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
52
58
from bzrlib.symbol_versioning import (
53
59
deprecated_function,
64
from bzrlib import symbol_versioning
67
# Cross platform wall-clock time functionality with decent resolution.
68
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
69
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
70
# synchronized with ``time.time()``, this is only meant to be used to find
71
# delta times by subtracting from another call to this function.
72
timer_func = time.time
73
if sys.platform == 'win32':
74
timer_func = time.clock
61
from bzrlib.trace import mutter
76
64
# On win32, O_BINARY is used to indicate the file should
77
65
# be opened in binary mode, rather than text mode.
78
66
# On other platforms, O_BINARY doesn't exist, because
79
67
# they always open in binary mode, so it is okay to
80
# OR with 0 on those platforms.
81
# O_NOINHERIT and O_TEXT exists only on win32 too.
68
# OR with 0 on those platforms
82
69
O_BINARY = getattr(os, 'O_BINARY', 0)
83
O_TEXT = getattr(os, 'O_TEXT', 0)
84
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
87
def get_unicode_argv():
89
user_encoding = get_user_encoding()
90
return [a.decode(user_encoding) for a in sys.argv[1:]]
91
except UnicodeDecodeError:
92
raise errors.BzrError("Parameter %r encoding is unsupported by %s "
93
"application locale." % (a, user_encoding))
96
72
def make_readonly(filename):
436
408
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
437
409
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
438
410
return shutil.rmtree(path, ignore_errors, onerror)
440
f = win32utils.get_unicode_argv # special function or None
444
411
elif sys.platform == 'darwin':
445
412
getcwd = _mac_getcwd
448
def get_terminal_encoding(trace=False):
415
def get_terminal_encoding():
449
416
"""Find the best encoding for printing to the screen.
451
418
This attempts to check both sys.stdout and sys.stdin to see
452
419
what encoding they are in, and if that fails it falls back to
453
osutils.get_user_encoding().
420
bzrlib.user_encoding.
454
421
The problem is that on Windows, locale.getpreferredencoding()
455
422
is not the same encoding as that used by the console:
456
423
http://mail.python.org/pipermail/python-list/2003-May/162357.html
458
425
On my standard US Windows XP, the preferred encoding is
459
426
cp1252, but the console is cp437
461
:param trace: If True trace the selected encoding via mutter().
463
from bzrlib.trace import mutter
464
428
output_encoding = getattr(sys.stdout, 'encoding', None)
465
429
if not output_encoding:
466
430
input_encoding = getattr(sys.stdin, 'encoding', None)
467
431
if not input_encoding:
468
output_encoding = get_user_encoding()
470
mutter('encoding stdout as osutils.get_user_encoding() %r',
432
output_encoding = bzrlib.user_encoding
433
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
473
435
output_encoding = input_encoding
475
mutter('encoding stdout as sys.stdin encoding %r',
436
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
479
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
438
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
480
439
if output_encoding == 'cp0':
481
440
# invalid encoding (cp0 means 'no codepage' on Windows)
482
output_encoding = get_user_encoding()
484
mutter('cp0 is invalid encoding.'
485
' encoding stdout as osutils.get_user_encoding() %r',
441
output_encoding = bzrlib.user_encoding
442
mutter('cp0 is invalid encoding.'
443
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
489
446
codecs.lookup(output_encoding)
742
651
:param timezone: How to display the time: 'utc', 'original' for the
743
652
timezone specified by offset, or 'local' for the process's current
745
:param date_fmt: strftime format.
746
:param show_offset: Whether to append the timezone.
748
(date_fmt, tt, offset_str) = \
749
_format_date(t, offset, timezone, date_fmt, show_offset)
750
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
751
date_str = time.strftime(date_fmt, tt)
752
return date_str + offset_str
755
# Cache of formatted offset strings
759
def format_date_with_offset_in_original_timezone(t, offset=0,
760
_cache=_offset_cache):
761
"""Return a formatted date string in the original timezone.
763
This routine may be faster then format_date.
765
:param t: Seconds since the epoch.
766
:param offset: Timezone offset in seconds east of utc.
770
tt = time.gmtime(t + offset)
771
date_fmt = _default_format_by_weekday_num[tt[6]]
772
date_str = time.strftime(date_fmt, tt)
773
offset_str = _cache.get(offset, None)
774
if offset_str is None:
775
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
776
_cache[offset] = offset_str
777
return date_str + offset_str
780
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
782
"""Return an unicode date string formatted according to the current locale.
784
:param t: Seconds since the epoch.
785
:param offset: Timezone offset in seconds east of utc.
