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
30
31
from ntpath import (abspath as _nt_abspath,
32
33
normpath as _nt_normpath,
49
49
from bzrlib import (
57
from bzrlib import symbol_versioning
57
58
from bzrlib.symbol_versioning import (
58
59
deprecated_function,
62
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
64
if sys.version_info < (2, 5):
65
import md5 as _mod_md5
67
import sha as _mod_sha
77
from bzrlib import symbol_versioning
80
# Cross platform wall-clock time functionality with decent resolution.
81
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
82
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
83
# synchronized with ``time.time()``, this is only meant to be used to find
84
# delta times by subtracting from another call to this function.
85
timer_func = time.time
86
if sys.platform == 'win32':
87
timer_func = time.clock
61
from bzrlib.trace import mutter
89
64
# On win32, O_BINARY is used to indicate the file should
90
65
# be opened in binary mode, rather than text mode.
91
66
# On other platforms, O_BINARY doesn't exist, because
92
67
# they always open in binary mode, so it is okay to
93
# OR with 0 on those platforms.
94
# O_NOINHERIT and O_TEXT exists only on win32 too.
68
# OR with 0 on those platforms
95
69
O_BINARY = getattr(os, 'O_BINARY', 0)
96
O_TEXT = getattr(os, 'O_TEXT', 0)
97
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
100
def get_unicode_argv():
102
user_encoding = get_user_encoding()
103
return [a.decode(user_encoding) for a in sys.argv[1:]]
104
except UnicodeDecodeError:
105
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
109
72
def make_readonly(filename):
209
195
def fancy_rename(old, new, rename_func, unlink_func):
210
196
"""A fancy rename, when you don't have atomic rename.
212
198
:param old: The old path, to rename from
213
199
:param new: The new path, to rename to
214
200
:param rename_func: The potentially non-atomic rename function
215
: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
218
204
# sftp rename doesn't allow overwriting, so play tricks:
219
206
base = os.path.basename(new)
220
207
dirname = os.path.dirname(new)
221
# callers use different encodings for the paths so the following MUST
222
# respect that. We rely on python upcasting to unicode if new is unicode
223
# and keeping a str if not.
224
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
225
os.getpid(), rand_chars(10))
208
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
226
209
tmp_name = pathjoin(dirname, tmp_name)
228
211
# Rename the file out of the way, but keep track if it didn't exist
361
340
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
364
def _add_rename_error_details(e, old, new):
365
new_e = OSError(e.errno, "failed to rename %s to %s: %s"
366
% (old, new, e.strerror))
368
new_e.to_filename = new
372
343
def _win32_rename(old, new):
373
344
"""We expect to be able to atomically replace 'new' with old.
375
346
On win32, if new exists, it must be moved out of the way first,
379
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)
380
351
except OSError, e:
381
352
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
382
# If we try to rename a non-existant file onto cwd, we get
383
# 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
384
355
# if the old path doesn't exist, sometimes we get EACCES
385
356
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
390
def _wrapped_rename(old, new):
391
"""Rename a file or directory"""
394
except (IOError, OSError), e:
395
# this is eventually called by all rename-like functions, so should
397
raise _add_rename_error_details(e, old, new)
400
361
def _mac_getcwd():
401
362
return unicodedata.normalize('NFC', os.getcwdu())
582
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
583
report_activity=None, direction='read'):
532
def pumpfile(fromfile, tofile):
584
533
"""Copy contents of one file to another.
586
The read_length can either be -1 to read to end-of-file (EOF) or
587
it can specify the maximum number of bytes to read.
589
The buff_size represents the maximum size for each read operation
590
performed on from_file.
592
:param report_activity: Call this as bytes are read, see
593
Transport._report_activity
594
:param direction: Will be passed to report_activity
596
535
:return: The number of bytes copied.
600
# read specified number of bytes
602
while read_length > 0:
603
num_bytes_to_read = min(read_length, buff_size)
605
block = from_file.read(num_bytes_to_read)
609
if report_activity is not None:
610
report_activity(len(block), direction)
613
actual_bytes_read = len(block)
614
read_length -= actual_bytes_read
615
length += actual_bytes_read
619
block = from_file.read(buff_size)
623
if report_activity is not None:
624
report_activity(len(block), direction)
540
b = fromfile.read(BUFSIZE)
630
def pump_string_file(bytes, file_handle, segment_size=None):
631
"""Write bytes to file_handle in many smaller writes.
