24
32
realpath as _nt_realpath,
25
33
splitdrive as _nt_splitdrive,
28
from os import listdir
33
from shutil import copyfile
35
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
36
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
43
from tempfile import (
56
from bzrlib.symbol_versioning import (
61
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
63
if sys.version_info < (2, 5):
64
import md5 as _mod_md5
66
import sha as _mod_sha
45
from bzrlib.errors import (BzrError,
46
BzrBadParameterNotUnicode,
51
from bzrlib.symbol_versioning import (deprecated_function,
53
from bzrlib.trace import mutter
76
from bzrlib import symbol_versioning
79
# Cross platform wall-clock time functionality with decent resolution.
80
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
81
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
82
# synchronized with ``time.time()``, this is only meant to be used to find
83
# delta times by subtracting from another call to this function.
84
timer_func = time.time
85
if sys.platform == 'win32':
86
timer_func = time.clock
56
88
# On win32, O_BINARY is used to indicate the file should
57
89
# be opened in binary mode, rather than text mode.
58
90
# On other platforms, O_BINARY doesn't exist, because
59
91
# they always open in binary mode, so it is okay to
60
# OR with 0 on those platforms
92
# OR with 0 on those platforms.
93
# O_NOINHERIT and O_TEXT exists only on win32 too.
61
94
O_BINARY = getattr(os, 'O_BINARY', 0)
95
O_TEXT = getattr(os, 'O_TEXT', 0)
96
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
99
def get_unicode_argv():
101
user_encoding = get_user_encoding()
102
return [a.decode(user_encoding) for a in sys.argv[1:]]
103
except UnicodeDecodeError:
104
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
64
108
def make_readonly(filename):
65
109
"""Make a filename read-only."""
66
mod = os.stat(filename).st_mode
68
os.chmod(filename, mod)
110
mod = os.lstat(filename).st_mode
111
if not stat.S_ISLNK(mod):
113
os.chmod(filename, mod)
71
116
def make_writable(filename):
72
mod = os.stat(filename).st_mode
74
os.chmod(filename, mod)
117
mod = os.lstat(filename).st_mode
118
if not stat.S_ISLNK(mod):
120
os.chmod(filename, mod)
123
def minimum_path_selection(paths):
124
"""Return the smallset subset of paths which are outside paths.
126
:param paths: A container (and hence not None) of paths.
127
:return: A set of paths sufficient to include everything in paths via
128
is_inside, drawn from the paths parameter.
134
return path.split('/')
135
sorted_paths = sorted(list(paths), key=sort_key)
137
search_paths = [sorted_paths[0]]
138
for path in sorted_paths[1:]:
139
if not is_inside(search_paths[-1], path):
140
# This path is unique, add it
141
search_paths.append(path)
143
return set(search_paths)
482
575
for dirname in dir_list:
483
576
if is_inside(dirname, fname) or is_inside(fname, dirname):
581
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
582
report_activity=None, direction='read'):
583
"""Copy contents of one file to another.
585
The read_length can either be -1 to read to end-of-file (EOF) or
586
it can specify the maximum number of bytes to read.
588
The buff_size represents the maximum size for each read operation
589
performed on from_file.
591
:param report_activity: Call this as bytes are read, see
592
Transport._report_activity
593
:param direction: Will be passed to report_activity
595
:return: The number of bytes copied.
599
# read specified number of bytes
601
while read_length > 0:
602
num_bytes_to_read = min(read_length, buff_size)
604
block = from_file.read(num_bytes_to_read)
608
if report_activity is not None:
609
report_activity(len(block), direction)
612
actual_bytes_read = len(block)
613
read_length -= actual_bytes_read
614
length += actual_bytes_read
489
def pumpfile(fromfile, tofile):
490
"""Copy contents of one file to another."""
493
b = fromfile.read(BUFSIZE)
618
block = from_file.read(buff_size)
622
if report_activity is not None:
623
report_activity(len(block), direction)
629
def pump_string_file(bytes, file_handle, segment_size=None):
630
"""Write bytes to file_handle in many smaller writes.
632
:param bytes: The string to write.
633
:param file_handle: The file to write to.
635
# Write data in chunks rather than all at once, because very large
636
# writes fail on some platforms (e.g. Windows with SMB mounted
639
segment_size = 5242880 # 5MB
640
segments = range(len(bytes) / segment_size + 1)
641
write = file_handle.write
642
for segment_index in segments:
643
segment = buffer(bytes, segment_index * segment_size, segment_size)
499
647
def file_iterator(input_file, readsize=32768):
555
731
def local_time_offset(t=None):
556
732
"""Return offset of local zone from GMT, either at present or at time t."""
557
# python2.3 localtime() can't take None
561
if time.localtime(t).tm_isdst and time.daylight:
564
return -time.timezone
567
def format_date(t, offset=0, timezone='original', date_fmt=None,
735
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
736
return offset.days * 86400 + offset.seconds
738
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
739
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
742
def format_date(t, offset=0, timezone='original', date_fmt=None,
568
743
show_offset=True):
569
## TODO: Perhaps a global option to use either universal or local time?
570
## Or perhaps just let people set $TZ?
571
assert isinstance(t, float)
744
"""Return a formatted date string.
