1 """Helper functions and classes for bup."""
3 from __future__ import absolute_import, division
4 from collections import namedtuple
5 from contextlib import contextmanager
6 from ctypes import sizeof, c_void_p
9 from subprocess import PIPE, Popen
10 import sys, os, subprocess, errno, select, mmap, stat, re, struct
11 import hashlib, heapq, math, operator, time, tempfile
13 from bup import _helpers
14 from bup import compat
15 from bup.compat import argv_bytes, byte_int, nullcontext, pending_raise
16 from bup.io import byte_stream, path_msg
17 # This function should really be in helpers, not in bup.options. But we
18 # want options.py to be standalone so people can include it in other projects.
19 from bup.options import _tty_width as tty_width
22 buglvl = int(os.environ.get('BUP_DEBUG', 0))
26 """Helper to deal with Python scoping issues"""
30 def nullcontext_if_not(manager):
31 return manager if manager is not None else nullcontext()
35 def finalized(enter_result=None, finalize=None):
39 except BaseException as ex:
40 with pending_raise(ex):
41 finalize(enter_result)
42 finalize(enter_result)
45 sc_page_size = os.sysconf('SC_PAGE_SIZE')
46 assert(sc_page_size > 0)
48 sc_arg_max = os.sysconf('SC_ARG_MAX')
49 if sc_arg_max == -1: # "no definite limit" - let's choose 2M
50 sc_arg_max = 2 * 1024 * 1024
54 for result in iterable:
59 _fdatasync = os.fdatasync
60 except AttributeError:
63 if sys.platform.startswith('darwin'):
64 # Apparently os.fsync on OS X doesn't guarantee to sync all the way down
68 return fcntl.fcntl(fd, fcntl.F_FULLFSYNC)
70 # Fallback for file systems (SMB) that do not support F_FULLFSYNC
71 if e.errno == errno.ENOTSUP:
76 fdatasync = _fdatasync
79 def partition(predicate, stream):
80 """Returns (leading_matches_it, rest_it), where leading_matches_it
81 must be completely exhausted before traversing rest_it.
86 ns.first_nonmatch = None
87 def leading_matches():
92 ns.first_nonmatch = (x,)
96 yield ns.first_nonmatch[0]
99 return (leading_matches(), rest())
109 def lines_until_sentinel(f, sentinel, ex_type):
110 # sentinel must end with \n and must contain only one \n
113 if not (line and line.endswith(b'\n')):
114 raise ex_type('Hit EOF while reading line')
120 def stat_if_exists(path):
124 if e.errno != errno.ENOENT:
129 # Write (blockingly) to sockets that may or may not be in blocking mode.
130 # We need this because our stderr is sometimes eaten by subprocesses
131 # (probably ssh) that sometimes make it nonblocking, if only temporarily,
132 # leading to race conditions. Ick. We'll do it the hard way.
133 def _hard_write(fd, buf):
135 (r,w,x) = select.select([], [fd], [], None)
137 raise IOError('select(fd) returned without being writable')
139 sz = os.write(fd, buf)
141 if e.errno != errno.EAGAIN:
149 """Print a log message to stderr."""
152 _hard_write(sys.stderr.fileno(), s if isinstance(s, bytes) else s.encode())
166 istty1 = os.isatty(1) or (int(os.environ.get('BUP_FORCE_TTY', 0)) & 1)
167 istty2 = os.isatty(2) or (int(os.environ.get('BUP_FORCE_TTY', 0)) & 2)
170 """Calls log() if stderr is a TTY. Does nothing otherwise."""
171 global _last_progress
178 """Calls progress() only if we haven't printed progress in a while.
180 This avoids overloading the stderr buffer with excess junk.
184 if now - _last_prog > 0.1:
190 """Calls progress() to redisplay the most recent progress message.
192 Useful after you've printed some other message that wipes out the
195 if _last_progress and _last_progress.endswith('\r'):
196 progress(_last_progress)
199 def mkdirp(d, mode=None):
200 """Recursively create directories on path 'd'.
202 Unlike os.makedirs(), it doesn't raise an exception if the last element of
203 the path already exists.
