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,
56
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
58
if sys.version_info < (2, 5):
59
import md5 as _mod_md5
61
import sha as _mod_sha
71
from bzrlib import symbol_versioning
74
# Cross platform wall-clock time functionality with decent resolution.
75
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
76
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
77
# synchronized with ``time.time()``, this is only meant to be used to find
78
# delta times by subtracting from another call to this function.
79
timer_func = time.time
80
if sys.platform == 'win32':
81
timer_func = time.clock
62
from bzrlib.trace import mutter
83
65
# On win32, O_BINARY is used to indicate the file should
84
66
# be opened in binary mode, rather than text mode.
85
67
# On other platforms, O_BINARY doesn't exist, because
86
68
# they always open in binary mode, so it is okay to
87
# OR with 0 on those platforms.
88
# O_NOINHERIT and O_TEXT exists only on win32 too.
69
# OR with 0 on those platforms
89
70
O_BINARY = getattr(os, 'O_BINARY', 0)
90
O_TEXT = getattr(os, 'O_TEXT', 0)
91
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
94
def get_unicode_argv():
96
user_encoding = get_user_encoding()
97
return [a.decode(user_encoding) for a in sys.argv[1:]]
98
except UnicodeDecodeError:
99
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
103
73
def make_readonly(filename):
104
74
"""Make a filename read-only."""
105
mod = os.lstat(filename).st_mode
106
if not stat.S_ISLNK(mod):
108
os.chmod(filename, mod)
75
mod = os.stat(filename).st_mode
77
os.chmod(filename, mod)
111
80
def make_writable(filename):
112
mod = os.lstat(filename).st_mode
113
if not stat.S_ISLNK(mod):
115
os.chmod(filename, mod)
118
def minimum_path_selection(paths):
119
"""Return the smallset subset of paths which are outside paths.
121
:param paths: A container (and hence not None) of paths.
122
:return: A set of paths sufficient to include everything in paths via
123
is_inside, drawn from the paths parameter.
129
return path.split('/')
130
sorted_paths = sorted(list(paths), key=sort_key)
132
search_paths = [sorted_paths[0]]
133
for path in sorted_paths[1:]:
134
if not is_inside(search_paths[-1], path):
135
# This path is unique, add it
136
search_paths.append(path)
138
return set(search_paths)
81
mod = os.stat(filename).st_mode
83
os.chmod(filename, mod)
426
383
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
427
384
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
428
385
return shutil.rmtree(path, ignore_errors, onerror)
430
f = win32utils.get_unicode_argv # special function or None
434
386
elif sys.platform == 'darwin':
435
387
getcwd = _mac_getcwd
438
def get_terminal_encoding(trace=False):
390
def get_terminal_encoding():
439
391
"""Find the best encoding for printing to the screen.
441
393
This attempts to check both sys.stdout and sys.stdin to see
442
394
what encoding they are in, and if that fails it falls back to
443
osutils.get_user_encoding().
395
bzrlib.user_encoding.
444
396
The problem is that on Windows, locale.getpreferredencoding()
445
397
is not the same encoding as that used by the console:
446
398
http://mail.python.org/pipermail/python-list/2003-May/162357.html
448
400
On my standard US Windows XP, the preferred encoding is
449
401
cp1252, but the console is cp437
451
:param trace: If True trace the selected encoding via mutter().
453
from bzrlib.trace import mutter
454
403
output_encoding = getattr(sys.stdout, 'encoding', None)
455
404
if not output_encoding:
456
405
input_encoding = getattr(sys.stdin, 'encoding', None)
457
406
if not input_encoding:
458
output_encoding = get_user_encoding()
460
mutter('encoding stdout as osutils.get_user_encoding() %r',
407
output_encoding = bzrlib.user_encoding
408
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
463
410
output_encoding = input_encoding
465
mutter('encoding stdout as sys.stdin encoding %r',
411
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
469
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
413
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
470
414
if output_encoding == 'cp0':
471
415
# invalid encoding (cp0 means 'no codepage' on Windows)
472
output_encoding = get_user_encoding()
474
mutter('cp0 is invalid encoding.'
475
' encoding stdout as osutils.get_user_encoding() %r',
416
output_encoding = bzrlib.user_encoding
417
mutter('cp0 is invalid encoding.'
418
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
479
421
codecs.lookup(output_encoding)
565
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
566
report_activity=None, direction='read'):
567
"""Copy contents of one file to another.
569
The read_length can either be -1 to read to end-of-file (EOF) or
570
it can specify the maximum number of bytes to read.
572
The buff_size represents the maximum size for each read operation
573
performed on from_file.
575
:param report_activity: Call this as bytes are read, see
576
Transport._report_activity
577
:param direction: Will be passed to report_activity
579
:return: The number of bytes copied.