746
:param t: Seconds since the epoch.
747
:param offset: Timezone offset in seconds east of utc.
748
:param timezone: How to display the time: 'utc', 'original' for the
749
timezone specified by offset, or 'local' for the process's current
751
:param date_fmt: strftime format.
752
:param show_offset: Whether to append the timezone.
754
(date_fmt, tt, offset_str) = \
755
_format_date(t, offset, timezone, date_fmt, show_offset)
756
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
757
date_str = time.strftime(date_fmt, tt)
758
return date_str + offset_str
761
# Cache of formatted offset strings
765
def format_date_with_offset_in_original_timezone(t, offset=0,
766
_cache=_offset_cache):
767
"""Return a formatted date string in the original timezone.
769
This routine may be faster then format_date.
771
:param t: Seconds since the epoch.
772
:param offset: Timezone offset in seconds east of utc.
776
tt = time.gmtime(t + offset)
777
date_fmt = _default_format_by_weekday_num[tt[6]]
778
date_str = time.strftime(date_fmt, tt)
779
offset_str = _cache.get(offset, None)
780
if offset_str is None:
781
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
782
_cache[offset] = offset_str
783
return date_str + offset_str
786
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
788
"""Return an unicode date string formatted according to the current locale.
790
:param t: Seconds since the epoch.
791
:param offset: Timezone offset in seconds east of utc.
792
:param timezone: How to display the time: 'utc', 'original' for the
793
timezone specified by offset, or 'local' for the process's current
795
:param date_fmt: strftime format.
796
:param show_offset: Whether to append the timezone.
798
(date_fmt, tt, offset_str) = \
799
_format_date(t, offset, timezone, date_fmt, show_offset)
800
date_str = time.strftime(date_fmt, tt)
801
if not isinstance(date_str, unicode):
802
date_str = date_str.decode(get_user_encoding(), 'replace')
803
return date_str + offset_str
806
def _format_date(t, offset, timezone, date_fmt, show_offset):
573
807
if timezone == 'utc':
574
808
tt = time.gmtime(t)
581
815
tt = time.localtime(t)
582
816
offset = local_time_offset(t)
584
raise BzrError("unsupported timezone format %r" % timezone,
585
['options are "utc", "original", "local"'])
818
raise errors.UnsupportedTimezoneFormat(timezone)
586
819
if date_fmt is None:
587
820
date_fmt = "%a %Y-%m-%d %H:%M:%S"
589
822
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
592
return (time.strftime(date_fmt, tt) + offset_str)
825
return (date_fmt, tt, offset_str)
595
828
def compact_date(when):
596
829
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
832
def format_delta(delta):
833
"""Get a nice looking string for a time delta.
835
:param delta: The time difference in seconds, can be positive or negative.
836
positive indicates time in the past, negative indicates time in the
837
future. (usually time.time() - stored_time)
838
:return: String formatted to show approximate resolution
844
direction = 'in the future'
848
if seconds < 90: # print seconds up to 90 seconds
850
return '%d second %s' % (seconds, direction,)
852
return '%d seconds %s' % (seconds, direction)
854
minutes = int(seconds / 60)
855
seconds -= 60 * minutes
860
if minutes < 90: # print minutes, seconds up to 90 minutes
862
return '%d minute, %d second%s %s' % (
863
minutes, seconds, plural_seconds, direction)
865
return '%d minutes, %d second%s %s' % (
866
minutes, seconds, plural_seconds, direction)
868
hours = int(minutes / 60)
869
minutes -= 60 * hours
876
return '%d hour, %d minute%s %s' % (hours, minutes,
877
plural_minutes, direction)
878
return '%d hours, %d minute%s %s' % (hours, minutes,
879
plural_minutes, direction)
601
882
"""Return size of given open file."""
669
raise BzrError("sorry, %r not allowed in path" % f)
934
raise errors.BzrError("sorry, %r not allowed in path" % f)
670
935
elif (f == '.') or (f == ''):
677
assert isinstance(p, list)
679
944
if (f == '..') or (f is None) or (f == ''):
680
raise BzrError("sorry, %r not allowed in path" % f)
945
raise errors.BzrError("sorry, %r not allowed in path" % f)
681
946
return pathjoin(*p)
684
@deprecated_function(zero_nine)
685
def appendpath(p1, p2):
689
return pathjoin(p1, p2)
949
def parent_directories(filename):
950
"""Return the list of parent directories, deepest first.
952
For example, parent_directories("a/b/c") -> ["a/b", "a"].
955
parts = splitpath(dirname(filename))
957
parents.append(joinpath(parts))
962
_extension_load_failures = []
965
def failed_to_load_extension(exception):
966
"""Handle failing to load a binary extension.
