13
15
# You should have received a copy of the GNU General Public License
14
16
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
from cStringIO import StringIO
21
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
28
from datetime import datetime
30
21
from ntpath import (abspath as _nt_abspath,
32
23
normpath as _nt_normpath,
33
24
realpath as _nt_realpath,
34
25
splitdrive as _nt_splitdrive,
28
from os import listdir
33
from shutil import copyfile
35
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
36
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
44
from tempfile import (
57
from bzrlib.symbol_versioning import (
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
89
# On win32, O_BINARY is used to indicate the file should
90
# be opened in binary mode, rather than text mode.
91
# On other platforms, O_BINARY doesn't exist, because
92
# 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.
95
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 "
45
from bzrlib.errors import (BzrError,
46
BzrBadParameterNotUnicode,
51
from bzrlib.symbol_versioning import (deprecated_function,
53
from bzrlib.trace import mutter
109
56
def make_readonly(filename):
110
57
"""Make a filename read-only."""
111
mod = os.lstat(filename).st_mode
112
if not stat.S_ISLNK(mod):
114
os.chmod(filename, mod)
58
mod = os.stat(filename).st_mode
60
os.chmod(filename, mod)
117
63
def make_writable(filename):
118
mod = os.lstat(filename).st_mode
119
if not stat.S_ISLNK(mod):
121
os.chmod(filename, mod)
124
def minimum_path_selection(paths):
125
"""Return the smallset subset of paths which are outside paths.
127
:param paths: A container (and hence not None) of paths.
128
:return: A set of paths sufficient to include everything in paths via
129
is_inside, drawn from the paths parameter.
135
return path.split('/')
136
sorted_paths = sorted(list(paths), key=sort_key)
138
search_paths = [sorted_paths[0]]
139
for path in sorted_paths[1:]:
140
if not is_inside(search_paths[-1], path):
141
# This path is unique, add it
142
search_paths.append(path)
144
return set(search_paths)
64
mod = os.stat(filename).st_mode
66
os.chmod(filename, mod)
714
537
def local_time_offset(t=None):
715
538
"""Return offset of local zone from GMT, either at present or at time t."""
539
# python2.3 localtime() can't take None
718
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
719
return offset.days * 86400 + offset.seconds
721
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
722
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
725
def format_date(t, offset=0, timezone='original', date_fmt=None,
543
if time.localtime(t).tm_isdst and time.daylight:
546
return -time.timezone
549
def format_date(t, offset=0, timezone='original', date_fmt=None,
726
550
show_offset=True):
727
"""Return a formatted date string.
729
:param t: Seconds since the epoch.
730
:param offset: Timezone offset in seconds east of utc.
731
:param timezone: How to display the time: 'utc', 'original' for the
732
timezone specified by offset, or 'local' for the process's current
734
:param date_fmt: strftime format.
735
:param show_offset: Whether to append the timezone.
737
(date_fmt, tt, offset_str) = \
738
_format_date(t, offset, timezone, date_fmt, show_offset)
739
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
740
date_str = time.strftime(date_fmt, tt)
741
return date_str + offset_str
744
# Cache of formatted offset strings
748
def format_date_with_offset_in_original_timezone(t, offset=0,
749
_cache=_offset_cache):
750
"""Return a formatted date string in the original timezone.
752
This routine may be faster then format_date.
754
:param t: Seconds since the epoch.
755
:param offset: Timezone offset in seconds east of utc.
759
tt = time.gmtime(t + offset)
760
date_fmt = _default_format_by_weekday_num[tt[6]]
761
date_str = time.strftime(date_fmt, tt)
762
offset_str = _cache.get(offset, None)
763
if offset_str is None:
764
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
765
_cache[offset] = offset_str
766
return date_str + offset_str
769
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
771
"""Return an unicode date string formatted according to the current locale.
773
:param t: Seconds since the epoch.
774
:param offset: Timezone offset in seconds east of utc.
775
:param timezone: How to display the time: 'utc', 'original' for the
776
timezone specified by offset, or 'local' for the process's current
778
:param date_fmt: strftime format.
779
:param show_offset: Whether to append the timezone.
