629
1130
avoids that problem.
632
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
633
' exceed the platform minimum length (which is %d)' %
1133
if len(base) < MIN_ABS_PATHLENGTH:
1134
# must have space for e.g. a drive letter
1135
raise ValueError('%r is too short to calculate a relative path'
635
1138
rp = abspath(path)
639
while len(head) >= len(base):
1143
if len(head) <= len(base) and head != base:
1144
raise errors.PathNotChild(rp, base)
640
1145
if head == base:
642
head, tail = os.path.split(head)
1147
head, tail = split(head)
646
# XXX This should raise a NotChildPath exception, as its not tied
648
raise PathNotChild(rp, base)
1152
return pathjoin(*reversed(s))
1157
def _cicp_canonical_relpath(base, path):
1158
"""Return the canonical path relative to base.
1160
Like relpath, but on case-insensitive-case-preserving file-systems, this
1161
will return the relpath as stored on the file-system rather than in the
1162
case specified in the input string, for all existing portions of the path.
1164
This will cause O(N) behaviour if called for every path in a tree; if you
1165
have a number of paths to convert, you should use canonical_relpaths().
1167
# TODO: it should be possible to optimize this for Windows by using the
1168
# win32 API FindFiles function to look for the specified name - but using
1169
# os.listdir() still gives us the correct, platform agnostic semantics in
1172
rel = relpath(base, path)
1173
# '.' will have been turned into ''
1177
abs_base = abspath(base)
1179
_listdir = os.listdir
1181
# use an explicit iterator so we can easily consume the rest on early exit.
1182
bit_iter = iter(rel.split('/'))
1183
for bit in bit_iter:
1186
next_entries = _listdir(current)
1187
except OSError: # enoent, eperm, etc
1188
# We can't find this in the filesystem, so just append the
1190
current = pathjoin(current, bit, *list(bit_iter))
1192
for look in next_entries:
1193
if lbit == look.lower():
1194
current = pathjoin(current, look)
1197
# got to the end, nothing matched, so we just return the
1198
# non-existing bits as they were specified (the filename may be
1199
# the target of a move, for example).
1200
current = pathjoin(current, bit, *list(bit_iter))
1202
return current[len(abs_base):].lstrip('/')
1204
# XXX - TODO - we need better detection/integration of case-insensitive
1205
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1206
# filesystems), for example, so could probably benefit from the same basic
1207
# support there. For now though, only Windows and OSX get that support, and
1208
# they get it for *all* file-systems!
1209
if sys.platform in ('win32', 'darwin'):
1210
canonical_relpath = _cicp_canonical_relpath
1212
canonical_relpath = relpath
1214
def canonical_relpaths(base, paths):
1215
"""Create an iterable to canonicalize a sequence of relative paths.
1217
The intent is for this implementation to use a cache, vastly speeding
1218
up multiple transformations in the same directory.
1220
# but for now, we haven't optimized...
1221
return [canonical_relpath(base, p) for p in paths]
656
1223
def safe_unicode(unicode_or_utf8_string):
657
1224
"""Coerce unicode_or_utf8_string into unicode.
659
1226
If it is unicode, it is returned.
660
Otherwise it is decoded from utf-8. If a decoding error
661
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
662
as a BzrBadParameter exception.
1227
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1228
wrapped in a BzrBadParameterNotUnicode exception.
664
1230
if isinstance(unicode_or_utf8_string, unicode):
665
1231
return unicode_or_utf8_string
667
1233
return unicode_or_utf8_string.decode('utf8')
668
1234
except UnicodeDecodeError:
669
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
1235
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1238
def safe_utf8(unicode_or_utf8_string):
1239
"""Coerce unicode_or_utf8_string to a utf8 string.
1241
If it is a str, it is returned.
1242
If it is Unicode, it is encoded into a utf-8 string.
1244
if isinstance(unicode_or_utf8_string, str):
1245
# TODO: jam 20070209 This is overkill, and probably has an impact on
1246
# performance if we are dealing with lots of apis that want a
1249
# Make sure it is a valid utf-8 string
1250
unicode_or_utf8_string.decode('utf-8')
1251
except UnicodeDecodeError:
1252
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1253
return unicode_or_utf8_string
1254
return unicode_or_utf8_string.encode('utf-8')
1257
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1258
' Revision id generators should be creating utf8'
1262
def safe_revision_id(unicode_or_utf8_string, warn=True):
1263
"""Revision ids should now be utf8, but at one point they were unicode.