786
:param timezone: How to display the time: 'utc', 'original' for the
787
timezone specified by offset, or 'local' for the process's current
789
:param date_fmt: strftime format.
790
:param show_offset: Whether to append the timezone.
792
(date_fmt, tt, offset_str) = \
793
_format_date(t, offset, timezone, date_fmt, show_offset)
794
date_str = time.strftime(date_fmt, tt)
795
if not isinstance(date_str, unicode):
796
date_str = date_str.decode(get_user_encoding(), 'replace')
797
return date_str + offset_str
800
def _format_date(t, offset, timezone, date_fmt, show_offset):
654
:param show_offset: Whether to append the timezone.
655
:param date_fmt: strftime format.
801
657
if timezone == 'utc':
802
658
tt = time.gmtime(t)
940
795
return pathjoin(*p)
943
def parent_directories(filename):
944
"""Return the list of parent directories, deepest first.
946
For example, parent_directories("a/b/c") -> ["a/b", "a"].
949
parts = splitpath(dirname(filename))
951
parents.append(joinpath(parts))
956
_extension_load_failures = []
959
def failed_to_load_extension(exception):
960
"""Handle failing to load a binary extension.
962
This should be called from the ImportError block guarding the attempt to
963
import the native extension. If this function returns, the pure-Python
964
implementation should be loaded instead::
967
>>> import bzrlib._fictional_extension_pyx
968
>>> except ImportError, e:
969
>>> bzrlib.osutils.failed_to_load_extension(e)
970
>>> import bzrlib._fictional_extension_py
972
# NB: This docstring is just an example, not a doctest, because doctest
973
# currently can't cope with the use of lazy imports in this namespace --
976
# This currently doesn't report the failure at the time it occurs, because
977
# they tend to happen very early in startup when we can't check config
978
# files etc, and also we want to report all failures but not spam the user
980
exception_str = str(exception)
981
if exception_str not in _extension_load_failures:
982
trace.mutter("failed to load compiled extension: %s" % exception_str)
983
_extension_load_failures.append(exception_str)
986
def report_extension_load_failures():
987
if not _extension_load_failures:
989
if config.GlobalStack().get('ignore_missing_extensions'):
991
# the warnings framework should by default show this only once
992
from bzrlib.trace import warning
994
"bzr: warning: some compiled extensions could not be loaded; "
995
"see <https://answers.launchpad.net/bzr/+faq/703>")
996
# we no longer show the specific missing extensions here, because it makes
997
# the message too long and scary - see
998
# https://bugs.launchpad.net/bzr/+bug/430529
1002
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1003
except ImportError, e:
1004
failed_to_load_extension(e)
1005
from bzrlib._chunks_to_lines_py import chunks_to_lines
1008
798
def split_lines(s):
1009
799
"""Split s into lines, but without removing the newline characters."""
1010
# Trivially convert a fulltext into a 'chunked' representation, and let
1011
# chunks_to_lines do the heavy lifting.
1012
if isinstance(s, str):
1013
# chunks_to_lines only supports 8-bit strings
1014
return chunks_to_lines([s])
1016
return _split_lines(s)
1019
def _split_lines(s):
1020
"""Split s into lines, but without removing the newline characters.
1022
This supports Unicode or plain string objects.
1024
800
lines = s.split('\n')
1025
801
result = [line + '\n' for line in lines[:-1]]
1165
if len(head) <= len(base) and head != base:
1166
raise errors.PathNotChild(rp, base)
900
while len(head) >= len(base):
1167
901
if head == base:
1169
head, tail = split(head)
903
head, tail = os.path.split(head)
907
raise errors.PathNotChild(rp, base)
1174
return pathjoin(*reversed(s))
1179
def _cicp_canonical_relpath(base, path):
1180
"""Return the canonical path relative to base.
1182
Like relpath, but on case-insensitive-case-preserving file-systems, this
1183
will return the relpath as stored on the file-system rather than in the
1184
case specified in the input string, for all existing portions of the path.
1186
This will cause O(N) behaviour if called for every path in a tree; if you
1187
have a number of paths to convert, you should use canonical_relpaths().