781
(date_fmt, tt, offset_str) = \
782
_format_date(t, offset, timezone, date_fmt, show_offset)
783
date_str = time.strftime(date_fmt, tt)
784
if not isinstance(date_str, unicode):
785
date_str = date_str.decode(get_user_encoding(), 'replace')
786
return date_str + offset_str
789
def _format_date(t, offset, timezone, date_fmt, show_offset):
551
## TODO: Perhaps a global option to use either universal or local time?
552
## Or perhaps just let people set $TZ?
553
assert isinstance(t, float)
790
555
if timezone == 'utc':
791
556
tt = time.gmtime(t)
793
558
elif timezone == 'original':
796
561
tt = time.gmtime(t + offset)
797
562
elif timezone == 'local':
798
563
tt = time.localtime(t)
799
564
offset = local_time_offset(t)
801
raise errors.UnsupportedTimezoneFormat(timezone)
566
raise BzrError("unsupported timezone format %r" % timezone,
567
['options are "utc", "original", "local"'])
802
568
if date_fmt is None:
803
569
date_fmt = "%a %Y-%m-%d %H:%M:%S"
805
571
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
808
return (date_fmt, tt, offset_str)
574
return (time.strftime(date_fmt, tt) + offset_str)
811
577
def compact_date(when):
812
578
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
815
def format_delta(delta):
816
"""Get a nice looking string for a time delta.
818
:param delta: The time difference in seconds, can be positive or negative.
819
positive indicates time in the past, negative indicates time in the
820
future. (usually time.time() - stored_time)
821
:return: String formatted to show approximate resolution
827
direction = 'in the future'
831
if seconds < 90: # print seconds up to 90 seconds
833
return '%d second %s' % (seconds, direction,)
835
return '%d seconds %s' % (seconds, direction)
837
minutes = int(seconds / 60)
838
seconds -= 60 * minutes
843
if minutes < 90: # print minutes, seconds up to 90 minutes
845
return '%d minute, %d second%s %s' % (
846
minutes, seconds, plural_seconds, direction)
848
return '%d minutes, %d second%s %s' % (
849
minutes, seconds, plural_seconds, direction)
851
hours = int(minutes / 60)
852
minutes -= 60 * hours
859
return '%d hour, %d minute%s %s' % (hours, minutes,
860
plural_minutes, direction)
861
return '%d hours, %d minute%s %s' % (hours, minutes,
862
plural_minutes, direction)
865
583
"""Return size of given open file."""
1138
741
os.path.commonprefix (python2.4) has a bad bug that it works just
1139
742
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1140
743
avoids that problem.
1142
NOTE: `base` should not have a trailing slash otherwise you'll get
1143
PathNotChild exceptions regardless of `path`.
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'
746
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
747
' exceed the platform minimum length (which is %d)' %
1151
750
rp = abspath(path)
1156
if len(head) <= len(base) and head != base:
1157
raise errors.PathNotChild(rp, base)
754
while len(head) >= len(base):
1158
755
if head == base:
1160
head, tail = split(head)
757
head, tail = os.path.split(head)
761
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]
1236
769
def safe_unicode(unicode_or_utf8_string):
1237
770
"""Coerce unicode_or_utf8_string into unicode.
1239
772
If it is unicode, it is returned.
1240
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1241
wrapped in a BzrBadParameterNotUnicode exception.
773
Otherwise it is decoded from utf-8. If a decoding error
774
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
775
as a BzrBadParameter exception.
1243
777
if isinstance(unicode_or_utf8_string, unicode):
1244
778
return unicode_or_utf8_string
1246
780
return unicode_or_utf8_string.decode('utf8')
1247
781
except UnicodeDecodeError:
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)
782
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
1316
785
_platform_normalizes_filenames = False
1357
826
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.
1403
# Keep some state so that terminal_width can detect if _terminal_size has
1404
# returned a different size since the process started. See docstring and
1405
# comments of terminal_width for details.
1406
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1407
_terminal_size_state = 'no_data'
1408
_first_terminal_size = None
1410
829
def terminal_width():
1411
"""Return terminal width.
1413
None is returned if the width can't established precisely.
1416
- if BZR_COLUMNS is set, returns its value
1417
- if there is no controlling terminal, returns None
1418
- query the OS, if the queried size has changed since the last query,
1420
- if COLUMNS is set, returns its value,
1421
- if the OS has a value (even though it's never changed), return its value.