968
This should be called from the ImportError block guarding the attempt to
969
import the native extension. If this function returns, the pure-Python
970
implementation should be loaded instead::
973
>>> import bzrlib._fictional_extension_pyx
974
>>> except ImportError, e:
975
>>> bzrlib.osutils.failed_to_load_extension(e)
976
>>> import bzrlib._fictional_extension_py
978
# NB: This docstring is just an example, not a doctest, because doctest
979
# currently can't cope with the use of lazy imports in this namespace --
982
# This currently doesn't report the failure at the time it occurs, because
983
# they tend to happen very early in startup when we can't check config
984
# files etc, and also we want to report all failures but not spam the user
986
from bzrlib import trace
987
exception_str = str(exception)
988
if exception_str not in _extension_load_failures:
989
trace.mutter("failed to load compiled extension: %s" % exception_str)
990
_extension_load_failures.append(exception_str)
993
def report_extension_load_failures():
994
if not _extension_load_failures:
996
from bzrlib.config import GlobalConfig
997
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
999
# the warnings framework should by default show this only once
1000
from bzrlib.trace import warning
1002
"bzr: warning: some compiled extensions could not be loaded; "
1003
"see <https://answers.launchpad.net/bzr/+faq/703>")
1004
# we no longer show the specific missing extensions here, because it makes
1005
# the message too long and scary - see
1006
# https://bugs.launchpad.net/bzr/+bug/430529
1010
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1011
except ImportError, e:
1012
failed_to_load_extension(e)
1013
from bzrlib._chunks_to_lines_py import chunks_to_lines
692
1016
def split_lines(s):
693
1017
"""Split s into lines, but without removing the newline characters."""
1018
# Trivially convert a fulltext into a 'chunked' representation, and let
1019
# chunks_to_lines do the heavy lifting.
1020
if isinstance(s, str):
1021
# chunks_to_lines only supports 8-bit strings
1022
return chunks_to_lines([s])
1024
return _split_lines(s)
1027
def _split_lines(s):
1028
"""Split s into lines, but without removing the newline characters.
1030
This supports Unicode or plain string objects.
694
1032
lines = s.split('\n')
695
1033
result = [line + '\n' for line in lines[:-1]]
761
1157
avoids that problem.
764
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
765
' exceed the platform minimum length (which is %d)' %
1160
if len(base) < MIN_ABS_PATHLENGTH:
1161
# must have space for e.g. a drive letter
1162
raise ValueError('%r is too short to calculate a relative path'
768
1165
rp = abspath(path)
772
while len(head) >= len(base):
1170
if len(head) <= len(base) and head != base:
1171
raise errors.PathNotChild(rp, base)
773
1172
if head == base:
775
head, tail = os.path.split(head)
1174
head, tail = split(head)
779
raise PathNotChild(rp, base)
1179
return pathjoin(*reversed(s))
1184
def _cicp_canonical_relpath(base, path):
1185
"""Return the canonical path relative to base.
1187
Like relpath, but on case-insensitive-case-preserving file-systems, this
1188
will return the relpath as stored on the file-system rather than in the
1189
case specified in the input string, for all existing portions of the path.
1191
This will cause O(N) behaviour if called for every path in a tree; if you
1192
have a number of paths to convert, you should use canonical_relpaths().
1194
# TODO: it should be possible to optimize this for Windows by using the
1195
# win32 API FindFiles function to look for the specified name - but using
1196
# os.listdir() still gives us the correct, platform agnostic semantics in
1199
rel = relpath(base, path)
1200
# '.' will have been turned into ''
1204
abs_base = abspath(base)
1206
_listdir = os.listdir
1208
# use an explicit iterator so we can easily consume the rest on early exit.
1209
bit_iter = iter(rel.split('/'))
1210
for bit in bit_iter:
1213
next_entries = _listdir(current)
1214
except OSError: # enoent, eperm, etc
1215
# We can't find this in the filesystem, so just append the
1217
current = pathjoin(current, bit, *list(bit_iter))
1219
for look in next_entries:
1220
if lbit == look.lower():
1221
current = pathjoin(current, look)
1224
# got to the end, nothing matched, so we just return the
1225
# non-existing bits as they were specified (the filename may be
1226
# the target of a move, for example).
1227
current = pathjoin(current, bit, *list(bit_iter))
1229
return current[len(abs_base):].lstrip('/')
1231
# XXX - TODO - we need better detection/integration of case-insensitive
1232
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1233
# filesystems), for example, so could probably benefit from the same basic
1234
# support there. For now though, only Windows and OSX get that support, and
1235
# they get it for *all* file-systems!
1236
if sys.platform in ('win32', 'darwin'):
1237
canonical_relpath = _cicp_canonical_relpath
1239
canonical_relpath = relpath
1241
def canonical_relpaths(base, paths):
1242
"""Create an iterable to canonicalize a sequence of relative paths.
1244
The intent is for this implementation to use a cache, vastly speeding
1245
up multiple transformations in the same directory.
1247
# but for now, we haven't optimized...
1248
return [canonical_relpath(base, p) for p in paths]
787
1250
def safe_unicode(unicode_or_utf8_string):
788
1251
"""Coerce unicode_or_utf8_string into unicode.
790
1253
If it is unicode, it is returned.
791
Otherwise it is decoded from utf-8. If a decoding error
792
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
793
as a BzrBadParameter exception.