1265
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1267
:param warn: Functions that are sanitizing user data can set warn=False
1268
:return: None or a utf8 revision id.
1270
if (unicode_or_utf8_string is None
1271
or unicode_or_utf8_string.__class__ == str):
1272
return unicode_or_utf8_string
1274
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1276
return cache_utf8.encode(unicode_or_utf8_string)
1279
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1280
' generators should be creating utf8 file ids.')
1283
def safe_file_id(unicode_or_utf8_string, warn=True):
1284
"""File ids should now be utf8, but at one point they were unicode.
1286
This is the same as safe_utf8, except it uses the cached encode functions
1287
to save a little bit of performance.
1289
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1291
:param warn: Functions that are sanitizing user data can set warn=False
1292
:return: None or a utf8 file id.
1294
if (unicode_or_utf8_string is None
1295
or unicode_or_utf8_string.__class__ == str):
1296
return unicode_or_utf8_string
1298
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1300
return cache_utf8.encode(unicode_or_utf8_string)
1303
_platform_normalizes_filenames = False
1304
if sys.platform == 'darwin':
1305
_platform_normalizes_filenames = True
1308
def normalizes_filenames():
1309
"""Return True if this platform normalizes unicode filenames.
1311
Mac OSX does, Windows/Linux do not.
1313
return _platform_normalizes_filenames
1316
def _accessible_normalized_filename(path):
1317
"""Get the unicode normalized path, and if you can access the file.
1319
On platforms where the system normalizes filenames (Mac OSX),
1320
you can access a file by any path which will normalize correctly.
1321
On platforms where the system does not normalize filenames
1322
(Windows, Linux), you have to access a file by its exact path.
1324
Internally, bzr only supports NFC normalization, since that is
1325
the standard for XML documents.
1327
So return the normalized path, and a flag indicating if the file
1328
can be accessed by that path.
1331
return unicodedata.normalize('NFC', unicode(path)), True
1334
def _inaccessible_normalized_filename(path):
1335
__doc__ = _accessible_normalized_filename.__doc__
1337
normalized = unicodedata.normalize('NFC', unicode(path))
1338
return normalized, normalized == path
1341
if _platform_normalizes_filenames:
1342
normalized_filename = _accessible_normalized_filename
1344
normalized_filename = _inaccessible_normalized_filename
1347
default_terminal_width = 80
1348
"""The default terminal width for ttys.
1350
This is defined so that higher levels can share a common fallback value when
1351
terminal_width() returns None.
672
1355
def terminal_width():
673
"""Return estimated terminal width."""
674
if sys.platform == 'win32':
675
import bzrlib.win32console
676
return bzrlib.win32console.get_console_size()[0]
1356
"""Return terminal width.
1358
None is returned if the width can't established precisely.
1361
- if BZR_COLUMNS is set, returns its value
1362
- if there is no controlling terminal, returns None
1363
- if COLUMNS is set, returns its value,
1365
From there, we need to query the OS to get the size of the controlling
1369
- get termios.TIOCGWINSZ
1370
- if an error occurs or a negative value is obtained, returns None
1374
- win32utils.get_console_size() decides,
1375
- returns None on error (provided default value)
1378
# If BZR_COLUMNS is set, take it, user is always right
1380
return int(os.environ['BZR_COLUMNS'])
1381
except (KeyError, ValueError):
1384
isatty = getattr(sys.stdout, 'isatty', None)
1385
if isatty is None or not isatty():
1386
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1389
# If COLUMNS is set, take it, the terminal knows better (even inside a
1390
# given terminal, the application can decide to set COLUMNS to a lower
1391
# value (splitted screen) or a bigger value (scroll bars))
1393
return int(os.environ['COLUMNS'])
1394
except (KeyError, ValueError):
1397
width, height = _terminal_size(None, None)
1399
# Consider invalid values as meaning no width
1405
def _win32_terminal_size(width, height):
1406
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1407
return width, height
1410
def _ioctl_terminal_size(width, height):
679
1412
import struct, fcntl, termios
680
1413
s = struct.pack('HHHH', 0, 0, 0, 0)
681
1414
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
682
width = struct.unpack('HHHH', x)[1]
1415
height, width = struct.unpack('HHHH', x)[0:2]
1416
except (IOError, AttributeError):
687
width = int(os.environ['COLUMNS'])
1418
return width, height
1420
_terminal_size = None
1421
"""Returns the terminal size as (width, height).