1423
From there, we need to query the OS to get the size of the controlling
1426
On Unices we query the OS by:
1427
- get termios.TIOCGWINSZ
1428
- if an error occurs or a negative value is obtained, returns None
1430
On Windows we query the OS by:
1431
- win32utils.get_console_size() decides,
1432
- returns None on error (provided default value)
1434
# Note to implementors: if changing the rules for determining the width,
1435
# make sure you've considered the behaviour in these cases:
1436
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1437
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1439
# - (add more interesting cases here, if you find any)
1440
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1441
# but we don't want to register a signal handler because it is impossible
1442
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1443
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1444
# time so we can notice if the reported size has changed, which should have
1447
# If BZR_COLUMNS is set, take it, user is always right
1449
return int(os.environ['BZR_COLUMNS'])
1450
except (KeyError, ValueError):
1453
isatty = getattr(sys.stdout, 'isatty', None)
1454
if isatty is None or not isatty():
1455
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1459
width, height = os_size = _terminal_size(None, None)
1460
global _first_terminal_size, _terminal_size_state
1461
if _terminal_size_state == 'no_data':
1462
_first_terminal_size = os_size
1463
_terminal_size_state = 'unchanged'
1464
elif (_terminal_size_state == 'unchanged' and
1465
_first_terminal_size != os_size):
1466
_terminal_size_state = 'changed'
1468
# If the OS claims to know how wide the terminal is, and this value has
1469
# ever changed, use that.
1470
if _terminal_size_state == 'changed':
1471
if width is not None and width > 0:
1474
# If COLUMNS is set, use it.
1476
return int(os.environ['COLUMNS'])
1477
except (KeyError, ValueError):
1480
# Finally, use an unchanged size from the OS, if we have one.
1481
if _terminal_size_state == 'unchanged':
1482
if width is not None and width > 0:
1485
# The width could not be determined.
1489
def _win32_terminal_size(width, height):
1490
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1491
return width, height
1494
def _ioctl_terminal_size(width, height):
830
"""Return estimated terminal width."""
831
if sys.platform == 'win32':
832
import bzrlib.win32console
833
return bzrlib.win32console.get_console_size()[0]
1496
836
import struct, fcntl, termios
1497
837
s = struct.pack('HHHH', 0, 0, 0, 0)
1498
838
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1499
height, width = struct.unpack('HHHH', x)[0:2]
1500
except (IOError, AttributeError):
839
width = struct.unpack('HHHH', x)[1]
1502
return width, height
1504
_terminal_size = None
1505
"""Returns the terminal size as (width, height).
1507
:param width: Default value for width.
1508
:param height: Default value for height.
1510
This is defined specifically for each OS and query the size of the controlling
1511
terminal. If any error occurs, the provided default values should be returned.
1513
if sys.platform == 'win32':
1514
_terminal_size = _win32_terminal_size
1516
_terminal_size = _ioctl_terminal_size
844
width = int(os.environ['COLUMNS'])
1519
852
def supports_executable():
1520
853
return sys.platform != "win32"
1523
def supports_posix_readonly():
1524
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1526
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1527
directory controls creation/deletion, etc.
1529
And under win32, readonly means that the directory itself cannot be
1530
deleted. The contents of a readonly directory can be changed, unlike POSIX
1531
where files in readonly directories cannot be added, deleted or renamed.
1533
return sys.platform != "win32"
1536
def set_or_unset_env(env_variable, value):
1537
"""Modify the environment, setting or removing the env_variable.
1539
:param env_variable: The environment variable in question
1540
:param value: The value to set the environment to. If None, then
1541
the variable will be removed.
1542
:return: The original value of the environment variable.
1544
orig_val = os.environ.get(env_variable)
1546
if orig_val is not None:
1547
del os.environ[env_variable]
1549
if isinstance(value, unicode):
1550
value = value.encode(get_user_encoding())
1551
os.environ[env_variable] = value
1555
856
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1558
859
def check_legal_path(path):
1559
"""Check whether the supplied path is legal.
860
"""Check whether the supplied path is legal.
1560
861
This is only required on Windows, so we don't test on other platforms
1563
864
if sys.platform != "win32":
1565
866
if _validWin32PathRE.match(path) is None:
1566
raise errors.IllegalPath(path)
1569
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1571
def _is_error_enotdir(e):
1572
"""Check if this exception represents ENOTDIR.
