48
48
from bzrlib import (
54
56
from bzrlib.symbol_versioning import (
55
57
deprecated_function,
58
from bzrlib.trace import mutter
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
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
61
88
# On win32, O_BINARY is used to indicate the file should
62
89
# be opened in binary mode, rather than text mode.
63
90
# On other platforms, O_BINARY doesn't exist, because
64
91
# they always open in binary mode, so it is okay to
65
# OR with 0 on those platforms
92
# OR with 0 on those platforms.
93
# O_NOINHERIT and O_TEXT exists only on win32 too.
66
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 "
69
108
def make_readonly(filename):
70
109
"""Make a filename read-only."""
71
mod = os.stat(filename).st_mode
73
os.chmod(filename, mod)
110
mod = os.lstat(filename).st_mode
111
if not stat.S_ISLNK(mod):
113
os.chmod(filename, mod)
76
116
def make_writable(filename):
77
mod = os.stat(filename).st_mode
79
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)
278
360
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
363
def _add_rename_error_details(e, old, new):
364
new_e = OSError(e.errno, "failed to rename %s to %s: %s"
365
% (old, new, e.strerror))
367
new_e.to_filename = new
281
371
def _win32_rename(old, new):
282
372
"""We expect to be able to atomically replace 'new' with old.
284
374
On win32, if new exists, it must be moved out of the way first,
288
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
378
fancy_rename(old, new, rename_func=_wrapped_rename, unlink_func=os.unlink)
289
379
except OSError, e:
290
380
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
291
# If we try to rename a non-existant file onto cwd, we get
292
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
381
# If we try to rename a non-existant file onto cwd, we get
382
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
293
383
# if the old path doesn't exist, sometimes we get EACCES
294
384
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
389
def _wrapped_rename(old, new):
390
"""Rename a file or directory"""
393
except (IOError, OSError), e:
394
# this is eventually called by all rename-like functions, so should
396
raise _add_rename_error_details(e, old, new)
299
399
def _mac_getcwd():
300
return unicodedata.normalize('NFKC', os.getcwdu())
400
return unicodedata.normalize('NFC', os.getcwdu())
303
403
# Default is to just use the python builtins, but these can be rebound on
475
575
for dirname in dir_list:
476
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
482
def pumpfile(fromfile, tofile):
483
"""Copy contents of one file to another."""
486
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)
492
647
def file_iterator(input_file, readsize=32768):
548
731
def local_time_offset(t=None):
549
732
"""Return offset of local zone from GMT, either at present or at time t."""
550
# python2.3 localtime() can't take None
554
if time.localtime(t).tm_isdst and time.daylight:
557
return -time.timezone
560
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,
561
743
show_offset=True):
562
## TODO: Perhaps a global option to use either universal or local time?
563
## Or perhaps just let people set $TZ?
564
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):
566
807
if timezone == 'utc':
567
808
tt = time.gmtime(t)
704
assert isinstance(p, list)
706
944
if (f == '..') or (f is None) or (f == ''):
707
945
raise errors.BzrError("sorry, %r not allowed in path" % f)
708
946
return pathjoin(*p)
711
@deprecated_function(zero_nine)
712
def appendpath(p1, p2):
716
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
719
1016
def split_lines(s):
720
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.
721
1032
lines = s.split('\n')
722
1033
result = [line + '\n' for line in lines[:-1]]
786
1155
os.path.commonprefix (python2.4) has a bad bug that it works just
787
1156
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
788
1157
avoids that problem.
1159
NOTE: `base` should not have a trailing slash otherwise you'll get
1160
PathNotChild exceptions regardless of `path`.
791
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
792
' exceed the platform minimum length (which is %d)' %
1163
if len(base) < MIN_ABS_PATHLENGTH:
1164
# must have space for e.g. a drive letter
1165
raise ValueError('%r is too short to calculate a relative path'
795
1168
rp = abspath(path)
799
while len(head) >= len(base):
1173
if len(head) <= len(base) and head != base:
1174
raise errors.PathNotChild(rp, base)
800
1175
if head == base:
802
head, tail = os.path.split(head)
1177
head, tail = split(head)
806
raise errors.PathNotChild(rp, base)
1182
return pathjoin(*reversed(s))
1187
def _cicp_canonical_relpath(base, path):
1188
"""Return the canonical path relative to base.
