169
274
unlink_func(tmp_name)
171
276
rename_func(tmp_name, new)
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
277
if failure_exc is not None:
278
raise failure_exc[0], failure_exc[1], failure_exc[2]
281
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
282
# choke on a Unicode string containing a relative path if
283
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
285
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
286
def _posix_abspath(path):
287
# jam 20060426 rather than encoding to fsencoding
288
# copy posixpath.abspath, but use os.getcwdu instead
289
if not posixpath.isabs(path):
290
path = posixpath.join(getcwd(), path)
291
return posixpath.normpath(path)
294
def _posix_realpath(path):
295
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
298
def _win32_fixdrive(path):
299
"""Force drive letters to be consistent.
301
win32 is inconsistent whether it returns lower or upper case
302
and even if it was consistent the user might type the other
303
so we force it to uppercase
304
running python.exe under cmd.exe return capital C:\\
305
running win32 python inside a cygwin shell returns lowercase c:\\
307
drive, path = _nt_splitdrive(path)
308
return drive.upper() + path
311
def _win32_abspath(path):
312
# Real _nt_abspath doesn't have a problem with a unicode cwd
313
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
316
def _win98_abspath(path):
317
"""Return the absolute version of a path.
318
Windows 98 safe implementation (python reimplementation
319
of Win32 API function GetFullPathNameW)
324
# \\HOST\path => //HOST/path
325
# //HOST/path => //HOST/path
326
# path => C:/cwd/path
329
# check for absolute path
330
drive = _nt_splitdrive(path)[0]
331
if drive == '' and path[:2] not in('//','\\\\'):
333
# we cannot simply os.path.join cwd and path
334
# because os.path.join('C:','/path') produce '/path'
335
# and this is incorrect
336
if path[:1] in ('/','\\'):
337
cwd = _nt_splitdrive(cwd)[0]
339
path = cwd + '\\' + path
340
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
343
def _win32_realpath(path):
344
# Real _nt_realpath doesn't have a problem with a unicode cwd
345
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
348
def _win32_pathjoin(*args):
349
return _nt_join(*args).replace('\\', '/')
352
def _win32_normpath(path):
353
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
357
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
360
def _win32_mkdtemp(*args, **kwargs):
361
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
364
def _win32_rename(old, new):
365
"""We expect to be able to atomically replace 'new' with old.
367
On win32, if new exists, it must be moved out of the way first,
371
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
373
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
374
# If we try to rename a non-existant file onto cwd, we get
375
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
376
# if the old path doesn't exist, sometimes we get EACCES
377
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
383
return unicodedata.normalize('NFC', os.getcwdu())
386
# Default is to just use the python builtins, but these can be rebound on
387
# particular platforms.
388
abspath = _posix_abspath
389
realpath = _posix_realpath
176
390
pathjoin = os.path.join
177
391
normpath = os.path.normpath
178
392
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
180
393
rename = os.rename
181
394
dirname = os.path.dirname
182
395
basename = os.path.basename
184
if os.name == "posix":
185
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
186
# choke on a Unicode string containing a relative path if
187
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
189
_fs_enc = sys.getfilesystemencoding()
191
return os.path.abspath(path.encode(_fs_enc)).decode(_fs_enc)
194
return os.path.realpath(path.encode(_fs_enc)).decode(_fs_enc)
396
split = os.path.split
397
splitext = os.path.splitext
398
# These were already imported into local scope
399
# mkdtemp = tempfile.mkdtemp
400
# rmtree = shutil.rmtree
402
MIN_ABS_PATHLENGTH = 1
196
405
if sys.platform == 'win32':
197
# We need to use the Unicode-aware os.path.abspath and
198
# os.path.realpath on Windows systems.
200
return os.path.abspath(path).replace('\\', '/')
203
return os.path.realpath(path).replace('\\', '/')
206
return os.path.join(*args).replace('\\', '/')
209
return os.path.normpath(path).replace('\\', '/')
212
return os.getcwdu().replace('\\', '/')
214
def mkdtemp(*args, **kwargs):
215
return tempfile.mkdtemp(*args, **kwargs).replace('\\', '/')
217
def rename(old, new):
218
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
406
if win32utils.winver == 'Windows 98':
407
abspath = _win98_abspath
409
abspath = _win32_abspath
410
realpath = _win32_realpath
411
pathjoin = _win32_pathjoin
412
normpath = _win32_normpath
413
getcwd = _win32_getcwd
414
mkdtemp = _win32_mkdtemp
415
rename = _win32_rename
417
MIN_ABS_PATHLENGTH = 3
419
def _win32_delete_readonly(function, path, excinfo):
420
"""Error handler for shutil.rmtree function [for win32]
421
Helps to remove files and dirs marked as read-only.
423
exception = excinfo[1]
424
if function in (os.remove, os.rmdir) \
425
and isinstance(exception, OSError) \
426
and exception.errno == errno.EACCES:
432
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
433
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
434
return shutil.rmtree(path, ignore_errors, onerror)
436
f = win32utils.get_unicode_argv # special function or None
440
elif sys.platform == 'darwin':
444
def get_terminal_encoding():
445
"""Find the best encoding for printing to the screen.