1574
Unfortunately, python is very inconsistent about the exception
1575
here. The cases are:
1576
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1577
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1578
which is the windows error code.
1579
3) Windows, Python2.5 uses errno == EINVAL and
1580
winerror == ERROR_DIRECTORY
1582
:param e: An Exception object (expected to be OSError with an errno
1583
attribute, but we should be able to cope with anything)
1584
:return: True if this represents an ENOTDIR error. False otherwise.
1586
en = getattr(e, 'errno', None)
1587
if (en == errno.ENOTDIR
1588
or (sys.platform == 'win32'
1589
and (en == _WIN32_ERROR_DIRECTORY
1590
or (en == errno.EINVAL
1591
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
867
raise IllegalPath(path)
1597
870
def walkdirs(top, prefix=""):
1598
871
"""Yield data about all the directories in a tree.
1600
873
This yields all the data about the contents of a directory at a time.
1601
874
After each directory has been yielded, if the caller has mutated the list
1602
875
to exclude some directories, they are then not descended into.
1604
877
The data yielded is of the form:
1605
((directory-relpath, directory-path-from-top),
1606
[(relpath, basename, kind, lstat, path-from-top), ...]),
1607
- directory-relpath is the relative path of the directory being returned
1608
with respect to top. prefix is prepended to this.
1609
- directory-path-from-root is the path including top for this directory.
1610
It is suitable for use with os functions.
1611
- relpath is the relative path within the subtree being walked.
1612
- basename is the basename of the path
1613
- kind is the kind of the file now. If unknown then the file is not
1614
present within the tree - but it may be recorded as versioned. See
1616
- lstat is the stat data *if* the file was statted.
1617
- planned, not implemented:
1618
path_from_tree_root is the path from the root of the tree.
878
[(relpath, basename, kind, lstat, path_from_top), ...]
1620
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
880
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1621
881
allows one to walk a subtree but get paths that are relative to a tree
1622
882
rooted higher up.
1623
883
:return: an iterator over the dirs.
1625
#TODO there is a bit of a smell where the results of the directory-
1626
# summary in this, and the path from the root, may not agree
1627
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1628
# potentially confusing output. We should make this more robust - but
1629
# not at a speed cost. RBC 20060731
1631
887
_directory = _directory_kind
1632
_listdir = os.listdir
1633
_kind_from_mode = file_kind_from_stat_mode
1634
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
889
pending = [(prefix, "", _directory, None, top)]
892
currentdir = pending.pop()
1636
893
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1637
relroot, _, _, _, top = pending.pop()
1639
relprefix = relroot + u'/'
1642
top_slash = top + u'/'
1645
append = dirblock.append
1647
names = sorted(_listdir(top))
1649
if not _is_error_enotdir(e):
1653
abspath = top_slash + name
1654
statvalue = _lstat(abspath)
1655
kind = _kind_from_mode(statvalue.st_mode)
1656
append((relprefix + name, name, kind, statvalue, abspath))
1657
yield (relroot, top), dirblock
1659
# push the user specified dirs from dirblock
1660
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1663
class DirReader(object):
1664
"""An interface for reading directories."""
1666
def top_prefix_to_starting_dir(self, top, prefix=""):
1667
"""Converts top and prefix to a starting dir entry
1669
:param top: A utf8 path
1670
:param prefix: An optional utf8 path to prefix output relative paths
1672
:return: A tuple starting with prefix, and ending with the native
1675
raise NotImplementedError(self.top_prefix_to_starting_dir)
1677
def read_dir(self, prefix, top):
1678
"""Read a specific dir.
1680
:param prefix: A utf8 prefix to be preprended to the path basenames.
1681
:param top: A natively encoded path to read.
1682
:return: A list of the directories contents. Each item contains:
1683
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1685
raise NotImplementedError(self.read_dir)
1688
_selected_dir_reader = None
1691
def _walkdirs_utf8(top, prefix=""):
1692
"""Yield data about all the directories in a tree.
1694
This yields the same information as walkdirs() only each entry is yielded
1695
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1696
are returned as exact byte-strings.
1698
:return: yields a tuple of (dir_info, [file_info])
1699
dir_info is (utf8_relpath, path-from-top)
1700
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1701
if top is an absolute path, path-from-top is also an absolute path.