1189
# TODO: it should be possible to optimize this for Windows by using the
1190
# win32 API FindFiles function to look for the specified name - but using
1191
# os.listdir() still gives us the correct, platform agnostic semantics in
1194
rel = relpath(base, path)
1195
# '.' will have been turned into ''
1199
abs_base = abspath(base)
1201
_listdir = os.listdir
1203
# use an explicit iterator so we can easily consume the rest on early exit.
1204
bit_iter = iter(rel.split('/'))
1205
for bit in bit_iter:
1208
next_entries = _listdir(current)
1209
except OSError: # enoent, eperm, etc
1210
# We can't find this in the filesystem, so just append the
1212
current = pathjoin(current, bit, *list(bit_iter))
1214
for look in next_entries:
1215
if lbit == look.lower():
1216
current = pathjoin(current, look)
1219
# got to the end, nothing matched, so we just return the
1220
# non-existing bits as they were specified (the filename may be
1221
# the target of a move, for example).
1222
current = pathjoin(current, bit, *list(bit_iter))
1224
return current[len(abs_base):].lstrip('/')
1226
# XXX - TODO - we need better detection/integration of case-insensitive
1227
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1228
# filesystems), for example, so could probably benefit from the same basic
1229
# support there. For now though, only Windows and OSX get that support, and
1230
# they get it for *all* file-systems!
1231
if sys.platform in ('win32', 'darwin'):
1232
canonical_relpath = _cicp_canonical_relpath
1234
canonical_relpath = relpath
1236
def canonical_relpaths(base, paths):
1237
"""Create an iterable to canonicalize a sequence of relative paths.
1239
The intent is for this implementation to use a cache, vastly speeding
1240
up multiple transformations in the same directory.
1242
# but for now, we haven't optimized...
1243
return [canonical_relpath(base, p) for p in paths]
1246
def decode_filename(filename):
1247
"""Decode the filename using the filesystem encoding
1249
If it is unicode, it is returned.
1250
Otherwise it is decoded from the the filesystem's encoding. If decoding
1251
fails, a errors.BadFilenameEncoding exception is raised.
1253
if type(filename) is unicode:
1256
return filename.decode(_fs_enc)
1257
except UnicodeDecodeError:
1258
raise errors.BadFilenameEncoding(filename, _fs_enc)
1261
915
def safe_unicode(unicode_or_utf8_string):
1262
916
"""Coerce unicode_or_utf8_string into unicode.
1264
918
If it is unicode, it is returned.
1265
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1266
wrapped in a BzrBadParameterNotUnicode exception.
919
Otherwise it is decoded from utf-8. If a decoding error
920
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
921
as a BzrBadParameter exception.
1268
923
if isinstance(unicode_or_utf8_string, unicode):
1269
924
return unicode_or_utf8_string
1382
1037
normalized_filename = _inaccessible_normalized_filename
1385
def set_signal_handler(signum, handler, restart_syscall=True):
1386
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1387
on platforms that support that.
1389
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1390
automatically restart (by calling `signal.siginterrupt(signum,
1391
False)`). May be ignored if the feature is not available on this
1392
platform or Python version.
1396
siginterrupt = signal.siginterrupt
1398
# This python implementation doesn't provide signal support, hence no
1401
except AttributeError:
1402
# siginterrupt doesn't exist on this platform, or for this version
1404
siginterrupt = lambda signum, flag: None
1406
def sig_handler(*args):
1407
# Python resets the siginterrupt flag when a signal is
1408
# received. <http://bugs.python.org/issue8354>
1409
# As a workaround for some cases, set it back the way we want it.
1410
siginterrupt(signum, False)
1411
# Now run the handler function passed to set_signal_handler.
1414
sig_handler = handler
1415
old_handler = signal.signal(signum, sig_handler)
1417
siginterrupt(signum, False)
1421
default_terminal_width = 80
1422
"""The default terminal width for ttys.
1424
This is defined so that higher levels can share a common fallback value when
1425
terminal_width() returns None.
1428
# Keep some state so that terminal_width can detect if _terminal_size has
1429
# returned a different size since the process started. See docstring and
1430
# comments of terminal_width for details.
1431
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1432
_terminal_size_state = 'no_data'
1433
_first_terminal_size = None
1435
1040
def terminal_width():
1436
"""Return terminal width.
1438
None is returned if the width can't established precisely.
1441
- if BZR_COLUMNS is set, returns its value
1442
- if there is no controlling terminal, returns None
1443
- query the OS, if the queried size has changed since the last query,
1445
- if COLUMNS is set, returns its value,
1446
- if the OS has a value (even though it's never changed), return its value.