447
This attempts to check both sys.stdout and sys.stdin to see
448
what encoding they are in, and if that fails it falls back to
449
osutils.get_user_encoding().
450
The problem is that on Windows, locale.getpreferredencoding()
451
is not the same encoding as that used by the console:
452
http://mail.python.org/pipermail/python-list/2003-May/162357.html
454
On my standard US Windows XP, the preferred encoding is
455
cp1252, but the console is cp437
457
from bzrlib.trace import mutter
458
output_encoding = getattr(sys.stdout, 'encoding', None)
459
if not output_encoding:
460
input_encoding = getattr(sys.stdin, 'encoding', None)
461
if not input_encoding:
462
output_encoding = get_user_encoding()
463
mutter('encoding stdout as osutils.get_user_encoding() %r',
466
output_encoding = input_encoding
467
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
469
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
470
if output_encoding == 'cp0':
471
# invalid encoding (cp0 means 'no codepage' on Windows)
472
output_encoding = get_user_encoding()
473
mutter('cp0 is invalid encoding.'
474
' encoding stdout as osutils.get_user_encoding() %r',
478
codecs.lookup(output_encoding)
480
sys.stderr.write('bzr: warning:'
481
' unknown terminal encoding %s.\n'
482
' Using encoding %s instead.\n'
483
% (output_encoding, get_user_encoding())
485
output_encoding = get_user_encoding()
487
return output_encoding
221
490
def normalizepath(f):
222
if hasattr(os.path, 'realpath'):
491
if getattr(os.path, 'realpath', None) is not None:
594
1138
os.path.commonprefix (python2.4) has a bad bug that it works just
595
1139
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
596
1140
avoids that problem.
1142
NOTE: `base` should not have a trailing slash otherwise you'll get
1143
PathNotChild exceptions regardless of `path`.
598
if sys.platform != "win32":
602
assert len(base) >= minlength, ('Length of base must be equal or exceed the'
603
' platform minimum length (which is %d)' % minlength)
1146
if len(base) < MIN_ABS_PATHLENGTH:
1147
# must have space for e.g. a drive letter
1148
raise ValueError('%r is too short to calculate a relative path'
604
1151
rp = abspath(path)
608
while len(head) >= len(base):
1156
if len(head) <= len(base) and head != base:
1157
raise errors.PathNotChild(rp, base)
609
1158
if head == base:
611
head, tail = os.path.split(head)
1160
head, tail = split(head)
615
# XXX This should raise a NotChildPath exception, as its not tied
617
raise PathNotChild(rp, base)
1165
return pathjoin(*reversed(s))
1170
def _cicp_canonical_relpath(base, path):
1171
"""Return the canonical path relative to base.
1173
Like relpath, but on case-insensitive-case-preserving file-systems, this
1174
will return the relpath as stored on the file-system rather than in the
1175
case specified in the input string, for all existing portions of the path.
1177
This will cause O(N) behaviour if called for every path in a tree; if you
1178
have a number of paths to convert, you should use canonical_relpaths().
1180
# TODO: it should be possible to optimize this for Windows by using the
1181
# win32 API FindFiles function to look for the specified name - but using
1182
# os.listdir() still gives us the correct, platform agnostic semantics in
1185
rel = relpath(base, path)
1186
# '.' will have been turned into ''
1190
abs_base = abspath(base)
1192
_listdir = os.listdir
1194
# use an explicit iterator so we can easily consume the rest on early exit.
1195
bit_iter = iter(rel.split('/'))
1196
for bit in bit_iter:
1199
next_entries = _listdir(current)
1200
except OSError: # enoent, eperm, etc
1201
# We can't find this in the filesystem, so just append the
1203
current = pathjoin(current, bit, *list(bit_iter))
1205
for look in next_entries:
1206
if lbit == look.lower():
1207
current = pathjoin(current, look)
1210
# got to the end, nothing matched, so we just return the
1211
# non-existing bits as they were specified (the filename may be
1212
# the target of a move, for example).
1213
current = pathjoin(current, bit, *list(bit_iter))
1215
return current[len(abs_base):].lstrip('/')
1217
# XXX - TODO - we need better detection/integration of case-insensitive
1218
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1219
# filesystems), for example, so could probably benefit from the same basic
1220
# support there. For now though, only Windows and OSX get that support, and
1221
# they get it for *all* file-systems!
1222
if sys.platform in ('win32', 'darwin'):
1223
canonical_relpath = _cicp_canonical_relpath
1225
canonical_relpath = relpath
1227
def canonical_relpaths(base, paths):
1228
"""Create an iterable to canonicalize a sequence of relative paths.
1230
The intent is for this implementation to use a cache, vastly speeding
1231
up multiple transformations in the same directory.
