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 encoding is unsupported by %s "
100
"application locale." % (a, user_encoding))
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)
440
383
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
441
384
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
442
385
return shutil.rmtree(path, ignore_errors, onerror)
444
f = win32utils.get_unicode_argv # special function or None
448
386
elif sys.platform == 'darwin':
449
387
getcwd = _mac_getcwd
452
def get_terminal_encoding(trace=False):
390
def get_terminal_encoding():
453
391
"""Find the best encoding for printing to the screen.
455
393
This attempts to check both sys.stdout and sys.stdin to see
456
394
what encoding they are in, and if that fails it falls back to
457
osutils.get_user_encoding().
395
bzrlib.user_encoding.
458
396
The problem is that on Windows, locale.getpreferredencoding()
459
397
is not the same encoding as that used by the console:
460
398
http://mail.python.org/pipermail/python-list/2003-May/162357.html
462
400
On my standard US Windows XP, the preferred encoding is
463
401
cp1252, but the console is cp437
465
:param trace: If True trace the selected encoding via mutter().
467
from bzrlib.trace import mutter
468
403
output_encoding = getattr(sys.stdout, 'encoding', None)
469
404
if not output_encoding:
470
405
input_encoding = getattr(sys.stdin, 'encoding', None)
471
406
if not input_encoding:
472
output_encoding = get_user_encoding()
474
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)
477
410
output_encoding = input_encoding
479
mutter('encoding stdout as sys.stdin encoding %r',
411
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
483
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
413
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
484
414
if output_encoding == 'cp0':
485
415
# invalid encoding (cp0 means 'no codepage' on Windows)
486
output_encoding = get_user_encoding()
488
mutter('cp0 is invalid encoding.'
489
' 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)
493
421
codecs.lookup(output_encoding)
579
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
580
report_activity=None, direction='read'):
581
"""Copy contents of one file to another.
583
The read_length can either be -1 to read to end-of-file (EOF) or
584
it can specify the maximum number of bytes to read.
586
The buff_size represents the maximum size for each read operation
587
performed on from_file.
589
:param report_activity: Call this as bytes are read, see
590
Transport._report_activity
591
:param direction: Will be passed to report_activity
593
:return: The number of bytes copied.
597
# read specified number of bytes
599
while read_length > 0:
600
num_bytes_to_read = min(read_length, buff_size)
602
block = from_file.read(num_bytes_to_read)
606
if report_activity is not None:
607
report_activity(len(block), direction)
610
actual_bytes_read = len(block)
611
read_length -= actual_bytes_read
612
length += actual_bytes_read
616
block = from_file.read(buff_size)
620
if report_activity is not None:
621
report_activity(len(block), direction)
627
def pump_string_file(bytes, file_handle, segment_size=None):
628
"""Write bytes to file_handle in many smaller writes.
630
:param bytes: The string to write.
631
:param file_handle: The file to write to.
633
# Write data in chunks rather than all at once, because very large
634
# writes fail on some platforms (e.g. Windows with SMB mounted
637
segment_size = 5242880 # 5MB
638
segments = range(len(bytes) / segment_size + 1)
639
write = file_handle.write
640
for segment_index in segments:
641
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)
645
545
def file_iterator(input_file, readsize=32768):
733
605
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
734
606
return offset.days * 86400 + offset.seconds
736
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
737
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
740
def format_date(t, offset=0, timezone='original', date_fmt=None,
609
def format_date(t, offset=0, timezone='original', date_fmt=None,
741
610
show_offset=True):
742
"""Return a formatted date string.
744
:param t: Seconds since the epoch.
745
:param offset: Timezone offset in seconds east of utc.
746
:param timezone: How to display the time: 'utc', 'original' for the
747
timezone specified by offset, or 'local' for the process's current
749
:param date_fmt: strftime format.
750
:param show_offset: Whether to append the timezone.
