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, pwd, subprocess, errno, socket, select, mmap, stat, re, struct
11 import hashlib, heapq, math, operator, time, grp, tempfile
13 from bup import _helpers
14 from bup import compat
15 from bup.compat import argv_bytes, byte_int
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
23 """Helper to deal with Python scoping issues"""
27 sc_page_size = os.sysconf('SC_PAGE_SIZE')
28 assert(sc_page_size > 0)
30 sc_arg_max = os.sysconf('SC_ARG_MAX')
31 if sc_arg_max == -1: # "no definite limit" - let's choose 2M
32 sc_arg_max = 2 * 1024 * 1024
36 for result in iterable:
42 """Convert s (ascii bytes) to an integer. Return 0 if s is not a number."""
50 """Convert s (ascii bytes) to a float. Return 0 if s is not a number."""
52 return float(s or b'0')
57 buglvl = atoi(os.environ.get('BUP_DEBUG', 0))
61 _fdatasync = os.fdatasync
62 except AttributeError:
65 if sys.platform.startswith('darwin'):
66 # Apparently os.fsync on OS X doesn't guarantee to sync all the way down
70 return fcntl.fcntl(fd, fcntl.F_FULLFSYNC)
72 # Fallback for file systems (SMB) that do not support F_FULLFSYNC
73 if e.errno == errno.ENOTSUP:
78 fdatasync = _fdatasync
81 def partition(predicate, stream):
82 """Returns (leading_matches_it, rest_it), where leading_matches_it
83 must be completely exhausted before traversing rest_it.
88 ns.first_nonmatch = None
89 def leading_matches():
94 ns.first_nonmatch = (x,)
98 yield ns.first_nonmatch[0]
101 return (leading_matches(), rest())
111 def lines_until_sentinel(f, sentinel, ex_type):
112 # sentinel must end with \n and must contain only one \n
115 if not (line and line.endswith(b'\n')):
116 raise ex_type('Hit EOF while reading line')
122 def stat_if_exists(path):
126 if e.errno != errno.ENOENT:
131 # Write (blockingly) to sockets that may or may not be in blocking mode.
132 # We need this because our stderr is sometimes eaten by subprocesses
133 # (probably ssh) that sometimes make it nonblocking, if only temporarily,
134 # leading to race conditions. Ick. We'll do it the hard way.
135 def _hard_write(fd, buf):
137 (r,w,x) = select.select([], [fd], [], None)
139 raise IOError('select(fd) returned without being writable')
141 sz = os.write(fd, buf)
143 if e.errno != errno.EAGAIN:
151 """Print a log message to stderr."""
154 _hard_write(sys.stderr.fileno(), s if isinstance(s, bytes) else s.encode())
168 istty1 = os.isatty(1) or (atoi(os.environ.get('BUP_FORCE_TTY')) & 1)
169 istty2 = os.isatty(2) or (atoi(os.environ.get('BUP_FORCE_TTY')) & 2)
172 """Calls log() if stderr is a TTY. Does nothing otherwise."""
173 global _last_progress
180 """Calls progress() only if we haven't printed progress in a while.
182 This avoids overloading the stderr buffer with excess junk.
186 if now - _last_prog > 0.1:
192 """Calls progress() to redisplay the most recent progress message.
194 Useful after you've printed some other message that wipes out the
197 if _last_progress and _last_progress.endswith('\r'):
198 progress(_last_progress)
201 def mkdirp(d, mode=None):
202 """Recursively create directories on path 'd'.
204 Unlike os.makedirs(), it doesn't raise an exception if the last element of
205 the path already exists.
213 if e.errno == errno.EEXIST:
220 def __init__(self, entry, read_it):
222 self.read_it = read_it
224 return self.entry < x.entry
226 def merge_iter(iters, pfreq, pfunc, pfinal, key=None):
228 samekey = lambda e, pe: getattr(e, key) == getattr(pe, key, None)
230 samekey = operator.eq
232 total = sum(len(it) for it in iters)
233 iters = (iter(it) for it in iters)
234 heap = ((next(it, None),it) for it in iters)
235 heap = [MergeIterItem(e, it) for e, it in heap if e]
240 if not count % pfreq:
242 e, it = heap[0].entry, heap[0].read_it
243 if not samekey(e, pe):
249 except StopIteration:
250 heapq.heappop(heap) # remove current
252 # shift current to new location
253 heapq.heapreplace(heap, MergeIterItem(e, it))
258 """Delete a file at path 'f' if it currently exists.