633
:param bytes: The string to write.
634
:param file_handle: The file to write to.
636
# Write data in chunks rather than all at once, because very large
637
# writes fail on some platforms (e.g. Windows with SMB mounted
640
segment_size = 5242880 # 5MB
641
segments = range(len(bytes) / segment_size + 1)
642
write = file_handle.write
643
for segment_index in segments:
644
segment = buffer(bytes, segment_index * segment_size, segment_size)
648
548
def file_iterator(input_file, readsize=32768):
650
550
b = input_file.read(readsize)
749
628
:param timezone: How to display the time: 'utc', 'original' for the
750
629
timezone specified by offset, or 'local' for the process's current
752
:param date_fmt: strftime format.
753
:param show_offset: Whether to append the timezone.
755
(date_fmt, tt, offset_str) = \
756
_format_date(t, offset, timezone, date_fmt, show_offset)
757
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
758
date_str = time.strftime(date_fmt, tt)
759
return date_str + offset_str
762
# Cache of formatted offset strings
766
def format_date_with_offset_in_original_timezone(t, offset=0,
767
_cache=_offset_cache):
768
"""Return a formatted date string in the original timezone.
770
This routine may be faster then format_date.
772
:param t: Seconds since the epoch.
773
:param offset: Timezone offset in seconds east of utc.
777
tt = time.gmtime(t + offset)
778
date_fmt = _default_format_by_weekday_num[tt[6]]
779
date_str = time.strftime(date_fmt, tt)
780
offset_str = _cache.get(offset, None)
781
if offset_str is None:
782
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
783
_cache[offset] = offset_str
784
return date_str + offset_str
787
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
789
"""Return an unicode date string formatted according to the current locale.
791
:param t: Seconds since the epoch.
792
:param offset: Timezone offset in seconds east of utc.
793
:param timezone: How to display the time: 'utc', 'original' for the
794
timezone specified by offset, or 'local' for the process's current
796
:param date_fmt: strftime format.
797
:param show_offset: Whether to append the timezone.
799
(date_fmt, tt, offset_str) = \
800
_format_date(t, offset, timezone, date_fmt, show_offset)
801
date_str = time.strftime(date_fmt, tt)
802
if not isinstance(date_str, unicode):
803
date_str = date_str.decode(get_user_encoding(), 'replace')
804
return date_str + offset_str
807
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.
808
634
if timezone == 'utc':
809
635
tt = time.gmtime(t)
947
772
return pathjoin(*p)
950
def parent_directories(filename):
951
"""Return the list of parent directories, deepest first.
953
For example, parent_directories("a/b/c") -> ["a/b", "a"].
956
parts = splitpath(dirname(filename))
958
parents.append(joinpath(parts))
963
_extension_load_failures = []
966
def failed_to_load_extension(exception):
967
"""Handle failing to load a binary extension.
969
This should be called from the ImportError block guarding the attempt to
970
import the native extension. If this function returns, the pure-Python
971
implementation should be loaded instead::
974
>>> import bzrlib._fictional_extension_pyx
975
>>> except ImportError, e:
976
>>> bzrlib.osutils.failed_to_load_extension(e)
977
>>> import bzrlib._fictional_extension_py
979
# NB: This docstring is just an example, not a doctest, because doctest
980
# currently can't cope with the use of lazy imports in this namespace --
983
# This currently doesn't report the failure at the time it occurs, because
984
# they tend to happen very early in startup when we can't check config
985
# files etc, and also we want to report all failures but not spam the user
987
from bzrlib import trace
988
exception_str = str(exception)
989
if exception_str not in _extension_load_failures:
990
trace.mutter("failed to load compiled extension: %s" % exception_str)
991
_extension_load_failures.append(exception_str)
994
def report_extension_load_failures():
995
if not _extension_load_failures:
997
from bzrlib.config import GlobalConfig
998
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1000
# the warnings framework should by default show this only once
1001
from bzrlib.trace import warning
1003
"bzr: warning: some compiled extensions could not be loaded; "
1004
"see <https://answers.launchpad.net/bzr/+faq/703>")
1005
# we no longer show the specific missing extensions here, because it makes
1006
# the message too long and scary - see
1007
# https://bugs.launchpad.net/bzr/+bug/430529
1011
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1012
except ImportError, e:
1013
failed_to_load_extension(e)
1014
from bzrlib._chunks_to_lines_py import chunks_to_lines
1017
775
def split_lines(s):
1018
776
"""Split s into lines, but without removing the newline characters."""