583
# read specified number of bytes
585
while read_length > 0:
586
num_bytes_to_read = min(read_length, buff_size)
588
block = from_file.read(num_bytes_to_read)
592
if report_activity is not None:
593
report_activity(len(block), direction)
596
actual_bytes_read = len(block)
597
read_length -= actual_bytes_read
598
length += actual_bytes_read
602
block = from_file.read(buff_size)
606
if report_activity is not None:
607
report_activity(len(block), direction)
613
def pump_string_file(bytes, file_handle, segment_size=None):
614
"""Write bytes to file_handle in many smaller writes.
616
:param bytes: The string to write.
617
:param file_handle: The file to write to.
619
# Write data in chunks rather than all at once, because very large
620
# writes fail on some platforms (e.g. Windows with SMB mounted
623
segment_size = 5242880 # 5MB
624
segments = range(len(bytes) / segment_size + 1)
625
write = file_handle.write
626
for segment_index in segments:
627
segment = buffer(bytes, segment_index * segment_size, segment_size)
535
def pumpfile(fromfile, tofile):
536
"""Copy contents of one file to another."""
539
b = fromfile.read(BUFSIZE)
631
545
def file_iterator(input_file, readsize=32768):
719
605
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
720
606
return offset.days * 86400 + offset.seconds
722
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
723
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
726
def format_date(t, offset=0, timezone='original', date_fmt=None,
609
def format_date(t, offset=0, timezone='original', date_fmt=None,
727
610
show_offset=True):
728
"""Return a formatted date string.
730
:param t: Seconds since the epoch.
731
:param offset: Timezone offset in seconds east of utc.
732
:param timezone: How to display the time: 'utc', 'original' for the
733
timezone specified by offset, or 'local' for the process's current
735
:param date_fmt: strftime format.
736
:param show_offset: Whether to append the timezone.
738
(date_fmt, tt, offset_str) = \
739
_format_date(t, offset, timezone, date_fmt, show_offset)
740
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
741
date_str = time.strftime(date_fmt, tt)
742
return date_str + offset_str
745
# Cache of formatted offset strings
749
def format_date_with_offset_in_original_timezone(t, offset=0,
750
_cache=_offset_cache):
751
"""Return a formatted date string in the original timezone.
753
This routine may be faster then format_date.
755
:param t: Seconds since the epoch.
756
:param offset: Timezone offset in seconds east of utc.
760
tt = time.gmtime(t + offset)
761
date_fmt = _default_format_by_weekday_num[tt[6]]
762
date_str = time.strftime(date_fmt, tt)
763
offset_str = _cache.get(offset, None)
764
if offset_str is None:
765
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
766
_cache[offset] = offset_str
767
return date_str + offset_str
770
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
772
"""Return an unicode date string formatted according to the current locale.
774
:param t: Seconds since the epoch.
775
:param offset: Timezone offset in seconds east of utc.
776
:param timezone: How to display the time: 'utc', 'original' for the
777
timezone specified by offset, or 'local' for the process's current
779
:param date_fmt: strftime format.
780
:param show_offset: Whether to append the timezone.
782
(date_fmt, tt, offset_str) = \
783
_format_date(t, offset, timezone, date_fmt, show_offset)
784
date_str = time.strftime(date_fmt, tt)
785
if not isinstance(date_str, unicode):
786
date_str = date_str.decode(get_user_encoding(), 'replace')
787
return date_str + offset_str
790
def _format_date(t, offset, timezone, date_fmt, show_offset):
611
## TODO: Perhaps a global option to use either universal or local time?
612
## Or perhaps just let people set $TZ?
613
assert isinstance(t, float)
791
615
if timezone == 'utc':
792
616
tt = time.gmtime(t)
753
assert isinstance(p, (list, tuple))
928
755
if (f == '..') or (f is None) or (f == ''):
929
756
raise errors.BzrError("sorry, %r not allowed in path" % f)
930
757
return pathjoin(*p)
933
def parent_directories(filename):
934
"""Return the list of parent directories, deepest first.
936
For example, parent_directories("a/b/c") -> ["a/b", "a"].
939
parts = splitpath(dirname(filename))
941
parents.append(joinpath(parts))
946
_extension_load_failures = []
949
def failed_to_load_extension(exception):
950
"""Handle failing to load a binary extension.