1254
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1255
wrapped in a BzrBadParameterNotUnicode exception.
795
1257
if isinstance(unicode_or_utf8_string, unicode):
796
1258
return unicode_or_utf8_string
798
1260
return unicode_or_utf8_string.decode('utf8')
799
1261
except UnicodeDecodeError:
800
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
1262
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1265
def safe_utf8(unicode_or_utf8_string):
1266
"""Coerce unicode_or_utf8_string to a utf8 string.
1268
If it is a str, it is returned.
1269
If it is Unicode, it is encoded into a utf-8 string.
1271
if isinstance(unicode_or_utf8_string, str):
1272
# TODO: jam 20070209 This is overkill, and probably has an impact on
1273
# performance if we are dealing with lots of apis that want a
1276
# Make sure it is a valid utf-8 string
1277
unicode_or_utf8_string.decode('utf-8')
1278
except UnicodeDecodeError:
1279
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1280
return unicode_or_utf8_string
1281
return unicode_or_utf8_string.encode('utf-8')
1284
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1285
' Revision id generators should be creating utf8'
1289
def safe_revision_id(unicode_or_utf8_string, warn=True):
1290
"""Revision ids should now be utf8, but at one point they were unicode.
1292
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1294
:param warn: Functions that are sanitizing user data can set warn=False
1295
:return: None or a utf8 revision id.
1297
if (unicode_or_utf8_string is None
1298
or unicode_or_utf8_string.__class__ == str):
1299
return unicode_or_utf8_string
1301
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1303
return cache_utf8.encode(unicode_or_utf8_string)
1306
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1307
' generators should be creating utf8 file ids.')
1310
def safe_file_id(unicode_or_utf8_string, warn=True):
1311
"""File ids should now be utf8, but at one point they were unicode.
1313
This is the same as safe_utf8, except it uses the cached encode functions
1314
to save a little bit of performance.
1316
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1318
:param warn: Functions that are sanitizing user data can set warn=False
1319
:return: None or a utf8 file id.
1321
if (unicode_or_utf8_string is None
1322
or unicode_or_utf8_string.__class__ == str):
1323
return unicode_or_utf8_string
1325
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1327
return cache_utf8.encode(unicode_or_utf8_string)
803
1330
_platform_normalizes_filenames = False
844
1371
normalized_filename = _inaccessible_normalized_filename
1374
def set_signal_handler(signum, handler, restart_syscall=True):
1375
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1376
on platforms that support that.
1378
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1379
automatically restart (by calling `signal.siginterrupt(signum,
1380
False)`). May be ignored if the feature is not available on this
1381
platform or Python version.
1385
siginterrupt = signal.siginterrupt
1387
# This python implementation doesn't provide signal support, hence no
1390
except AttributeError:
1391
# siginterrupt doesn't exist on this platform, or for this version
1393
siginterrupt = lambda signum, flag: None
1395
def sig_handler(*args):
1396
# Python resets the siginterrupt flag when a signal is
1397
# received. <http://bugs.python.org/issue8354>
1398
# As a workaround for some cases, set it back the way we want it.
1399
siginterrupt(signum, False)
1400
# Now run the handler function passed to set_signal_handler.
1403
sig_handler = handler
1404
old_handler = signal.signal(signum, sig_handler)
1406
siginterrupt(signum, False)
1410
default_terminal_width = 80
1411
"""The default terminal width for ttys.
1413
This is defined so that higher levels can share a common fallback value when
1414
terminal_width() returns None.
847
1418
def terminal_width():
848
"""Return estimated terminal width."""
849
if sys.platform == 'win32':
850
import bzrlib.win32console
851
return bzrlib.win32console.get_console_size()[0]
1419
"""Return terminal width.
1421
None is returned if the width can't established precisely.
1424
- if BZR_COLUMNS is set, returns its value
1425
- if there is no controlling terminal, returns None
1426
- if COLUMNS is set, returns its value,
1428
From there, we need to query the OS to get the size of the controlling
1432
- get termios.TIOCGWINSZ
1433
- if an error occurs or a negative value is obtained, returns None
1437
- win32utils.get_console_size() decides,
1438
- returns None on error (provided default value)
1441
# If BZR_COLUMNS is set, take it, user is always right
1443
return int(os.environ['BZR_COLUMNS'])
1444
except (KeyError, ValueError):
1447
isatty = getattr(sys.stdout, 'isatty', None)
1448
if isatty is None or not isatty():
1449
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1452
# If COLUMNS is set, take it, the terminal knows better (even inside a
1453
# given terminal, the application can decide to set COLUMNS to a lower
1454
# value (splitted screen) or a bigger value (scroll bars))
1456
return int(os.environ['COLUMNS'])
1457
except (KeyError, ValueError):
1460
width, height = _terminal_size(None, None)
1462
# Consider invalid values as meaning no width
1468
def _win32_terminal_size(width, height):
1469
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1470
return width, height
1473
def _ioctl_terminal_size(width, height):
854
1475
import struct, fcntl, termios
855
1476
s = struct.pack('HHHH', 0, 0, 0, 0)
856
1477
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
857
width = struct.unpack('HHHH', x)[1]
1478
height, width = struct.unpack('HHHH', x)[0:2]
1479
except (IOError, AttributeError):
1481
return width, height
1483
_terminal_size = None
1484
"""Returns the terminal size as (width, height).