1448
From there, we need to query the OS to get the size of the controlling
1451
On Unices we query the OS by:
1452
- get termios.TIOCGWINSZ
1453
- if an error occurs or a negative value is obtained, returns None
1455
On Windows we query the OS by:
1456
- win32utils.get_console_size() decides,
1457
- returns None on error (provided default value)
1459
# Note to implementors: if changing the rules for determining the width,
1460
# make sure you've considered the behaviour in these cases:
1461
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1462
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1464
# - (add more interesting cases here, if you find any)
1465
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1466
# but we don't want to register a signal handler because it is impossible
1467
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1468
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1469
# time so we can notice if the reported size has changed, which should have
1472
# If BZR_COLUMNS is set, take it, user is always right
1473
# Except if they specified 0 in which case, impose no limit here
1475
width = int(os.environ['BZR_COLUMNS'])
1476
except (KeyError, ValueError):
1478
if width is not None:
1484
isatty = getattr(sys.stdout, 'isatty', None)
1485
if isatty is None or not isatty():
1486
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1490
width, height = os_size = _terminal_size(None, None)
1491
global _first_terminal_size, _terminal_size_state
1492
if _terminal_size_state == 'no_data':
1493
_first_terminal_size = os_size
1494
_terminal_size_state = 'unchanged'
1495
elif (_terminal_size_state == 'unchanged' and
1496
_first_terminal_size != os_size):
1497
_terminal_size_state = 'changed'
1499
# If the OS claims to know how wide the terminal is, and this value has
1500
# ever changed, use that.
1501
if _terminal_size_state == 'changed':
1502
if width is not None and width > 0:
1505
# If COLUMNS is set, use it.
1507
return int(os.environ['COLUMNS'])
1508
except (KeyError, ValueError):
1511
# Finally, use an unchanged size from the OS, if we have one.
1512
if _terminal_size_state == 'unchanged':
1513
if width is not None and width > 0:
1516
# The width could not be determined.
1520
def _win32_terminal_size(width, height):
1521
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1522
return width, height
1525
def _ioctl_terminal_size(width, height):
1041
"""Return estimated terminal width."""
1042
if sys.platform == 'win32':
1043
return win32utils.get_console_size()[0]
1527
1046
import struct, fcntl, termios
1528
1047
s = struct.pack('HHHH', 0, 0, 0, 0)
1529
1048
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1530
height, width = struct.unpack('HHHH', x)[0:2]
1531
except (IOError, AttributeError):
1049
width = struct.unpack('HHHH', x)[1]
1533
return width, height
1535
_terminal_size = None
1536
"""Returns the terminal size as (width, height).
1538
:param width: Default value for width.
1539
:param height: Default value for height.
1541
This is defined specifically for each OS and query the size of the controlling
1542
terminal. If any error occurs, the provided default values should be returned.
1544
if sys.platform == 'win32':
1545
_terminal_size = _win32_terminal_size
1547
_terminal_size = _ioctl_terminal_size
1054
width = int(os.environ['COLUMNS'])
1550
1063
def supports_executable():
1597
1110
raise errors.IllegalPath(path)
1600
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1602
def _is_error_enotdir(e):
1603
"""Check if this exception represents ENOTDIR.
1605
Unfortunately, python is very inconsistent about the exception
1606
here. The cases are:
1607
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1608
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1609
which is the windows error code.
1610
3) Windows, Python2.5 uses errno == EINVAL and
1611
winerror == ERROR_DIRECTORY
1613
:param e: An Exception object (expected to be OSError with an errno
1614
attribute, but we should be able to cope with anything)
1615
:return: True if this represents an ENOTDIR error. False otherwise.
1617
en = getattr(e, 'errno', None)
1618
if (en == errno.ENOTDIR
1619
or (sys.platform == 'win32'
1620
and (en == _WIN32_ERROR_DIRECTORY
1621
or (en == errno.EINVAL
1622
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1628
1113
def walkdirs(top, prefix=""):
1629
1114
"""Yield data about all the directories in a tree.
1631
1116
This yields all the data about the contents of a directory at a time.
1632
1117
After each directory has been yielded, if the caller has mutated the list
1633
1118
to exclude some directories, they are then not descended into.
1635
1120
The data yielded is of the form:
1636
1121
((directory-relpath, directory-path-from-top),
1637
1122
[(relpath, basename, kind, lstat, path-from-top), ...]),
1638
1123
- directory-relpath is the relative path of the directory being returned
1639
1124
with respect to top. prefix is prepended to this.
1640
- directory-path-from-root is the path including top for this directory.
1125
- directory-path-from-root is the path including top for this directory.
1641
1126
It is suitable for use with os functions.
1642
1127
- relpath is the relative path within the subtree being walked.
1643
1128
- basename is the basename of the path
1676
1161
append = dirblock.append
1678
names = sorted(map(decode_filename, _listdir(top)))
1680
if not _is_error_enotdir(e):
1684
abspath = top_slash + name
1685
statvalue = _lstat(abspath)
1686
kind = _kind_from_mode(statvalue.st_mode)
1687
append((relprefix + name, name, kind, statvalue, abspath))
1162
for name in sorted(_listdir(top)):
1163
abspath = top_slash + name
1164
statvalue = _lstat(abspath)
1165
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1166
append((relprefix + name, name, kind, statvalue, abspath))
1688
1167
yield (relroot, top), dirblock
1690
1169
# push the user specified dirs from dirblock
1691
1170
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1694
class DirReader(object):
1695
"""An interface for reading directories."""