1233
# but for now, we haven't optimized...
1234
return [canonical_relpath(base, p) for p in paths]
625
1236
def safe_unicode(unicode_or_utf8_string):
626
1237
"""Coerce unicode_or_utf8_string into unicode.
628
1239
If it is unicode, it is returned.
629
Otherwise it is decoded from utf-8. If a decoding error
630
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
631
as a BzrBadParameter exception.
1240
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1241
wrapped in a BzrBadParameterNotUnicode exception.
633
1243
if isinstance(unicode_or_utf8_string, unicode):
634
1244
return unicode_or_utf8_string
636
1246
return unicode_or_utf8_string.decode('utf8')
637
1247
except UnicodeDecodeError:
638
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
1248
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1251
def safe_utf8(unicode_or_utf8_string):
1252
"""Coerce unicode_or_utf8_string to a utf8 string.
1254
If it is a str, it is returned.
1255
If it is Unicode, it is encoded into a utf-8 string.
1257
if isinstance(unicode_or_utf8_string, str):
1258
# TODO: jam 20070209 This is overkill, and probably has an impact on
1259
# performance if we are dealing with lots of apis that want a
1262
# Make sure it is a valid utf-8 string
1263
unicode_or_utf8_string.decode('utf-8')
1264
except UnicodeDecodeError:
1265
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1266
return unicode_or_utf8_string
1267
return unicode_or_utf8_string.encode('utf-8')
1270
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1271
' Revision id generators should be creating utf8'
1275
def safe_revision_id(unicode_or_utf8_string, warn=True):
1276
"""Revision ids should now be utf8, but at one point they were unicode.
1278
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1280
:param warn: Functions that are sanitizing user data can set warn=False
1281
:return: None or a utf8 revision id.
1283
if (unicode_or_utf8_string is None
1284
or unicode_or_utf8_string.__class__ == str):
1285
return unicode_or_utf8_string
1287
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1289
return cache_utf8.encode(unicode_or_utf8_string)
1292
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1293
' generators should be creating utf8 file ids.')
1296
def safe_file_id(unicode_or_utf8_string, warn=True):
1297
"""File ids should now be utf8, but at one point they were unicode.
1299
This is the same as safe_utf8, except it uses the cached encode functions
1300
to save a little bit of performance.
1302
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1304
:param warn: Functions that are sanitizing user data can set warn=False
1305
:return: None or a utf8 file id.
1307
if (unicode_or_utf8_string is None
1308
or unicode_or_utf8_string.__class__ == str):
1309
return unicode_or_utf8_string
1311
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1313
return cache_utf8.encode(unicode_or_utf8_string)
1316
_platform_normalizes_filenames = False
1317
if sys.platform == 'darwin':
1318
_platform_normalizes_filenames = True
1321
def normalizes_filenames():
1322
"""Return True if this platform normalizes unicode filenames.
1324
Mac OSX does, Windows/Linux do not.
1326
return _platform_normalizes_filenames
1329
def _accessible_normalized_filename(path):
1330
"""Get the unicode normalized path, and if you can access the file.
1332
On platforms where the system normalizes filenames (Mac OSX),
1333
you can access a file by any path which will normalize correctly.
1334
On platforms where the system does not normalize filenames
1335
(Windows, Linux), you have to access a file by its exact path.
1337
Internally, bzr only supports NFC normalization, since that is
1338
the standard for XML documents.
1340
So return the normalized path, and a flag indicating if the file
1341
can be accessed by that path.
1344
return unicodedata.normalize('NFC', unicode(path)), True
1347
def _inaccessible_normalized_filename(path):
1348
__doc__ = _accessible_normalized_filename.__doc__
1350
normalized = unicodedata.normalize('NFC', unicode(path))
1351
return normalized, normalized == path
1354
if _platform_normalizes_filenames:
1355
normalized_filename = _accessible_normalized_filename
1357
normalized_filename = _inaccessible_normalized_filename
1360
def set_signal_handler(signum, handler, restart_syscall=True):
1361
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1362
on platforms that support that.
1364
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1365
automatically restart (by calling `signal.siginterrupt(signum,
1366
False)`). May be ignored if the feature is not available on this
1367
platform or Python version.
1371
siginterrupt = signal.siginterrupt
1373
# This python implementation doesn't provide signal support, hence no
1376
except AttributeError:
1377
# siginterrupt doesn't exist on this platform, or for this version
1379
siginterrupt = lambda signum, flag: None
1381
def sig_handler(*args):
1382
# Python resets the siginterrupt flag when a signal is
1383
# received. <http://bugs.python.org/issue8354>
1384
# As a workaround for some cases, set it back the way we want it.
1385
siginterrupt(signum, False)
1386
# Now run the handler function passed to set_signal_handler.
1389
sig_handler = handler
1390
old_handler = signal.signal(signum, sig_handler)
1392
siginterrupt(signum, False)
1396
default_terminal_width = 80
1397
"""The default terminal width for ttys.
1399
This is defined so that higher levels can share a common fallback value when
1400
terminal_width() returns None.
641
1404
def terminal_width():
642
"""Return estimated terminal width."""