1190
Like relpath, but on case-insensitive-case-preserving file-systems, this
1191
will return the relpath as stored on the file-system rather than in the
1192
case specified in the input string, for all existing portions of the path.
1194
This will cause O(N) behaviour if called for every path in a tree; if you
1195
have a number of paths to convert, you should use canonical_relpaths().
1197
# TODO: it should be possible to optimize this for Windows by using the
1198
# win32 API FindFiles function to look for the specified name - but using
1199
# os.listdir() still gives us the correct, platform agnostic semantics in
1202
rel = relpath(base, path)
1203
# '.' will have been turned into ''
1207
abs_base = abspath(base)
1209
_listdir = os.listdir
1211
# use an explicit iterator so we can easily consume the rest on early exit.
1212
bit_iter = iter(rel.split('/'))
1213
for bit in bit_iter:
1216
next_entries = _listdir(current)
1217
except OSError: # enoent, eperm, etc
1218
# We can't find this in the filesystem, so just append the
1220
current = pathjoin(current, bit, *list(bit_iter))
1222
for look in next_entries:
1223
if lbit == look.lower():
1224
current = pathjoin(current, look)
1227
# got to the end, nothing matched, so we just return the
1228
# non-existing bits as they were specified (the filename may be
1229
# the target of a move, for example).
1230
current = pathjoin(current, bit, *list(bit_iter))
1232
return current[len(abs_base):].lstrip('/')
1234
# XXX - TODO - we need better detection/integration of case-insensitive
1235
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1236
# filesystems), for example, so could probably benefit from the same basic
1237
# support there. For now though, only Windows and OSX get that support, and
1238
# they get it for *all* file-systems!
1239
if sys.platform in ('win32', 'darwin'):
1240
canonical_relpath = _cicp_canonical_relpath
1242
canonical_relpath = relpath
1244
def canonical_relpaths(base, paths):
1245
"""Create an iterable to canonicalize a sequence of relative paths.
1247
The intent is for this implementation to use a cache, vastly speeding
1248
up multiple transformations in the same directory.
1250
# but for now, we haven't optimized...
1251
return [canonical_relpath(base, p) for p in paths]
814
1253
def safe_unicode(unicode_or_utf8_string):
815
1254
"""Coerce unicode_or_utf8_string into unicode.
817
1256
If it is unicode, it is returned.
818
Otherwise it is decoded from utf-8. If a decoding error
819
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
820
as a BzrBadParameter exception.
1257
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1258
wrapped in a BzrBadParameterNotUnicode exception.
822
1260
if isinstance(unicode_or_utf8_string, unicode):
823
1261
return unicode_or_utf8_string
827
1265
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1268
def safe_utf8(unicode_or_utf8_string):
1269
"""Coerce unicode_or_utf8_string to a utf8 string.
1271
If it is a str, it is returned.
1272
If it is Unicode, it is encoded into a utf-8 string.
1274
if isinstance(unicode_or_utf8_string, str):
1275
# TODO: jam 20070209 This is overkill, and probably has an impact on
1276
# performance if we are dealing with lots of apis that want a
1279
# Make sure it is a valid utf-8 string
1280
unicode_or_utf8_string.decode('utf-8')
1281
except UnicodeDecodeError:
1282
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1283
return unicode_or_utf8_string
1284
return unicode_or_utf8_string.encode('utf-8')
1287
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1288
' Revision id generators should be creating utf8'
1292
def safe_revision_id(unicode_or_utf8_string, warn=True):
1293
"""Revision ids should now be utf8, but at one point they were unicode.
1295
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1297
:param warn: Functions that are sanitizing user data can set warn=False
1298
:return: None or a utf8 revision id.
1300
if (unicode_or_utf8_string is None
1301
or unicode_or_utf8_string.__class__ == str):
1302
return unicode_or_utf8_string
1304
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1306
return cache_utf8.encode(unicode_or_utf8_string)
1309
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1310
' generators should be creating utf8 file ids.')
1313
def safe_file_id(unicode_or_utf8_string, warn=True):
1314
"""File ids should now be utf8, but at one point they were unicode.