1019
# Trivially convert a fulltext into a 'chunked' representation, and let
1020
# chunks_to_lines do the heavy lifting.
1021
if isinstance(s, str):
1022
# chunks_to_lines only supports 8-bit strings
1023
return chunks_to_lines([s])
1025
return _split_lines(s)
1028
def _split_lines(s):
1029
"""Split s into lines, but without removing the newline characters.
1031
This supports Unicode or plain string objects.
1033
777
lines = s.split('\n')
1034
778
result = [line + '\n' for line in lines[:-1]]
1054
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.
1057
806
def delete_any(path):
1058
"""Delete a file, symlink or directory.
1060
Will delete even if readonly.
1063
_delete_file_or_dir(path)
1064
except (OSError, IOError), e:
1065
if e.errno in (errno.EPERM, errno.EACCES):
1066
# make writable and try again
1069
except (OSError, IOError):
1071
_delete_file_or_dir(path)
1076
def _delete_file_or_dir(path):
1077
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1078
# Forgiveness than Permission (EAFP) because:
1079
# - root can damage a solaris file system by using unlink,
1080
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1081
# EACCES, OSX: EPERM) when invoked on a directory.
807
"""Delete a file or directory."""
1082
808
if isdir(path): # Takes care of symlinks
1174
if len(head) <= len(base) and head != base:
1175
raise errors.PathNotChild(rp, base)
877
while len(head) >= len(base):
1176
878
if head == base:
1178
head, tail = split(head)
880
head, tail = os.path.split(head)
884
raise errors.PathNotChild(rp, base)
1183
return pathjoin(*reversed(s))
1188
def _cicp_canonical_relpath(base, path):
1189
"""Return the canonical path relative to base.
1191
Like relpath, but on case-insensitive-case-preserving file-systems, this
1192
will return the relpath as stored on the file-system rather than in the
1193
case specified in the input string, for all existing portions of the path.
1195
This will cause O(N) behaviour if called for every path in a tree; if you
1196
have a number of paths to convert, you should use canonical_relpaths().
1198
# TODO: it should be possible to optimize this for Windows by using the
1199
# win32 API FindFiles function to look for the specified name - but using
1200
# os.listdir() still gives us the correct, platform agnostic semantics in
1203
rel = relpath(base, path)
1204
# '.' will have been turned into ''
1208
abs_base = abspath(base)
1210
_listdir = os.listdir
1212
# use an explicit iterator so we can easily consume the rest on early exit.
1213
bit_iter = iter(rel.split('/'))
1214
for bit in bit_iter:
1217
next_entries = _listdir(current)
1218
except OSError: # enoent, eperm, etc
1219
# We can't find this in the filesystem, so just append the
1221
current = pathjoin(current, bit, *list(bit_iter))
1223
for look in next_entries:
1224
if lbit == look.lower():
1225
current = pathjoin(current, look)
1228
# got to the end, nothing matched, so we just return the
1229
# non-existing bits as they were specified (the filename may be
1230
# the target of a move, for example).
1231
current = pathjoin(current, bit, *list(bit_iter))
1233
return current[len(abs_base):].lstrip('/')
1235
# XXX - TODO - we need better detection/integration of case-insensitive
1236
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1237
# filesystems), for example, so could probably benefit from the same basic
1238
# support there. For now though, only Windows and OSX get that support, and
1239
# they get it for *all* file-systems!
1240
if sys.platform in ('win32', 'darwin'):
1241
canonical_relpath = _cicp_canonical_relpath
1243
canonical_relpath = relpath
1245
def canonical_relpaths(base, paths):
1246
"""Create an iterable to canonicalize a sequence of relative paths.