1486
:param width: Default value for width.
1487
:param height: Default value for height.
1489
This is defined specifically for each OS and query the size of the controlling
1490
terminal. If any error occurs, the provided default values should be returned.
1492
if sys.platform == 'win32':
1493
_terminal_size = _win32_terminal_size
1495
_terminal_size = _ioctl_terminal_size
1498
def _terminal_size_changed(signum, frame):
1499
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1500
width, height = _terminal_size(None, None)
1501
if width is not None:
1502
os.environ['COLUMNS'] = str(width)
1505
_registered_sigwinch = False
1506
def watch_sigwinch():
1507
"""Register for SIGWINCH, once and only once.
1509
Do nothing if the signal module is not available.
1511
global _registered_sigwinch
1512
if not _registered_sigwinch:
862
width = int(os.environ['COLUMNS'])
1515
if getattr(signal, "SIGWINCH", None) is not None:
1516
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1518
# python doesn't provide signal support, nothing we can do about it
1520
_registered_sigwinch = True
871
1523
def supports_executable():
872
1524
return sys.platform != "win32"
1527
def supports_posix_readonly():
1528
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1530
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1531
directory controls creation/deletion, etc.
1533
And under win32, readonly means that the directory itself cannot be
1534
deleted. The contents of a readonly directory can be changed, unlike POSIX
1535
where files in readonly directories cannot be added, deleted or renamed.
1537
return sys.platform != "win32"
875
1540
def set_or_unset_env(env_variable, value):
876
1541
"""Modify the environment, setting or removing the env_variable.
897
1562
def check_legal_path(path):
898
"""Check whether the supplied path is legal.
1563
"""Check whether the supplied path is legal.
899
1564
This is only required on Windows, so we don't test on other platforms
902
1567
if sys.platform != "win32":
904
1569
if _validWin32PathRE.match(path) is None:
905
raise IllegalPath(path)
1570
raise errors.IllegalPath(path)
1573
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1575
def _is_error_enotdir(e):
1576
"""Check if this exception represents ENOTDIR.
1578
Unfortunately, python is very inconsistent about the exception
1579
here. The cases are:
1580
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1581
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1582
which is the windows error code.
1583
3) Windows, Python2.5 uses errno == EINVAL and
1584
winerror == ERROR_DIRECTORY
1586
:param e: An Exception object (expected to be OSError with an errno
1587
attribute, but we should be able to cope with anything)
1588
:return: True if this represents an ENOTDIR error. False otherwise.
1590
en = getattr(e, 'errno', None)
1591
if (en == errno.ENOTDIR
1592
or (sys.platform == 'win32'
1593
and (en == _WIN32_ERROR_DIRECTORY
1594
or (en == errno.EINVAL
1595
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
908
1601
def walkdirs(top, prefix=""):
909
1602
"""Yield data about all the directories in a tree.
911
1604
This yields all the data about the contents of a directory at a time.
912
1605
After each directory has been yielded, if the caller has mutated the list
913
1606
to exclude some directories, they are then not descended into.
915
1608
The data yielded is of the form:
916
1609
((directory-relpath, directory-path-from-top),
917
[(relpath, basename, kind, lstat), ...]),
1610
[(relpath, basename, kind, lstat, path-from-top), ...]),
918
1611
- directory-relpath is the relative path of the directory being returned
919
1612
with respect to top. prefix is prepended to this.
920
- directory-path-from-root is the path including top for this directory.
1613
- directory-path-from-root is the path including top for this directory.
921
1614
It is suitable for use with os functions.
922
1615
- relpath is the relative path within the subtree being walked.
923
1616
- basename is the basename of the path
925
1618
present within the tree - but it may be recorded as versioned. See
927
1620
- lstat is the stat data *if* the file was statted.
928
- planned, not implemented:
1621
- planned, not implemented:
929
1622
path_from_tree_root is the path from the root of the tree.
931
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1624
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
932
1625
allows one to walk a subtree but get paths that are relative to a tree
933
1626
rooted higher up.
934
1627
:return: an iterator over the dirs.
936
1629
#TODO there is a bit of a smell where the results of the directory-
937
# summary in this, and the path from the root, may not agree
1630
# summary in this, and the path from the root, may not agree
938
1631
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
939
1632
# potentially confusing output. We should make this more robust - but
940
1633
# not at a speed cost. RBC 20060731
943
1635
_directory = _directory_kind
945
pending = [(prefix, "", _directory, None, top)]
1636
_listdir = os.listdir
1637
_kind_from_mode = file_kind_from_stat_mode
1638
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
948
currentdir = pending.pop()
949
1640
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
952
relroot = currentdir[0] + '/'
1641
relroot, _, _, _, top = pending.pop()
1643
relprefix = relroot + u'/'
1646
top_slash = top + u'/'
1649
append = dirblock.append
1651
names = sorted(_listdir(top))
1653
if not _is_error_enotdir(e):
1657
abspath = top_slash + name
1658
statvalue = _lstat(abspath)
1659
kind = _kind_from_mode(statvalue.st_mode)
1660
append((relprefix + name, name, kind, statvalue, abspath))
1661
yield (relroot, top), dirblock
1663
# push the user specified dirs from dirblock
1664
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1667
class DirReader(object):
1668
"""An interface for reading directories."""