1316
This is the same as safe_utf8, except it uses the cached encode functions
1317
to save a little bit of performance.
1319
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1321
:param warn: Functions that are sanitizing user data can set warn=False
1322
:return: None or a utf8 file id.
1324
if (unicode_or_utf8_string is None
1325
or unicode_or_utf8_string.__class__ == str):
1326
return unicode_or_utf8_string
1328
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1330
return cache_utf8.encode(unicode_or_utf8_string)
830
1333
_platform_normalizes_filenames = False
831
1334
if sys.platform == 'darwin':
832
1335
_platform_normalizes_filenames = True
871
1374
normalized_filename = _inaccessible_normalized_filename
1377
def set_signal_handler(signum, handler, restart_syscall=True):
1378
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1379
on platforms that support that.
1381
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1382
automatically restart (by calling `signal.siginterrupt(signum,
1383
False)`). May be ignored if the feature is not available on this
1384
platform or Python version.
1388
siginterrupt = signal.siginterrupt
1390
# This python implementation doesn't provide signal support, hence no
1393
except AttributeError:
1394
# siginterrupt doesn't exist on this platform, or for this version
1396
siginterrupt = lambda signum, flag: None
1398
def sig_handler(*args):
1399
# Python resets the siginterrupt flag when a signal is
1400
# received. <http://bugs.python.org/issue8354>
1401
# As a workaround for some cases, set it back the way we want it.
1402
siginterrupt(signum, False)
1403
# Now run the handler function passed to set_signal_handler.
1406
sig_handler = handler
1407
old_handler = signal.signal(signum, sig_handler)
1409
siginterrupt(signum, False)
1413
default_terminal_width = 80
1414
"""The default terminal width for ttys.
1416
This is defined so that higher levels can share a common fallback value when
1417
terminal_width() returns None.
874
1421
def terminal_width():
875
"""Return estimated terminal width."""
876
if sys.platform == 'win32':
877
import bzrlib.win32console
878
return bzrlib.win32console.get_console_size()[0]
1422
"""Return terminal width.
1424
None is returned if the width can't established precisely.
1427
- if BZR_COLUMNS is set, returns its value
1428
- if there is no controlling terminal, returns None
1429
- if COLUMNS is set, returns its value,
1431
From there, we need to query the OS to get the size of the controlling
1435
- get termios.TIOCGWINSZ
1436
- if an error occurs or a negative value is obtained, returns None
1440
- win32utils.get_console_size() decides,
1441
- returns None on error (provided default value)
1444
# If BZR_COLUMNS is set, take it, user is always right
1446
return int(os.environ['BZR_COLUMNS'])
1447
except (KeyError, ValueError):
1450
isatty = getattr(sys.stdout, 'isatty', None)
1451
if isatty is None or not isatty():
1452
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1455
# If COLUMNS is set, take it, the terminal knows better (even inside a
1456
# given terminal, the application can decide to set COLUMNS to a lower
1457
# value (splitted screen) or a bigger value (scroll bars))
1459
return int(os.environ['COLUMNS'])
1460
except (KeyError, ValueError):
1463
width, height = _terminal_size(None, None)
1465
# Consider invalid values as meaning no width
1471
def _win32_terminal_size(width, height):
1472
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1473
return width, height
1476
def _ioctl_terminal_size(width, height):
881
1478
import struct, fcntl, termios
882
1479
s = struct.pack('HHHH', 0, 0, 0, 0)
883
1480
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
884
width = struct.unpack('HHHH', x)[1]
1481
height, width = struct.unpack('HHHH', x)[0:2]
1482
except (IOError, AttributeError):
1484
return width, height
1486
_terminal_size = None
1487
"""Returns the terminal size as (width, height).
1489
:param width: Default value for width.
1490
:param height: Default value for height.
1492
This is defined specifically for each OS and query the size of the controlling
1493
terminal. If any error occurs, the provided default values should be returned.
1495
if sys.platform == 'win32':
1496
_terminal_size = _win32_terminal_size
1498
_terminal_size = _ioctl_terminal_size
1501
def _terminal_size_changed(signum, frame):
1502
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1503
width, height = _terminal_size(None, None)
1504
if width is not None:
1505
os.environ['COLUMNS'] = str(width)
1508
_registered_sigwinch = False
1509
def watch_sigwinch():
1510
"""Register for SIGWINCH, once and only once.