1248
The intent is for this implementation to use a cache, vastly speeding
1249
up multiple transformations in the same directory.
1251
# but for now, we haven't optimized...
1252
return [canonical_relpath(base, p) for p in paths]
1254
892
def safe_unicode(unicode_or_utf8_string):
1255
893
"""Coerce unicode_or_utf8_string into unicode.
1257
895
If it is unicode, it is returned.
1258
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1259
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.
1261
900
if isinstance(unicode_or_utf8_string, unicode):
1262
901
return unicode_or_utf8_string
1375
1014
normalized_filename = _inaccessible_normalized_filename
1378
def set_signal_handler(signum, handler, restart_syscall=True):
1379
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1380
on platforms that support that.
1382
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1383
automatically restart (by calling `signal.siginterrupt(signum,
1384
False)`). May be ignored if the feature is not available on this
1385
platform or Python version.
1389
siginterrupt = signal.siginterrupt
1391
# This python implementation doesn't provide signal support, hence no
1394
except AttributeError:
1395
# siginterrupt doesn't exist on this platform, or for this version
1397
siginterrupt = lambda signum, flag: None
1399
def sig_handler(*args):
1400
# Python resets the siginterrupt flag when a signal is
1401
# received. <http://bugs.python.org/issue8354>
1402
# As a workaround for some cases, set it back the way we want it.
1403
siginterrupt(signum, False)
1404
# Now run the handler function passed to set_signal_handler.
1407
sig_handler = handler
1408
old_handler = signal.signal(signum, sig_handler)
1410
siginterrupt(signum, False)
1414
default_terminal_width = 80
1415
"""The default terminal width for ttys.
1417
This is defined so that higher levels can share a common fallback value when
1418
terminal_width() returns None.
1422
1017
def terminal_width():
1423
"""Return terminal width.
1425
None is returned if the width can't established precisely.
1428
- if BZR_COLUMNS is set, returns its value
1429
- if there is no controlling terminal, returns None
1430
- if COLUMNS is set, returns its value,
1432
From there, we need to query the OS to get the size of the controlling
1436
- get termios.TIOCGWINSZ
1437
- if an error occurs or a negative value is obtained, returns None
1441
- win32utils.get_console_size() decides,
1442
- returns None on error (provided default value)
1445
# If BZR_COLUMNS is set, take it, user is always right
1447
return int(os.environ['BZR_COLUMNS'])
1448
except (KeyError, ValueError):
1451
isatty = getattr(sys.stdout, 'isatty', None)
1452
if isatty is None or not isatty():
1453
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1456
# If COLUMNS is set, take it, the terminal knows better (even inside a
1457
# given terminal, the application can decide to set COLUMNS to a lower
1458
# value (splitted screen) or a bigger value (scroll bars))
1460
return int(os.environ['COLUMNS'])
1461
except (KeyError, ValueError):
1464
width, height = _terminal_size(None, None)
1466
# Consider invalid values as meaning no width
1472
def _win32_terminal_size(width, height):
1473
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1474
return width, height
1477
def _ioctl_terminal_size(width, height):
1018
"""Return estimated terminal width."""
1019
if sys.platform == 'win32':
1020
return win32utils.get_console_size()[0]
1479
1023
import struct, fcntl, termios
1480
1024
s = struct.pack('HHHH', 0, 0, 0, 0)
1481
1025
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1482
height, width = struct.unpack('HHHH', x)[0:2]
1483
except (IOError, AttributeError):
1026
width = struct.unpack('HHHH', x)[1]
1485
return width, height
1487
_terminal_size = None
1488
"""Returns the terminal size as (width, height).
1490
:param width: Default value for width.
1491
:param height: Default value for height.
1493
This is defined specifically for each OS and query the size of the controlling
1494
terminal. If any error occurs, the provided default values should be returned.
