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
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
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
29
31
from ntpath import (abspath as _nt_abspath,
48
49
from bzrlib import (
57
from bzrlib import symbol_versioning
56
58
from bzrlib.symbol_versioning import (
57
59
deprecated_function,
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
from bzrlib.trace import mutter
88
64
# On win32, O_BINARY is used to indicate the file should
89
65
# be opened in binary mode, rather than text mode.
90
66
# On other platforms, O_BINARY doesn't exist, because
91
67
# they always open in binary mode, so it is okay to
92
# OR with 0 on those platforms.
93
# O_NOINHERIT and O_TEXT exists only on win32 too.
68
# OR with 0 on those platforms
94
69
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 "
108
72
def make_readonly(filename):
208
195
def fancy_rename(old, new, rename_func, unlink_func):
209
196
"""A fancy rename, when you don't have atomic rename.
211
198
:param old: The old path, to rename from
212
199
:param new: The new path, to rename to
213
200
:param rename_func: The potentially non-atomic rename function
214
:param unlink_func: A way to delete the target file if the full rename
201
:param unlink_func: A way to delete the target file if the full rename succeeds
217
204
# sftp rename doesn't allow overwriting, so play tricks:
218
206
base = os.path.basename(new)
219
207
dirname = os.path.dirname(new)
220
# callers use different encodings for the paths so the following MUST
221
# respect that. We rely on python upcasting to unicode if new is unicode
222
# and keeping a str if not.
223
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
224
os.getpid(), rand_chars(10))
208
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
225
209
tmp_name = pathjoin(dirname, tmp_name)
227
211
# Rename the file out of the way, but keep track if it didn't exist
360
340
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
371
343
def _win32_rename(old, new):
372
344
"""We expect to be able to atomically replace 'new' with old.
374
346
On win32, if new exists, it must be moved out of the way first,
378
fancy_rename(old, new, rename_func=_wrapped_rename, unlink_func=os.unlink)
350
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
379
351
except OSError, e:
380
352
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
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
353
# If we try to rename a non-existant file onto cwd, we get
354
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
383
355
# if the old path doesn't exist, sometimes we get EACCES
384
356
# 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)
399
361
def _mac_getcwd():
400
362
return unicodedata.normalize('NFC', os.getcwdu())
581
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
582
report_activity=None, direction='read'):
532
def pumpfile(fromfile, tofile):
583
533
"""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
535
: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
618
block = from_file.read(buff_size)
622
if report_activity is not None:
623
report_activity(len(block), direction)
540
b = fromfile.read(BUFSIZE)
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)
647
548
def file_iterator(input_file, readsize=32768):
649
550
b = input_file.read(readsize)
748
628
:param timezone: How to display the time: 'utc', 'original' for the
749
629
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):
631
:param show_offset: Whether to append the timezone.
632
:param date_fmt: strftime format.
807
634
if timezone == 'utc':
808
635
tt = time.gmtime(t)
946
772
return pathjoin(*p)
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
1016
775
def split_lines(s):
1017
776
"""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.
1032
777
lines = s.split('\n')
1033
778
result = [line + '\n' for line in lines[:-1]]
1053
798
shutil.copyfile(src, dest)
801
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
802
# Forgiveness than Permission (EAFP) because:
803
# - root can damage a solaris file system by using unlink,
804
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
805
# EACCES, OSX: EPERM) when invoked on a directory.
1056
806
def delete_any(path):
1057
"""Delete a file, symlink or directory.
1059
Will delete even if readonly.
1062
_delete_file_or_dir(path)
1063
except (OSError, IOError), e:
1064
if e.errno in (errno.EPERM, errno.EACCES):
1065
# make writable and try again
1068
except (OSError, IOError):
1070
_delete_file_or_dir(path)
1075
def _delete_file_or_dir(path):
1076
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1077
# Forgiveness than Permission (EAFP) because:
1078
# - root can damage a solaris file system by using unlink,
1079
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1080
# EACCES, OSX: EPERM) when invoked on a directory.
807
"""Delete a file or directory."""