752
(date_fmt, tt, offset_str) = \
753
_format_date(t, offset, timezone, date_fmt, show_offset)
754
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
755
date_str = time.strftime(date_fmt, tt)
756
return date_str + offset_str
759
# Cache of formatted offset strings
763
def format_date_with_offset_in_original_timezone(t, offset=0,
764
_cache=_offset_cache):
765
"""Return a formatted date string in the original timezone.
767
This routine may be faster then format_date.
769
:param t: Seconds since the epoch.
770
:param offset: Timezone offset in seconds east of utc.
774
tt = time.gmtime(t + offset)
775
date_fmt = _default_format_by_weekday_num[tt[6]]
776
date_str = time.strftime(date_fmt, tt)
777
offset_str = _cache.get(offset, None)
778
if offset_str is None:
779
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
780
_cache[offset] = offset_str
781
return date_str + offset_str
784
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
786
"""Return an unicode date string formatted according to the current locale.
788
:param t: Seconds since the epoch.
789
:param offset: Timezone offset in seconds east of utc.
790
:param timezone: How to display the time: 'utc', 'original' for the
791
timezone specified by offset, or 'local' for the process's current
793
:param date_fmt: strftime format.
794
:param show_offset: Whether to append the timezone.
796
(date_fmt, tt, offset_str) = \
797
_format_date(t, offset, timezone, date_fmt, show_offset)
798
date_str = time.strftime(date_fmt, tt)
799
if not isinstance(date_str, unicode):
800
date_str = date_str.decode(get_user_encoding(), 'replace')
801
return date_str + offset_str
804
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)
805
615
if timezone == 'utc':
806
616
tt = time.gmtime(t)
753
assert isinstance(p, (list, tuple))
942
755
if (f == '..') or (f is None) or (f == ''):
943
756
raise errors.BzrError("sorry, %r not allowed in path" % f)
944
757
return pathjoin(*p)
947
def parent_directories(filename):
948
"""Return the list of parent directories, deepest first.
950
For example, parent_directories("a/b/c") -> ["a/b", "a"].
953
parts = splitpath(dirname(filename))
955
parents.append(joinpath(parts))
960
_extension_load_failures = []
963
def failed_to_load_extension(exception):
964
"""Handle failing to load a binary extension.
966
This should be called from the ImportError block guarding the attempt to
967
import the native extension. If this function returns, the pure-Python
968
implementation should be loaded instead::
971
>>> import bzrlib._fictional_extension_pyx
972
>>> except ImportError, e:
973
>>> bzrlib.osutils.failed_to_load_extension(e)
974
>>> import bzrlib._fictional_extension_py
976
# NB: This docstring is just an example, not a doctest, because doctest
977
# currently can't cope with the use of lazy imports in this namespace --
980
# This currently doesn't report the failure at the time it occurs, because
981
# they tend to happen very early in startup when we can't check config
982
# files etc, and also we want to report all failures but not spam the user
984
exception_str = str(exception)
985
if exception_str not in _extension_load_failures:
986
trace.mutter("failed to load compiled extension: %s" % exception_str)
987
_extension_load_failures.append(exception_str)
990
def report_extension_load_failures():
991
if not _extension_load_failures:
993
from bzrlib.config import GlobalConfig
994
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
996
# the warnings framework should by default show this only once
997
from bzrlib.trace import warning
999
"bzr: warning: some compiled extensions could not be loaded; "
1000
"see <https://answers.launchpad.net/bzr/+faq/703>")
1001
# we no longer show the specific missing extensions here, because it makes
1002
# the message too long and scary - see
1003
# https://bugs.launchpad.net/bzr/+bug/430529
1007
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1008
except ImportError, e:
1009
failed_to_load_extension(e)
1010
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)
1013
768
def split_lines(s):
1014
769
"""Split s into lines, but without removing the newline characters."""
1015
# Trivially convert a fulltext into a 'chunked' representation, and let
1016
# chunks_to_lines do the heavy lifting.
1017
if isinstance(s, str):
1018
# chunks_to_lines only supports 8-bit strings
1019
return chunks_to_lines([s])
1021
return _split_lines(s)
1024
def _split_lines(s):
1025
"""Split s into lines, but without removing the newline characters.