1496
if sys.platform == 'win32':
1497
_terminal_size = _win32_terminal_size
1499
_terminal_size = _ioctl_terminal_size
1502
def _terminal_size_changed(signum, frame):
1503
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1504
width, height = _terminal_size(None, None)
1505
if width is not None:
1506
os.environ['COLUMNS'] = str(width)
1509
_registered_sigwinch = False
1510
def watch_sigwinch():
1511
"""Register for SIGWINCH, once and only once.
1513
Do nothing if the signal module is not available.
1515
global _registered_sigwinch
1516
if not _registered_sigwinch:
1519
if getattr(signal, "SIGWINCH", None) is not None:
1520
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1522
# python doesn't provide signal support, nothing we can do about it
1031
width = int(os.environ['COLUMNS'])
1524
_registered_sigwinch = True
1527
1040
def supports_executable():
1574
1087
raise errors.IllegalPath(path)
1577
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1579
def _is_error_enotdir(e):
1580
"""Check if this exception represents ENOTDIR.
1582
Unfortunately, python is very inconsistent about the exception
1583
here. The cases are:
1584
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1585
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1586
which is the windows error code.
1587
3) Windows, Python2.5 uses errno == EINVAL and
1588
winerror == ERROR_DIRECTORY
1590
:param e: An Exception object (expected to be OSError with an errno
1591
attribute, but we should be able to cope with anything)
1592
:return: True if this represents an ENOTDIR error. False otherwise.
1594
en = getattr(e, 'errno', None)
1595
if (en == errno.ENOTDIR
1596
or (sys.platform == 'win32'
1597
and (en == _WIN32_ERROR_DIRECTORY
1598
or (en == errno.EINVAL
1599
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1605
1090
def walkdirs(top, prefix=""):
1606
1091
"""Yield data about all the directories in a tree.
1608
1093
This yields all the data about the contents of a directory at a time.
1609
1094
After each directory has been yielded, if the caller has mutated the list
1610
1095
to exclude some directories, they are then not descended into.
1612
1097
The data yielded is of the form:
1613
1098
((directory-relpath, directory-path-from-top),
1614
1099
[(relpath, basename, kind, lstat, path-from-top), ...]),
1615
1100
- directory-relpath is the relative path of the directory being returned
1616
1101
with respect to top. prefix is prepended to this.
1617
- 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.
1618
1103
It is suitable for use with os functions.
1619
1104
- relpath is the relative path within the subtree being walked.
1620
1105
- basename is the basename of the path
1622
1107
present within the tree - but it may be recorded as versioned. See
1623
1108
versioned_kind.
1624
1109
- lstat is the stat data *if* the file was statted.
1625
- planned, not implemented:
1110
- planned, not implemented:
1626
1111
path_from_tree_root is the path from the root of the tree.
1628
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1113
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1629
1114
allows one to walk a subtree but get paths that are relative to a tree
1630
1115
rooted higher up.
1631
1116
:return: an iterator over the dirs.
1633
1118
#TODO there is a bit of a smell where the results of the directory-
1634
# 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
1635
1120
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1636
1121
# potentially confusing output. We should make this more robust - but
1637
1122
# not at a speed cost. RBC 20060731
1638
1123
_lstat = os.lstat
1639
1124
_directory = _directory_kind
1640
1125
_listdir = os.listdir
1641
_kind_from_mode = file_kind_from_stat_mode
1126
_kind_from_mode = _formats.get
1642
1127
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1644
1129
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1653
1138
append = dirblock.append
1655
names = sorted(_listdir(top))
1657
if not _is_error_enotdir(e):
1661
abspath = top_slash + name
1662
statvalue = _lstat(abspath)
1663
kind = _kind_from_mode(statvalue.st_mode)
1664
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))
1665
1144
yield (relroot, top), dirblock
1667
1146
# push the user specified dirs from dirblock
1668
1147
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1671
class DirReader(object):
1672
"""An interface for reading directories."""
1674
def top_prefix_to_starting_dir(self, top, prefix=""):
1675
"""Converts top and prefix to a starting dir entry
1677
:param top: A utf8 path
1678
:param prefix: An optional utf8 path to prefix output relative paths
1680
:return: A tuple starting with prefix, and ending with the native
1683
raise NotImplementedError(self.top_prefix_to_starting_dir)
1685
def read_dir(self, prefix, top):
1686
"""Read a specific dir.