1512
Do nothing if the signal module is not available.
1514
global _registered_sigwinch
1515
if not _registered_sigwinch:
889
width = int(os.environ['COLUMNS'])
1518
if getattr(signal, "SIGWINCH", None) is not None:
1519
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1521
# python doesn't provide signal support, nothing we can do about it
1523
_registered_sigwinch = True
898
1526
def supports_executable():
899
1527
return sys.platform != "win32"
1530
def supports_posix_readonly():
1531
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1533
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1534
directory controls creation/deletion, etc.
1536
And under win32, readonly means that the directory itself cannot be
1537
deleted. The contents of a readonly directory can be changed, unlike POSIX
1538
where files in readonly directories cannot be added, deleted or renamed.
1540
return sys.platform != "win32"
902
1543
def set_or_unset_env(env_variable, value):
903
1544
"""Modify the environment, setting or removing the env_variable.
932
1573
raise errors.IllegalPath(path)
1576
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1578
def _is_error_enotdir(e):
1579
"""Check if this exception represents ENOTDIR.
1581
Unfortunately, python is very inconsistent about the exception
1582
here. The cases are:
1583
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1584
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1585
which is the windows error code.
1586
3) Windows, Python2.5 uses errno == EINVAL and
1587
winerror == ERROR_DIRECTORY
1589
:param e: An Exception object (expected to be OSError with an errno
1590
attribute, but we should be able to cope with anything)
1591
:return: True if this represents an ENOTDIR error. False otherwise.
1593
en = getattr(e, 'errno', None)
1594
if (en == errno.ENOTDIR
1595
or (sys.platform == 'win32'
1596
and (en == _WIN32_ERROR_DIRECTORY
1597
or (en == errno.EINVAL
1598
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
935
1604
def walkdirs(top, prefix=""):
936
1605
"""Yield data about all the directories in a tree.
938
1607
This yields all the data about the contents of a directory at a time.
939
1608
After each directory has been yielded, if the caller has mutated the list
940
1609
to exclude some directories, they are then not descended into.
942
1611
The data yielded is of the form:
943
1612
((directory-relpath, directory-path-from-top),
944
[(relpath, basename, kind, lstat), ...]),
1613
[(relpath, basename, kind, lstat, path-from-top), ...]),
945
1614
- directory-relpath is the relative path of the directory being returned
946
1615
with respect to top. prefix is prepended to this.
947
- directory-path-from-root is the path including top for this directory.
1616
- directory-path-from-root is the path including top for this directory.
948
1617
It is suitable for use with os functions.
949
1618
- relpath is the relative path within the subtree being walked.
950
1619
- basename is the basename of the path
952
1621
present within the tree - but it may be recorded as versioned. See
954
1623
- lstat is the stat data *if* the file was statted.
955
- planned, not implemented:
1624
- planned, not implemented:
956
1625
path_from_tree_root is the path from the root of the tree.
958
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1627
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
959
1628
allows one to walk a subtree but get paths that are relative to a tree
960
1629
rooted higher up.
961
1630
:return: an iterator over the dirs.
963
1632
#TODO there is a bit of a smell where the results of the directory-
964
# summary in this, and the path from the root, may not agree
1633
# summary in this, and the path from the root, may not agree
965
1634
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
966
1635
# potentially confusing output. We should make this more robust - but
967
1636
# not at a speed cost. RBC 20060731
970
1638
_directory = _directory_kind
971
1639
_listdir = os.listdir
972
pending = [(prefix, "", _directory, None, top)]
1640
_kind_from_mode = file_kind_from_stat_mode
1641
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
975
currentdir = pending.pop()
976
1643
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
979
relroot = currentdir[0] + '/'
1644
relroot, _, _, _, top = pending.pop()
1646
relprefix = relroot + u'/'
1649
top_slash = top + u'/'
1652
append = dirblock.append
1654
names = sorted(_listdir(top))
1656
if not _is_error_enotdir(e):
1660
abspath = top_slash + name
1661
statvalue = _lstat(abspath)
1662
kind = _kind_from_mode(statvalue.st_mode)
1663
append((relprefix + name, name, kind, statvalue, abspath))
1664
yield (relroot, top), dirblock
1666
# push the user specified dirs from dirblock
1667
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1670
class DirReader(object):
1671
"""An interface for reading directories."""