644
# TODO: Do something smart on Windows?
646
# TODO: Is there anything that gets a better update when the window
647
# is resized while the program is running? We could use the Python termcap
1405
"""Return terminal width.
1407
None is returned if the width can't established precisely.
1410
- if BZR_COLUMNS is set, returns its value
1411
- if there is no controlling terminal, returns None
1412
- if COLUMNS is set, returns its value,
1414
From there, we need to query the OS to get the size of the controlling
1418
- get termios.TIOCGWINSZ
1419
- if an error occurs or a negative value is obtained, returns None
1423
- win32utils.get_console_size() decides,
1424
- returns None on error (provided default value)
1427
# If BZR_COLUMNS is set, take it, user is always right
1429
return int(os.environ['BZR_COLUMNS'])
1430
except (KeyError, ValueError):
1433
isatty = getattr(sys.stdout, 'isatty', None)
1434
if isatty is None or not isatty():
1435
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1438
# If COLUMNS is set, take it, the terminal knows better (even inside a
1439
# given terminal, the application can decide to set COLUMNS to a lower
1440
# value (splitted screen) or a bigger value (scroll bars))
650
1442
return int(os.environ['COLUMNS'])
651
except (IndexError, KeyError, ValueError):
1443
except (KeyError, ValueError):
1446
width, height = _terminal_size(None, None)
1448
# Consider invalid values as meaning no width
1454
def _win32_terminal_size(width, height):
1455
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1456
return width, height
1459
def _ioctl_terminal_size(width, height):
1461
import struct, fcntl, termios
1462
s = struct.pack('HHHH', 0, 0, 0, 0)
1463
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1464
height, width = struct.unpack('HHHH', x)[0:2]
1465
except (IOError, AttributeError):
1467
return width, height
1469
_terminal_size = None
1470
"""Returns the terminal size as (width, height).
1472
:param width: Default value for width.
1473
:param height: Default value for height.
1475
This is defined specifically for each OS and query the size of the controlling
1476
terminal. If any error occurs, the provided default values should be returned.
1478
if sys.platform == 'win32':
1479
_terminal_size = _win32_terminal_size
1481
_terminal_size = _ioctl_terminal_size
1484
def _terminal_size_changed(signum, frame):
1485
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1486
width, height = _terminal_size(None, None)
1487
if width is not None:
1488
os.environ['COLUMNS'] = str(width)
1491
_registered_sigwinch = False
1492
def watch_sigwinch():
1493
"""Register for SIGWINCH, once and only once.
1495
Do nothing if the signal module is not available.
1497
global _registered_sigwinch
1498
if not _registered_sigwinch:
1501
if getattr(signal, "SIGWINCH", None) is not None:
1502
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1504
# python doesn't provide signal support, nothing we can do about it
1506
_registered_sigwinch = True
654
1509
def supports_executable():
655
1510
return sys.platform != "win32"
1513
def supports_posix_readonly():
1514
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1516
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1517
directory controls creation/deletion, etc.
1519
And under win32, readonly means that the directory itself cannot be
1520
deleted. The contents of a readonly directory can be changed, unlike POSIX
1521
where files in readonly directories cannot be added, deleted or renamed.
1523
return sys.platform != "win32"
1526
def set_or_unset_env(env_variable, value):
1527
"""Modify the environment, setting or removing the env_variable.
1529
:param env_variable: The environment variable in question
1530
:param value: The value to set the environment to. If None, then
1531
the variable will be removed.
1532
:return: The original value of the environment variable.
1534
orig_val = os.environ.get(env_variable)
1536
if orig_val is not None:
1537
del os.environ[env_variable]
1539
if isinstance(value, unicode):
1540
value = value.encode(get_user_encoding())
1541
os.environ[env_variable] = value
1545
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1548
def check_legal_path(path):
1549
"""Check whether the supplied path is legal.
1550
This is only required on Windows, so we don't test on other platforms
1553
if sys.platform != "win32":
1555
if _validWin32PathRE.match(path) is None:
1556
raise errors.IllegalPath(path)
1559
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1561
def _is_error_enotdir(e):
1562
"""Check if this exception represents ENOTDIR.
1564
Unfortunately, python is very inconsistent about the exception
1565
here. The cases are:
1566
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1567
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1568
which is the windows error code.
1569
3) Windows, Python2.5 uses errno == EINVAL and
1570
winerror == ERROR_DIRECTORY
1572
:param e: An Exception object (expected to be OSError with an errno
1573
attribute, but we should be able to cope with anything)
1574
:return: True if this represents an ENOTDIR error. False otherwise.
1576
en = getattr(e, 'errno', None)
1577
if (en == errno.ENOTDIR
1578
or (sys.platform == 'win32'
1579
and (en == _WIN32_ERROR_DIRECTORY
1580
or (en == errno.EINVAL
1581
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1587
def walkdirs(top, prefix=""):
1588
"""Yield data about all the directories in a tree.
1590
This yields all the data about the contents of a directory at a time.