1423
:param width: Default value for width.
1424
:param height: Default value for height.
1426
This is defined specifically for each OS and query the size of the controlling
1427
terminal. If any error occurs, the provided default values should be returned.
1429
if sys.platform == 'win32':
1430
_terminal_size = _win32_terminal_size
1432
_terminal_size = _ioctl_terminal_size
1435
def _terminal_size_changed(signum, frame):
1436
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1437
width, height = _terminal_size(None, None)
1438
if width is not None:
1439
os.environ['COLUMNS'] = str(width)
1441
if sys.platform == 'win32':
1442
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from ReadConsoleInput but
1443
# I've no idea how to plug that in the current design -- vila 20091216
1446
signal.signal(signal.SIGWINCH, _terminal_size_changed)
695
1449
def supports_executable():
696
1450
return sys.platform != "win32"
699
def strip_trailing_slash(path):
700
"""Strip trailing slash, except for root paths.
701
The definition of 'root path' is platform-dependent.
703
if len(path) != MIN_ABS_PATHLENGTH and path[-1] == '/':
1453
def supports_posix_readonly():
1454
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1456
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1457
directory controls creation/deletion, etc.
1459
And under win32, readonly means that the directory itself cannot be
1460
deleted. The contents of a readonly directory can be changed, unlike POSIX
1461
where files in readonly directories cannot be added, deleted or renamed.
1463
return sys.platform != "win32"
1466
def set_or_unset_env(env_variable, value):
1467
"""Modify the environment, setting or removing the env_variable.
1469
:param env_variable: The environment variable in question
1470
:param value: The value to set the environment to. If None, then
1471
the variable will be removed.
1472
:return: The original value of the environment variable.
1474
orig_val = os.environ.get(env_variable)
1476
if orig_val is not None:
1477
del os.environ[env_variable]
1479
if isinstance(value, unicode):
1480
value = value.encode(get_user_encoding())
1481
os.environ[env_variable] = value
709
1485
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
712
1488
def check_legal_path(path):
713
"""Check whether the supplied path is legal.
1489
"""Check whether the supplied path is legal.
714
1490
This is only required on Windows, so we don't test on other platforms
717
1493
if sys.platform != "win32":
719
1495
if _validWin32PathRE.match(path) is None:
720
raise IllegalPath(path)
1496
raise errors.IllegalPath(path)
1499
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1501
def _is_error_enotdir(e):
1502
"""Check if this exception represents ENOTDIR.
1504
Unfortunately, python is very inconsistent about the exception
1505
here. The cases are:
1506
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1507
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1508
which is the windows error code.
1509
3) Windows, Python2.5 uses errno == EINVAL and
1510
winerror == ERROR_DIRECTORY
1512
:param e: An Exception object (expected to be OSError with an errno
1513
attribute, but we should be able to cope with anything)
1514
:return: True if this represents an ENOTDIR error. False otherwise.
1516
en = getattr(e, 'errno', None)
1517
if (en == errno.ENOTDIR
1518
or (sys.platform == 'win32'
1519
and (en == _WIN32_ERROR_DIRECTORY
1520
or (en == errno.EINVAL
1521
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1527
def walkdirs(top, prefix=""):
1528
"""Yield data about all the directories in a tree.
1530
This yields all the data about the contents of a directory at a time.
1531
After each directory has been yielded, if the caller has mutated the list
1532
to exclude some directories, they are then not descended into.
1534
The data yielded is of the form:
1535
((directory-relpath, directory-path-from-top),
1536
[(relpath, basename, kind, lstat, path-from-top), ...]),
1537
- directory-relpath is the relative path of the directory being returned
1538
with respect to top. prefix is prepended to this.