1673
def top_prefix_to_starting_dir(self, top, prefix=""):
1674
"""Converts top and prefix to a starting dir entry
1676
:param top: A utf8 path
1677
:param prefix: An optional utf8 path to prefix output relative paths
1679
:return: A tuple starting with prefix, and ending with the native
1682
raise NotImplementedError(self.top_prefix_to_starting_dir)
1684
def read_dir(self, prefix, top):
1685
"""Read a specific dir.
1687
:param prefix: A utf8 prefix to be preprended to the path basenames.
1688
:param top: A natively encoded path to read.
1689
:return: A list of the directories contents. Each item contains:
1690
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1692
raise NotImplementedError(self.read_dir)
1695
_selected_dir_reader = None
1698
def _walkdirs_utf8(top, prefix=""):
1699
"""Yield data about all the directories in a tree.
1701
This yields the same information as walkdirs() only each entry is yielded
1702
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1703
are returned as exact byte-strings.
1705
:return: yields a tuple of (dir_info, [file_info])
1706
dir_info is (utf8_relpath, path-from-top)
1707
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1708
if top is an absolute path, path-from-top is also an absolute path.
1709
path-from-top might be unicode or utf8, but it is the correct path to
1710
pass to os functions to affect the file in question. (such as os.lstat)
1712
global _selected_dir_reader
1713
if _selected_dir_reader is None:
1714
fs_encoding = _fs_enc.upper()
1715
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1716
# Win98 doesn't have unicode apis like FindFirstFileW
1717
# TODO: We possibly could support Win98 by falling back to the
1718
# original FindFirstFile, and using TCHAR instead of WCHAR,
1719
# but that gets a bit tricky, and requires custom compiling
1722
from bzrlib._walkdirs_win32 import Win32ReadDir
1723
_selected_dir_reader = Win32ReadDir()
1726
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1727
# ANSI_X3.4-1968 is a form of ASCII
1729
from bzrlib._readdir_pyx import UTF8DirReader
1730
_selected_dir_reader = UTF8DirReader()
1731
except ImportError, e:
1732
failed_to_load_extension(e)
1735
if _selected_dir_reader is None:
1736
# Fallback to the python version
1737
_selected_dir_reader = UnicodeDirReader()
1739
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1740
# But we don't actually uses 1-3 in pending, so set them to None
1741
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1742
read_dir = _selected_dir_reader.read_dir
1743
_directory = _directory_kind
1745
relroot, _, _, _, top = pending[-1].pop()
1748
dirblock = sorted(read_dir(relroot, top))
1749
yield (relroot, top), dirblock
1750
# push the user specified dirs from dirblock
1751
next = [d for d in reversed(dirblock) if d[2] == _directory]
1753
pending.append(next)
1756
class UnicodeDirReader(DirReader):
1757
"""A dir reader for non-utf8 file systems, which transcodes."""
1759
__slots__ = ['_utf8_encode']
1762
self._utf8_encode = codecs.getencoder('utf8')
1764
def top_prefix_to_starting_dir(self, top, prefix=""):
1765
"""See DirReader.top_prefix_to_starting_dir."""
1766
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1768
def read_dir(self, prefix, top):
1769
"""Read a single directory from a non-utf8 file system.
1771
top, and the abspath element in the output are unicode, all other paths
1772
are utf8. Local disk IO is done via unicode calls to listdir etc.
1774
This is currently the fallback code path when the filesystem encoding is
1775
not UTF-8. It may be better to implement an alternative so that we can
1776
safely handle paths that are not properly decodable in the current
1779
See DirReader.read_dir for details.