1670
def top_prefix_to_starting_dir(self, top, prefix=""):
1671
"""Converts top and prefix to a starting dir entry
1673
:param top: A utf8 path
1674
:param prefix: An optional utf8 path to prefix output relative paths
1676
:return: A tuple starting with prefix, and ending with the native
1679
raise NotImplementedError(self.top_prefix_to_starting_dir)
1681
def read_dir(self, prefix, top):
1682
"""Read a specific dir.
1684
:param prefix: A utf8 prefix to be preprended to the path basenames.
1685
:param top: A natively encoded path to read.
1686
:return: A list of the directories contents. Each item contains:
1687
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1689
raise NotImplementedError(self.read_dir)
1692
_selected_dir_reader = None
1695
def _walkdirs_utf8(top, prefix=""):
1696
"""Yield data about all the directories in a tree.
1698
This yields the same information as walkdirs() only each entry is yielded
1699
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1700
are returned as exact byte-strings.
1702
:return: yields a tuple of (dir_info, [file_info])
1703
dir_info is (utf8_relpath, path-from-top)
1704
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1705
if top is an absolute path, path-from-top is also an absolute path.
1706
path-from-top might be unicode or utf8, but it is the correct path to
1707
pass to os functions to affect the file in question. (such as os.lstat)
1709
global _selected_dir_reader
1710
if _selected_dir_reader is None:
1711
fs_encoding = _fs_enc.upper()
1712
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1713
# Win98 doesn't have unicode apis like FindFirstFileW
1714
# TODO: We possibly could support Win98 by falling back to the
1715
# original FindFirstFile, and using TCHAR instead of WCHAR,
1716
# but that gets a bit tricky, and requires custom compiling
1719
from bzrlib._walkdirs_win32 import Win32ReadDir
1720
_selected_dir_reader = Win32ReadDir()
1723
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1724
# ANSI_X3.4-1968 is a form of ASCII
1726
from bzrlib._readdir_pyx import UTF8DirReader
1727
_selected_dir_reader = UTF8DirReader()
1728
except ImportError, e:
1729
failed_to_load_extension(e)
1732
if _selected_dir_reader is None:
1733
# Fallback to the python version
1734
_selected_dir_reader = UnicodeDirReader()
1736
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1737
# But we don't actually uses 1-3 in pending, so set them to None
1738
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1739
read_dir = _selected_dir_reader.read_dir
1740
_directory = _directory_kind
1742
relroot, _, _, _, top = pending[-1].pop()
1745
dirblock = sorted(read_dir(relroot, top))
1746
yield (relroot, top), dirblock
1747
# push the user specified dirs from dirblock
1748
next = [d for d in reversed(dirblock) if d[2] == _directory]
1750
pending.append(next)
1753
class UnicodeDirReader(DirReader):
1754
"""A dir reader for non-utf8 file systems, which transcodes."""
1756
__slots__ = ['_utf8_encode']
1759
self._utf8_encode = codecs.getencoder('utf8')
1761
def top_prefix_to_starting_dir(self, top, prefix=""):
1762
"""See DirReader.top_prefix_to_starting_dir."""
1763
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1765
def read_dir(self, prefix, top):
1766
"""Read a single directory from a non-utf8 file system.
1768
top, and the abspath element in the output are unicode, all other paths
1769
are utf8. Local disk IO is done via unicode calls to listdir etc.
1771
This is currently the fallback code path when the filesystem encoding is
1772
not UTF-8. It may be better to implement an alternative so that we can
1773
safely handle paths that are not properly decodable in the current
1776
See DirReader.read_dir for details.
1778
_utf8_encode = self._utf8_encode
1780
_listdir = os.listdir
1781
_kind_from_mode = file_kind_from_stat_mode
1784
relprefix = prefix + '/'
1787
top_slash = top + u'/'
1790
append = dirblock.append
955
1791
for name in sorted(_listdir(top)):
956
abspath = top + '/' + name
957
statvalue = lstat(abspath)
958
dirblock.append((relroot + name, name,
959
file_kind_from_stat_mode(statvalue.st_mode),
961
yield (currentdir[0], top), dirblock
962
# push the user specified dirs from dirblock
963
for dir in reversed(dirblock):
964
if dir[2] == _directory:
1793
name_utf8 = _utf8_encode(name)[0]
1794
except UnicodeDecodeError:
1795
raise errors.BadFilenameEncoding(
1796
_utf8_encode(relprefix)[0] + name, _fs_enc)
1797
abspath = top_slash + name
1798
statvalue = _lstat(abspath)
1799
kind = _kind_from_mode(statvalue.st_mode)
1800
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
968
1804
def copy_tree(from_path, to_path, handlers={}):
969
1805
"""Copy all of the entries in from_path into to_path.