1688
:param prefix: A utf8 prefix to be preprended to the path basenames.
1689
:param top: A natively encoded path to read.
1690
:return: A list of the directories contents. Each item contains:
1691
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1693
raise NotImplementedError(self.read_dir)
1696
_selected_dir_reader = None
1699
1150
def _walkdirs_utf8(top, prefix=""):
1700
1151
"""Yield data about all the directories in a tree.
1710
1161
path-from-top might be unicode or utf8, but it is the correct path to
1711
1162
pass to os functions to affect the file in question. (such as os.lstat)
1713
global _selected_dir_reader
1714
if _selected_dir_reader is None:
1715
fs_encoding = _fs_enc.upper()
1716
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1717
# Win98 doesn't have unicode apis like FindFirstFileW
1718
# TODO: We possibly could support Win98 by falling back to the
1719
# original FindFirstFile, and using TCHAR instead of WCHAR,
1720
# but that gets a bit tricky, and requires custom compiling
1723
from bzrlib._walkdirs_win32 import Win32ReadDir
1724
_selected_dir_reader = Win32ReadDir()
1727
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1728
# ANSI_X3.4-1968 is a form of ASCII
1730
from bzrlib._readdir_pyx import UTF8DirReader
1731
_selected_dir_reader = UTF8DirReader()
1732
except ImportError, e:
1733
failed_to_load_extension(e)
1736
if _selected_dir_reader is None:
1737
# Fallback to the python version
1738
_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
1740
1183
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1741
1184
# But we don't actually uses 1-3 in pending, so set them to None
1742
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1743
read_dir = _selected_dir_reader.read_dir
1744
_directory = _directory_kind
1185
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1746
relroot, _, _, _, top = pending[-1].pop()
1749
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))
1750
1201
yield (relroot, top), dirblock
1751
1203
# push the user specified dirs from dirblock
1752
next = [d for d in reversed(dirblock) if d[2] == _directory]
1754
pending.append(next)
1757
class UnicodeDirReader(DirReader):
1758
"""A dir reader for non-utf8 file systems, which transcodes."""
1760
__slots__ = ['_utf8_encode']
1763
self._utf8_encode = codecs.getencoder('utf8')
1765
def top_prefix_to_starting_dir(self, top, prefix=""):
1766
"""See DirReader.top_prefix_to_starting_dir."""
1767
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1769
def read_dir(self, prefix, top):
1770
"""Read a single directory from a non-utf8 file system.
1772
top, and the abspath element in the output are unicode, all other paths
1773
are utf8. Local disk IO is done via unicode calls to listdir etc.
1775
This is currently the fallback code path when the filesystem encoding is
1776
not UTF-8. It may be better to implement an alternative so that we can
1777
safely handle paths that are not properly decodable in the current
1780
See DirReader.read_dir for details.
1782
_utf8_encode = self._utf8_encode
1784
_listdir = os.listdir
1785
_kind_from_mode = file_kind_from_stat_mode
1788
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 + '/'
1791
1230
top_slash = top + u'/'
1794
1233
append = dirblock.append
1795
1234
for name in sorted(_listdir(top)):
1797
name_utf8 = _utf8_encode(name)[0]
1798
except UnicodeDecodeError:
1799
raise errors.BadFilenameEncoding(
1800
_utf8_encode(relprefix)[0] + name, _fs_enc)
1235
name_utf8 = _utf8_encode(name)[0]
1801
1236
abspath = top_slash + name
1802
1237
statvalue = _lstat(abspath)
1803
kind = _kind_from_mode(statvalue.st_mode)
1238
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1804
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)
1808
1246
def copy_tree(from_path, to_path, handlers={}):
1809
1247
"""Copy all of the entries in from_path into to_path.
1811
:param from_path: The base directory to copy.
1249
:param from_path: The base directory to copy.
1812
1250
:param to_path: The target directory. If it does not exist, it will
1814
1252
:param handlers: A dictionary of functions, which takes a source and
1960
1364
return user_encoding
1963
def get_host_name():
1964
"""Return the current unicode host name.