1702
path-from-top might be unicode or utf8, but it is the correct path to
1703
pass to os functions to affect the file in question. (such as os.lstat)
1705
global _selected_dir_reader
1706
if _selected_dir_reader is None:
1707
fs_encoding = _fs_enc.upper()
1708
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1709
# Win98 doesn't have unicode apis like FindFirstFileW
1710
# TODO: We possibly could support Win98 by falling back to the
1711
# original FindFirstFile, and using TCHAR instead of WCHAR,
1712
# but that gets a bit tricky, and requires custom compiling
1715
from bzrlib._walkdirs_win32 import Win32ReadDir
1716
_selected_dir_reader = Win32ReadDir()
1719
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1720
# ANSI_X3.4-1968 is a form of ASCII
1722
from bzrlib._readdir_pyx import UTF8DirReader
1723
_selected_dir_reader = UTF8DirReader()
1724
except ImportError, e:
1725
failed_to_load_extension(e)
1728
if _selected_dir_reader is None:
1729
# Fallback to the python version
1730
_selected_dir_reader = UnicodeDirReader()
1732
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1733
# But we don't actually uses 1-3 in pending, so set them to None
1734
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1735
read_dir = _selected_dir_reader.read_dir
1736
_directory = _directory_kind
1738
relroot, _, _, _, top = pending[-1].pop()
1741
dirblock = sorted(read_dir(relroot, top))
1742
yield (relroot, top), dirblock
1743
# push the user specified dirs from dirblock
1744
next = [d for d in reversed(dirblock) if d[2] == _directory]
1746
pending.append(next)
1749
class UnicodeDirReader(DirReader):
1750
"""A dir reader for non-utf8 file systems, which transcodes."""
1752
__slots__ = ['_utf8_encode']
1755
self._utf8_encode = codecs.getencoder('utf8')
1757
def top_prefix_to_starting_dir(self, top, prefix=""):
1758
"""See DirReader.top_prefix_to_starting_dir."""
1759
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1761
def read_dir(self, prefix, top):
1762
"""Read a single directory from a non-utf8 file system.
1764
top, and the abspath element in the output are unicode, all other paths
1765
are utf8. Local disk IO is done via unicode calls to listdir etc.
1767
This is currently the fallback code path when the filesystem encoding is
1768
not UTF-8. It may be better to implement an alternative so that we can
1769
safely handle paths that are not properly decodable in the current
1772
See DirReader.read_dir for details.
1774
_utf8_encode = self._utf8_encode
1776
_listdir = os.listdir
1777
_kind_from_mode = file_kind_from_stat_mode
1780
relprefix = prefix + '/'
1783
top_slash = top + u'/'
1786
append = dirblock.append
896
relroot = currentdir[0] + '/'
1787
899
for name in sorted(_listdir(top)):
1789
name_utf8 = _utf8_encode(name)[0]
1790
except UnicodeDecodeError:
1791
raise errors.BadFilenameEncoding(
1792
_utf8_encode(relprefix)[0] + name, _fs_enc)
1793
abspath = top_slash + name
1794
statvalue = _lstat(abspath)
1795
kind = _kind_from_mode(statvalue.st_mode)
1796
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1800
def copy_tree(from_path, to_path, handlers={}):
1801
"""Copy all of the entries in from_path into to_path.
1803
:param from_path: The base directory to copy.
1804
:param to_path: The target directory. If it does not exist, it will
1806
:param handlers: A dictionary of functions, which takes a source and
1807
destinations for files, directories, etc.
1808
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1809
'file', 'directory', and 'symlink' should always exist.
1810
If they are missing, they will be replaced with 'os.mkdir()',
1811
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1813
# Now, just copy the existing cached tree to the new location
1814
# We use a cheap trick here.
1815
# Absolute paths are prefixed with the first parameter
1816
# relative paths are prefixed with the second.
1817
# So we can get both the source and target returned
1818
# without any extra work.
1820
def copy_dir(source, dest):
1823
def copy_link(source, dest):
1824
"""Copy the contents of a symlink"""
1825
link_to = os.readlink(source)
1826
os.symlink(link_to, dest)
1828
real_handlers = {'file':shutil.copy2,
1829
'symlink':copy_link,
1830
'directory':copy_dir,
1832
real_handlers.update(handlers)
1834
if not os.path.exists(to_path):
1835
real_handlers['directory'](from_path, to_path)
1837
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1838
for relpath, name, kind, st, abspath in entries:
1839
real_handlers[kind](abspath, relpath)
1842
def copy_ownership_from_path(dst, src=None):
1843
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1845
If src is None, the containing directory is used as source. If chown
1846
fails, the error is ignored and a warning is printed.