211 if e.errno == errno.EEXIST:
218 def __init__(self, entry, read_it):
220 self.read_it = read_it
222 return self.entry < x.entry
224 def merge_iter(iters, pfreq, pfunc, pfinal, key=None):
226 samekey = lambda e, pe: getattr(e, key) == getattr(pe, key, None)
228 samekey = operator.eq
230 total = sum(len(it) for it in iters)
231 iters = (iter(it) for it in iters)
232 heap = ((next(it, None),it) for it in iters)
233 heap = [MergeIterItem(e, it) for e, it in heap if e]
238 if not count % pfreq:
240 e, it = heap[0].entry, heap[0].read_it
241 if not samekey(e, pe):
247 except StopIteration:
248 heapq.heappop(heap) # remove current
250 # shift current to new location
251 heapq.heapreplace(heap, MergeIterItem(e, it))
256 """Delete a file at path 'f' if it currently exists.
258 Unlike os.unlink(), does not throw an exception if the file didn't already
264 if e.errno != errno.ENOENT:
268 _bq_simple_id_rx = re.compile(br'^[-_./a-zA-Z0-9]+$')
269 _sq_simple_id_rx = re.compile(r'^[-_./a-zA-Z0-9]+$')
274 if _bq_simple_id_rx.match(x):
276 return b"'%s'" % x.replace(b"'", b"'\"'\"'")
281 if _sq_simple_id_rx.match(x):
283 return "'%s'" % x.replace("'", "'\"'\"'")
286 if isinstance(x, bytes):
288 if isinstance(x, compat.str_type):
291 # some versions of pylint get confused
295 """Return a shell quoted string for cmd if it's a sequence, else cmd.
297 cmd must be a string, bytes, or a sequence of one or the other,
298 and the assumption is that if cmd is a string or bytes, then it's
299 already quoted (because it's what's actually being passed to
300 call() and friends. e.g. log(shstr(cmd)); call(cmd)
303 if isinstance(cmd, (bytes, compat.str_type)):
305 elif all(isinstance(x, bytes) for x in cmd):
306 return b' '.join(map(bquote, cmd))
307 elif all(isinstance(x, compat.str_type) for x in cmd):
308 return ' '.join(map(squote, cmd))
309 raise TypeError('unsupported shstr argument: ' + repr(cmd))
312 exc = subprocess.check_call
323 assert stdin in (None, PIPE)
326 stdin=stdin, stdout=PIPE, stderr=stderr,
328 preexec_fn=preexec_fn,
330 out, err = p.communicate(input)
331 if check and p.returncode != 0:
332 raise Exception('subprocess %r failed with status %d%s'
333 % (b' '.join(map(quote, cmd)), p.returncode,
334 ', stderr: %r' % err if err else ''))
337 def readpipe(argv, preexec_fn=None, shell=False):
338 """Run a subprocess and return its output."""
339 return exo(argv, preexec_fn=preexec_fn, shell=shell)[0]
342 def _argmax_base(command):
345 base_size += len(command) + 1
346 for k, v in compat.items(environ):
347 base_size += len(k) + len(v) + 2 + sizeof(c_void_p)
351 def _argmax_args_size(args):
352 return sum(len(x) + 1 + sizeof(c_void_p) for x in args)
355 def batchpipe(command, args, preexec_fn=None, arg_max=sc_arg_max):
356 """If args is not empty, yield the output produced by calling the
357 command list with args as a sequence of strings (It may be necessary
358 to return multiple strings in order to respect ARG_MAX)."""
359 # The optional arg_max arg is a workaround for an issue with the
360 # current wvtest behavior.
361 base_size = _argmax_base(command)
363 room = arg_max - base_size
366 next_size = _argmax_args_size(args[i:i+1])
367 if room - next_size < 0:
373 assert(len(sub_args))
374 yield readpipe(command + sub_args, preexec_fn=preexec_fn)
377 def resolve_parent(p):
378 """Return the absolute path of a file without following any final symlink.
380 Behaves like os.path.realpath, but doesn't follow a symlink for the last
381 element. (ie. if 'p' itself is a symlink, this one won't follow it, but it
382 will follow symlinks in p's directory)
388 if st and stat.S_ISLNK(st.st_mode):
389 (dir, name) = os.path.split(p)
390 dir = os.path.realpath(dir)
391 out = os.path.join(dir, name)
393 out = os.path.realpath(p)
394 #log('realpathing:%r,%r\n' % (p, out))
398 def detect_fakeroot():
399 "Return True if we appear to be running under fakeroot."