260 Unlike os.unlink(), does not throw an exception if the file didn't already
266 if e.errno != errno.ENOENT:
270 _bq_simple_id_rx = re.compile(br'^[-_./a-zA-Z0-9]+$')
271 _sq_simple_id_rx = re.compile(r'^[-_./a-zA-Z0-9]+$')
276 if _bq_simple_id_rx.match(x):
278 return b"'%s'" % x.replace(b"'", b"'\"'\"'")
283 if _sq_simple_id_rx.match(x):
285 return "'%s'" % x.replace("'", "'\"'\"'")
288 if isinstance(x, bytes):
290 if isinstance(x, compat.str_type):
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 = socket.getfqdn().encode('iso-8859-1')
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 while tail != b'BUPMUX':
582 b = os.read(infd, (len(tail) < 6) and (6-len(tail)) or 1)
584 raise IOError('demux: unexpected EOF during initialization')
586 byte_stream(sys.stderr).write(tail[:-6]) # pre-mux log messages
593 def write(self, data):
595 BaseConn.write(self, data)
597 def _next_packet(self, timeout):
598 if self.closed: return False
599 rl, wl, xl = select.select([self.infd], [], [], timeout)
600 if not rl: return False
601 assert(rl[0] == self.infd)
602 ns = b''.join(checked_reader(self.infd, 5))
603 n, fdw = struct.unpack('!IB', ns)
604 assert(n <= MAX_PACKET)
606 self.reader = checked_reader(self.infd, n)
608 for buf in checked_reader(self.infd, n):
609 byte_stream(sys.stderr).write(buf)
612 debug2("DemuxConn: marked closed\n")
615 def _load_buf(self, timeout):
616 if self.buf is not None:
618 while not self.closed:
619 while not self.reader:
620 if not self._next_packet(timeout):
623 self.buf = next(self.reader)
625 except StopIteration:
629 def _read_parts(self, ix_fn):
630 while self._load_buf(None):
631 assert(self.buf is not None)
633 if i is None or i == len(self.buf):
638 self.buf = self.buf[i:]
646 return buf.index(b'\n')+1
649 return b''.join(self._read_parts(find_eol))
651 def _read(self, size):
653 def until_size(buf): # Closes on csize
654 if len(buf) < csize[0]:
659 return b''.join(self._read_parts(until_size))
662 return self._load_buf(0)
666 """Generate a list of input lines from 'f' without terminating newlines."""
674 def chunkyreader(f, count = None):
675 """Generate a list of chunks of data read from 'f'.
677 If count is None, read until EOF is reached.
679 If count is a positive integer, read 'count' bytes from 'f'. If EOF is
680 reached while reading, raise IOError.
684 b = f.read(min(count, 65536))
686 raise IOError('EOF with %d bytes remaining' % count)
697 def atomically_replaced_file(name, mode='w', buffering=-1):
698 """Yield a file that will be atomically renamed name when leaving the block.
700 This contextmanager yields an open file object that is backed by a
701 temporary file which will be renamed (atomically) to the target
702 name if everything succeeds.
704 The mode and buffering arguments are handled exactly as with open,
705 and the yielded file will have very restrictive permissions, as
710 with atomically_replaced_file('foo.txt', 'w') as f:
711 f.write('hello jack.')
715 (ffd, tempname) = tempfile.mkstemp(dir=os.path.dirname(name),
716 text=('b' not in mode))
719 f = os.fdopen(ffd, mode, buffering)
727 os.rename(tempname, name)
729 unlink(tempname) # nonexistant file is ignored
733 """Append "/" to 's' if it doesn't aleady end in "/"."""
734 assert isinstance(s, bytes)
735 if s and not s.endswith(b'/'):
741 def _mmap_do(f, sz, flags, prot, close):
743 st = os.fstat(f.fileno())
746 # trying to open a zero-length map gives an error, but an empty
747 # string has all the same behaviour of a zero-length map, ie. it has
750 map = mmap.mmap(f.fileno(), sz, flags, prot)
752 f.close() # map will persist beyond file close
756 def mmap_read(f, sz = 0, close=True):
757 """Create a read-only memory mapped region on file 'f'.