952
This should be called from the ImportError block guarding the attempt to
953
import the native extension. If this function returns, the pure-Python
954
implementation should be loaded instead::
957
>>> import bzrlib._fictional_extension_pyx
958
>>> except ImportError, e:
959
>>> bzrlib.osutils.failed_to_load_extension(e)
960
>>> import bzrlib._fictional_extension_py
962
# NB: This docstring is just an example, not a doctest, because doctest
963
# currently can't cope with the use of lazy imports in this namespace --
966
# This currently doesn't report the failure at the time it occurs, because
967
# they tend to happen very early in startup when we can't check config
968
# files etc, and also we want to report all failures but not spam the user
970
exception_str = str(exception)
971
if exception_str not in _extension_load_failures:
972
trace.mutter("failed to load compiled extension: %s" % exception_str)
973
_extension_load_failures.append(exception_str)
976
def report_extension_load_failures():
977
if not _extension_load_failures:
979
from bzrlib.config import GlobalConfig
980
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
982
# the warnings framework should by default show this only once
983
from bzrlib.trace import warning
985
"bzr: warning: some compiled extensions could not be loaded; "
986
"see <https://answers.launchpad.net/bzr/+faq/703>")
987
# we no longer show the specific missing extensions here, because it makes
988
# the message too long and scary - see
989
# https://bugs.launchpad.net/bzr/+bug/430529
993
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
994
except ImportError, e:
995
failed_to_load_extension(e)
996
from bzrlib._chunks_to_lines_py import chunks_to_lines
760
@deprecated_function(zero_nine)
761
def appendpath(p1, p2):
765
return pathjoin(p1, p2)
999
768
def split_lines(s):
1000
769
"""Split s into lines, but without removing the newline characters."""
1001
# Trivially convert a fulltext into a 'chunked' representation, and let
1002
# chunks_to_lines do the heavy lifting.
1003
if isinstance(s, str):
1004
# chunks_to_lines only supports 8-bit strings
1005
return chunks_to_lines([s])
1007
return _split_lines(s)
1010
def _split_lines(s):
1011
"""Split s into lines, but without removing the newline characters.
1013
This supports Unicode or plain string objects.
1015
770
lines = s.split('\n')
1016
771
result = [line + '\n' for line in lines[:-1]]
1138
846
os.path.commonprefix (python2.4) has a bad bug that it works just
1139
847
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1140
848
avoids that problem.
1142
NOTE: `base` should not have a trailing slash otherwise you'll get
1143
PathNotChild exceptions regardless of `path`.
1146
if len(base) < MIN_ABS_PATHLENGTH:
1147
# must have space for e.g. a drive letter
1148
raise ValueError('%r is too short to calculate a relative path'
851
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
852
' exceed the platform minimum length (which is %d)' %
1151
855
rp = abspath(path)
1156
if len(head) <= len(base) and head != base:
1157
raise errors.PathNotChild(rp, base)
859
while len(head) >= len(base):
1158
860
if head == base:
1160
head, tail = split(head)
862
head, tail = os.path.split(head)
866
raise errors.PathNotChild(rp, base)
1165
return pathjoin(*reversed(s))
1170
def _cicp_canonical_relpath(base, path):
1171
"""Return the canonical path relative to base.
1173
Like relpath, but on case-insensitive-case-preserving file-systems, this
1174
will return the relpath as stored on the file-system rather than in the
1175
case specified in the input string, for all existing portions of the path.
1177
This will cause O(N) behaviour if called for every path in a tree; if you
1178
have a number of paths to convert, you should use canonical_relpaths().
1180
# TODO: it should be possible to optimize this for Windows by using the
1181
# win32 API FindFiles function to look for the specified name - but using
1182
# os.listdir() still gives us the correct, platform agnostic semantics in
1185
rel = relpath(base, path)
1186
# '.' will have been turned into ''
1190
abs_base = abspath(base)
1192
_listdir = os.listdir
1194
# use an explicit iterator so we can easily consume the rest on early exit.
1195
bit_iter = iter(rel.split('/'))
1196
for bit in bit_iter:
1199
next_entries = _listdir(current)
1200
except OSError: # enoent, eperm, etc
1201
# We can't find this in the filesystem, so just append the
1203
current = pathjoin(current, bit, *list(bit_iter))
1205
for look in next_entries:
1206
if lbit == look.lower():
1207
current = pathjoin(current, look)
1210
# got to the end, nothing matched, so we just return the
1211
# non-existing bits as they were specified (the filename may be
1212
# the target of a move, for example).
1213
current = pathjoin(current, bit, *list(bit_iter))
1215
return current[len(abs_base):].lstrip('/')
1217
# XXX - TODO - we need better detection/integration of case-insensitive
1218
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1219
# filesystems), for example, so could probably benefit from the same basic
1220
# support there. For now though, only Windows and OSX get that support, and
1221
# they get it for *all* file-systems!
1222
if sys.platform in ('win32', 'darwin'):
1223
canonical_relpath = _cicp_canonical_relpath
1225
canonical_relpath = relpath
1227
def canonical_relpaths(base, paths):
1228
"""Create an iterable to canonicalize a sequence of relative paths.
1230
The intent is for this implementation to use a cache, vastly speeding
1231
up multiple transformations in the same directory.
1233
# but for now, we haven't optimized...
1234
return [canonical_relpath(base, p) for p in paths]
1237
def decode_filename(filename):
1238
"""Decode the filename using the filesystem encoding
1240
If it is unicode, it is returned.
1241
Otherwise it is decoded from the the filesystem's encoding. If decoding
1242
fails, a errors.BadFilenameEncoding exception is raised.