400 return os.getenv("FAKEROOTKEY") != None
403 if sys.platform.startswith('cygwin'):
405 # https://cygwin.com/ml/cygwin/2015-02/msg00057.html
406 groups = os.getgroups()
407 return 544 in groups or 0 in groups
410 return os.geteuid() == 0
413 def cache_key_value(get_value, key, cache):
414 """Return (value, was_cached). If there is a value in the cache
415 for key, use that, otherwise, call get_value(key) which should
416 throw a KeyError if there is no value -- in which case the cached
417 and returned value will be None.
419 try: # Do we already have it (or know there wasn't one)?
426 cache[key] = value = get_value(key)
434 """Get the FQDN of this machine."""
437 _hostname = _helpers.gethostname()
441 def format_filesize(size):
446 exponent = int(math.log(size) // math.log(unit))
447 size_prefix = "KMGTPE"[exponent - 1]
448 return "%.1f%s" % (size / math.pow(unit, exponent), size_prefix)
451 class NotOk(Exception):
456 def __init__(self, outp):
460 while self._read(65536): pass
462 def _read(self, size):
463 raise NotImplementedError("Subclasses must implement _read")
465 def read(self, size):
466 """Read 'size' bytes from input stream."""
468 return self._read(size)
470 def _readline(self, size):
471 raise NotImplementedError("Subclasses must implement _readline")
474 """Read from input stream until a newline is found."""
476 return self._readline()
478 def write(self, data):
479 """Write 'data' to output stream."""
480 #log('%d writing: %d bytes\n' % (os.getpid(), len(data)))
481 self.outp.write(data)
484 """Return true if input stream is readable."""
485 raise NotImplementedError("Subclasses must implement has_input")
488 """Indicate end of output from last sent command."""
489 self.write(b'\nok\n')
492 """Indicate server error to the client."""
493 s = re.sub(br'\s+', b' ', s)
494 self.write(b'\nerror %s\n' % s)
496 def _check_ok(self, onempty):
499 for rl in linereader(self):
500 #log('%d got line: %r\n' % (os.getpid(), rl))
501 if not rl: # empty line
505 elif rl.startswith(b'error '):
506 #log('client: error: %s\n' % rl[6:])
510 raise Exception('server exited unexpectedly; see errors above')
512 def drain_and_check_ok(self):
513 """Remove all data for the current command from input stream."""
516 return self._check_ok(onempty)
519 """Verify that server action completed successfully."""
521 raise Exception('expected "ok", got %r' % rl)
522 return self._check_ok(onempty)
525 class Conn(BaseConn):
526 def __init__(self, inp, outp):
527 BaseConn.__init__(self, outp)
530 def _read(self, size):
531 return self.inp.read(size)
534 return self.inp.readline()
537 [rl, wl, xl] = select.select([self.inp.fileno()], [], [], 0)
539 assert(rl[0] == self.inp.fileno())
545 def checked_reader(fd, n):
547 rl, _, _ = select.select([fd], [], [])
550 if not buf: raise Exception("Unexpected EOF reading %d more bytes" % n)
555 MAX_PACKET = 128 * 1024
556 def mux(p, outfd, outr, errr):
559 while p.poll() is None:
560 rl, _, _ = select.select(fds, [], [])
563 buf = os.read(outr, MAX_PACKET)
565 os.write(outfd, struct.pack('!IB', len(buf), 1) + buf)
567 buf = os.read(errr, 1024)
569 os.write(outfd, struct.pack('!IB', len(buf), 2) + buf)
571 os.write(outfd, struct.pack('!IB', 0, 3))
574 class DemuxConn(BaseConn):
575 """A helper class for bup's client-server protocol."""
576 def __init__(self, infd, outp):
577 BaseConn.__init__(self, outp)
578 # Anything that comes through before the sync string was not
579 # multiplexed and can be assumed to be debug/log before mux init.