1781
_utf8_encode = self._utf8_encode
1783
_listdir = os.listdir
1784
_kind_from_mode = file_kind_from_stat_mode
1787
relprefix = prefix + '/'
1790
top_slash = top + u'/'
1793
append = dirblock.append
982
1794
for name in sorted(_listdir(top)):
983
abspath = top + '/' + name
984
statvalue = lstat(abspath)
985
dirblock.append((relroot + name, name,
986
file_kind_from_stat_mode(statvalue.st_mode),
988
yield (currentdir[0], top), dirblock
989
# push the user specified dirs from dirblock
990
for dir in reversed(dirblock):
991
if dir[2] == _directory:
1796
name_utf8 = _utf8_encode(name)[0]
1797
except UnicodeDecodeError:
1798
raise errors.BadFilenameEncoding(
1799
_utf8_encode(relprefix)[0] + name, _fs_enc)
1800
abspath = top_slash + name
1801
statvalue = _lstat(abspath)
1802
kind = _kind_from_mode(statvalue.st_mode)
1803
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
995
1807
def copy_tree(from_path, to_path, handlers={}):
996
1808
"""Copy all of the entries in from_path into to_path.
998
:param from_path: The base directory to copy.
1810
:param from_path: The base directory to copy.
999
1811
:param to_path: The target directory. If it does not exist, it will
1001
1813
:param handlers: A dictionary of functions, which takes a source and
1052
1886
_cached_user_encoding = None
1055
def get_user_encoding():
1889
def get_user_encoding(use_cache=True):
1056
1890
"""Find out what the preferred user encoding is.
1058
1892
This is generally the encoding that is used for command line parameters
1059
1893
and file contents. This may be different from the terminal encoding
1060
1894
or the filesystem encoding.
1896
:param use_cache: Enable cache for detected encoding.
1897
(This parameter is turned on by default,
1898
and required only for selftesting)
1062
1900
:return: A string defining the preferred user encoding
1064
1902
global _cached_user_encoding
1065
if _cached_user_encoding is not None:
1903
if _cached_user_encoding is not None and use_cache:
1066
1904
return _cached_user_encoding
1068
1906
if sys.platform == 'darwin':
1069
# work around egregious python 2.4 bug
1907
# python locale.getpreferredencoding() always return
1908
# 'mac-roman' on darwin. That's a lie.
1070
1909
sys.platform = 'posix'
1911
if os.environ.get('LANG', None) is None:
1912
# If LANG is not set, we end up with 'ascii', which is bad
1913
# ('mac-roman' is more than ascii), so we set a default which
1914
# will give us UTF-8 (which appears to work in all cases on
1915
# OSX). Users are still free to override LANG of course, as
1916
# long as it give us something meaningful. This work-around
1917
# *may* not be needed with python 3k and/or OSX 10.5, but will
1918
# work with them too -- vila 20080908
1919
os.environ['LANG'] = 'en_US.UTF-8'
1074
1922
sys.platform = 'darwin'
1084
1932
' doesn\'t support the locale set by $LANG (%s)\n'
1085
1933
" Continuing with ascii encoding.\n"
1086
1934
% (e, os.environ.get('LANG')))
1088
if _cached_user_encoding is None:
1089
_cached_user_encoding = 'ascii'
1090
return _cached_user_encoding
1093
def recv_all(socket, bytes):
1935
user_encoding = 'ascii'
1937
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1938
# treat that as ASCII, and not support printing unicode characters to the
1941
# For python scripts run under vim, we get '', so also treat that as ASCII
1942
if user_encoding in (None, 'cp0', ''):
1943
user_encoding = 'ascii'
1947
codecs.lookup(user_encoding)
1949
sys.stderr.write('bzr: warning:'
1950
' unknown encoding %s.'
1951
' Continuing with ascii encoding.\n'
1954
user_encoding = 'ascii'
1957
_cached_user_encoding = user_encoding
1959
return user_encoding
1962
def get_host_name():
1963
"""Return the current unicode host name.
1965
This is meant to be used in place of socket.gethostname() because that
1966
behaves inconsistently on different platforms.
1968
if sys.platform == "win32":
1970
return win32utils.get_host_name()
1973
return socket.gethostname().decode(get_user_encoding())
1976
# We must not read/write any more than 64k at a time from/to a socket so we
1977
# don't risk "no buffer space available" errors on some platforms. Windows in
1978
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1980
MAX_SOCKET_CHUNK = 64 * 1024
1982
def read_bytes_from_socket(sock, report_activity=None,
1983
max_read_size=MAX_SOCKET_CHUNK):
1984
"""Read up to max_read_size of bytes from sock and notify of progress.