1848
chown = getattr(os, 'chown', None)
1853
src = os.path.dirname(dst)
1859
chown(dst, s.st_uid, s.st_gid)
1861
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
900
abspath = top + '/' + name
901
statvalue = lstat(abspath)
902
dirblock.append ((relroot + name, name, file_kind_from_stat_mode(statvalue.st_mode), statvalue, abspath))
904
# push the user specified dirs from dirblock
905
for dir in reversed(dirblock):
906
if dir[2] == _directory:
1864
910
def path_prefix_key(path):
1874
920
key_a = path_prefix_key(path_a)
1875
921
key_b = path_prefix_key(path_b)
1876
922
return cmp(key_a, key_b)
1879
_cached_user_encoding = None
1882
def get_user_encoding(use_cache=True):
1883
"""Find out what the preferred user encoding is.
1885
This is generally the encoding that is used for command line parameters
1886
and file contents. This may be different from the terminal encoding
1887
or the filesystem encoding.
1889
:param use_cache: Enable cache for detected encoding.
1890
(This parameter is turned on by default,
1891
and required only for selftesting)
1893
:return: A string defining the preferred user encoding
1895
global _cached_user_encoding
1896
if _cached_user_encoding is not None and use_cache:
1897
return _cached_user_encoding
1899
if sys.platform == 'darwin':
1900
# python locale.getpreferredencoding() always return
1901
# 'mac-roman' on darwin. That's a lie.
1902
sys.platform = 'posix'
1904
if os.environ.get('LANG', None) is None:
1905
# If LANG is not set, we end up with 'ascii', which is bad
1906
# ('mac-roman' is more than ascii), so we set a default which
1907
# will give us UTF-8 (which appears to work in all cases on
1908
# OSX). Users are still free to override LANG of course, as
1909
# long as it give us something meaningful. This work-around
1910
# *may* not be needed with python 3k and/or OSX 10.5, but will
1911
# work with them too -- vila 20080908
1912
os.environ['LANG'] = 'en_US.UTF-8'
1915
sys.platform = 'darwin'
1920
user_encoding = locale.getpreferredencoding()
1921
except locale.Error, e:
1922
sys.stderr.write('bzr: warning: %s\n'
1923
' Could not determine what text encoding to use.\n'
1924
' This error usually means your Python interpreter\n'
1925
' doesn\'t support the locale set by $LANG (%s)\n'
1926
" Continuing with ascii encoding.\n"
1927
% (e, os.environ.get('LANG')))
1928
user_encoding = 'ascii'
1930
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1931
# treat that as ASCII, and not support printing unicode characters to the
1934
# For python scripts run under vim, we get '', so also treat that as ASCII
1935
if user_encoding in (None, 'cp0', ''):
1936
user_encoding = 'ascii'
1940
codecs.lookup(user_encoding)
1942
sys.stderr.write('bzr: warning:'
1943
' unknown encoding %s.'
1944
' Continuing with ascii encoding.\n'
1947
user_encoding = 'ascii'
1950
_cached_user_encoding = user_encoding
1952
return user_encoding
1955
def get_host_name():
1956
"""Return the current unicode host name.
1958
This is meant to be used in place of socket.gethostname() because that
1959
behaves inconsistently on different platforms.
1961
if sys.platform == "win32":
1963
return win32utils.get_host_name()
1966
return socket.gethostname().decode(get_user_encoding())
1969
# We must not read/write any more than 64k at a time from/to a socket so we
1970
# don't risk "no buffer space available" errors on some platforms. Windows in
1971
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1973
MAX_SOCKET_CHUNK = 64 * 1024
1975
def read_bytes_from_socket(sock, report_activity=None,
1976
max_read_size=MAX_SOCKET_CHUNK):
1977
"""Read up to max_read_size of bytes from sock and notify of progress.