581 stderr = byte_stream(sys.stderr)
582 while tail != b'BUPMUX':
583 # Make sure to write all pre-BUPMUX output to stderr
584 b = os.read(infd, (len(tail) < 6) and (6-len(tail)) or 1)
586 ex = IOError('demux: unexpected EOF during initialization')
587 with pending_raise(ex):
591 stderr.write(tail[:-6])
599 def write(self, data):
601 BaseConn.write(self, data)
603 def _next_packet(self, timeout):
604 if self.closed: return False
605 rl, wl, xl = select.select([self.infd], [], [], timeout)
606 if not rl: return False
607 assert(rl[0] == self.infd)
608 ns = b''.join(checked_reader(self.infd, 5))
609 n, fdw = struct.unpack('!IB', ns)
611 # assume that something went wrong and print stuff
612 ns += os.read(self.infd, 1024)
613 stderr = byte_stream(sys.stderr)
616 raise Exception("Connection broken")
618 self.reader = checked_reader(self.infd, n)
620 for buf in checked_reader(self.infd, n):
621 byte_stream(sys.stderr).write(buf)
624 debug2("DemuxConn: marked closed\n")
627 def _load_buf(self, timeout):
628 if self.buf is not None:
630 while not self.closed:
631 while not self.reader:
632 if not self._next_packet(timeout):
635 self.buf = next(self.reader)
637 except StopIteration:
641 def _read_parts(self, ix_fn):
642 while self._load_buf(None):
643 assert(self.buf is not None)
645 if i is None or i == len(self.buf):
650 self.buf = self.buf[i:]
658 return buf.index(b'\n')+1
661 return b''.join(self._read_parts(find_eol))
663 def _read(self, size):
665 def until_size(buf): # Closes on csize
666 if len(buf) < csize[0]:
671 return b''.join(self._read_parts(until_size))
674 return self._load_buf(0)
678 """Generate a list of input lines from 'f' without terminating newlines."""
686 def chunkyreader(f, count = None):
687 """Generate a list of chunks of data read from 'f'.
689 If count is None, read until EOF is reached.
691 If count is a positive integer, read 'count' bytes from 'f'. If EOF is
692 reached while reading, raise IOError.
696 b = f.read(min(count, 65536))
698 raise IOError('EOF with %d bytes remaining' % count)
709 def atomically_replaced_file(name, mode='w', buffering=-1):
710 """Yield a file that will be atomically renamed name when leaving the block.
712 This contextmanager yields an open file object that is backed by a
713 temporary file which will be renamed (atomically) to the target
714 name if everything succeeds.
716 The mode and buffering arguments are handled exactly as with open,
717 and the yielded file will have very restrictive permissions, as
722 with atomically_replaced_file('foo.txt', 'w') as f:
723 f.write('hello jack.')
727 (ffd, tempname) = tempfile.mkstemp(dir=os.path.dirname(name),
728 text=('b' not in mode))
731 f = os.fdopen(ffd, mode, buffering)
739 os.rename(tempname, name)
741 unlink(tempname) # nonexistant file is ignored
745 """Append "/" to 's' if it doesn't aleady end in "/"."""
746 assert isinstance(s, bytes)
747 if s and not s.endswith(b'/'):
753 def _mmap_do(f, sz, flags, prot, close):
755 st = os.fstat(f.fileno())
758 # trying to open a zero-length map gives an error, but an empty
759 # string has all the same behaviour of a zero-length map, ie. it has
762 map = compat.mmap(f.fileno(), sz, flags, prot)
764 f.close() # map will persist beyond file close
768 def mmap_read(f, sz = 0, close=True):
769 """Create a read-only memory mapped region on file 'f'.
770 If sz is 0, the region will cover the entire file.
772 return _mmap_do(f, sz, mmap.MAP_PRIVATE, mmap.PROT_READ, close)
775 def mmap_readwrite(f, sz = 0, close=True):
776 """Create a read-write memory mapped region on file 'f'.
777 If sz is 0, the region will cover the entire file.
779 return _mmap_do(f, sz, mmap.MAP_SHARED, mmap.PROT_READ|mmap.PROT_WRITE,
783 def mmap_readwrite_private(f, sz = 0, close=True):
784 """Create a read-write memory mapped region on file 'f'.
785 If sz is 0, the region will cover the entire file.
786 The map is private, which means the changes are never flushed back to the
789 return _mmap_do(f, sz, mmap.MAP_PRIVATE, mmap.PROT_READ|mmap.PROT_WRITE,
793 _mincore = getattr(_helpers, 'mincore', None)
795 # ./configure ensures that we're on Linux if MINCORE_INCORE isn't defined.