971
:param from_path: The base directory to copy.
1807
:param from_path: The base directory to copy.
972
1808
:param to_path: The target directory. If it does not exist, it will
974
1810
:param handlers: A dictionary of functions, which takes a source and
1057
1929
' doesn\'t support the locale set by $LANG (%s)\n'
1058
1930
" Continuing with ascii encoding.\n"
1059
1931
% (e, os.environ.get('LANG')))
1061
if _cached_user_encoding is None:
1062
_cached_user_encoding = 'ascii'
1063
return _cached_user_encoding
1932
user_encoding = 'ascii'
1934
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1935
# treat that as ASCII, and not support printing unicode characters to the
1938
# For python scripts run under vim, we get '', so also treat that as ASCII
1939
if user_encoding in (None, 'cp0', ''):
1940
user_encoding = 'ascii'
1944
codecs.lookup(user_encoding)
1946
sys.stderr.write('bzr: warning:'
1947
' unknown encoding %s.'
1948
' Continuing with ascii encoding.\n'
1951
user_encoding = 'ascii'
1954
_cached_user_encoding = user_encoding
1956
return user_encoding
1959
def get_host_name():
1960
"""Return the current unicode host name.
1962
This is meant to be used in place of socket.gethostname() because that
1963
behaves inconsistently on different platforms.
1965
if sys.platform == "win32":
1967
return win32utils.get_host_name()
1970
return socket.gethostname().decode(get_user_encoding())
1973
# We must not read/write any more than 64k at a time from/to a socket so we
1974
# don't risk "no buffer space available" errors on some platforms. Windows in
1975
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1977
MAX_SOCKET_CHUNK = 64 * 1024
1979
def read_bytes_from_socket(sock, report_activity=None,
1980
max_read_size=MAX_SOCKET_CHUNK):
1981
"""Read up to max_read_size of bytes from sock and notify of progress.
1983
Translates "Connection reset by peer" into file-like EOF (return an
1984
empty string rather than raise an error), and repeats the recv if
1985
interrupted by a signal.
1989
bytes = sock.recv(max_read_size)
1990
except socket.error, e:
1992
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1993
# The connection was closed by the other side. Callers expect
1994
# an empty string to signal end-of-stream.
1996
elif eno == errno.EINTR:
1997
# Retry the interrupted recv.
2001
if report_activity is not None:
2002
report_activity(len(bytes), 'read')
2006
def recv_all(socket, count):
2007
"""Receive an exact number of bytes.
2009
Regular Socket.recv() may return less than the requested number of bytes,
2010
depending on what's in the OS buffer. MSG_WAITALL is not available
2011
on all platforms, but this should work everywhere. This will return
2012
less than the requested amount if the remote end closes.
2014
This isn't optimized and is intended mostly for use in testing.
2017
while len(b) < count:
2018
new = read_bytes_from_socket(socket, None, count - len(b))
2025
def send_all(sock, bytes, report_activity=None):
2026
"""Send all bytes on a socket.
2028
Breaks large blocks in smaller chunks to avoid buffering limitations on
2029
some platforms, and catches EINTR which may be thrown if the send is
2030
interrupted by a signal.
2032
This is preferred to socket.sendall(), because it avoids portability bugs
2033
and provides activity reporting.
2035
:param report_activity: Call this as bytes are read, see
2036
Transport._report_activity
2039
byte_count = len(bytes)
2040
while sent_total < byte_count:
2042
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2043
except socket.error, e:
2044
if e.args[0] != errno.EINTR:
2048
report_activity(sent, 'write')
2051
def dereference_path(path):
2052
"""Determine the real path to a file.
2054
All parent elements are dereferenced. But the file itself is not
2056
:param path: The original path. May be absolute or relative.
2057
:return: the real path *to* the file
2059
parent, base = os.path.split(path)
2060
# The pathjoin for '.' is a workaround for Python bug #1213894.
2061
# (initial path components aren't dereferenced)
2062
return pathjoin(realpath(pathjoin('.', parent)), base)
2065
def supports_mapi():
2066
"""Return True if we can use MAPI to launch a mail client."""
2067
return sys.platform == "win32"
2070
def resource_string(package, resource_name):
2071
"""Load a resource from a package and return it as a string.
2073
Note: Only packages that start with bzrlib are currently supported.
2075
This is designed to be a lightweight implementation of resource
2076
loading in a way which is API compatible with the same API from
2078
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2079
If and when pkg_resources becomes a standard library, this routine
2082
# Check package name is within bzrlib
2083
if package == "bzrlib":
2084
resource_relpath = resource_name
2085
elif package.startswith("bzrlib."):
2086
package = package[len("bzrlib."):].replace('.', os.sep)
2087
resource_relpath = pathjoin(package, resource_name)
2089
raise errors.BzrError('resource package %s not in bzrlib' % package)
2091
# Map the resource to a file and read its contents
2092
base = dirname(bzrlib.__file__)
2093
if getattr(sys, 'frozen', None): # bzr.exe
2094
base = abspath(pathjoin(base, '..', '..'))