1539
- directory-path-from-root is the path including top for this directory.
1540
It is suitable for use with os functions.
1541
- relpath is the relative path within the subtree being walked.
1542
- basename is the basename of the path
1543
- kind is the kind of the file now. If unknown then the file is not
1544
present within the tree - but it may be recorded as versioned. See
1546
- lstat is the stat data *if* the file was statted.
1547
- planned, not implemented:
1548
path_from_tree_root is the path from the root of the tree.
1550
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1551
allows one to walk a subtree but get paths that are relative to a tree
1553
:return: an iterator over the dirs.
1555
#TODO there is a bit of a smell where the results of the directory-
1556
# summary in this, and the path from the root, may not agree
1557
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1558
# potentially confusing output. We should make this more robust - but
1559
# not at a speed cost. RBC 20060731
1561
_directory = _directory_kind
1562
_listdir = os.listdir
1563
_kind_from_mode = file_kind_from_stat_mode
1564
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1566
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1567
relroot, _, _, _, top = pending.pop()
1569
relprefix = relroot + u'/'
1572
top_slash = top + u'/'
1575
append = dirblock.append
1577
names = sorted(_listdir(top))
1579
if not _is_error_enotdir(e):
1583
abspath = top_slash + name
1584
statvalue = _lstat(abspath)
1585
kind = _kind_from_mode(statvalue.st_mode)
1586
append((relprefix + name, name, kind, statvalue, abspath))
1587
yield (relroot, top), dirblock
1589
# push the user specified dirs from dirblock
1590
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1593
class DirReader(object):
1594
"""An interface for reading directories."""
1596
def top_prefix_to_starting_dir(self, top, prefix=""):
1597
"""Converts top and prefix to a starting dir entry
1599
:param top: A utf8 path
1600
:param prefix: An optional utf8 path to prefix output relative paths
1602
:return: A tuple starting with prefix, and ending with the native
1605
raise NotImplementedError(self.top_prefix_to_starting_dir)
1607
def read_dir(self, prefix, top):
1608
"""Read a specific dir.
1610
:param prefix: A utf8 prefix to be preprended to the path basenames.
1611
:param top: A natively encoded path to read.
1612
:return: A list of the directories contents. Each item contains:
1613
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1615
raise NotImplementedError(self.read_dir)
1618
_selected_dir_reader = None
1621
def _walkdirs_utf8(top, prefix=""):
1622
"""Yield data about all the directories in a tree.
1624
This yields the same information as walkdirs() only each entry is yielded
1625
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1626
are returned as exact byte-strings.
1628
:return: yields a tuple of (dir_info, [file_info])
1629
dir_info is (utf8_relpath, path-from-top)
1630
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1631
if top is an absolute path, path-from-top is also an absolute path.
1632
path-from-top might be unicode or utf8, but it is the correct path to
1633
pass to os functions to affect the file in question. (such as os.lstat)
1635
global _selected_dir_reader
1636
if _selected_dir_reader is None:
1637
fs_encoding = _fs_enc.upper()
1638
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1639
# Win98 doesn't have unicode apis like FindFirstFileW
1640
# TODO: We possibly could support Win98 by falling back to the
1641
# original FindFirstFile, and using TCHAR instead of WCHAR,
1642
# but that gets a bit tricky, and requires custom compiling
1645
from bzrlib._walkdirs_win32 import Win32ReadDir
1646
_selected_dir_reader = Win32ReadDir()
1649
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1650
# ANSI_X3.4-1968 is a form of ASCII
1652
from bzrlib._readdir_pyx import UTF8DirReader
1653
_selected_dir_reader = UTF8DirReader()
1654
except ImportError, e:
1655
failed_to_load_extension(e)
1658
if _selected_dir_reader is None:
1659
# Fallback to the python version
1660
_selected_dir_reader = UnicodeDirReader()
1662
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1663
# But we don't actually uses 1-3 in pending, so set them to None
1664
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1665
read_dir = _selected_dir_reader.read_dir
1666
_directory = _directory_kind
1668
relroot, _, _, _, top = pending[-1].pop()
1671
dirblock = sorted(read_dir(relroot, top))
1672
yield (relroot, top), dirblock
1673
# push the user specified dirs from dirblock
1674
next = [d for d in reversed(dirblock) if d[2] == _directory]
1676
pending.append(next)
1679
class UnicodeDirReader(DirReader):
1680
"""A dir reader for non-utf8 file systems, which transcodes."""