1027
This supports Unicode or plain string objects.
1029
770
lines = s.split('\n')
1030
771
result = [line + '\n' for line in lines[:-1]]
1152
846
os.path.commonprefix (python2.4) has a bad bug that it works just
1153
847
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1154
848
avoids that problem.
1156
NOTE: `base` should not have a trailing slash otherwise you'll get
1157
PathNotChild exceptions regardless of `path`.
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'
851
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
852
' exceed the platform minimum length (which is %d)' %
1165
855
rp = abspath(path)
1170
if len(head) <= len(base) and head != base:
1171
raise errors.PathNotChild(rp, base)
859
while len(head) >= len(base):
1172
860
if head == base:
1174
head, tail = split(head)
862
head, tail = os.path.split(head)
866
raise errors.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]
1251
def decode_filename(filename):
1252
"""Decode the filename using the filesystem encoding
1254
If it is unicode, it is returned.
1255
Otherwise it is decoded from the the filesystem's encoding. If decoding
1256
fails, a errors.BadFilenameEncoding exception is raised.
1258
if type(filename) is unicode:
1261
return filename.decode(_fs_enc)
1262
except UnicodeDecodeError:
1263
raise errors.BadFilenameEncoding(filename, _fs_enc)
1266
874
def safe_unicode(unicode_or_utf8_string):
1267
875
"""Coerce unicode_or_utf8_string into unicode.
1269
877
If it is unicode, it is returned.
1270
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1271
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.
1273
882
if isinstance(unicode_or_utf8_string, unicode):
1274
883
return unicode_or_utf8_string
1387
996
normalized_filename = _inaccessible_normalized_filename
1390
def set_signal_handler(signum, handler, restart_syscall=True):
1391
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1392
on platforms that support that.
1394
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1395
automatically restart (by calling `signal.siginterrupt(signum,
1396
False)`). May be ignored if the feature is not available on this
1397
platform or Python version.
1401
siginterrupt = signal.siginterrupt
1403
# This python implementation doesn't provide signal support, hence no
1406
except AttributeError:
1407
# siginterrupt doesn't exist on this platform, or for this version
1409
siginterrupt = lambda signum, flag: None
1411
def sig_handler(*args):
1412
# Python resets the siginterrupt flag when a signal is
1413
# received. <http://bugs.python.org/issue8354>
1414
# As a workaround for some cases, set it back the way we want it.
1415
siginterrupt(signum, False)
1416
# Now run the handler function passed to set_signal_handler.
1419
sig_handler = handler
1420
old_handler = signal.signal(signum, sig_handler)
1422
siginterrupt(signum, False)
1426
default_terminal_width = 80
1427
"""The default terminal width for ttys.
1429
This is defined so that higher levels can share a common fallback value when
1430
terminal_width() returns None.
1433
# Keep some state so that terminal_width can detect if _terminal_size has
1434
# returned a different size since the process started. See docstring and
1435
# comments of terminal_width for details.
1436
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1437
_terminal_size_state = 'no_data'
1438
_first_terminal_size = None
1440
999
def terminal_width():
1441
"""Return terminal width.
1443
None is returned if the width can't established precisely.
1446
- if BZR_COLUMNS is set, returns its value
1447
- if there is no controlling terminal, returns None
1448
- query the OS, if the queried size has changed since the last query,
1450
- if COLUMNS is set, returns its value,
1451
- if the OS has a value (even though it's never changed), return its value.