796 MINCORE_INCORE = getattr(_helpers, 'MINCORE_INCORE', 1)
798 _fmincore_chunk_size = None
799 def _set_fmincore_chunk_size():
800 global _fmincore_chunk_size
801 pref_chunk_size = 64 * 1024 * 1024
802 chunk_size = sc_page_size
803 if (sc_page_size < pref_chunk_size):
804 chunk_size = sc_page_size * (pref_chunk_size // sc_page_size)
805 _fmincore_chunk_size = chunk_size
808 """Return the mincore() data for fd as a bytearray whose values can be
809 tested via MINCORE_INCORE, or None if fd does not fully
810 support the operation."""
812 if (st.st_size == 0):
814 if not _fmincore_chunk_size:
815 _set_fmincore_chunk_size()
816 pages_per_chunk = _fmincore_chunk_size // sc_page_size;
817 page_count = (st.st_size + sc_page_size - 1) // sc_page_size;
818 chunk_count = (st.st_size + _fmincore_chunk_size - 1) // _fmincore_chunk_size
819 result = bytearray(page_count)
820 for ci in compat.range(chunk_count):
821 pos = _fmincore_chunk_size * ci;
822 msize = min(_fmincore_chunk_size, st.st_size - pos)
824 m = compat.mmap(fd, msize, mmap.MAP_PRIVATE, 0, 0, pos)
825 except mmap.error as ex:
826 if ex.errno == errno.EINVAL or ex.errno == errno.ENODEV:
827 # Perhaps the file was a pipe, i.e. "... | bup split ..."
831 _mincore(m, msize, 0, result, ci * pages_per_chunk)
832 except OSError as ex:
833 if ex.errno == errno.ENOSYS:
839 def parse_timestamp(epoch_str):
840 """Return the number of nanoseconds since the epoch that are described
841 by epoch_str (100ms, 100ns, ...); when epoch_str cannot be parsed,
842 throw a ValueError that may contain additional information."""
843 ns_per = {'s' : 1000000000,
847 match = re.match(r'^((?:[-+]?[0-9]+)?)(s|ms|us|ns)$', epoch_str)
849 if re.match(r'^([-+]?[0-9]+)$', epoch_str):
850 raise ValueError('must include units, i.e. 100ns, 100ms, ...')
852 (n, units) = match.group(1, 2)
856 return n * ns_per[units]
860 """Parse string or bytes as a possibly unit suffixed number.
863 199.2k means 203981 bytes
864 1GB means 1073741824 bytes
865 2.1 tb means 2199023255552 bytes
867 if isinstance(s, bytes):
868 # FIXME: should this raise a ValueError for UnicodeDecodeError
869 # (perhaps with the latter as the context).
870 s = s.decode('ascii')
871 g = re.match(r'([-+\d.e]+)\s*(\w*)', str(s))
873 raise ValueError("can't parse %r as a number" % s)
874 (val, unit) = g.groups()
877 if unit in ['t', 'tb']:
878 mult = 1024*1024*1024*1024
879 elif unit in ['g', 'gb']:
880 mult = 1024*1024*1024
881 elif unit in ['m', 'mb']:
883 elif unit in ['k', 'kb']:
885 elif unit in ['', 'b']:
888 raise ValueError("invalid unit %r in number %r" % (unit, s))
894 """Append an error message to the list of saved errors.
896 Once processing is able to stop and output the errors, the saved errors are
897 accessible in the module variable helpers.saved_errors.
899 saved_errors.append(e)
908 def die_if_errors(msg=None, status=1):
912 msg = 'warning: %d errors encountered\n' % len(saved_errors)
918 """Replace the default exception handler for KeyboardInterrupt (Ctrl-C).
920 The new exception handler will make sure that bup will exit without an ugly
921 stacktrace when Ctrl-C is hit.
923 oldhook = sys.excepthook
924 def newhook(exctype, value, traceback):
925 if exctype == KeyboardInterrupt:
926 log('\nInterrupted.\n')
928 oldhook(exctype, value, traceback)
929 sys.excepthook = newhook
932 def columnate(l, prefix):
933 """Format elements of 'l' in columns with 'prefix' leading each line.
935 The number of columns is determined automatically based on the string
938 binary = isinstance(prefix, bytes)
939 nothing = b'' if binary else ''
940 nl = b'\n' if binary else '\n'
944 clen = max(len(s) for s in l)
945 ncols = (tty_width() - len(prefix)) // (clen + 2)
950 while len(l) % ncols:
952 rows = len(l) // ncols
953 for s in compat.range(0, len(l), rows):
954 cols.append(l[s:s+rows])
956 fmt = b'%-*s' if binary else '%-*s'
957 for row in zip(*cols):
958 out += prefix + nothing.join((fmt % (clen+2, s)) for s in row) + nl
962 def parse_date_or_fatal(str, fatal):
963 """Parses the given date or calls Option.fatal().