758 If sz is 0, the region will cover the entire file.
760 return _mmap_do(f, sz, mmap.MAP_PRIVATE, mmap.PROT_READ, close)
763 def mmap_readwrite(f, sz = 0, close=True):
764 """Create a read-write memory mapped region on file 'f'.
765 If sz is 0, the region will cover the entire file.
767 return _mmap_do(f, sz, mmap.MAP_SHARED, mmap.PROT_READ|mmap.PROT_WRITE,
771 def mmap_readwrite_private(f, sz = 0, close=True):
772 """Create a read-write memory mapped region on file 'f'.
773 If sz is 0, the region will cover the entire file.
774 The map is private, which means the changes are never flushed back to the
777 return _mmap_do(f, sz, mmap.MAP_PRIVATE, mmap.PROT_READ|mmap.PROT_WRITE,
781 _mincore = getattr(_helpers, 'mincore', None)
783 # ./configure ensures that we're on Linux if MINCORE_INCORE isn't defined.
784 MINCORE_INCORE = getattr(_helpers, 'MINCORE_INCORE', 1)
786 _fmincore_chunk_size = None
787 def _set_fmincore_chunk_size():
788 global _fmincore_chunk_size
789 pref_chunk_size = 64 * 1024 * 1024
790 chunk_size = sc_page_size
791 if (sc_page_size < pref_chunk_size):
792 chunk_size = sc_page_size * (pref_chunk_size // sc_page_size)
793 _fmincore_chunk_size = chunk_size
796 """Return the mincore() data for fd as a bytearray whose values can be
797 tested via MINCORE_INCORE, or None if fd does not fully
798 support the operation."""
800 if (st.st_size == 0):
802 if not _fmincore_chunk_size:
803 _set_fmincore_chunk_size()
804 pages_per_chunk = _fmincore_chunk_size // sc_page_size;
805 page_count = (st.st_size + sc_page_size - 1) // sc_page_size;
806 chunk_count = page_count // _fmincore_chunk_size
809 result = bytearray(page_count)
810 for ci in compat.range(chunk_count):
811 pos = _fmincore_chunk_size * ci;
812 msize = min(_fmincore_chunk_size, st.st_size - pos)
814 m = mmap.mmap(fd, msize, mmap.MAP_PRIVATE, 0, 0, pos)
815 except mmap.error as ex:
816 if ex.errno == errno.EINVAL or ex.errno == errno.ENODEV:
817 # Perhaps the file was a pipe, i.e. "... | bup split ..."
821 _mincore(m, msize, 0, result, ci * pages_per_chunk)
822 except OSError as ex:
823 if ex.errno == errno.ENOSYS:
829 def parse_timestamp(epoch_str):
830 """Return the number of nanoseconds since the epoch that are described
831 by epoch_str (100ms, 100ns, ...); when epoch_str cannot be parsed,
832 throw a ValueError that may contain additional information."""
833 ns_per = {'s' : 1000000000,
837 match = re.match(r'^((?:[-+]?[0-9]+)?)(s|ms|us|ns)$', epoch_str)
839 if re.match(r'^([-+]?[0-9]+)$', epoch_str):
840 raise ValueError('must include units, i.e. 100ns, 100ms, ...')
842 (n, units) = match.group(1, 2)
846 return n * ns_per[units]
850 """Parse string or bytes as a possibly unit suffixed number.
853 199.2k means 203981 bytes
854 1GB means 1073741824 bytes
855 2.1 tb means 2199023255552 bytes
857 if isinstance(s, bytes):
858 # FIXME: should this raise a ValueError for UnicodeDecodeError
859 # (perhaps with the latter as the context).
860 s = s.decode('ascii')
861 g = re.match(r'([-+\d.e]+)\s*(\w*)', str(s))
863 raise ValueError("can't parse %r as a number" % s)
864 (val, unit) = g.groups()
867 if unit in ['t', 'tb']:
868 mult = 1024*1024*1024*1024
869 elif unit in ['g', 'gb']:
870 mult = 1024*1024*1024
871 elif unit in ['m', 'mb']:
873 elif unit in ['k', 'kb']:
875 elif unit in ['', 'b']:
878 raise ValueError("invalid unit %r in number %r" % (unit, s))
884 """Append an error message to the list of saved errors.
886 Once processing is able to stop and output the errors, the saved errors are
887 accessible in the module variable helpers.saved_errors.