1697
def top_prefix_to_starting_dir(self, top, prefix=""):
1698
"""Converts top and prefix to a starting dir entry
1700
:param top: A utf8 path
1701
:param prefix: An optional utf8 path to prefix output relative paths
1703
:return: A tuple starting with prefix, and ending with the native
1706
raise NotImplementedError(self.top_prefix_to_starting_dir)
1708
def read_dir(self, prefix, top):
1709
"""Read a specific dir.
1711
:param prefix: A utf8 prefix to be preprended to the path basenames.
1712
:param top: A natively encoded path to read.
1713
:return: A list of the directories contents. Each item contains:
1714
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1716
raise NotImplementedError(self.read_dir)
1719
_selected_dir_reader = None
1722
1173
def _walkdirs_utf8(top, prefix=""):
1723
1174
"""Yield data about all the directories in a tree.
1733
1184
path-from-top might be unicode or utf8, but it is the correct path to
1734
1185
pass to os functions to affect the file in question. (such as os.lstat)
1736
global _selected_dir_reader
1737
if _selected_dir_reader is None:
1738
fs_encoding = _fs_enc.upper()
1739
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1740
# Win98 doesn't have unicode apis like FindFirstFileW
1741
# TODO: We possibly could support Win98 by falling back to the
1742
# original FindFirstFile, and using TCHAR instead of WCHAR,
1743
# but that gets a bit tricky, and requires custom compiling
1746
from bzrlib._walkdirs_win32 import Win32ReadDir
1747
_selected_dir_reader = Win32ReadDir()
1750
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1751
# ANSI_X3.4-1968 is a form of ASCII
1753
from bzrlib._readdir_pyx import UTF8DirReader
1754
_selected_dir_reader = UTF8DirReader()
1755
except ImportError, e:
1756
failed_to_load_extension(e)
1759
if _selected_dir_reader is None:
1760
# Fallback to the python version
1761
_selected_dir_reader = UnicodeDirReader()
1187
fs_encoding = _fs_enc.upper()
1188
if (sys.platform == 'win32' or
1189
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1190
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1192
return _walkdirs_fs_utf8(top, prefix=prefix)
1195
def _walkdirs_fs_utf8(top, prefix=""):
1196
"""See _walkdirs_utf8.
1198
This sub-function is called when we know the filesystem is already in utf8
1199
encoding. So we don't need to transcode filenames.
1202
_directory = _directory_kind
1203
_listdir = os.listdir
1204
_kind_from_mode = _formats.get
1763
1206
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1764
1207
# But we don't actually uses 1-3 in pending, so set them to None
1765
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1766
read_dir = _selected_dir_reader.read_dir
1767
_directory = _directory_kind
1208
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1769
relroot, _, _, _, top = pending[-1].pop()
1772
dirblock = sorted(read_dir(relroot, top))
1210
relroot, _, _, _, top = pending.pop()
1212
relprefix = relroot + '/'
1215
top_slash = top + '/'
1218
append = dirblock.append
1219
for name in sorted(_listdir(top)):
1220
abspath = top_slash + name
1221
statvalue = _lstat(abspath)
1222
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1223
append((relprefix + name, name, kind, statvalue, abspath))
1773
1224
yield (relroot, top), dirblock
1774
1226
# push the user specified dirs from dirblock
1775
next = [d for d in reversed(dirblock) if d[2] == _directory]
1777
pending.append(next)
1780
class UnicodeDirReader(DirReader):
1781
"""A dir reader for non-utf8 file systems, which transcodes."""
1783
__slots__ = ['_utf8_encode']
1786
self._utf8_encode = codecs.getencoder('utf8')
1788
def top_prefix_to_starting_dir(self, top, prefix=""):
1789
"""See DirReader.top_prefix_to_starting_dir."""
1790
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1792
def read_dir(self, prefix, top):
1793
"""Read a single directory from a non-utf8 file system.
1795
top, and the abspath element in the output are unicode, all other paths
1796
are utf8. Local disk IO is done via unicode calls to listdir etc.
1798
This is currently the fallback code path when the filesystem encoding is
1799
not UTF-8. It may be better to implement an alternative so that we can
1800
safely handle paths that are not properly decodable in the current
1803
See DirReader.read_dir for details.