1591
After each directory has been yielded, if the caller has mutated the list
1592
to exclude some directories, they are then not descended into.
1594
The data yielded is of the form:
1595
((directory-relpath, directory-path-from-top),
1596
[(relpath, basename, kind, lstat, path-from-top), ...]),
1597
- directory-relpath is the relative path of the directory being returned
1598
with respect to top. prefix is prepended to this.
1599
- directory-path-from-root is the path including top for this directory.
1600
It is suitable for use with os functions.
1601
- relpath is the relative path within the subtree being walked.
1602
- basename is the basename of the path
1603
- kind is the kind of the file now. If unknown then the file is not
1604
present within the tree - but it may be recorded as versioned. See
1606
- lstat is the stat data *if* the file was statted.
1607
- planned, not implemented:
1608
path_from_tree_root is the path from the root of the tree.
1610
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1611
allows one to walk a subtree but get paths that are relative to a tree
1613
:return: an iterator over the dirs.
1615
#TODO there is a bit of a smell where the results of the directory-
1616
# summary in this, and the path from the root, may not agree
1617
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1618
# potentially confusing output. We should make this more robust - but
1619
# not at a speed cost. RBC 20060731
1621
_directory = _directory_kind
1622
_listdir = os.listdir
1623
_kind_from_mode = file_kind_from_stat_mode
1624
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1626
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1627
relroot, _, _, _, top = pending.pop()
1629
relprefix = relroot + u'/'
1632
top_slash = top + u'/'
1635
append = dirblock.append
1637
names = sorted(_listdir(top))
1639
if not _is_error_enotdir(e):
1643
abspath = top_slash + name
1644
statvalue = _lstat(abspath)
1645
kind = _kind_from_mode(statvalue.st_mode)
1646
append((relprefix + name, name, kind, statvalue, abspath))
1647
yield (relroot, top), dirblock
1649
# push the user specified dirs from dirblock
1650
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1653
class DirReader(object):
1654
"""An interface for reading directories."""
1656
def top_prefix_to_starting_dir(self, top, prefix=""):
1657
"""Converts top and prefix to a starting dir entry
1659
:param top: A utf8 path
1660
:param prefix: An optional utf8 path to prefix output relative paths
1662
:return: A tuple starting with prefix, and ending with the native
1665
raise NotImplementedError(self.top_prefix_to_starting_dir)
1667
def read_dir(self, prefix, top):
1668
"""Read a specific dir.
1670
:param prefix: A utf8 prefix to be preprended to the path basenames.
1671
:param top: A natively encoded path to read.
1672
:return: A list of the directories contents. Each item contains:
1673
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1675
raise NotImplementedError(self.read_dir)
1678
_selected_dir_reader = None
1681
def _walkdirs_utf8(top, prefix=""):
1682
"""Yield data about all the directories in a tree.
1684
This yields the same information as walkdirs() only each entry is yielded
1685
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1686
are returned as exact byte-strings.
1688
:return: yields a tuple of (dir_info, [file_info])
1689
dir_info is (utf8_relpath, path-from-top)
1690
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1691
if top is an absolute path, path-from-top is also an absolute path.
1692
path-from-top might be unicode or utf8, but it is the correct path to
1693
pass to os functions to affect the file in question. (such as os.lstat)
1695
global _selected_dir_reader
1696
if _selected_dir_reader is None:
1697
fs_encoding = _fs_enc.upper()
1698
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1699
# Win98 doesn't have unicode apis like FindFirstFileW
1700
# TODO: We possibly could support Win98 by falling back to the
1701
# original FindFirstFile, and using TCHAR instead of WCHAR,
1702
# but that gets a bit tricky, and requires custom compiling
1705
from bzrlib._walkdirs_win32 import Win32ReadDir
1706
_selected_dir_reader = Win32ReadDir()
1709
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1710
# ANSI_X3.4-1968 is a form of ASCII
1712
from bzrlib._readdir_pyx import UTF8DirReader
1713
_selected_dir_reader = UTF8DirReader()
1714
except ImportError, e:
1715
failed_to_load_extension(e)
1718
if _selected_dir_reader is None:
1719
# Fallback to the python version
1720
_selected_dir_reader = UnicodeDirReader()
1722
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1723
# But we don't actually uses 1-3 in pending, so set them to None
1724
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1725
read_dir = _selected_dir_reader.read_dir
1726
_directory = _directory_kind
1728
relroot, _, _, _, top = pending[-1].pop()
1731
dirblock = sorted(read_dir(relroot, top))
1732
yield (relroot, top), dirblock
1733
# push the user specified dirs from dirblock
1734
next = [d for d in reversed(dirblock) if d[2] == _directory]
1736
pending.append(next)
1739
class UnicodeDirReader(DirReader):
1740
"""A dir reader for non-utf8 file systems, which transcodes."""
1742
__slots__ = ['_utf8_encode']
1745
self._utf8_encode = codecs.getencoder('utf8')
1747
def top_prefix_to_starting_dir(self, top, prefix=""):
1748
"""See DirReader.top_prefix_to_starting_dir."""