1244
if type(filename) is unicode:
1247
return filename.decode(_fs_enc)
1248
except UnicodeDecodeError:
1249
raise errors.BadFilenameEncoding(filename, _fs_enc)
1252
874
def safe_unicode(unicode_or_utf8_string):
1253
875
"""Coerce unicode_or_utf8_string into unicode.
1255
877
If it is unicode, it is returned.
1256
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1257
wrapped in a BzrBadParameterNotUnicode exception.
878
Otherwise it is decoded from utf-8. If a decoding error
879
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
880
as a BzrBadParameter exception.
1259
882
if isinstance(unicode_or_utf8_string, unicode):
1260
883
return unicode_or_utf8_string
1373
996
normalized_filename = _inaccessible_normalized_filename
1376
def set_signal_handler(signum, handler, restart_syscall=True):
1377
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1378
on platforms that support that.
1380
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1381
automatically restart (by calling `signal.siginterrupt(signum,
1382
False)`). May be ignored if the feature is not available on this
1383
platform or Python version.
1387
siginterrupt = signal.siginterrupt
1389
# This python implementation doesn't provide signal support, hence no
1392
except AttributeError:
1393
# siginterrupt doesn't exist on this platform, or for this version
1395
siginterrupt = lambda signum, flag: None
1397
def sig_handler(*args):
1398
# Python resets the siginterrupt flag when a signal is
1399
# received. <http://bugs.python.org/issue8354>
1400
# As a workaround for some cases, set it back the way we want it.
1401
siginterrupt(signum, False)
1402
# Now run the handler function passed to set_signal_handler.
1405
sig_handler = handler
1406
old_handler = signal.signal(signum, sig_handler)
1408
siginterrupt(signum, False)
1412
default_terminal_width = 80
1413
"""The default terminal width for ttys.
1415
This is defined so that higher levels can share a common fallback value when
1416
terminal_width() returns None.
1419
# Keep some state so that terminal_width can detect if _terminal_size has
1420
# returned a different size since the process started. See docstring and
1421
# comments of terminal_width for details.
1422
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1423
_terminal_size_state = 'no_data'
1424
_first_terminal_size = None
1426
999
def terminal_width():
1427
"""Return terminal width.
1429
None is returned if the width can't established precisely.
1432
- if BZR_COLUMNS is set, returns its value
1433
- if there is no controlling terminal, returns None
1434
- query the OS, if the queried size has changed since the last query,
1436
- if COLUMNS is set, returns its value,
1437
- if the OS has a value (even though it's never changed), return its value.
1439
From there, we need to query the OS to get the size of the controlling
1442
On Unices we query the OS by:
1443
- get termios.TIOCGWINSZ
1444
- if an error occurs or a negative value is obtained, returns None
1446
On Windows we query the OS by:
1447
- win32utils.get_console_size() decides,
1448
- returns None on error (provided default value)
1450
# Note to implementors: if changing the rules for determining the width,
1451
# make sure you've considered the behaviour in these cases:
1452
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1453
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1455
# - (add more interesting cases here, if you find any)
1456
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1457
# but we don't want to register a signal handler because it is impossible
1458
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1459
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1460
# time so we can notice if the reported size has changed, which should have
1463
# If BZR_COLUMNS is set, take it, user is always right
1464
# Except if they specified 0 in which case, impose no limit here
1466
width = int(os.environ['BZR_COLUMNS'])
1467
except (KeyError, ValueError):
1469
if width is not None:
1475
isatty = getattr(sys.stdout, 'isatty', None)
1476
if isatty is None or not isatty():
1477
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1481
width, height = os_size = _terminal_size(None, None)
1482
global _first_terminal_size, _terminal_size_state
1483
if _terminal_size_state == 'no_data':
1484
_first_terminal_size = os_size
1485
_terminal_size_state = 'unchanged'
1486
elif (_terminal_size_state == 'unchanged' and
1487
_first_terminal_size != os_size):
1488
_terminal_size_state = 'changed'
1490
# If the OS claims to know how wide the terminal is, and this value has
1491
# ever changed, use that.
1492
if _terminal_size_state == 'changed':
1493
if width is not None and width > 0:
1496
# If COLUMNS is set, use it.
1498
return int(os.environ['COLUMNS'])
1499
except (KeyError, ValueError):
1502
# Finally, use an unchanged size from the OS, if we have one.
1503
if _terminal_size_state == 'unchanged':
1504
if width is not None and width > 0:
1507
# The width could not be determined.
1511
def _win32_terminal_size(width, height):
1512
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1513
return width, height
1516
def _ioctl_terminal_size(width, height):
1000
"""Return estimated terminal width."""
1001
if sys.platform == 'win32':
1002
return win32utils.get_console_size()[0]
1518
1005
import struct, fcntl, termios
1519
1006
s = struct.pack('HHHH', 0, 0, 0, 0)
1520
1007
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1521
height, width = struct.unpack('HHHH', x)[0:2]
1522
except (IOError, AttributeError):
1008
width = struct.unpack('HHHH', x)[1]
1524
return width, height
1526
_terminal_size = None
1527
"""Returns the terminal size as (width, height).