1966
This is meant to be used in place of socket.gethostname() because that
1967
behaves inconsistently on different platforms.
1969
if sys.platform == "win32":
1971
return win32utils.get_host_name()
1974
return socket.gethostname().decode(get_user_encoding())
1977
# We must not read/write any more than 64k at a time from/to a socket so we
1978
# don't risk "no buffer space available" errors on some platforms. Windows in
1979
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1981
MAX_SOCKET_CHUNK = 64 * 1024
1983
def read_bytes_from_socket(sock, report_activity=None,
1984
max_read_size=MAX_SOCKET_CHUNK):
1985
"""Read up to max_read_size of bytes from sock and notify of progress.
1987
Translates "Connection reset by peer" into file-like EOF (return an
1988
empty string rather than raise an error), and repeats the recv if
1989
interrupted by a signal.
1993
bytes = sock.recv(max_read_size)
1994
except socket.error, e:
1996
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1997
# The connection was closed by the other side. Callers expect
1998
# an empty string to signal end-of-stream.
2000
elif eno == errno.EINTR:
2001
# Retry the interrupted recv.
2005
if report_activity is not None:
2006
report_activity(len(bytes), 'read')
2010
def recv_all(socket, count):
1367
def recv_all(socket, bytes):
2011
1368
"""Receive an exact number of bytes.
2013
1370
Regular Socket.recv() may return less than the requested number of bytes,
2014
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
2015
1372
on all platforms, but this should work everywhere. This will return
2016
1373
less than the requested amount if the remote end closes.
2018
1375
This isn't optimized and is intended mostly for use in testing.
2021
while len(b) < count:
2022
new = read_bytes_from_socket(socket, None, count - len(b))
1378
while len(b) < bytes:
1379
new = socket.recv(bytes - len(b))
2029
def send_all(sock, bytes, report_activity=None):
1386
def send_all(socket, bytes):
2030
1387
"""Send all bytes on a socket.
2032
Breaks large blocks in smaller chunks to avoid buffering limitations on
2033
some platforms, and catches EINTR which may be thrown if the send is
2034
interrupted by a signal.
2036
This is preferred to socket.sendall(), because it avoids portability bugs
2037
and provides activity reporting.
2039
:param report_activity: Call this as bytes are read, see
2040
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.
2043
byte_count = len(bytes)
2044
while sent_total < byte_count:
2046
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2047
except socket.error, e:
2048
if e.args[0] != errno.EINTR:
2052
report_activity(sent, 'write')
1393
for pos in xrange(0, len(bytes), chunk_size):
1394
socket.sendall(bytes[pos:pos+chunk_size])
2055
1397
def dereference_path(path):
2098
1440
base = abspath(pathjoin(base, '..', '..'))
2099
1441
filename = pathjoin(base, resource_relpath)
2100
1442
return open(filename, 'rU').read()
2103
def file_kind_from_stat_mode_thunk(mode):
2104
global file_kind_from_stat_mode
2105
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2107
from bzrlib._readdir_pyx import UTF8DirReader
2108
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2109
except ImportError, e:
2110
# This is one time where we won't warn that an extension failed to
2111
# load. The extension is never available on Windows anyway.
2112
from bzrlib._readdir_py import (
2113
_kind_from_mode as file_kind_from_stat_mode
2115
return file_kind_from_stat_mode(mode)
2116
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2119
def file_kind(f, _lstat=os.lstat):
2121
return file_kind_from_stat_mode(_lstat(f).st_mode)
2123
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2124
raise errors.NoSuchFile(f)
2128
def until_no_eintr(f, *a, **kw):
2129
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2131
WARNING: you must be certain that it is safe to retry the call repeatedly
2132
if EINTR does occur. This is typically only true for low-level operations
2133
like os.read. If in any doubt, don't use this.
2135
Keep in mind that this is not a complete solution to EINTR. There is
2136
probably code in the Python standard library and other dependencies that
2137
may encounter EINTR if a signal arrives (and there is signal handler for
2138
that signal). So this function can reduce the impact for IO that bzrlib
2139
directly controls, but it is not a complete solution.