1749
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1751
def read_dir(self, prefix, top):
1752
"""Read a single directory from a non-utf8 file system.
1754
top, and the abspath element in the output are unicode, all other paths
1755
are utf8. Local disk IO is done via unicode calls to listdir etc.
1757
This is currently the fallback code path when the filesystem encoding is
1758
not UTF-8. It may be better to implement an alternative so that we can
1759
safely handle paths that are not properly decodable in the current
1762
See DirReader.read_dir for details.
1764
_utf8_encode = self._utf8_encode
1766
_listdir = os.listdir
1767
_kind_from_mode = file_kind_from_stat_mode
1770
relprefix = prefix + '/'
1773
top_slash = top + u'/'
1776
append = dirblock.append
1777
for name in sorted(_listdir(top)):
1779
name_utf8 = _utf8_encode(name)[0]
1780
except UnicodeDecodeError:
1781
raise errors.BadFilenameEncoding(
1782
_utf8_encode(relprefix)[0] + name, _fs_enc)
1783
abspath = top_slash + name
1784
statvalue = _lstat(abspath)
1785
kind = _kind_from_mode(statvalue.st_mode)
1786
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1790
def copy_tree(from_path, to_path, handlers={}):
1791
"""Copy all of the entries in from_path into to_path.
1793
:param from_path: The base directory to copy.
1794
:param to_path: The target directory. If it does not exist, it will
1796
:param handlers: A dictionary of functions, which takes a source and
1797
destinations for files, directories, etc.
1798
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1799
'file', 'directory', and 'symlink' should always exist.
1800
If they are missing, they will be replaced with 'os.mkdir()',
1801
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1803
# Now, just copy the existing cached tree to the new location
1804
# We use a cheap trick here.
1805
# Absolute paths are prefixed with the first parameter
1806
# relative paths are prefixed with the second.
1807
# So we can get both the source and target returned
1808
# without any extra work.
1810
def copy_dir(source, dest):
1813
def copy_link(source, dest):
1814
"""Copy the contents of a symlink"""
1815
link_to = os.readlink(source)
1816
os.symlink(link_to, dest)
1818
real_handlers = {'file':shutil.copy2,
1819
'symlink':copy_link,
1820
'directory':copy_dir,
1822
real_handlers.update(handlers)
1824
if not os.path.exists(to_path):
1825
real_handlers['directory'](from_path, to_path)
1827
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1828
for relpath, name, kind, st, abspath in entries:
1829
real_handlers[kind](abspath, relpath)
1832
def copy_ownership_from_path(dst, src=None):
1833
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1835
If src is None, the containing directory is used as source. If chown
1836
fails, the error is ignored and a warning is printed.
1838
chown = getattr(os, 'chown', None)
1843
src = os.path.dirname(dst)
1849
chown(dst, s.st_uid, s.st_gid)
1851
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1854
def path_prefix_key(path):
1855
"""Generate a prefix-order path key for path.
1857
This can be used to sort paths in the same way that walkdirs does.
1859
return (dirname(path) , path)
1862
def compare_paths_prefix_order(path_a, path_b):
1863
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1864
key_a = path_prefix_key(path_a)
1865
key_b = path_prefix_key(path_b)
1866
return cmp(key_a, key_b)
1869
_cached_user_encoding = None
1872
def get_user_encoding(use_cache=True):
1873
"""Find out what the preferred user encoding is.
1875
This is generally the encoding that is used for command line parameters
1876
and file contents. This may be different from the terminal encoding
1877
or the filesystem encoding.
1879
:param use_cache: Enable cache for detected encoding.
1880
(This parameter is turned on by default,
1881
and required only for selftesting)
1883
:return: A string defining the preferred user encoding
1885
global _cached_user_encoding
1886
if _cached_user_encoding is not None and use_cache:
1887
return _cached_user_encoding
1889
if sys.platform == 'darwin':
1890
# python locale.getpreferredencoding() always return
1891
# 'mac-roman' on darwin. That's a lie.
1892
sys.platform = 'posix'
1894
if os.environ.get('LANG', None) is None:
1895
# If LANG is not set, we end up with 'ascii', which is bad
1896
# ('mac-roman' is more than ascii), so we set a default which
1897
# will give us UTF-8 (which appears to work in all cases on
1898
# OSX). Users are still free to override LANG of course, as
1899
# long as it give us something meaningful. This work-around
1900
# *may* not be needed with python 3k and/or OSX 10.5, but will
1901
# work with them too -- vila 20080908
1902
os.environ['LANG'] = 'en_US.UTF-8'
1905
sys.platform = 'darwin'
1910
user_encoding = locale.getpreferredencoding()
1911
except locale.Error, e:
1912
sys.stderr.write('bzr: warning: %s\n'
1913
' Could not determine what text encoding to use.\n'
1914
' This error usually means your Python interpreter\n'
1915
' doesn\'t support the locale set by $LANG (%s)\n'
1916
" Continuing with ascii encoding.\n"
1917
% (e, os.environ.get('LANG')))
1918
user_encoding = 'ascii'
1920
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1921
# treat that as ASCII, and not support printing unicode characters to the
1924
# For python scripts run under vim, we get '', so also treat that as ASCII
1925
if user_encoding in (None, 'cp0', ''):
1926
user_encoding = 'ascii'
1930
codecs.lookup(user_encoding)
1932
sys.stderr.write('bzr: warning:'
1933
' unknown encoding %s.'