1529
:param width: Default value for width.
1530
:param height: Default value for height.
1532
This is defined specifically for each OS and query the size of the controlling
1533
terminal. If any error occurs, the provided default values should be returned.
1535
if sys.platform == 'win32':
1536
_terminal_size = _win32_terminal_size
1538
_terminal_size = _ioctl_terminal_size
1013
width = int(os.environ['COLUMNS'])
1541
1022
def supports_executable():
1588
1069
raise errors.IllegalPath(path)
1591
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1593
def _is_error_enotdir(e):
1594
"""Check if this exception represents ENOTDIR.
1596
Unfortunately, python is very inconsistent about the exception
1597
here. The cases are:
1598
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1599
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1600
which is the windows error code.
1601
3) Windows, Python2.5 uses errno == EINVAL and
1602
winerror == ERROR_DIRECTORY
1604
:param e: An Exception object (expected to be OSError with an errno
1605
attribute, but we should be able to cope with anything)
1606
:return: True if this represents an ENOTDIR error. False otherwise.
1608
en = getattr(e, 'errno', None)
1609
if (en == errno.ENOTDIR
1610
or (sys.platform == 'win32'
1611
and (en == _WIN32_ERROR_DIRECTORY
1612
or (en == errno.EINVAL
1613
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1619
1072
def walkdirs(top, prefix=""):
1620
1073
"""Yield data about all the directories in a tree.
1622
1075
This yields all the data about the contents of a directory at a time.
1623
1076
After each directory has been yielded, if the caller has mutated the list
1624
1077
to exclude some directories, they are then not descended into.
1626
1079
The data yielded is of the form:
1627
1080
((directory-relpath, directory-path-from-top),
1628
[(relpath, basename, kind, lstat, path-from-top), ...]),
1081
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1629
1082
- directory-relpath is the relative path of the directory being returned
1630
1083
with respect to top. prefix is prepended to this.
1631
- directory-path-from-root is the path including top for this directory.
1084
- directory-path-from-root is the path including top for this directory.
1632
1085
It is suitable for use with os functions.
1633
1086
- relpath is the relative path within the subtree being walked.
1634
1087
- basename is the basename of the path
1667
1120
append = dirblock.append
1669
names = sorted(map(decode_filename, _listdir(top)))
1671
if not _is_error_enotdir(e):
1675
abspath = top_slash + name
1676
statvalue = _lstat(abspath)
1677
kind = _kind_from_mode(statvalue.st_mode)
1678
append((relprefix + name, name, kind, statvalue, abspath))
1121
for name in sorted(_listdir(top)):
1122
abspath = top_slash + name
1123
statvalue = _lstat(abspath)
1124
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1125
append((relprefix + name, name, kind, statvalue, abspath))
1679
1126
yield (relroot, top), dirblock
1681
1128
# push the user specified dirs from dirblock
1682
1129
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1685
class DirReader(object):
1686
"""An interface for reading directories."""
1688
def top_prefix_to_starting_dir(self, top, prefix=""):
1689
"""Converts top and prefix to a starting dir entry
1691
:param top: A utf8 path
1692
:param prefix: An optional utf8 path to prefix output relative paths
1694
:return: A tuple starting with prefix, and ending with the native
1697
raise NotImplementedError(self.top_prefix_to_starting_dir)
1699
def read_dir(self, prefix, top):
1700
"""Read a specific dir.
1702
:param prefix: A utf8 prefix to be preprended to the path basenames.
1703
:param top: A natively encoded path to read.
1704
:return: A list of the directories contents. Each item contains:
1705
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1707
raise NotImplementedError(self.read_dir)
1710
_selected_dir_reader = None
1713
1132
def _walkdirs_utf8(top, prefix=""):
1714
1133
"""Yield data about all the directories in a tree.
1724
1143
path-from-top might be unicode or utf8, but it is the correct path to
1725
1144
pass to os functions to affect the file in question. (such as os.lstat)
1727
global _selected_dir_reader
1728
if _selected_dir_reader is None:
1729
fs_encoding = _fs_enc.upper()
1730
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1731
# Win98 doesn't have unicode apis like FindFirstFileW
1732
# TODO: We possibly could support Win98 by falling back to the
1733
# original FindFirstFile, and using TCHAR instead of WCHAR,
1734
# but that gets a bit tricky, and requires custom compiling
1737
from bzrlib._walkdirs_win32 import Win32ReadDir
1738
_selected_dir_reader = Win32ReadDir()
1741
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1742
# ANSI_X3.4-1968 is a form of ASCII
1744
from bzrlib._readdir_pyx import UTF8DirReader
1745
_selected_dir_reader = UTF8DirReader()
1746
except ImportError, e:
1747
failed_to_load_extension(e)
1750
if _selected_dir_reader is None:
1751
# Fallback to the python version
1752
_selected_dir_reader = UnicodeDirReader()
1146
fs_encoding = sys.getfilesystemencoding()
1147
if (sys.platform == 'win32' or
1148
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1149
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1151
return _walkdirs_fs_utf8(top, prefix=prefix)
1154
def _walkdirs_fs_utf8(top, prefix=""):
1155
"""See _walkdirs_utf8.