964 For now we expect a string that contains a float."""
967 except ValueError as e:
968 raise fatal('invalid date format (should be a float): %r' % e)
973 def parse_excludes(options, fatal):
974 """Traverse the options and extract all excludes, or call Option.fatal()."""
978 (option, parameter) = flag
979 if option == '--exclude':
980 excluded_paths.append(resolve_parent(argv_bytes(parameter)))
981 elif option == '--exclude-from':
983 f = open(resolve_parent(argv_bytes(parameter)), 'rb')
985 raise fatal("couldn't read %r" % parameter)
986 for exclude_path in f.readlines():
987 # FIXME: perhaps this should be rstrip('\n')
988 exclude_path = resolve_parent(exclude_path.strip())
990 excluded_paths.append(exclude_path)
991 return sorted(frozenset(excluded_paths))
994 def parse_rx_excludes(options, fatal):
995 """Traverse the options and extract all rx excludes, or call
997 excluded_patterns = []
1000 (option, parameter) = flag
1001 if option == '--exclude-rx':
1003 excluded_patterns.append(re.compile(argv_bytes(parameter)))
1004 except re.error as ex:
1005 fatal('invalid --exclude-rx pattern (%r): %s' % (parameter, ex))
1006 elif option == '--exclude-rx-from':
1008 f = open(resolve_parent(parameter), 'rb')
1009 except IOError as e:
1010 raise fatal("couldn't read %r" % parameter)
1011 for pattern in f.readlines():
1012 spattern = pattern.rstrip(b'\n')
1016 excluded_patterns.append(re.compile(spattern))
1017 except re.error as ex:
1018 fatal('invalid --exclude-rx pattern (%r): %s' % (spattern, ex))
1019 return excluded_patterns
1022 def should_rx_exclude_path(path, exclude_rxs):
1023 """Return True if path matches a regular expression in exclude_rxs."""
1024 for rx in exclude_rxs:
1026 debug1('Skipping %r: excluded by rx pattern %r.\n'
1027 % (path, rx.pattern))
1032 # FIXME: Carefully consider the use of functions (os.path.*, etc.)
1033 # that resolve against the current filesystem in the strip/graft
1034 # functions for example, but elsewhere as well. I suspect bup's not
1035 # always being careful about that. For some cases, the contents of
1036 # the current filesystem should be irrelevant, and consulting it might
1037 # produce the wrong result, perhaps via unintended symlink resolution,
1040 def path_components(path):
1041 """Break path into a list of pairs of the form (name,
1042 full_path_to_name). Path must start with '/'.
1044 '/home/foo' -> [('', '/'), ('home', '/home'), ('foo', '/home/foo')]"""
1045 if not path.startswith(b'/'):
1046 raise Exception('path must start with "/": %s' % path_msg(path))
1047 # Since we assume path startswith('/'), we can skip the first element.
1048 result = [(b'', b'/')]
1049 norm_path = os.path.abspath(path)
1050 if norm_path == b'/':
1053 for p in norm_path.split(b'/')[1:]:
1054 full_path += b'/' + p
1055 result.append((p, full_path))
1059 def stripped_path_components(path, strip_prefixes):
1060 """Strip any prefix in strip_prefixes from path and return a list
1061 of path components where each component is (name,
1062 none_or_full_fs_path_to_name). Assume path startswith('/').
1063 See thelpers.py for examples."""
1064 normalized_path = os.path.abspath(path)
1065 sorted_strip_prefixes = sorted(strip_prefixes, key=len, reverse=True)
1066 for bp in sorted_strip_prefixes:
1067 normalized_bp = os.path.abspath(bp)
1068 if normalized_bp == b'/':
1070 if normalized_path.startswith(normalized_bp):
1071 prefix = normalized_path[:len(normalized_bp)]
1073 for p in normalized_path[len(normalized_bp):].split(b'/'):
1077 result.append((p, prefix))
1080 return path_components(path)
1083 def grafted_path_components(graft_points, path):
1084 # Create a result that consists of some number of faked graft
1085 # directories before the graft point, followed by all of the real
1086 # directories from path that are after the graft point. Arrange
1087 # for the directory at the graft point in the result to correspond
1088 # to the "orig" directory in --graft orig=new. See t/thelpers.py
1089 # for some examples.