1682
__slots__ = ['_utf8_encode']
1685
self._utf8_encode = codecs.getencoder('utf8')
1687
def top_prefix_to_starting_dir(self, top, prefix=""):
1688
"""See DirReader.top_prefix_to_starting_dir."""
1689
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1691
def read_dir(self, prefix, top):
1692
"""Read a single directory from a non-utf8 file system.
1694
top, and the abspath element in the output are unicode, all other paths
1695
are utf8. Local disk IO is done via unicode calls to listdir etc.
1697
This is currently the fallback code path when the filesystem encoding is
1698
not UTF-8. It may be better to implement an alternative so that we can
1699
safely handle paths that are not properly decodable in the current
1702
See DirReader.read_dir for details.
1704
_utf8_encode = self._utf8_encode
1706
_listdir = os.listdir
1707
_kind_from_mode = file_kind_from_stat_mode
1710
relprefix = prefix + '/'
1713
top_slash = top + u'/'
1716
append = dirblock.append
1717
for name in sorted(_listdir(top)):
1719
name_utf8 = _utf8_encode(name)[0]
1720
except UnicodeDecodeError:
1721
raise errors.BadFilenameEncoding(
1722
_utf8_encode(relprefix)[0] + name, _fs_enc)
1723
abspath = top_slash + name
1724
statvalue = _lstat(abspath)
1725
kind = _kind_from_mode(statvalue.st_mode)
1726
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1730
def copy_tree(from_path, to_path, handlers={}):
1731
"""Copy all of the entries in from_path into to_path.
1733
:param from_path: The base directory to copy.
1734
:param to_path: The target directory. If it does not exist, it will
1736
:param handlers: A dictionary of functions, which takes a source and
1737
destinations for files, directories, etc.
1738
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1739
'file', 'directory', and 'symlink' should always exist.
1740
If they are missing, they will be replaced with 'os.mkdir()',
1741
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1743
# Now, just copy the existing cached tree to the new location
1744
# We use a cheap trick here.
1745
# Absolute paths are prefixed with the first parameter
1746
# relative paths are prefixed with the second.
1747
# So we can get both the source and target returned
1748
# without any extra work.
1750
def copy_dir(source, dest):
1753
def copy_link(source, dest):
1754
"""Copy the contents of a symlink"""
1755
link_to = os.readlink(source)
1756
os.symlink(link_to, dest)
1758
real_handlers = {'file':shutil.copy2,
1759
'symlink':copy_link,
1760
'directory':copy_dir,
1762
real_handlers.update(handlers)
1764
if not os.path.exists(to_path):
1765
real_handlers['directory'](from_path, to_path)
1767
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1768
for relpath, name, kind, st, abspath in entries:
1769
real_handlers[kind](abspath, relpath)
1772
def path_prefix_key(path):
1773
"""Generate a prefix-order path key for path.
1775
This can be used to sort paths in the same way that walkdirs does.
1777
return (dirname(path) , path)
1780
def compare_paths_prefix_order(path_a, path_b):
1781
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1782
key_a = path_prefix_key(path_a)
1783
key_b = path_prefix_key(path_b)
1784
return cmp(key_a, key_b)
1787
_cached_user_encoding = None
1790
def get_user_encoding(use_cache=True):
1791
"""Find out what the preferred user encoding is.
1793
This is generally the encoding that is used for command line parameters
1794
and file contents. This may be different from the terminal encoding
1795
or the filesystem encoding.
1797
:param use_cache: Enable cache for detected encoding.