1081
808
if isdir(path): # Takes care of symlinks
1173
if len(head) <= len(base) and head != base:
1174
raise errors.PathNotChild(rp, base)
877
while len(head) >= len(base):
1175
878
if head == base:
1177
head, tail = split(head)
880
head, tail = os.path.split(head)
884
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]
1253
892
def safe_unicode(unicode_or_utf8_string):
1254
893
"""Coerce unicode_or_utf8_string into unicode.
1256
895
If it is unicode, it is returned.
1257
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1258
wrapped in a BzrBadParameterNotUnicode exception.
896
Otherwise it is decoded from utf-8. If a decoding error
897
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
898
as a BzrBadParameter exception.
1260
900
if isinstance(unicode_or_utf8_string, unicode):
1261
901
return unicode_or_utf8_string
1374
1014
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.
1421
1017
def terminal_width():
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):
1018
"""Return estimated terminal width."""
1019
if sys.platform == 'win32':
1020
return win32utils.get_console_size()[0]
1478
1023
import struct, fcntl, termios
1479
1024
s = struct.pack('HHHH', 0, 0, 0, 0)
1480
1025
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1481
height, width = struct.unpack('HHHH', x)[0:2]
1482
except (IOError, AttributeError):
1026
width = struct.unpack('HHHH', x)[1]
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:
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
1031
width = int(os.environ['COLUMNS'])
1523
_registered_sigwinch = True
1526
1040
def supports_executable():
1573
1087
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)
1604
1090
def walkdirs(top, prefix=""):
1605
1091
"""Yield data about all the directories in a tree.
1607
1093
This yields all the data about the contents of a directory at a time.
1608
1094
After each directory has been yielded, if the caller has mutated the list
1609
1095
to exclude some directories, they are then not descended into.
1611
1097
The data yielded is of the form:
1612
1098
((directory-relpath, directory-path-from-top),
1613
1099
[(relpath, basename, kind, lstat, path-from-top), ...]),
1614
1100
- directory-relpath is the relative path of the directory being returned
1615
1101
with respect to top. prefix is prepended to this.
1616
- directory-path-from-root is the path including top for this directory.
1102
- directory-path-from-root is the path including top for this directory.
1617
1103
It is suitable for use with os functions.
1618
1104
- relpath is the relative path within the subtree being walked.
1619
1105
- basename is the basename of the path
1621
1107
present within the tree - but it may be recorded as versioned. See
1622
1108
versioned_kind.
1623
1109
- lstat is the stat data *if* the file was statted.
1624
- planned, not implemented:
1110
- planned, not implemented:
1625
1111
path_from_tree_root is the path from the root of the tree.
1627
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1113
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1628
1114
allows one to walk a subtree but get paths that are relative to a tree
1629
1115
rooted higher up.
1630
1116
:return: an iterator over the dirs.
1632
1118
#TODO there is a bit of a smell where the results of the directory-
1633
# summary in this, and the path from the root, may not agree
1119
# summary in this, and the path from the root, may not agree
1634
1120
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1635
1121
# potentially confusing output. We should make this more robust - but
1636
1122
# not at a speed cost. RBC 20060731
1637
1123
_lstat = os.lstat
1638
1124
_directory = _directory_kind
1639
1125
_listdir = os.listdir
1640
_kind_from_mode = file_kind_from_stat_mode
1126
_kind_from_mode = _formats.get
1641
1127
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1643
1129
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1652
1138
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))
1139
for name in sorted(_listdir(top)):
1140
abspath = top_slash + name
1141
statvalue = _lstat(abspath)
1142
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1143
append((relprefix + name, name, kind, statvalue, abspath))
1664
1144
yield (relroot, top), dirblock
1666
1146
# push the user specified dirs from dirblock
1667
1147
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
1150
def _walkdirs_utf8(top, prefix=""):
1699
1151
"""Yield data about all the directories in a tree.
1709
1161
path-from-top might be unicode or utf8, but it is the correct path to
1710
1162
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()
1164
fs_encoding = _fs_enc.upper()
1165
if (sys.platform == 'win32' or
1166
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1167
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1169
return _walkdirs_fs_utf8(top, prefix=prefix)
1172
def _walkdirs_fs_utf8(top, prefix=""):
1173
"""See _walkdirs_utf8.