1453
From there, we need to query the OS to get the size of the controlling
1456
On Unices we query the OS by:
1457
- get termios.TIOCGWINSZ
1458
- if an error occurs or a negative value is obtained, returns None
1460
On Windows we query the OS by:
1461
- win32utils.get_console_size() decides,
1462
- returns None on error (provided default value)
1464
# Note to implementors: if changing the rules for determining the width,
1465
# make sure you've considered the behaviour in these cases:
1466
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1467
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1469
# - (add more interesting cases here, if you find any)
1470
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1471
# but we don't want to register a signal handler because it is impossible
1472
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1473
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1474
# time so we can notice if the reported size has changed, which should have
1477
# If BZR_COLUMNS is set, take it, user is always right
1478
# Except if they specified 0 in which case, impose no limit here
1480
width = int(os.environ['BZR_COLUMNS'])
1481
except (KeyError, ValueError):
1483
if width is not None:
1489
isatty = getattr(sys.stdout, 'isatty', None)
1490
if isatty is None or not isatty():
1491
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1495
width, height = os_size = _terminal_size(None, None)
1496
global _first_terminal_size, _terminal_size_state
1497
if _terminal_size_state == 'no_data':
1498
_first_terminal_size = os_size
1499
_terminal_size_state = 'unchanged'
1500
elif (_terminal_size_state == 'unchanged' and
1501
_first_terminal_size != os_size):
1502
_terminal_size_state = 'changed'
1504
# If the OS claims to know how wide the terminal is, and this value has
1505
# ever changed, use that.
1506
if _terminal_size_state == 'changed':
1507
if width is not None and width > 0:
1510
# If COLUMNS is set, use it.
1512
return int(os.environ['COLUMNS'])
1513
except (KeyError, ValueError):
1516
# Finally, use an unchanged size from the OS, if we have one.
1517
if _terminal_size_state == 'unchanged':
1518
if width is not None and width > 0:
1521
# The width could not be determined.
1525
def _win32_terminal_size(width, height):
1526
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1527
return width, height
1530
def _ioctl_terminal_size(width, height):
1000
"""Return estimated terminal width."""
1001
if sys.platform == 'win32':
1002
return win32utils.get_console_size()[0]
1532
1005
import struct, fcntl, termios
1533
1006
s = struct.pack('HHHH', 0, 0, 0, 0)
1534
1007
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1535
height, width = struct.unpack('HHHH', x)[0:2]
1536
except (IOError, AttributeError):
1008
width = struct.unpack('HHHH', x)[1]
1538
return width, height
1540
_terminal_size = None
1541
"""Returns the terminal size as (width, height).
1543
:param width: Default value for width.
1544
:param height: Default value for height.
1546
This is defined specifically for each OS and query the size of the controlling
1547
terminal. If any error occurs, the provided default values should be returned.
1549
if sys.platform == 'win32':
1550
_terminal_size = _win32_terminal_size
1552
_terminal_size = _ioctl_terminal_size
1013
width = int(os.environ['COLUMNS'])
1555
1022
def supports_executable():
1602
1069
raise errors.IllegalPath(path)
1605
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1607
def _is_error_enotdir(e):
1608
"""Check if this exception represents ENOTDIR.
1610
Unfortunately, python is very inconsistent about the exception
1611
here. The cases are:
1612
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1613
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1614
which is the windows error code.
1615
3) Windows, Python2.5 uses errno == EINVAL and
1616
winerror == ERROR_DIRECTORY
1618
:param e: An Exception object (expected to be OSError with an errno
1619
attribute, but we should be able to cope with anything)
1620
:return: True if this represents an ENOTDIR error. False otherwise.
1622
en = getattr(e, 'errno', None)
1623
if (en == errno.ENOTDIR
1624
or (sys.platform == 'win32'
1625
and (en == _WIN32_ERROR_DIRECTORY
1626
or (en == errno.EINVAL
1627
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1633
1072
def walkdirs(top, prefix=""):
1634
1073
"""Yield data about all the directories in a tree.
1636
1075
This yields all the data about the contents of a directory at a time.
1637
1076
After each directory has been yielded, if the caller has mutated the list
1638
1077
to exclude some directories, they are then not descended into.
1640
1079
The data yielded is of the form:
1641
1080
((directory-relpath, directory-path-from-top),
1642
[(relpath, basename, kind, lstat, path-from-top), ...]),
1081
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1643
1082
- directory-relpath is the relative path of the directory being returned
1644
1083
with respect to top. prefix is prepended to this.