1798
(This parameter is turned on by default,
1799
and required only for selftesting)
1801
:return: A string defining the preferred user encoding
1803
global _cached_user_encoding
1804
if _cached_user_encoding is not None and use_cache:
1805
return _cached_user_encoding
1807
if sys.platform == 'darwin':
1808
# python locale.getpreferredencoding() always return
1809
# 'mac-roman' on darwin. That's a lie.
1810
sys.platform = 'posix'
1812
if os.environ.get('LANG', None) is None:
1813
# If LANG is not set, we end up with 'ascii', which is bad
1814
# ('mac-roman' is more than ascii), so we set a default which
1815
# will give us UTF-8 (which appears to work in all cases on
1816
# OSX). Users are still free to override LANG of course, as
1817
# long as it give us something meaningful. This work-around
1818
# *may* not be needed with python 3k and/or OSX 10.5, but will
1819
# work with them too -- vila 20080908
1820
os.environ['LANG'] = 'en_US.UTF-8'
1823
sys.platform = 'darwin'
1828
user_encoding = locale.getpreferredencoding()
1829
except locale.Error, e:
1830
sys.stderr.write('bzr: warning: %s\n'
1831
' Could not determine what text encoding to use.\n'
1832
' This error usually means your Python interpreter\n'
1833
' doesn\'t support the locale set by $LANG (%s)\n'
1834
" Continuing with ascii encoding.\n"
1835
% (e, os.environ.get('LANG')))
1836
user_encoding = 'ascii'
1838
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1839
# treat that as ASCII, and not support printing unicode characters to the
1842
# For python scripts run under vim, we get '', so also treat that as ASCII
1843
if user_encoding in (None, 'cp0', ''):
1844
user_encoding = 'ascii'
1848
codecs.lookup(user_encoding)
1850
sys.stderr.write('bzr: warning:'
1851
' unknown encoding %s.'
1852
' Continuing with ascii encoding.\n'
1855
user_encoding = 'ascii'
1858
_cached_user_encoding = user_encoding
1860
return user_encoding
1863
def get_host_name():
1864
"""Return the current unicode host name.
1866
This is meant to be used in place of socket.gethostname() because that
1867
behaves inconsistently on different platforms.
1869
if sys.platform == "win32":
1871
return win32utils.get_host_name()
1874
return socket.gethostname().decode(get_user_encoding())
1877
def recv_all(socket, bytes):
1878
"""Receive an exact number of bytes.
1880
Regular Socket.recv() may return less than the requested number of bytes,
1881
dependning on what's in the OS buffer. MSG_WAITALL is not available
1882
on all platforms, but this should work everywhere. This will return
1883
less than the requested amount if the remote end closes.
1885
This isn't optimized and is intended mostly for use in testing.
1888
while len(b) < bytes:
1889
new = until_no_eintr(socket.recv, bytes - len(b))
1896
def send_all(socket, bytes, report_activity=None):
1897
"""Send all bytes on a socket.
1899
Regular socket.sendall() can give socket error 10053 on Windows. This
1900
implementation sends no more than 64k at a time, which avoids this problem.
1902
:param report_activity: Call this as bytes are read, see
1903
Transport._report_activity
1906
for pos in xrange(0, len(bytes), chunk_size):
1907
block = bytes[pos:pos+chunk_size]
1908
if report_activity is not None:
1909
report_activity(len(block), 'write')
1910
until_no_eintr(socket.sendall, block)
1913
def dereference_path(path):
1914
"""Determine the real path to a file.
1916
All parent elements are dereferenced. But the file itself is not
1918
:param path: The original path. May be absolute or relative.
1919
:return: the real path *to* the file
1921
parent, base = os.path.split(path)
1922
# The pathjoin for '.' is a workaround for Python bug #1213894.
1923
# (initial path components aren't dereferenced)
1924
return pathjoin(realpath(pathjoin('.', parent)), base)
1927
def supports_mapi():
1928
"""Return True if we can use MAPI to launch a mail client."""
1929
return sys.platform == "win32"
1932
def resource_string(package, resource_name):
1933
"""Load a resource from a package and return it as a string.
1935
Note: Only packages that start with bzrlib are currently supported.
1937
This is designed to be a lightweight implementation of resource
1938
loading in a way which is API compatible with the same API from
1940
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1941
If and when pkg_resources becomes a standard library, this routine
1944
# Check package name is within bzrlib
1945
if package == "bzrlib":
1946
resource_relpath = resource_name
1947
elif package.startswith("bzrlib."):
1948
package = package[len("bzrlib."):].replace('.', os.sep)
1949
resource_relpath = pathjoin(package, resource_name)
1951
raise errors.BzrError('resource package %s not in bzrlib' % package)
1953
# Map the resource to a file and read its contents
1954
base = dirname(bzrlib.__file__)
1955
if getattr(sys, 'frozen', None): # bzr.exe
1956
base = abspath(pathjoin(base, '..', '..'))