1157
This sub-function is called when we know the filesystem is already in utf8
1158
encoding. So we don't need to transcode filenames.
1161
_directory = _directory_kind
1162
_listdir = os.listdir
1163
_kind_from_mode = _formats.get
1754
1165
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1755
1166
# But we don't actually uses 1-3 in pending, so set them to None
1756
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1757
read_dir = _selected_dir_reader.read_dir
1758
_directory = _directory_kind
1167
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1760
relroot, _, _, _, top = pending[-1].pop()
1763
dirblock = sorted(read_dir(relroot, top))
1169
relroot, _, _, _, top = pending.pop()
1171
relprefix = relroot + '/'
1174
top_slash = top + '/'
1177
append = dirblock.append
1178
for name in sorted(_listdir(top)):
1179
abspath = top_slash + name
1180
statvalue = _lstat(abspath)
1181
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1182
append((relprefix + name, name, kind, statvalue, abspath))
1764
1183
yield (relroot, top), dirblock
1765
1185
# push the user specified dirs from dirblock
1766
next = [d for d in reversed(dirblock) if d[2] == _directory]
1768
pending.append(next)
1771
class UnicodeDirReader(DirReader):
1772
"""A dir reader for non-utf8 file systems, which transcodes."""
1774
__slots__ = ['_utf8_encode']
1777
self._utf8_encode = codecs.getencoder('utf8')
1779
def top_prefix_to_starting_dir(self, top, prefix=""):
1780
"""See DirReader.top_prefix_to_starting_dir."""
1781
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1783
def read_dir(self, prefix, top):
1784
"""Read a single directory from a non-utf8 file system.
1786
top, and the abspath element in the output are unicode, all other paths
1787
are utf8. Local disk IO is done via unicode calls to listdir etc.
1789
This is currently the fallback code path when the filesystem encoding is
1790
not UTF-8. It may be better to implement an alternative so that we can
1791
safely handle paths that are not properly decodable in the current
1794
See DirReader.read_dir for details.
1796
_utf8_encode = self._utf8_encode
1798
_listdir = os.listdir
1799
_kind_from_mode = file_kind_from_stat_mode
1802
relprefix = prefix + '/'
1186
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1189
def _walkdirs_unicode_to_utf8(top, prefix=""):
1190
"""See _walkdirs_utf8
1192
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1194
This is currently the fallback code path when the filesystem encoding is
1195
not UTF-8. It may be better to implement an alternative so that we can
1196
safely handle paths that are not properly decodable in the current
1199
_utf8_encode = codecs.getencoder('utf8')
1201
_directory = _directory_kind
1202
_listdir = os.listdir
1203
_kind_from_mode = _formats.get
1205
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1207
relroot, _, _, _, top = pending.pop()
1209
relprefix = relroot + '/'
1805
1212
top_slash = top + u'/'
1977
1346
return user_encoding
1980
def get_diff_header_encoding():
1981
return get_terminal_encoding()
1984
def get_host_name():
1985
"""Return the current unicode host name.
1987
This is meant to be used in place of socket.gethostname() because that
1988
behaves inconsistently on different platforms.
1990
if sys.platform == "win32":
1992
return win32utils.get_host_name()
1995
return socket.gethostname().decode(get_user_encoding())
1998
# We must not read/write any more than 64k at a time from/to a socket so we
1999
# don't risk "no buffer space available" errors on some platforms. Windows in
2000
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2002
MAX_SOCKET_CHUNK = 64 * 1024
2004
_end_of_stream_errors = [errno.ECONNRESET]
2005
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2006
_eno = getattr(errno, _eno, None)
2007
if _eno is not None:
2008
_end_of_stream_errors.append(_eno)
2012
def read_bytes_from_socket(sock, report_activity=None,
2013
max_read_size=MAX_SOCKET_CHUNK):
2014
"""Read up to max_read_size of bytes from sock and notify of progress.
2016
Translates "Connection reset by peer" into file-like EOF (return an
2017
empty string rather than raise an error), and repeats the recv if
2018
interrupted by a signal.
2022
bytes = sock.recv(max_read_size)
2023
except socket.error, e:
2025
if eno in _end_of_stream_errors:
2026
# The connection was closed by the other side. Callers expect
2027
# an empty string to signal end-of-stream.
2029
elif eno == errno.EINTR:
2030
# Retry the interrupted recv.
2034
if report_activity is not None:
2035
report_activity(len(bytes), 'read')
2039
def recv_all(socket, count):
1349
def recv_all(socket, bytes):
2040
1350
"""Receive an exact number of bytes.