1986
Translates "Connection reset by peer" into file-like EOF (return an
1987
empty string rather than raise an error), and repeats the recv if
1988
interrupted by a signal.
1992
bytes = sock.recv(max_read_size)
1993
except socket.error, e:
1995
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1996
# The connection was closed by the other side. Callers expect
1997
# an empty string to signal end-of-stream.
1999
elif eno == errno.EINTR:
2000
# Retry the interrupted recv.
2004
if report_activity is not None:
2005
report_activity(len(bytes), 'read')
2009
def recv_all(socket, count):
1094
2010
"""Receive an exact number of bytes.
1096
2012
Regular Socket.recv() may return less than the requested number of bytes,
1097
dependning on what's in the OS buffer. MSG_WAITALL is not available
2013
depending on what's in the OS buffer. MSG_WAITALL is not available
1098
2014
on all platforms, but this should work everywhere. This will return
1099
2015
less than the requested amount if the remote end closes.
1101
2017
This isn't optimized and is intended mostly for use in testing.
1104
while len(b) < bytes:
1105
new = socket.recv(bytes - len(b))
2020
while len(b) < count:
2021
new = read_bytes_from_socket(socket, None, count - len(b))
2028
def send_all(sock, bytes, report_activity=None):
2029
"""Send all bytes on a socket.
2031
Breaks large blocks in smaller chunks to avoid buffering limitations on
2032
some platforms, and catches EINTR which may be thrown if the send is
2033
interrupted by a signal.
2035
This is preferred to socket.sendall(), because it avoids portability bugs
2036
and provides activity reporting.
2038
:param report_activity: Call this as bytes are read, see
2039
Transport._report_activity
2042
byte_count = len(bytes)
2043
while sent_total < byte_count:
2045
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2046
except socket.error, e:
2047
if e.args[0] != errno.EINTR:
2051
report_activity(sent, 'write')
1111
2054
def dereference_path(path):
1112
2055
"""Determine the real path to a file.
1120
2063
# The pathjoin for '.' is a workaround for Python bug #1213894.
1121
2064
# (initial path components aren't dereferenced)
1122
2065
return pathjoin(realpath(pathjoin('.', parent)), base)
2068
def supports_mapi():
2069
"""Return True if we can use MAPI to launch a mail client."""
2070
return sys.platform == "win32"
2073
def resource_string(package, resource_name):
2074
"""Load a resource from a package and return it as a string.
2076
Note: Only packages that start with bzrlib are currently supported.
2078
This is designed to be a lightweight implementation of resource
2079
loading in a way which is API compatible with the same API from
2081
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2082
If and when pkg_resources becomes a standard library, this routine
2085
# Check package name is within bzrlib
2086
if package == "bzrlib":
2087
resource_relpath = resource_name
2088
elif package.startswith("bzrlib."):
2089
package = package[len("bzrlib."):].replace('.', os.sep)
2090
resource_relpath = pathjoin(package, resource_name)
2092
raise errors.BzrError('resource package %s not in bzrlib' % package)
2094
# Map the resource to a file and read its contents
2095
base = dirname(bzrlib.__file__)
2096
if getattr(sys, 'frozen', None): # bzr.exe
2097
base = abspath(pathjoin(base, '..', '..'))
2098
filename = pathjoin(base, resource_relpath)
2099
return open(filename, 'rU').read()
2102
def file_kind_from_stat_mode_thunk(mode):
2103
global file_kind_from_stat_mode
2104
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2106
from bzrlib._readdir_pyx import UTF8DirReader
2107
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2108
except ImportError, e:
2109
# This is one time where we won't warn that an extension failed to
2110
# load. The extension is never available on Windows anyway.
2111
from bzrlib._readdir_py import (
2112
_kind_from_mode as file_kind_from_stat_mode
2114
return file_kind_from_stat_mode(mode)
2115
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2118
def file_kind(f, _lstat=os.lstat):
2120
return file_kind_from_stat_mode(_lstat(f).st_mode)
2122
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2123
raise errors.NoSuchFile(f)
2127
def until_no_eintr(f, *a, **kw):
2128
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2130
WARNING: you must be certain that it is safe to retry the call repeatedly
2131
if EINTR does occur. This is typically only true for low-level operations
2132
like os.read. If in any doubt, don't use this.