1645
- 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.
1646
1085
It is suitable for use with os functions.
1647
1086
- relpath is the relative path within the subtree being walked.
1648
1087
- basename is the basename of the path
1681
1120
append = dirblock.append
1683
names = sorted(map(decode_filename, _listdir(top)))
1685
if not _is_error_enotdir(e):
1689
abspath = top_slash + name
1690
statvalue = _lstat(abspath)
1691
kind = _kind_from_mode(statvalue.st_mode)
1692
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))
1693
1126
yield (relroot, top), dirblock
1695
1128
# push the user specified dirs from dirblock
1696
1129
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1699
class DirReader(object):
1700
"""An interface for reading directories."""
1702
def top_prefix_to_starting_dir(self, top, prefix=""):
1703
"""Converts top and prefix to a starting dir entry
1705
:param top: A utf8 path
1706
:param prefix: An optional utf8 path to prefix output relative paths
1708
:return: A tuple starting with prefix, and ending with the native
1711
raise NotImplementedError(self.top_prefix_to_starting_dir)
1713
def read_dir(self, prefix, top):
1714
"""Read a specific dir.
1716
:param prefix: A utf8 prefix to be preprended to the path basenames.
1717
:param top: A natively encoded path to read.
1718
:return: A list of the directories contents. Each item contains:
1719
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1721
raise NotImplementedError(self.read_dir)
1724
_selected_dir_reader = None
1727
1132
def _walkdirs_utf8(top, prefix=""):
1728
1133
"""Yield data about all the directories in a tree.
1738
1143
path-from-top might be unicode or utf8, but it is the correct path to
1739
1144
pass to os functions to affect the file in question. (such as os.lstat)
1741
global _selected_dir_reader
1742
if _selected_dir_reader is None:
1743
fs_encoding = _fs_enc.upper()
1744
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1745
# Win98 doesn't have unicode apis like FindFirstFileW
1746
# TODO: We possibly could support Win98 by falling back to the
1747
# original FindFirstFile, and using TCHAR instead of WCHAR,
1748
# but that gets a bit tricky, and requires custom compiling
1751
from bzrlib._walkdirs_win32 import Win32ReadDir
1752
_selected_dir_reader = Win32ReadDir()
1755
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1756
# ANSI_X3.4-1968 is a form of ASCII
1758
from bzrlib._readdir_pyx import UTF8DirReader
1759
_selected_dir_reader = UTF8DirReader()
1760
except ImportError, e:
1761
failed_to_load_extension(e)
1764
if _selected_dir_reader is None:
1765
# Fallback to the python version
1766
_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
1768
1165
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1769
1166
# But we don't actually uses 1-3 in pending, so set them to None
1770
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1771
read_dir = _selected_dir_reader.read_dir
1772
_directory = _directory_kind
1167
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1774
relroot, _, _, _, top = pending[-1].pop()
1777
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))
1778
1183
yield (relroot, top), dirblock
1779
1185
# push the user specified dirs from dirblock
1780
next = [d for d in reversed(dirblock) if d[2] == _directory]
1782
pending.append(next)
1785
class UnicodeDirReader(DirReader):
1786
"""A dir reader for non-utf8 file systems, which transcodes."""
1788
__slots__ = ['_utf8_encode']
1791
self._utf8_encode = codecs.getencoder('utf8')
1793
def top_prefix_to_starting_dir(self, top, prefix=""):
1794
"""See DirReader.top_prefix_to_starting_dir."""
1795
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1797
def read_dir(self, prefix, top):
1798
"""Read a single directory from a non-utf8 file system.
1800
top, and the abspath element in the output are unicode, all other paths
1801
are utf8. Local disk IO is done via unicode calls to listdir etc.
1803
This is currently the fallback code path when the filesystem encoding is
1804
not UTF-8. It may be better to implement an alternative so that we can
1805
safely handle paths that are not properly decodable in the current
1808
See DirReader.read_dir for details.