1805
_utf8_encode = self._utf8_encode
1807
_listdir = os.listdir
1808
_kind_from_mode = file_kind_from_stat_mode
1811
relprefix = prefix + '/'
1227
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1230
def _walkdirs_unicode_to_utf8(top, prefix=""):
1231
"""See _walkdirs_utf8
1233
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1235
This is currently the fallback code path when the filesystem encoding is
1236
not UTF-8. It may be better to implement an alternative so that we can
1237
safely handle paths that are not properly decodable in the current
1240
_utf8_encode = codecs.getencoder('utf8')
1242
_directory = _directory_kind
1243
_listdir = os.listdir
1244
_kind_from_mode = _formats.get
1246
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1248
relroot, _, _, _, top = pending.pop()
1250
relprefix = relroot + '/'
1814
1253
top_slash = top + u'/'
1986
1389
return user_encoding
1989
def get_diff_header_encoding():
1990
return get_terminal_encoding()
1993
def get_host_name():
1994
"""Return the current unicode host name.
1996
This is meant to be used in place of socket.gethostname() because that
1997
behaves inconsistently on different platforms.
1999
if sys.platform == "win32":
2001
return win32utils.get_host_name()
2004
return socket.gethostname().decode(get_user_encoding())
2007
# We must not read/write any more than 64k at a time from/to a socket so we
2008
# don't risk "no buffer space available" errors on some platforms. Windows in
2009
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2011
MAX_SOCKET_CHUNK = 64 * 1024
2013
_end_of_stream_errors = [errno.ECONNRESET]
2014
for _eno in ['WSAECONNRESET', 'WSAECONNABORTED']:
2015
_eno = getattr(errno, _eno, None)
2016
if _eno is not None:
2017
_end_of_stream_errors.append(_eno)
2021
def read_bytes_from_socket(sock, report_activity=None,
2022
max_read_size=MAX_SOCKET_CHUNK):
2023
"""Read up to max_read_size of bytes from sock and notify of progress.
2025
Translates "Connection reset by peer" into file-like EOF (return an
2026
empty string rather than raise an error), and repeats the recv if
2027
interrupted by a signal.
2031
bytes = sock.recv(max_read_size)
2032
except socket.error, e:
2034
if eno in _end_of_stream_errors:
2035
# The connection was closed by the other side. Callers expect
2036
# an empty string to signal end-of-stream.
2038
elif eno == errno.EINTR:
2039
# Retry the interrupted recv.
2043
if report_activity is not None:
2044
report_activity(len(bytes), 'read')
2048
def recv_all(socket, count):
1392
def recv_all(socket, bytes):
2049
1393
"""Receive an exact number of bytes.
2051
1395
Regular Socket.recv() may return less than the requested number of bytes,
2052
depending on what's in the OS buffer. MSG_WAITALL is not available
1396
dependning on what's in the OS buffer. MSG_WAITALL is not available
2053
1397
on all platforms, but this should work everywhere. This will return
2054
1398
less than the requested amount if the remote end closes.
2056
1400
This isn't optimized and is intended mostly for use in testing.
2059
while len(b) < count:
2060
new = read_bytes_from_socket(socket, None, count - len(b))
1403
while len(b) < bytes:
1404
new = socket.recv(bytes - len(b))
2067
def send_all(sock, bytes, report_activity=None):
1411
def send_all(socket, bytes):
2068
1412
"""Send all bytes on a socket.
2070
Breaks large blocks in smaller chunks to avoid buffering limitations on
2071
some platforms, and catches EINTR which may be thrown if the send is
2072
interrupted by a signal.
2074
This is preferred to socket.sendall(), because it avoids portability bugs
2075
and provides activity reporting.
2077
:param report_activity: Call this as bytes are read, see
2078
Transport._report_activity
1414
Regular socket.sendall() can give socket error 10053 on Windows. This
1415
implementation sends no more than 64k at a time, which avoids this problem.
2081
byte_count = len(bytes)
2082
while sent_total < byte_count:
2084
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2085
except socket.error, e:
2086
if e.args[0] != errno.EINTR:
2090
report_activity(sent, 'write')
2093
def connect_socket(address):
2094
# Slight variation of the socket.create_connection() function (provided by
2095
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2096
# provide it for previous python versions. Also, we don't use the timeout
2097
# parameter (provided by the python implementation) so we don't implement
2099
err = socket.error('getaddrinfo returns an empty list')
2100
host, port = address
2101
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2102
af, socktype, proto, canonname, sa = res
2105
sock = socket.socket(af, socktype, proto)
2109
except socket.error, err:
2110
# 'err' is now the most recent error
2111
if sock is not None:
1418
for pos in xrange(0, len(bytes), chunk_size):
1419
socket.sendall(bytes[pos:pos+chunk_size])
2116
1422
def dereference_path(path):
2157
1463
base = dirname(bzrlib.__file__)
2158
1464
if getattr(sys, 'frozen', None): # bzr.exe
2159
1465
base = abspath(pathjoin(base, '..', '..'))