2095
filename = pathjoin(base, resource_relpath)
2096
return open(filename, 'rU').read()
2099
def file_kind_from_stat_mode_thunk(mode):
2100
global file_kind_from_stat_mode
2101
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2103
from bzrlib._readdir_pyx import UTF8DirReader
2104
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2105
except ImportError, e:
2106
# This is one time where we won't warn that an extension failed to
2107
# load. The extension is never available on Windows anyway.
2108
from bzrlib._readdir_py import (
2109
_kind_from_mode as file_kind_from_stat_mode
2111
return file_kind_from_stat_mode(mode)
2112
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2115
def file_kind(f, _lstat=os.lstat):
2117
return file_kind_from_stat_mode(_lstat(f).st_mode)
2119
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2120
raise errors.NoSuchFile(f)
2124
def until_no_eintr(f, *a, **kw):
2125
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2127
WARNING: you must be certain that it is safe to retry the call repeatedly
2128
if EINTR does occur. This is typically only true for low-level operations
2129
like os.read. If in any doubt, don't use this.
2131
Keep in mind that this is not a complete solution to EINTR. There is
2132
probably code in the Python standard library and other dependencies that
2133
may encounter EINTR if a signal arrives (and there is signal handler for
2134
that signal). So this function can reduce the impact for IO that bzrlib
2135
directly controls, but it is not a complete solution.
2137
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2141
except (IOError, OSError), e:
2142
if e.errno == errno.EINTR:
2147
def re_compile_checked(re_string, flags=0, where=""):
2148
"""Return a compiled re, or raise a sensible error.
2150
This should only be used when compiling user-supplied REs.
2152
:param re_string: Text form of regular expression.
2153
:param flags: eg re.IGNORECASE
2154
:param where: Message explaining to the user the context where
2155
it occurred, eg 'log search filter'.
2157
# from https://bugs.launchpad.net/bzr/+bug/251352
2159
re_obj = re.compile(re_string, flags)
2164
where = ' in ' + where
2165
# despite the name 'error' is a type
2166
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2167
% (where, re_string, e))
2170
if sys.platform == "win32":
2173
return msvcrt.getch()
2178
fd = sys.stdin.fileno()
2179
settings = termios.tcgetattr(fd)
2182
ch = sys.stdin.read(1)
2184
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2188
if sys.platform == 'linux2':
2189
def _local_concurrency():
2191
prefix = 'processor'
2192
for line in file('/proc/cpuinfo', 'rb'):
2193
if line.startswith(prefix):
2194
concurrency = int(line[line.find(':')+1:]) + 1
2196
elif sys.platform == 'darwin':
2197
def _local_concurrency():
2198
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2199
stdout=subprocess.PIPE).communicate()[0]
2200
elif sys.platform[0:7] == 'freebsd':
2201
def _local_concurrency():
2202
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2203
stdout=subprocess.PIPE).communicate()[0]
2204
elif sys.platform == 'sunos5':
2205
def _local_concurrency():
2206
return subprocess.Popen(['psrinfo', '-p',],
2207
stdout=subprocess.PIPE).communicate()[0]
2208
elif sys.platform == "win32":
2209
def _local_concurrency():
2210
# This appears to return the number of cores.
2211
return os.environ.get('NUMBER_OF_PROCESSORS')
2213
def _local_concurrency():
2218
_cached_local_concurrency = None
2220
def local_concurrency(use_cache=True):
2221
"""Return how many processes can be run concurrently.
2223
Rely on platform specific implementations and default to 1 (one) if
2224
anything goes wrong.
2226
global _cached_local_concurrency
2228
if _cached_local_concurrency is not None and use_cache:
2229
return _cached_local_concurrency
2231
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2232
if concurrency is None:
2234
concurrency = _local_concurrency()
2235
except (OSError, IOError):
2238
concurrency = int(concurrency)
2239
except (TypeError, ValueError):
2242
_cached_concurrency = concurrency
2246
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2247
"""A stream writer that doesn't decode str arguments."""
2249
def __init__(self, encode, stream, errors='strict'):
2250
codecs.StreamWriter.__init__(self, stream, errors)
2251
self.encode = encode
2253
def write(self, object):
2254
if type(object) is str:
2255
self.stream.write(object)
2257
data, _ = self.encode(object, self.errors)
2258
self.stream.write(data)
2260
if sys.platform == 'win32':
2261
def open_file(filename, mode='r', bufsize=-1):
2262
"""This function is used to override the ``open`` builtin.
2264
But it uses O_NOINHERIT flag so the file handle is not inherited by
2265
child processes. Deleting or renaming a closed file opened with this
2266
function is not blocking child processes.
2268
writing = 'w' in mode
2269
appending = 'a' in mode
2270
updating = '+' in mode
2271
binary = 'b' in mode
2274
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2275
# for flags for each modes.
2285
flags |= os.O_WRONLY
2286
flags |= os.O_CREAT | os.O_TRUNC
2291
flags |= os.O_WRONLY
2292
flags |= os.O_CREAT | os.O_APPEND
2297
flags |= os.O_RDONLY
2299
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