1810
_utf8_encode = self._utf8_encode
1812
_listdir = os.listdir
1813
_kind_from_mode = file_kind_from_stat_mode
1816
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 + '/'
1819
1212
top_slash = top + u'/'
1991
1346
return user_encoding
1994
def get_diff_header_encoding():
1995
return get_terminal_encoding()
1998
def get_host_name():
1999
"""Return the current unicode host name.
2001
This is meant to be used in place of socket.gethostname() because that
2002
behaves inconsistently on different platforms.
2004
if sys.platform == "win32":
2006
return win32utils.get_host_name()
2009
return socket.gethostname().decode(get_user_encoding())
2012
# We must not read/write any more than 64k at a time from/to a socket so we
2013
# don't risk "no buffer space available" errors on some platforms. Windows in
2014
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2016
MAX_SOCKET_CHUNK = 64 * 1024
2018
_end_of_stream_errors = [errno.ECONNRESET]
2019
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2020
_eno = getattr(errno, _eno, None)
2021
if _eno is not None:
2022
_end_of_stream_errors.append(_eno)
2026
def read_bytes_from_socket(sock, report_activity=None,
2027
max_read_size=MAX_SOCKET_CHUNK):
2028
"""Read up to max_read_size of bytes from sock and notify of progress.
2030
Translates "Connection reset by peer" into file-like EOF (return an
2031
empty string rather than raise an error), and repeats the recv if
2032
interrupted by a signal.
2036
bytes = sock.recv(max_read_size)
2037
except socket.error, e:
2039
if eno in _end_of_stream_errors:
2040
# The connection was closed by the other side. Callers expect
2041
# an empty string to signal end-of-stream.
2043
elif eno == errno.EINTR:
2044
# Retry the interrupted recv.
2048
if report_activity is not None:
2049
report_activity(len(bytes), 'read')
2053
def recv_all(socket, count):
1349
def recv_all(socket, bytes):
2054
1350
"""Receive an exact number of bytes.
2056
1352
Regular Socket.recv() may return less than the requested number of bytes,
2057
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
2058
1354
on all platforms, but this should work everywhere. This will return
2059
1355
less than the requested amount if the remote end closes.
2061
1357
This isn't optimized and is intended mostly for use in testing.
2064
while len(b) < count:
2065
new = read_bytes_from_socket(socket, None, count - len(b))
1360
while len(b) < bytes:
1361
new = socket.recv(bytes - len(b))
2072
def send_all(sock, bytes, report_activity=None):
2073
"""Send all bytes on a socket.
2075
Breaks large blocks in smaller chunks to avoid buffering limitations on
2076
some platforms, and catches EINTR which may be thrown if the send is
2077
interrupted by a signal.
2079
This is preferred to socket.sendall(), because it avoids portability bugs
2080
and provides activity reporting.
2082
:param report_activity: Call this as bytes are read, see
2083
Transport._report_activity
2086
byte_count = len(bytes)
2087
while sent_total < byte_count:
2089
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2090
except socket.error, e:
2091
if e.args[0] != errno.EINTR:
2095
report_activity(sent, 'write')
2098
def connect_socket(address):
2099
# Slight variation of the socket.create_connection() function (provided by
2100
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2101
# provide it for previous python versions. Also, we don't use the timeout
2102
# parameter (provided by the python implementation) so we don't implement
2104
err = socket.error('getaddrinfo returns an empty list')
2105
host, port = address
2106
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2107
af, socktype, proto, canonname, sa = res
2110
sock = socket.socket(af, socktype, proto)
2114
except socket.error, err:
2115
# 'err' is now the most recent error
2116
if sock is not None:
2121
1367
def dereference_path(path):
2122
1368
"""Determine the real path to a file.
2130
1376
# The pathjoin for '.' is a workaround for Python bug #1213894.
2131
1377
# (initial path components aren't dereferenced)
2132
1378
return pathjoin(realpath(pathjoin('.', parent)), base)
2135
def supports_mapi():
2136
"""Return True if we can use MAPI to launch a mail client."""