2042
1352
Regular Socket.recv() may return less than the requested number of bytes,
2043
depending on what's in the OS buffer. MSG_WAITALL is not available
1353
dependning on what's in the OS buffer. MSG_WAITALL is not available
2044
1354
on all platforms, but this should work everywhere. This will return
2045
1355
less than the requested amount if the remote end closes.
2047
1357
This isn't optimized and is intended mostly for use in testing.
2050
while len(b) < count:
2051
new = read_bytes_from_socket(socket, None, count - len(b))
1360
while len(b) < bytes:
1361
new = socket.recv(bytes - len(b))
2058
def send_all(sock, bytes, report_activity=None):
2059
"""Send all bytes on a socket.
2061
Breaks large blocks in smaller chunks to avoid buffering limitations on
2062
some platforms, and catches EINTR which may be thrown if the send is
2063
interrupted by a signal.
2065
This is preferred to socket.sendall(), because it avoids portability bugs
2066
and provides activity reporting.
2068
:param report_activity: Call this as bytes are read, see
2069
Transport._report_activity
2072
byte_count = len(bytes)
2073
while sent_total < byte_count:
2075
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2076
except socket.error, e:
2077
if e.args[0] != errno.EINTR:
2081
report_activity(sent, 'write')
2084
def connect_socket(address):
2085
# Slight variation of the socket.create_connection() function (provided by
2086
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2087
# provide it for previous python versions. Also, we don't use the timeout
2088
# parameter (provided by the python implementation) so we don't implement
2090
err = socket.error('getaddrinfo returns an empty list')
2091
host, port = address
2092
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2093
af, socktype, proto, canonname, sa = res
2096
sock = socket.socket(af, socktype, proto)
2100
except socket.error, err:
2101
# 'err' is now the most recent error
2102
if sock is not None:
2107
1367
def dereference_path(path):
2108
1368
"""Determine the real path to a file.
2116
1376
# The pathjoin for '.' is a workaround for Python bug #1213894.
2117
1377
# (initial path components aren't dereferenced)
2118
1378
return pathjoin(realpath(pathjoin('.', parent)), base)
2121
def supports_mapi():
2122
"""Return True if we can use MAPI to launch a mail client."""
2123
return sys.platform == "win32"
2126
def resource_string(package, resource_name):
2127
"""Load a resource from a package and return it as a string.
2129
Note: Only packages that start with bzrlib are currently supported.
2131
This is designed to be a lightweight implementation of resource
2132
loading in a way which is API compatible with the same API from
2134
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2135
If and when pkg_resources becomes a standard library, this routine
2138
# Check package name is within bzrlib
2139
if package == "bzrlib":
2140
resource_relpath = resource_name
2141
elif package.startswith("bzrlib."):
2142
package = package[len("bzrlib."):].replace('.', os.sep)
2143
resource_relpath = pathjoin(package, resource_name)
2145
raise errors.BzrError('resource package %s not in bzrlib' % package)
2147
# Map the resource to a file and read its contents
2148
base = dirname(bzrlib.__file__)
2149
if getattr(sys, 'frozen', None): # bzr.exe
2150
base = abspath(pathjoin(base, '..', '..'))
2151
f = file(pathjoin(base, resource_relpath), "rU")
2157
def file_kind_from_stat_mode_thunk(mode):
2158
global file_kind_from_stat_mode
2159
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2161
from bzrlib._readdir_pyx import UTF8DirReader
2162
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2163
except ImportError, e:
2164
# This is one time where we won't warn that an extension failed to
2165
# load. The extension is never available on Windows anyway.
2166
from bzrlib._readdir_py import (
2167
_kind_from_mode as file_kind_from_stat_mode
2169
return file_kind_from_stat_mode(mode)
2170
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2173
def file_kind(f, _lstat=os.lstat):
2175
return file_kind_from_stat_mode(_lstat(f).st_mode)
2177
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2178
raise errors.NoSuchFile(f)
2182
def until_no_eintr(f, *a, **kw):
2183
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2185
WARNING: you must be certain that it is safe to retry the call repeatedly
2186
if EINTR does occur. This is typically only true for low-level operations
2187
like os.read. If in any doubt, don't use this.
2189
Keep in mind that this is not a complete solution to EINTR. There is
2190
probably code in the Python standard library and other dependencies that
2191
may encounter EINTR if a signal arrives (and there is signal handler for
2192
that signal). So this function can reduce the impact for IO that bzrlib
2193
directly controls, but it is not a complete solution.
2195
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2199
except (IOError, OSError), e:
2200
if e.errno == errno.EINTR:
2205
@deprecated_function(deprecated_in((2, 2, 0)))
2206
def re_compile_checked(re_string, flags=0, where=""):
2207
"""Return a compiled re, or raise a sensible error.
2209
This should only be used when compiling user-supplied REs.
2211
:param re_string: Text form of regular expression.
2212
:param flags: eg re.IGNORECASE
2213
:param where: Message explaining to the user the context where
2214
it occurred, eg 'log search filter'.