2160
f = file(pathjoin(base, resource_relpath), "rU")
2166
def file_kind_from_stat_mode_thunk(mode):
2167
global file_kind_from_stat_mode
2168
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2170
from bzrlib._readdir_pyx import UTF8DirReader
2171
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2172
except ImportError, e:
2173
# This is one time where we won't warn that an extension failed to
2174
# load. The extension is never available on Windows anyway.
2175
from bzrlib._readdir_py import (
2176
_kind_from_mode as file_kind_from_stat_mode
2178
return file_kind_from_stat_mode(mode)
2179
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2181
def file_stat(f, _lstat=os.lstat):
2186
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2187
raise errors.NoSuchFile(f)
2190
def file_kind(f, _lstat=os.lstat):
2191
stat_value = file_stat(f, _lstat)
2192
return file_kind_from_stat_mode(stat_value.st_mode)
2194
def until_no_eintr(f, *a, **kw):
2195
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2197
WARNING: you must be certain that it is safe to retry the call repeatedly
2198
if EINTR does occur. This is typically only true for low-level operations
2199
like os.read. If in any doubt, don't use this.
2201
Keep in mind that this is not a complete solution to EINTR. There is
2202
probably code in the Python standard library and other dependencies that
2203
may encounter EINTR if a signal arrives (and there is signal handler for
2204
that signal). So this function can reduce the impact for IO that bzrlib
2205
directly controls, but it is not a complete solution.
2207
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2211
except (IOError, OSError), e:
2212
if e.errno == errno.EINTR:
2217
@deprecated_function(deprecated_in((2, 2, 0)))
2218
def re_compile_checked(re_string, flags=0, where=""):
2219
"""Return a compiled re, or raise a sensible error.
2221
This should only be used when compiling user-supplied REs.
2223
:param re_string: Text form of regular expression.
2224
:param flags: eg re.IGNORECASE
2225
:param where: Message explaining to the user the context where
2226
it occurred, eg 'log search filter'.
2228
# from https://bugs.launchpad.net/bzr/+bug/251352
2230
re_obj = re.compile(re_string, flags)
2233
except errors.InvalidPattern, e:
2235
where = ' in ' + where
2236
# despite the name 'error' is a type
2237
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2241
if sys.platform == "win32":
2244
return msvcrt.getch()
2249
fd = sys.stdin.fileno()
2250
settings = termios.tcgetattr(fd)
2253
ch = sys.stdin.read(1)
2255
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2258
if sys.platform.startswith('linux'):
2259
def _local_concurrency():
2261
return os.sysconf('SC_NPROCESSORS_ONLN')
2262
except (ValueError, OSError, AttributeError):
2264
elif sys.platform == 'darwin':
2265
def _local_concurrency():
2266
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2267
stdout=subprocess.PIPE).communicate()[0]
2268
elif "bsd" in sys.platform:
2269
def _local_concurrency():
2270
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2271
stdout=subprocess.PIPE).communicate()[0]
2272
elif sys.platform == 'sunos5':
2273
def _local_concurrency():
2274
return subprocess.Popen(['psrinfo', '-p',],
2275
stdout=subprocess.PIPE).communicate()[0]
2276
elif sys.platform == "win32":
2277
def _local_concurrency():
2278
# This appears to return the number of cores.
2279
return os.environ.get('NUMBER_OF_PROCESSORS')
2281
def _local_concurrency():
2286
_cached_local_concurrency = None
2288
def local_concurrency(use_cache=True):
2289
"""Return how many processes can be run concurrently.
2291
Rely on platform specific implementations and default to 1 (one) if
2292
anything goes wrong.
2294
global _cached_local_concurrency
2296
if _cached_local_concurrency is not None and use_cache:
2297
return _cached_local_concurrency
2299
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2300
if concurrency is None:
2302
import multiprocessing
2304
# multiprocessing is only available on Python >= 2.6
2306
concurrency = _local_concurrency()
2307
except (OSError, IOError):
2310
concurrency = multiprocessing.cpu_count()
2312
concurrency = int(concurrency)
2313
except (TypeError, ValueError):
2316
_cached_concurrency = concurrency
2320
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2321
"""A stream writer that doesn't decode str arguments."""