1957
filename = pathjoin(base, resource_relpath)
1958
return open(filename, 'rU').read()
1961
def file_kind_from_stat_mode_thunk(mode):
1962
global file_kind_from_stat_mode
1963
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1965
from bzrlib._readdir_pyx import UTF8DirReader
1966
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1967
except ImportError, e:
1968
# This is one time where we won't warn that an extension failed to
1969
# load. The extension is never available on Windows anyway.
1970
from bzrlib._readdir_py import (
1971
_kind_from_mode as file_kind_from_stat_mode
1973
return file_kind_from_stat_mode(mode)
1974
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1977
def file_kind(f, _lstat=os.lstat):
1979
return file_kind_from_stat_mode(_lstat(f).st_mode)
1981
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1982
raise errors.NoSuchFile(f)
1986
def until_no_eintr(f, *a, **kw):
1987
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1988
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1992
except (IOError, OSError), e:
1993
if e.errno == errno.EINTR:
1997
def re_compile_checked(re_string, flags=0, where=""):
1998
"""Return a compiled re, or raise a sensible error.
2000
This should only be used when compiling user-supplied REs.
2002
:param re_string: Text form of regular expression.
2003
:param flags: eg re.IGNORECASE
2004
:param where: Message explaining to the user the context where
2005
it occurred, eg 'log search filter'.
2007
# from https://bugs.launchpad.net/bzr/+bug/251352
2009
re_obj = re.compile(re_string, flags)
2014
where = ' in ' + where
2015
# despite the name 'error' is a type
2016
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2017
% (where, re_string, e))
2020
if sys.platform == "win32":
2023
return msvcrt.getch()
2028
fd = sys.stdin.fileno()
2029
settings = termios.tcgetattr(fd)
2032
ch = sys.stdin.read(1)
2034
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2038
if sys.platform == 'linux2':
2039
def _local_concurrency():
2041
prefix = 'processor'
2042
for line in file('/proc/cpuinfo', 'rb'):
2043
if line.startswith(prefix):
2044
concurrency = int(line[line.find(':')+1:]) + 1
2046
elif sys.platform == 'darwin':
2047
def _local_concurrency():
2048
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2049
stdout=subprocess.PIPE).communicate()[0]
2050
elif sys.platform[0:7] == 'freebsd':
2051
def _local_concurrency():
2052
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2053
stdout=subprocess.PIPE).communicate()[0]
2054
elif sys.platform == 'sunos5':
2055
def _local_concurrency():
2056
return subprocess.Popen(['psrinfo', '-p',],
2057
stdout=subprocess.PIPE).communicate()[0]
2058
elif sys.platform == "win32":
2059
def _local_concurrency():
2060
# This appears to return the number of cores.
2061
return os.environ.get('NUMBER_OF_PROCESSORS')
2063
def _local_concurrency():
2068
_cached_local_concurrency = None
2070
def local_concurrency(use_cache=True):
2071
"""Return how many processes can be run concurrently.
2073
Rely on platform specific implementations and default to 1 (one) if
2074
anything goes wrong.
2076
global _cached_local_concurrency
2078
if _cached_local_concurrency is not None and use_cache:
2079
return _cached_local_concurrency
2081
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2082
if concurrency is None:
2084
concurrency = _local_concurrency()
2085
except (OSError, IOError):
2088
concurrency = int(concurrency)
2089
except (TypeError, ValueError):
2092
_cached_concurrency = concurrency
2096
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2097
"""A stream writer that doesn't decode str arguments."""
2099
def __init__(self, encode, stream, errors='strict'):
2100
codecs.StreamWriter.__init__(self, stream, errors)
2101
self.encode = encode
2103
def write(self, object):
2104
if type(object) is str:
2105
self.stream.write(object)
2107
data, _ = self.encode(object, self.errors)
2108
self.stream.write(data)