2216
# from https://bugs.launchpad.net/bzr/+bug/251352
2218
re_obj = re.compile(re_string, flags)
2221
except errors.InvalidPattern, e:
2223
where = ' in ' + where
2224
# despite the name 'error' is a type
2225
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2229
if sys.platform == "win32":
2232
return msvcrt.getch()
2237
fd = sys.stdin.fileno()
2238
settings = termios.tcgetattr(fd)
2241
ch = sys.stdin.read(1)
2243
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2247
if sys.platform == 'linux2':
2248
def _local_concurrency():
2250
prefix = 'processor'
2251
for line in file('/proc/cpuinfo', 'rb'):
2252
if line.startswith(prefix):
2253
concurrency = int(line[line.find(':')+1:]) + 1
2255
elif sys.platform == 'darwin':
2256
def _local_concurrency():
2257
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2258
stdout=subprocess.PIPE).communicate()[0]
2259
elif sys.platform[0:7] == 'freebsd':
2260
def _local_concurrency():
2261
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2262
stdout=subprocess.PIPE).communicate()[0]
2263
elif sys.platform == 'sunos5':
2264
def _local_concurrency():
2265
return subprocess.Popen(['psrinfo', '-p',],
2266
stdout=subprocess.PIPE).communicate()[0]
2267
elif sys.platform == "win32":
2268
def _local_concurrency():
2269
# This appears to return the number of cores.
2270
return os.environ.get('NUMBER_OF_PROCESSORS')
2272
def _local_concurrency():
2277
_cached_local_concurrency = None
2279
def local_concurrency(use_cache=True):
2280
"""Return how many processes can be run concurrently.
2282
Rely on platform specific implementations and default to 1 (one) if
2283
anything goes wrong.
2285
global _cached_local_concurrency
2287
if _cached_local_concurrency is not None and use_cache:
2288
return _cached_local_concurrency
2290
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2291
if concurrency is None:
2293
concurrency = _local_concurrency()
2294
except (OSError, IOError):
2297
concurrency = int(concurrency)
2298
except (TypeError, ValueError):
2301
_cached_concurrency = concurrency
2305
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2306
"""A stream writer that doesn't decode str arguments."""
2308
def __init__(self, encode, stream, errors='strict'):
2309
codecs.StreamWriter.__init__(self, stream, errors)
2310
self.encode = encode
2312
def write(self, object):
2313
if type(object) is str:
2314
self.stream.write(object)
2316
data, _ = self.encode(object, self.errors)
2317
self.stream.write(data)
2319
if sys.platform == 'win32':
2320
def open_file(filename, mode='r', bufsize=-1):
2321
"""This function is used to override the ``open`` builtin.
2323
But it uses O_NOINHERIT flag so the file handle is not inherited by
2324
child processes. Deleting or renaming a closed file opened with this
2325
function is not blocking child processes.
2327
writing = 'w' in mode
2328
appending = 'a' in mode
2329
updating = '+' in mode
2330
binary = 'b' in mode
2333
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2334
# for flags for each modes.
2344
flags |= os.O_WRONLY
2345
flags |= os.O_CREAT | os.O_TRUNC
2350
flags |= os.O_WRONLY
2351
flags |= os.O_CREAT | os.O_APPEND
2356
flags |= os.O_RDONLY
2358
return os.fdopen(os.open(filename, flags), mode, bufsize)
2363
def getuser_unicode():
2364
"""Return the username as unicode.
2367
user_encoding = get_user_encoding()
2368
username = getpass.getuser().decode(user_encoding)
2369
except UnicodeDecodeError:
2370
raise errors.BzrError("Can't decode username as %s." % \
2375
def available_backup_name(base, exists):
2376
"""Find a non-existing backup file name.
2378
This will *not* create anything, this only return a 'free' entry. This
2379
should be used for checking names in a directory below a locked
2380
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2381
Leap) and generally discouraged.
2383
:param base: The base name.
2385
:param exists: A callable returning True if the path parameter exists.
2388
name = "%s.~%d~" % (base, counter)
2391
name = "%s.~%d~" % (base, counter)
2395
def set_fd_cloexec(fd):
2396
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2397
support for this is not available.
2401
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2402
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2403
except (ImportError, AttributeError):
2404
# Either the fcntl module or specific constants are not present
2408
def find_executable_on_path(name):
2409
"""Finds an executable on the PATH.
2411
On Windows, this will try to append each extension in the PATHEXT
2412
environment variable to the name, if it cannot be found with the name
2415
:param name: The base name of the executable.
2416
:return: The path to the executable found or None.
2418
path = os.environ.get('PATH')
2421
path = path.split(os.pathsep)
2422
if sys.platform == 'win32':
2423
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2424
exts = [ext.lower() for ext in exts]
2425
base, ext = os.path.splitext(name)
2427
if ext.lower() not in exts:
2435
f = os.path.join(d, name) + ext
2436
if os.access(f, os.X_OK):
added
removed
bzrlib/osutils.py