1979
Translates "Connection reset by peer" into file-like EOF (return an
1980
empty string rather than raise an error), and repeats the recv if
1981
interrupted by a signal.
1985
bytes = sock.recv(max_read_size)
1986
except socket.error, e:
1988
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1989
# The connection was closed by the other side. Callers expect
1990
# an empty string to signal end-of-stream.
1992
elif eno == errno.EINTR:
1993
# Retry the interrupted recv.
1997
if report_activity is not None:
1998
report_activity(len(bytes), 'read')
2002
def recv_all(socket, count):
2003
"""Receive an exact number of bytes.
2005
Regular Socket.recv() may return less than the requested number of bytes,
2006
depending on what's in the OS buffer. MSG_WAITALL is not available
2007
on all platforms, but this should work everywhere. This will return
2008
less than the requested amount if the remote end closes.
2010
This isn't optimized and is intended mostly for use in testing.
2013
while len(b) < count:
2014
new = read_bytes_from_socket(socket, None, count - len(b))
2021
def send_all(sock, bytes, report_activity=None):
2022
"""Send all bytes on a socket.
2024
Breaks large blocks in smaller chunks to avoid buffering limitations on
2025
some platforms, and catches EINTR which may be thrown if the send is
2026
interrupted by a signal.
2028
This is preferred to socket.sendall(), because it avoids portability bugs
2029
and provides activity reporting.
2031
:param report_activity: Call this as bytes are read, see
2032
Transport._report_activity
2035
byte_count = len(bytes)
2036
while sent_total < byte_count:
2038
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2039
except socket.error, e:
2040
if e.args[0] != errno.EINTR:
2044
report_activity(sent, 'write')
2047
def dereference_path(path):
2048
"""Determine the real path to a file.
2050
All parent elements are dereferenced. But the file itself is not
2052
:param path: The original path. May be absolute or relative.
2053
:return: the real path *to* the file
2055
parent, base = os.path.split(path)
2056
# The pathjoin for '.' is a workaround for Python bug #1213894.
2057
# (initial path components aren't dereferenced)
2058
return pathjoin(realpath(pathjoin('.', parent)), base)
2061
def supports_mapi():
2062
"""Return True if we can use MAPI to launch a mail client."""
2063
return sys.platform == "win32"
2066
def resource_string(package, resource_name):
2067
"""Load a resource from a package and return it as a string.
2069
Note: Only packages that start with bzrlib are currently supported.
2071
This is designed to be a lightweight implementation of resource
2072
loading in a way which is API compatible with the same API from
2074
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2075
If and when pkg_resources becomes a standard library, this routine
2078
# Check package name is within bzrlib
2079
if package == "bzrlib":
2080
resource_relpath = resource_name
2081
elif package.startswith("bzrlib."):
2082
package = package[len("bzrlib."):].replace('.', os.sep)
2083
resource_relpath = pathjoin(package, resource_name)
2085
raise errors.BzrError('resource package %s not in bzrlib' % package)
2087
# Map the resource to a file and read its contents
2088
base = dirname(bzrlib.__file__)
2089
if getattr(sys, 'frozen', None): # bzr.exe
2090
base = abspath(pathjoin(base, '..', '..'))
2091
f = file(pathjoin(base, resource_relpath), "rU")
2097
def file_kind_from_stat_mode_thunk(mode):
2098
global file_kind_from_stat_mode
2099
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2101
from bzrlib._readdir_pyx import UTF8DirReader
2102
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2103
except ImportError, e:
2104
# This is one time where we won't warn that an extension failed to
2105
# load. The extension is never available on Windows anyway.
2106
from bzrlib._readdir_py import (
2107
_kind_from_mode as file_kind_from_stat_mode
2109
return file_kind_from_stat_mode(mode)
2110
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2113
def file_kind(f, _lstat=os.lstat):
2115
return file_kind_from_stat_mode(_lstat(f).st_mode)
2117
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2118
raise errors.NoSuchFile(f)
2122
def until_no_eintr(f, *a, **kw):
2123
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2125
WARNING: you must be certain that it is safe to retry the call repeatedly
2126
if EINTR does occur. This is typically only true for low-level operations
2127
like os.read. If in any doubt, don't use this.
2129
Keep in mind that this is not a complete solution to EINTR. There is
2130
probably code in the Python standard library and other dependencies that
2131
may encounter EINTR if a signal arrives (and there is signal handler for
2132
that signal). So this function can reduce the impact for IO that bzrlib
2133
directly controls, but it is not a complete solution.