1091 # Note that given --graft orig=new, orig and new have *nothing* to
1092 # do with each other, even if some of their component names
1093 # match. i.e. --graft /foo/bar/baz=/foo/bar/bax is semantically
1094 # equivalent to --graft /foo/bar/baz=/x/y/z, or even
1097 # FIXME: This can't be the best solution...
1098 clean_path = os.path.abspath(path)
1099 for graft_point in graft_points:
1100 old_prefix, new_prefix = graft_point
1101 # Expand prefixes iff not absolute paths.
1102 old_prefix = os.path.normpath(old_prefix)
1103 new_prefix = os.path.normpath(new_prefix)
1104 if clean_path.startswith(old_prefix):
1105 escaped_prefix = re.escape(old_prefix)
1106 grafted_path = re.sub(br'^' + escaped_prefix, new_prefix, clean_path)
1107 # Handle /foo=/ (at least) -- which produces //whatever.
1108 grafted_path = b'/' + grafted_path.lstrip(b'/')
1109 clean_path_components = path_components(clean_path)
1110 # Count the components that were stripped.
1111 strip_count = 0 if old_prefix == b'/' else old_prefix.count(b'/')
1112 new_prefix_parts = new_prefix.split(b'/')
1113 result_prefix = grafted_path.split(b'/')[:new_prefix.count(b'/')]
1114 result = [(p, None) for p in result_prefix] \
1115 + clean_path_components[strip_count:]
1116 # Now set the graft point name to match the end of new_prefix.
1117 graft_point = len(result_prefix)
1118 result[graft_point] = \
1119 (new_prefix_parts[-1], clean_path_components[strip_count][1])
1120 if new_prefix == b'/': # --graft ...=/ is a special case.
1123 return path_components(clean_path)
1129 _localtime = getattr(_helpers, 'localtime', None)
1132 bup_time = namedtuple('bup_time', ['tm_year', 'tm_mon', 'tm_mday',
1133 'tm_hour', 'tm_min', 'tm_sec',
1134 'tm_wday', 'tm_yday',
1135 'tm_isdst', 'tm_gmtoff', 'tm_zone'])
1137 # Define a localtime() that returns bup_time when possible. Note:
1138 # this means that any helpers.localtime() results may need to be
1139 # passed through to_py_time() before being passed to python's time
1140 # module, which doesn't appear willing to ignore the extra items.
1142 def localtime(time):
1143 return bup_time(*_helpers.localtime(int(floor(time))))
1144 def utc_offset_str(t):
1145 """Return the local offset from UTC as "+hhmm" or "-hhmm" for time t.
1146 If the current UTC offset does not represent an integer number
1147 of minutes, the fractional component will be truncated."""
1148 off = localtime(t).tm_gmtoff
1149 # Note: // doesn't truncate like C for negative values, it rounds down.
1150 offmin = abs(off) // 60
1152 h = (offmin - m) // 60
1153 return b'%+03d%02d' % (-h if off < 0 else h, m)
1155 if isinstance(x, time.struct_time):
1157 return time.struct_time(x[:9])
1159 localtime = time.localtime
1160 def utc_offset_str(t):
1161 return time.strftime(b'%z', localtime(t))
1166 _some_invalid_save_parts_rx = re.compile(br'[\[ ~^:?*\\]|\.\.|//|@{')
1168 def valid_save_name(name):
1169 # Enforce a superset of the restrictions in git-check-ref-format(1)
1171 or name.startswith(b'/') or name.endswith(b'/') \
1172 or name.endswith(b'.'):
1174 if _some_invalid_save_parts_rx.search(name):
1177 if byte_int(c) < 0x20 or byte_int(c) == 0x7f:
1179 for part in name.split(b'/'):
1180 if part.startswith(b'.') or part.endswith(b'.lock'):
1185 _period_rx = re.compile(br'^([0-9]+)(s|min|h|d|w|m|y)$')
1187 def period_as_secs(s):
1190 match = _period_rx.match(s)
1193 mag = int(match.group(1))
1194 scale = match.group(2)
1195 return mag * {b's': 1,
1199 b'w': 60 * 60 * 24 * 7,
1200 b'm': 60 * 60 * 24 * 31,
1201 b'y': 60 * 60 * 24 * 366}[scale]