2323
def __init__(self, encode, stream, errors='strict'):
2324
codecs.StreamWriter.__init__(self, stream, errors)
2325
self.encode = encode
2327
def write(self, object):
2328
if type(object) is str:
2329
self.stream.write(object)
2331
data, _ = self.encode(object, self.errors)
2332
self.stream.write(data)
2334
if sys.platform == 'win32':
2335
def open_file(filename, mode='r', bufsize=-1):
2336
"""This function is used to override the ``open`` builtin.
2338
But it uses O_NOINHERIT flag so the file handle is not inherited by
2339
child processes. Deleting or renaming a closed file opened with this
2340
function is not blocking child processes.
2342
writing = 'w' in mode
2343
appending = 'a' in mode
2344
updating = '+' in mode
2345
binary = 'b' in mode
2348
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2349
# for flags for each modes.
2359
flags |= os.O_WRONLY
2360
flags |= os.O_CREAT | os.O_TRUNC
2365
flags |= os.O_WRONLY
2366
flags |= os.O_CREAT | os.O_APPEND
2371
flags |= os.O_RDONLY
2373
return os.fdopen(os.open(filename, flags), mode, bufsize)
2378
def getuser_unicode():
2379
"""Return the username as unicode.
2382
user_encoding = get_user_encoding()
2383
username = getpass.getuser().decode(user_encoding)
2384
except UnicodeDecodeError:
2385
raise errors.BzrError("Can't decode username as %s." % \
2387
except ImportError, e:
2388
if sys.platform != 'win32':
2390
if str(e) != 'No module named pwd':
2392
# https://bugs.launchpad.net/bzr/+bug/660174
2393
# getpass.getuser() is unable to return username on Windows
2394
# if there is no USERNAME environment variable set.
2395
# That could be true if bzr is running as a service,
2396
# e.g. running `bzr serve` as a service on Windows.
2397
# We should not fail with traceback in this case.
2398
username = u'UNKNOWN'
2402
def available_backup_name(base, exists):
2403
"""Find a non-existing backup file name.
2405
This will *not* create anything, this only return a 'free' entry. This
2406
should be used for checking names in a directory below a locked
2407
tree/branch/repo to avoid race conditions. This is LBYL (Look Before You
2408
Leap) and generally discouraged.
2410
:param base: The base name.
2412
:param exists: A callable returning True if the path parameter exists.
2415
name = "%s.~%d~" % (base, counter)
2418
name = "%s.~%d~" % (base, counter)
2422
def set_fd_cloexec(fd):
2423
"""Set a Unix file descriptor's FD_CLOEXEC flag. Do nothing if platform
2424
support for this is not available.
2428
old = fcntl.fcntl(fd, fcntl.F_GETFD)
2429
fcntl.fcntl(fd, fcntl.F_SETFD, old | fcntl.FD_CLOEXEC)
2430
except (ImportError, AttributeError):
2431
# Either the fcntl module or specific constants are not present
2435
def find_executable_on_path(name):
2436
"""Finds an executable on the PATH.
2438
On Windows, this will try to append each extension in the PATHEXT
2439
environment variable to the name, if it cannot be found with the name
2442
:param name: The base name of the executable.
2443
:return: The path to the executable found or None.
2445
path = os.environ.get('PATH')
2448
path = path.split(os.pathsep)
2449
if sys.platform == 'win32':
2450
exts = os.environ.get('PATHEXT', '').split(os.pathsep)
2451
exts = [ext.lower() for ext in exts]
2452
base, ext = os.path.splitext(name)
2454
if ext.lower() not in exts:
2462
f = os.path.join(d, name) + ext
2463
if os.access(f, os.X_OK):
2468
def _posix_is_local_pid_dead(pid):
2469
"""True if pid doesn't correspond to live process on this machine"""
2471
# Special meaning of unix kill: just check if it's there.
2474
if e.errno == errno.ESRCH:
2475
# On this machine, and really not found: as sure as we can be
2478
elif e.errno == errno.EPERM:
2479
# exists, though not ours
2482
mutter("os.kill(%d, 0) failed: %s" % (pid, e))
2483
# Don't really know.
2486
# Exists and our process: not dead.
2489
if sys.platform == "win32":
2490
is_local_pid_dead = win32utils.is_local_pid_dead
2492
is_local_pid_dead = _posix_is_local_pid_dead
2495
def fdatasync(fileno):
2496
"""Flush file contents to disk if possible.
2498
:param fileno: Integer OS file handle.
2499
:raises TransportNotPossible: If flushing to disk is not possible.
2501
fn = getattr(os, 'fdatasync', getattr(os, 'fsync', None))
1466
filename = pathjoin(base, resource_relpath)
1467
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