2137
return sys.platform == "win32"
2140
def resource_string(package, resource_name):
2141
"""Load a resource from a package and return it as a string.
2143
Note: Only packages that start with bzrlib are currently supported.
2145
This is designed to be a lightweight implementation of resource
2146
loading in a way which is API compatible with the same API from
2148
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2149
If and when pkg_resources becomes a standard library, this routine
2152
# Check package name is within bzrlib
2153
if package == "bzrlib":
2154
resource_relpath = resource_name
2155
elif package.startswith("bzrlib."):
2156
package = package[len("bzrlib."):].replace('.', os.sep)
2157
resource_relpath = pathjoin(package, resource_name)
2159
raise errors.BzrError('resource package %s not in bzrlib' % package)
2161
# Map the resource to a file and read its contents
2162
base = dirname(bzrlib.__file__)
2163
if getattr(sys, 'frozen', None): # bzr.exe
2164
base = abspath(pathjoin(base, '..', '..'))
2165
f = file(pathjoin(base, resource_relpath), "rU")
2171
def file_kind_from_stat_mode_thunk(mode):
2172
global file_kind_from_stat_mode
2173
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2175
from bzrlib._readdir_pyx import UTF8DirReader
2176
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2177
except ImportError, e:
2178
# This is one time where we won't warn that an extension failed to
2179
# load. The extension is never available on Windows anyway.
2180
from bzrlib._readdir_py import (
2181
_kind_from_mode as file_kind_from_stat_mode
2183
return file_kind_from_stat_mode(mode)
2184
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2187
def file_kind(f, _lstat=os.lstat):
2189
return file_kind_from_stat_mode(_lstat(f).st_mode)
2191
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2192
raise errors.NoSuchFile(f)
2196
def until_no_eintr(f, *a, **kw):
2197
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2199
WARNING: you must be certain that it is safe to retry the call repeatedly
2200
if EINTR does occur. This is typically only true for low-level operations
2201
like os.read. If in any doubt, don't use this.
2203
Keep in mind that this is not a complete solution to EINTR. There is
2204
probably code in the Python standard library and other dependencies that
2205
may encounter EINTR if a signal arrives (and there is signal handler for
2206
that signal). So this function can reduce the impact for IO that bzrlib
2207
directly controls, but it is not a complete solution.
2209
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2213
except (IOError, OSError), e:
2214
if e.errno == errno.EINTR:
2219
@deprecated_function(deprecated_in((2, 2, 0)))
2220
def re_compile_checked(re_string, flags=0, where=""):
2221
"""Return a compiled re, or raise a sensible error.
2223
This should only be used when compiling user-supplied REs.
2225
:param re_string: Text form of regular expression.
2226
:param flags: eg re.IGNORECASE
2227
:param where: Message explaining to the user the context where
2228
it occurred, eg 'log search filter'.
2230
# from https://bugs.launchpad.net/bzr/+bug/251352
2232
re_obj = re.compile(re_string, flags)
2235
except errors.InvalidPattern, e:
2237
where = ' in ' + where
2238
# despite the name 'error' is a type
2239
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2243
if sys.platform == "win32":
2246
return msvcrt.getch()
2251
fd = sys.stdin.fileno()
2252
settings = termios.tcgetattr(fd)
2255
ch = sys.stdin.read(1)
2257
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2260
if sys.platform == 'linux2':
2261
def _local_concurrency():
2263
return os.sysconf('SC_NPROCESSORS_ONLN')
2264
except (ValueError, OSError, AttributeError):
2266
elif sys.platform == 'darwin':
2267
def _local_concurrency():
2268
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2269
stdout=subprocess.PIPE).communicate()[0]
2270
elif "bsd" in sys.platform:
2271
def _local_concurrency():
2272
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2273
stdout=subprocess.PIPE).communicate()[0]
2274
elif sys.platform == 'sunos5':
2275
def _local_concurrency():
2276
return subprocess.Popen(['psrinfo', '-p',],
2277
stdout=subprocess.PIPE).communicate()[0]
2278
elif sys.platform == "win32":
2279
def _local_concurrency():
2280
# This appears to return the number of cores.
2281
return os.environ.get('NUMBER_OF_PROCESSORS')
2283
def _local_concurrency():
2288
_cached_local_concurrency = None
2290
def local_concurrency(use_cache=True):
2291
"""Return how many processes can be run concurrently.