1934
' Continuing with ascii encoding.\n'
1937
user_encoding = 'ascii'
1940
_cached_user_encoding = user_encoding
1942
return user_encoding
1945
def get_host_name():
1946
"""Return the current unicode host name.
1948
This is meant to be used in place of socket.gethostname() because that
1949
behaves inconsistently on different platforms.
1951
if sys.platform == "win32":
1953
return win32utils.get_host_name()
1956
return socket.gethostname().decode(get_user_encoding())
1959
# We must not read/write any more than 64k at a time from/to a socket so we
1960
# don't risk "no buffer space available" errors on some platforms. Windows in
1961
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1963
MAX_SOCKET_CHUNK = 64 * 1024
1965
def read_bytes_from_socket(sock, report_activity=None,
1966
max_read_size=MAX_SOCKET_CHUNK):
1967
"""Read up to max_read_size of bytes from sock and notify of progress.
1969
Translates "Connection reset by peer" into file-like EOF (return an
1970
empty string rather than raise an error), and repeats the recv if
1971
interrupted by a signal.
1975
bytes = sock.recv(max_read_size)
1976
except socket.error, e:
1978
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1979
# The connection was closed by the other side. Callers expect
1980
# an empty string to signal end-of-stream.
1982
elif eno == errno.EINTR:
1983
# Retry the interrupted recv.
1987
if report_activity is not None:
1988
report_activity(len(bytes), 'read')
1992
def recv_all(socket, count):
1993
"""Receive an exact number of bytes.
1995
Regular Socket.recv() may return less than the requested number of bytes,
1996
depending on what's in the OS buffer. MSG_WAITALL is not available
1997
on all platforms, but this should work everywhere. This will return
1998
less than the requested amount if the remote end closes.
2000
This isn't optimized and is intended mostly for use in testing.
2003
while len(b) < count:
2004
new = read_bytes_from_socket(socket, None, count - len(b))
2011
def send_all(sock, bytes, report_activity=None):
2012
"""Send all bytes on a socket.
2014
Breaks large blocks in smaller chunks to avoid buffering limitations on
2015
some platforms, and catches EINTR which may be thrown if the send is
2016
interrupted by a signal.
2018
This is preferred to socket.sendall(), because it avoids portability bugs
2019
and provides activity reporting.
2021
:param report_activity: Call this as bytes are read, see
2022
Transport._report_activity
2025
byte_count = len(bytes)
2026
while sent_total < byte_count:
2028
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2029
except socket.error, e:
2030
if e.args[0] != errno.EINTR:
2034
report_activity(sent, 'write')
2037
def dereference_path(path):
2038
"""Determine the real path to a file.
2040
All parent elements are dereferenced. But the file itself is not
2042
:param path: The original path. May be absolute or relative.
2043
:return: the real path *to* the file
2045
parent, base = os.path.split(path)
2046
# The pathjoin for '.' is a workaround for Python bug #1213894.
2047
# (initial path components aren't dereferenced)
2048
return pathjoin(realpath(pathjoin('.', parent)), base)
2051
def supports_mapi():
2052
"""Return True if we can use MAPI to launch a mail client."""
2053
return sys.platform == "win32"
2056
def resource_string(package, resource_name):
2057
"""Load a resource from a package and return it as a string.
2059
Note: Only packages that start with bzrlib are currently supported.
2061
This is designed to be a lightweight implementation of resource
2062
loading in a way which is API compatible with the same API from
2064
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2065
If and when pkg_resources becomes a standard library, this routine
2068
# Check package name is within bzrlib
2069
if package == "bzrlib":
2070
resource_relpath = resource_name
2071
elif package.startswith("bzrlib."):
2072
package = package[len("bzrlib."):].replace('.', os.sep)
2073
resource_relpath = pathjoin(package, resource_name)
2075
raise errors.BzrError('resource package %s not in bzrlib' % package)
2077
# Map the resource to a file and read its contents
2078
base = dirname(bzrlib.__file__)
2079
if getattr(sys, 'frozen', None): # bzr.exe
2080
base = abspath(pathjoin(base, '..', '..'))