2135
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2139
except (IOError, OSError), e:
2140
if e.errno == errno.EINTR:
2145
def re_compile_checked(re_string, flags=0, where=""):
2146
"""Return a compiled re, or raise a sensible error.
2148
This should only be used when compiling user-supplied REs.
2150
:param re_string: Text form of regular expression.
2151
:param flags: eg re.IGNORECASE
2152
:param where: Message explaining to the user the context where
2153
it occurred, eg 'log search filter'.
2155
# from https://bugs.launchpad.net/bzr/+bug/251352
2157
re_obj = re.compile(re_string, flags)
2162
where = ' in ' + where
2163
# despite the name 'error' is a type
2164
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2165
% (where, re_string, e))
2168
if sys.platform == "win32":
2171
return msvcrt.getch()
2176
fd = sys.stdin.fileno()
2177
settings = termios.tcgetattr(fd)
2180
ch = sys.stdin.read(1)
2182
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2186
if sys.platform == 'linux2':
2187
def _local_concurrency():
2189
prefix = 'processor'
2190
for line in file('/proc/cpuinfo', 'rb'):
2191
if line.startswith(prefix):
2192
concurrency = int(line[line.find(':')+1:]) + 1
2194
elif sys.platform == 'darwin':
2195
def _local_concurrency():
2196
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2197
stdout=subprocess.PIPE).communicate()[0]
2198
elif sys.platform[0:7] == 'freebsd':
2199
def _local_concurrency():
2200
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2201
stdout=subprocess.PIPE).communicate()[0]
2202
elif sys.platform == 'sunos5':
2203
def _local_concurrency():
2204
return subprocess.Popen(['psrinfo', '-p',],
2205
stdout=subprocess.PIPE).communicate()[0]
2206
elif sys.platform == "win32":
2207
def _local_concurrency():
2208
# This appears to return the number of cores.
2209
return os.environ.get('NUMBER_OF_PROCESSORS')
2211
def _local_concurrency():
2216
_cached_local_concurrency = None
2218
def local_concurrency(use_cache=True):
2219
"""Return how many processes can be run concurrently.
2221
Rely on platform specific implementations and default to 1 (one) if
2222
anything goes wrong.
2224
global _cached_local_concurrency
2226
if _cached_local_concurrency is not None and use_cache:
2227
return _cached_local_concurrency
2229
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2230
if concurrency is None:
2232
concurrency = _local_concurrency()
2233
except (OSError, IOError):
2236
concurrency = int(concurrency)
2237
except (TypeError, ValueError):
2240
_cached_concurrency = concurrency
2244
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2245
"""A stream writer that doesn't decode str arguments."""
2247
def __init__(self, encode, stream, errors='strict'):
2248
codecs.StreamWriter.__init__(self, stream, errors)
2249
self.encode = encode
2251
def write(self, object):
2252
if type(object) is str:
2253
self.stream.write(object)
2255
data, _ = self.encode(object, self.errors)
2256
self.stream.write(data)
2258
if sys.platform == 'win32':
2259
def open_file(filename, mode='r', bufsize=-1):
2260
"""This function is used to override the ``open`` builtin.
2262
But it uses O_NOINHERIT flag so the file handle is not inherited by
2263
child processes. Deleting or renaming a closed file opened with this
2264
function is not blocking child processes.
2266
writing = 'w' in mode
2267
appending = 'a' in mode
2268
updating = '+' in mode
2269
binary = 'b' in mode
2272
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2273
# for flags for each modes.
2283
flags |= os.O_WRONLY
2284
flags |= os.O_CREAT | os.O_TRUNC
2289
flags |= os.O_WRONLY
2290
flags |= os.O_CREAT | os.O_APPEND
2295
flags |= os.O_RDONLY
2297
return os.fdopen(os.open(filename, flags), mode, bufsize)
2302
def getuser_unicode():
2303
"""Return the username as unicode.
2306
user_encoding = get_user_encoding()
2307
username = getpass.getuser().decode(user_encoding)
2308
except UnicodeDecodeError:
2309
raise errors.BzrError("Can't decode username as %s." % \
added
removed
bzrlib/osutils.py