2293
Rely on platform specific implementations and default to 1 (one) if
2294
anything goes wrong.
2296
global _cached_local_concurrency
2298
if _cached_local_concurrency is not None and use_cache:
2299
return _cached_local_concurrency
2301
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2302
if concurrency is None:
2304
import multiprocessing
2306
# multiprocessing is only available on Python >= 2.6
2308
concurrency = _local_concurrency()
2309
except (OSError, IOError):
2312
concurrency = multiprocessing.cpu_count()
2314
concurrency = int(concurrency)
2315
except (TypeError, ValueError):
2318
_cached_concurrency = concurrency
2322
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2323
"""A stream writer that doesn't decode str arguments."""
2325
def __init__(self, encode, stream, errors='strict'):
2326
codecs.StreamWriter.__init__(self, stream, errors)
2327
self.encode = encode
2329
def write(self, object):
2330
if type(object) is str:
2331
self.stream.write(object)
2333
data, _ = self.encode(object, self.errors)
2334
self.stream.write(data)
2336
if sys.platform == 'win32':
2337
def open_file(filename, mode='r', bufsize=-1):
2338
"""This function is used to override the ``open`` builtin.
2340
But it uses O_NOINHERIT flag so the file handle is not inherited by
2341
child processes. Deleting or renaming a closed file opened with this
2342
function is not blocking child processes.
2344
writing = 'w' in mode
2345
appending = 'a' in mode
2346
updating = '+' in mode
2347
binary = 'b' in mode
2350
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2351
# for flags for each modes.
2361
flags |= os.O_WRONLY
2362
flags |= os.O_CREAT | os.O_TRUNC
2367
flags |= os.O_WRONLY
2368
flags |= os.O_CREAT | os.O_APPEND
2373
flags |= os.O_RDONLY
2375
return os.fdopen(os.open(filename, flags), mode, bufsize)
2380
def getuser_unicode():
2381
"""Return the username as unicode.
2384
user_encoding = get_user_encoding()
2385
username = getpass.getuser().decode(user_encoding)
2386
except UnicodeDecodeError:
2387
raise errors.BzrError("Can't decode username as %s." % \
2392
def available_backup_name(base, exists):
2393
"""Find a non-existing backup file name.
2395
This will *not* create anything, this only return a 'free' entry. This
2396
should be used for checking names in a directory below a locked
2397
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2398
Leap) and generally discouraged.
2400
:param base: The base name.
2402
:param exists: A callable returning True if the path parameter exists.
2405
name = "%s.~%d~" % (base, counter)
2408
name = "%s.~%d~" % (base, counter)
2412
def set_fd_cloexec(fd):
2413
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2414
support for this is not available.
2418
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2419
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2420
except (ImportError, AttributeError):
2421
# Either the fcntl module or specific constants are not present
2425
def find_executable_on_path(name):
2426
"""Finds an executable on the PATH.
2428
On Windows, this will try to append each extension in the PATHEXT
2429
environment variable to the name, if it cannot be found with the name
2432
:param name: The base name of the executable.
2433
:return: The path to the executable found or None.
2435
path = os.environ.get('PATH')
2438
path = path.split(os.pathsep)
2439
if sys.platform == 'win32':
2440
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2441
exts = [ext.lower() for ext in exts]
2442
base, ext = os.path.splitext(name)
2444
if ext.lower() not in exts:
2452
f = os.path.join(d, name) + ext
2453
if os.access(f, os.X_OK):
added
removed
bzrlib/osutils.py