2081
f = file(pathjoin(base, resource_relpath), "rU")
2087
def file_kind_from_stat_mode_thunk(mode):
2088
global file_kind_from_stat_mode
2089
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2091
from bzrlib._readdir_pyx import UTF8DirReader
2092
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2093
except ImportError, e:
2094
# This is one time where we won't warn that an extension failed to
2095
# load. The extension is never available on Windows anyway.
2096
from bzrlib._readdir_py import (
2097
_kind_from_mode as file_kind_from_stat_mode
2099
return file_kind_from_stat_mode(mode)
2100
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2103
def file_kind(f, _lstat=os.lstat):
2105
return file_kind_from_stat_mode(_lstat(f).st_mode)
2107
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2108
raise errors.NoSuchFile(f)
2112
def until_no_eintr(f, *a, **kw):
2113
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2115
WARNING: you must be certain that it is safe to retry the call repeatedly
2116
if EINTR does occur. This is typically only true for low-level operations
2117
like os.read. If in any doubt, don't use this.
2119
Keep in mind that this is not a complete solution to EINTR. There is
2120
probably code in the Python standard library and other dependencies that
2121
may encounter EINTR if a signal arrives (and there is signal handler for
2122
that signal). So this function can reduce the impact for IO that bzrlib
2123
directly controls, but it is not a complete solution.
2125
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2129
except (IOError, OSError), e:
2130
if e.errno == errno.EINTR:
2135
def re_compile_checked(re_string, flags=0, where=""):
2136
"""Return a compiled re, or raise a sensible error.
2138
This should only be used when compiling user-supplied REs.
2140
:param re_string: Text form of regular expression.
2141
:param flags: eg re.IGNORECASE
2142
:param where: Message explaining to the user the context where
2143
it occurred, eg 'log search filter'.
2145
# from https://bugs.launchpad.net/bzr/+bug/251352
2147
re_obj = re.compile(re_string, flags)
2152
where = ' in ' + where
2153
# despite the name 'error' is a type
2154
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2155
% (where, re_string, e))
2158
if sys.platform == "win32":
2161
return msvcrt.getch()
2166
fd = sys.stdin.fileno()
2167
settings = termios.tcgetattr(fd)
2170
ch = sys.stdin.read(1)
2172
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2176
if sys.platform == 'linux2':
2177
def _local_concurrency():
2179
prefix = 'processor'
2180
for line in file('/proc/cpuinfo', 'rb'):
2181
if line.startswith(prefix):
2182
concurrency = int(line[line.find(':')+1:]) + 1
2184
elif sys.platform == 'darwin':
2185
def _local_concurrency():
2186
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2187
stdout=subprocess.PIPE).communicate()[0]
2188
elif sys.platform[0:7] == 'freebsd':
2189
def _local_concurrency():
2190
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2191
stdout=subprocess.PIPE).communicate()[0]
2192
elif sys.platform == 'sunos5':
2193
def _local_concurrency():
2194
return subprocess.Popen(['psrinfo', '-p',],
2195
stdout=subprocess.PIPE).communicate()[0]
2196
elif sys.platform == "win32":
2197
def _local_concurrency():
2198
# This appears to return the number of cores.
2199
return os.environ.get('NUMBER_OF_PROCESSORS')
2201
def _local_concurrency():
2206
_cached_local_concurrency = None
2208
def local_concurrency(use_cache=True):
2209
"""Return how many processes can be run concurrently.
2211
Rely on platform specific implementations and default to 1 (one) if
2212
anything goes wrong.
2214
global _cached_local_concurrency
2216
if _cached_local_concurrency is not None and use_cache:
2217
return _cached_local_concurrency
2219
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2220
if concurrency is None:
2222
concurrency = _local_concurrency()
2223
except (OSError, IOError):
2226
concurrency = int(concurrency)
2227
except (TypeError, ValueError):
2230
_cached_concurrency = concurrency
2234
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2235
"""A stream writer that doesn't decode str arguments."""
2237
def __init__(self, encode, stream, errors='strict'):
2238
codecs.StreamWriter.__init__(self, stream, errors)
2239
self.encode = encode
2241
def write(self, object):
2242
if type(object) is str:
2243
self.stream.write(object)
2245
data, _ = self.encode(object, self.errors)
2246
self.stream.write(data)
2248
if sys.platform == 'win32':
2249
def open_file(filename, mode='r', bufsize=-1):
2250
"""This function is used to override the ``open`` builtin.
2252
But it uses O_NOINHERIT flag so the file handle is not inherited by
2253
child processes. Deleting or renaming a closed file opened with this
2254
function is not blocking child processes.
2256
writing = 'w' in mode
2257
appending = 'a' in mode
2258
updating = '+' in mode
2259
binary = 'b' in mode
2262
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2263
# for flags for each modes.
2273
flags |= os.O_WRONLY
2274
flags |= os.O_CREAT | os.O_TRUNC
2279
flags |= os.O_WRONLY
2280
flags |= os.O_CREAT | os.O_APPEND
2285
flags |= os.O_RDONLY
2287
return os.fdopen(os.open(filename, flags), mode, bufsize)
2292
def getuser_unicode():
2293
"""Return the username as unicode.
2296
user_encoding = get_user_encoding()
2297
username = getpass.getuser().decode(user_encoding)
2298
except UnicodeDecodeError:
2299
raise errors.BzrError("Can't decode username as %s." % \