1175
This sub-function is called when we know the filesystem is already in utf8
1176
encoding. So we don't need to transcode filenames.
1179
_directory = _directory_kind
1180
_listdir = os.listdir
1181
_kind_from_mode = _formats.get
1739
1183
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1740
1184
# 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
1185
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1745
relroot, _, _, _, top = pending[-1].pop()
1748
dirblock = sorted(read_dir(relroot, top))
1187
relroot, _, _, _, top = pending.pop()
1189
relprefix = relroot + '/'
1192
top_slash = top + '/'
1195
append = dirblock.append
1196
for name in sorted(_listdir(top)):
1197
abspath = top_slash + name
1198
statvalue = _lstat(abspath)
1199
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1200
append((relprefix + name, name, kind, statvalue, abspath))
1749
1201
yield (relroot, top), dirblock
1750
1203
# 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 + '/'
1204
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1207
def _walkdirs_unicode_to_utf8(top, prefix=""):
1208
"""See _walkdirs_utf8
1210
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1212
This is currently the fallback code path when the filesystem encoding is
1213
not UTF-8. It may be better to implement an alternative so that we can
1214
safely handle paths that are not properly decodable in the current
1217
_utf8_encode = codecs.getencoder('utf8')
1219
_directory = _directory_kind
1220
_listdir = os.listdir
1221
_kind_from_mode = _formats.get
1223
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1225
relroot, _, _, _, top = pending.pop()
1227
relprefix = relroot + '/'
1790
1230
top_slash = top + u'/'
1793
1233
append = dirblock.append
1794
1234
for name in sorted(_listdir(top)):
1796
name_utf8 = _utf8_encode(name)[0]
1797
except UnicodeDecodeError:
1798
raise errors.BadFilenameEncoding(
1799
_utf8_encode(relprefix)[0] + name, _fs_enc)
1235
name_utf8 = _utf8_encode(name)[0]
1800
1236
abspath = top_slash + name
1801
1237
statvalue = _lstat(abspath)
1802
kind = _kind_from_mode(statvalue.st_mode)
1238
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1803
1239
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1240
yield (relroot, top), dirblock
1242
# push the user specified dirs from dirblock
1243
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1807
1246
def copy_tree(from_path, to_path, handlers={}):
1808
1247
"""Copy all of the entries in from_path into to_path.
1810
:param from_path: The base directory to copy.
1249
:param from_path: The base directory to copy.
1811
1250
:param to_path: The target directory. If it does not exist, it will
1813
1252
:param handlers: A dictionary of functions, which takes a source and
1959
1364
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):
1367
def recv_all(socket, bytes):
2010
1368
"""Receive an exact number of bytes.
2012
1370
Regular Socket.recv() may return less than the requested number of bytes,
2013
depending on what's in the OS buffer. MSG_WAITALL is not available
1371
dependning on what's in the OS buffer. MSG_WAITALL is not available
2014
1372
on all platforms, but this should work everywhere. This will return
2015
1373
less than the requested amount if the remote end closes.
2017
1375
This isn't optimized and is intended mostly for use in testing.
2020
while len(b) < count:
2021
new = read_bytes_from_socket(socket, None, count - len(b))
1378
while len(b) < bytes:
1379
new = socket.recv(bytes - len(b))
2028
def send_all(sock, bytes, report_activity=None):
1386
def send_all(socket, bytes):
2029
1387
"""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
1389
Regular socket.sendall() can give socket error 10053 on Windows. This
1390
implementation sends no more than 64k at a time, which avoids this problem.
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')
1393
for pos in xrange(0, len(bytes), chunk_size):
1394
socket.sendall(bytes[pos:pos+chunk_size])
2054
1397
def dereference_path(path):
2097
1440
base = abspath(pathjoin(base, '..', '..'))
2098
1441
filename = pathjoin(base, resource_relpath)
2099
1442
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