2141
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2145
except (IOError, OSError), e:
2146
if e.errno == errno.EINTR:
2151
def re_compile_checked(re_string, flags=0, where=""):
2152
"""Return a compiled re, or raise a sensible error.
2154
This should only be used when compiling user-supplied REs.
2156
:param re_string: Text form of regular expression.
2157
:param flags: eg re.IGNORECASE
2158
:param where: Message explaining to the user the context where
2159
it occurred, eg 'log search filter'.
2161
# from https://bugs.launchpad.net/bzr/+bug/251352
2163
re_obj = re.compile(re_string, flags)
2168
where = ' in ' + where
2169
# despite the name 'error' is a type
2170
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2171
% (where, re_string, e))
2174
if sys.platform == "win32":
2177
return msvcrt.getch()
2182
fd = sys.stdin.fileno()
2183
settings = termios.tcgetattr(fd)
2186
ch = sys.stdin.read(1)
2188
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2192
if sys.platform == 'linux2':
2193
def _local_concurrency():
2195
prefix = 'processor'
2196
for line in file('/proc/cpuinfo', 'rb'):
2197
if line.startswith(prefix):
2198
concurrency = int(line[line.find(':')+1:]) + 1
2200
elif sys.platform == 'darwin':
2201
def _local_concurrency():
2202
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2203
stdout=subprocess.PIPE).communicate()[0]
2204
elif sys.platform[0:7] == 'freebsd':
2205
def _local_concurrency():
2206
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2207
stdout=subprocess.PIPE).communicate()[0]
2208
elif sys.platform == 'sunos5':
2209
def _local_concurrency():
2210
return subprocess.Popen(['psrinfo', '-p',],
2211
stdout=subprocess.PIPE).communicate()[0]
2212
elif sys.platform == "win32":
2213
def _local_concurrency():
2214
# This appears to return the number of cores.
2215
return os.environ.get('NUMBER_OF_PROCESSORS')
2217
def _local_concurrency():
2222
_cached_local_concurrency = None
2224
def local_concurrency(use_cache=True):
2225
"""Return how many processes can be run concurrently.
2227
Rely on platform specific implementations and default to 1 (one) if
2228
anything goes wrong.
2230
global _cached_local_concurrency
2232
if _cached_local_concurrency is not None and use_cache:
2233
return _cached_local_concurrency
2235
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2236
if concurrency is None:
2238
concurrency = _local_concurrency()
2239
except (OSError, IOError):
2242
concurrency = int(concurrency)
2243
except (TypeError, ValueError):
2246
_cached_concurrency = concurrency
2250
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2251
"""A stream writer that doesn't decode str arguments."""
2253
def __init__(self, encode, stream, errors='strict'):
2254
codecs.StreamWriter.__init__(self, stream, errors)
2255
self.encode = encode
2257
def write(self, object):
2258
if type(object) is str:
2259
self.stream.write(object)
2261
data, _ = self.encode(object, self.errors)
2262
self.stream.write(data)
2264
if sys.platform == 'win32':
2265
def open_file(filename, mode='r', bufsize=-1):
2266
"""This function is used to override the ``open`` builtin.
2268
But it uses O_NOINHERIT flag so the file handle is not inherited by
2269
child processes. Deleting or renaming a closed file opened with this
2270
function is not blocking child processes.
2272
writing = 'w' in mode
2273
appending = 'a' in mode
2274
updating = '+' in mode
2275
binary = 'b' in mode
2278
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2279
# for flags for each modes.
2289
flags |= os.O_WRONLY
2290
flags |= os.O_CREAT | os.O_TRUNC
2295
flags |= os.O_WRONLY
2296
flags |= os.O_CREAT | os.O_APPEND
2301
flags |= os.O_RDONLY
2303
return os.fdopen(os.open(filename, flags), mode, bufsize)
2308
def getuser_unicode():
2309
"""Return the username as unicode.
2312
user_encoding = get_user_encoding()
2313
username = getpass.getuser().decode(user_encoding)
2314
except UnicodeDecodeError:
2315
raise errors.BzrError("Can't decode username as %s." % \
added
removed
bzrlib/osutils.py
