
if sys.platform == 'win32':
2447
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2448
exts = [ext.lower() for ext in exts]
2449
base, ext = os.path.splitext(name)
2451
if ext.lower() not in exts:
2459
f = os.path.join(d, name) + ext
2460
if os.access(f, os.X_OK):
2465
def _posix_is_local_pid_dead(pid):
2466
"""True if pid doesn't correspond to live process on this machine"""
2468
# Special meaning of unix kill: just check if it's there.
2471
if e.errno == errno.ESRCH:
2472
# On this machine, and really not found: as sure as we can be
2475
elif e.errno == errno.EPERM:
2476
# exists, though not ours
2479
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2480
# Don't really know.
2483
# Exists and our process: not dead.
2486
if sys.platform == "win32":
2487
is_local_pid_dead = win32utils.is_local_pid_dead
2489
is_local_pid_dead = _posix_is_local_pid_dead
added
removed
bzrlib/osutils.py