2134
Keep in mind that this is not a complete solution to EINTR. There is
2135
probably code in the Python standard library and other dependencies that
2136
may encounter EINTR if a signal arrives (and there is signal handler for
2137
that signal). So this function can reduce the impact for IO that bzrlib
2138
directly controls, but it is not a complete solution.
2140
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2144
except (IOError, OSError), e:
2145
if e.errno == errno.EINTR:
2150
def re_compile_checked(re_string, flags=0, where=""):
2151
"""Return a compiled re, or raise a sensible error.
2153
This should only be used when compiling user-supplied REs.
2155
:param re_string: Text form of regular expression.
2156
:param flags: eg re.IGNORECASE
2157
:param where: Message explaining to the user the context where
2158
it occurred, eg 'log search filter'.
2160
# from https://bugs.launchpad.net/bzr/+bug/251352
2162
re_obj = re.compile(re_string, flags)
2167
where = ' in ' + where
2168
# despite the name 'error' is a type
2169
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2170
% (where, re_string, e))
2173
if sys.platform == "win32":
2176
return msvcrt.getch()
2181
fd = sys.stdin.fileno()
2182
settings = termios.tcgetattr(fd)
2185
ch = sys.stdin.read(1)
2187
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2191
if sys.platform == 'linux2':
2192
def _local_concurrency():
2194
prefix = 'processor'
2195
for line in file('/proc/cpuinfo', 'rb'):
2196
if line.startswith(prefix):
2197
concurrency = int(line[line.find(':')+1:]) + 1
2199
elif sys.platform == 'darwin':
2200
def _local_concurrency():
2201
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2202
stdout=subprocess.PIPE).communicate()[0]
2203
elif sys.platform[0:7] == 'freebsd':
2204
def _local_concurrency():
2205
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2206
stdout=subprocess.PIPE).communicate()[0]
2207
elif sys.platform == 'sunos5':
2208
def _local_concurrency():
2209
return subprocess.Popen(['psrinfo', '-p',],
2210
stdout=subprocess.PIPE).communicate()[0]
2211
elif sys.platform == "win32":
2212
def _local_concurrency():
2213
# This appears to return the number of cores.
2214
return os.environ.get('NUMBER_OF_PROCESSORS')
2216
def _local_concurrency():
2221
_cached_local_concurrency = None
2223
def local_concurrency(use_cache=True):
2224
"""Return how many processes can be run concurrently.
2226
Rely on platform specific implementations and default to 1 (one) if
2227
anything goes wrong.
2229
global _cached_local_concurrency
2231
if _cached_local_concurrency is not None and use_cache:
2232
return _cached_local_concurrency
2234
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2235
if concurrency is None:
2237
concurrency = _local_concurrency()
2238
except (OSError, IOError):
2241
concurrency = int(concurrency)
2242
except (TypeError, ValueError):
2245
_cached_concurrency = concurrency
2249
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2250
"""A stream writer that doesn't decode str arguments."""
2252
def __init__(self, encode, stream, errors='strict'):
2253
codecs.StreamWriter.__init__(self, stream, errors)
2254
self.encode = encode
2256
def write(self, object):
2257
if type(object) is str:
2258
self.stream.write(object)
2260
data, _ = self.encode(object, self.errors)
2261
self.stream.write(data)
2263
if sys.platform == 'win32':
2264
def open_file(filename, mode='r', bufsize=-1):
2265
"""This function is used to override the ``open`` builtin.
2267
But it uses O_NOINHERIT flag so the file handle is not inherited by
2268
child processes. Deleting or renaming a closed file opened with this
2269
function is not blocking child processes.
2271
writing = 'w' in mode
2272
appending = 'a' in mode
2273
updating = '+' in mode
2274
binary = 'b' in mode
2277
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2278
# for flags for each modes.
2288
flags |= os.O_WRONLY
2289
flags |= os.O_CREAT | os.O_TRUNC
2294
flags |= os.O_WRONLY
2295
flags |= os.O_CREAT | os.O_APPEND
2300
flags |= os.O_RDONLY
2302
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