889 saved_errors.append(e)
898 def die_if_errors(msg=None, status=1):
902 msg = 'warning: %d errors encountered\n' % len(saved_errors)
908 """Replace the default exception handler for KeyboardInterrupt (Ctrl-C).
910 The new exception handler will make sure that bup will exit without an ugly
911 stacktrace when Ctrl-C is hit.
913 oldhook = sys.excepthook
914 def newhook(exctype, value, traceback):
915 if exctype == KeyboardInterrupt:
916 log('\nInterrupted.\n')
918 return oldhook(exctype, value, traceback)
919 sys.excepthook = newhook
922 def columnate(l, prefix):
923 """Format elements of 'l' in columns with 'prefix' leading each line.
925 The number of columns is determined automatically based on the string
928 binary = isinstance(prefix, bytes)
929 nothing = b'' if binary else ''
930 nl = b'\n' if binary else '\n'
934 clen = max(len(s) for s in l)
935 ncols = (tty_width() - len(prefix)) // (clen + 2)
940 while len(l) % ncols:
942 rows = len(l) // ncols
943 for s in compat.range(0, len(l), rows):
944 cols.append(l[s:s+rows])
946 fmt = b'%-*s' if binary else '%-*s'
947 for row in zip(*cols):
948 out += prefix + nothing.join((fmt % (clen+2, s)) for s in row) + nl
952 def parse_date_or_fatal(str, fatal):
953 """Parses the given date or calls Option.fatal().
954 For now we expect a string that contains a float."""
957 except ValueError as e:
958 raise fatal('invalid date format (should be a float): %r' % e)
963 def parse_excludes(options, fatal):
964 """Traverse the options and extract all excludes, or call Option.fatal()."""
968 (option, parameter) = flag
969 if option == '--exclude':
970 excluded_paths.append(resolve_parent(argv_bytes(parameter)))
971 elif option == '--exclude-from':
973 f = open(resolve_parent(argv_bytes(parameter)), 'rb')
975 raise fatal("couldn't read %r" % parameter)
976 for exclude_path in f.readlines():
977 # FIXME: perhaps this should be rstrip('\n')
978 exclude_path = resolve_parent(exclude_path.strip())
980 excluded_paths.append(exclude_path)
981 return sorted(frozenset(excluded_paths))
984 def parse_rx_excludes(options, fatal):
985 """Traverse the options and extract all rx excludes, or call
987 excluded_patterns = []
990 (option, parameter) = flag
991 if option == '--exclude-rx':
993 excluded_patterns.append(re.compile(argv_bytes(parameter)))
994 except re.error as ex:
995 fatal('invalid --exclude-rx pattern (%r): %s' % (parameter, ex))
996 elif option == '--exclude-rx-from':
998 f = open(resolve_parent(parameter), 'rb')
1000 raise fatal("couldn't read %r" % parameter)
1001 for pattern in f.readlines():
1002 spattern = pattern.rstrip(b'\n')
1006 excluded_patterns.append(re.compile(spattern))
1007 except re.error as ex:
1008 fatal('invalid --exclude-rx pattern (%r): %s' % (spattern, ex))
1009 return excluded_patterns
1012 def should_rx_exclude_path(path, exclude_rxs):
1013 """Return True if path matches a regular expression in exclude_rxs."""
1014 for rx in exclude_rxs:
1016 debug1('Skipping %r: excluded by rx pattern %r.\n'
1017 % (path, rx.pattern))
1022 # FIXME: Carefully consider the use of functions (os.path.*, etc.)
1023 # that resolve against the current filesystem in the strip/graft
1024 # functions for example, but elsewhere as well. I suspect bup's not
1025 # always being careful about that. For some cases, the contents of
1026 # the current filesystem should be irrelevant, and consulting it might
1027 # produce the wrong result, perhaps via unintended symlink resolution,
1030 def path_components(path):
1031 """Break path into a list of pairs of the form (name,
1032 full_path_to_name). Path must start with '/'.
1034 '/home/foo' -> [('', '/'), ('home', '/home'), ('foo', '/home/foo')]"""
1035 if not path.startswith(b'/'):
1036 raise Exception('path must start with "/": %s' % path_msg(path))
1037 # Since we assume path startswith('/'), we can skip the first element.
1038 result = [(b'', b'/')]
1039 norm_path = os.path.abspath(path)
1040 if norm_path == b'/':
1043 for p in norm_path.split(b'/')[1:]:
1044 full_path += b'/' + p
1045 result.append((p, full_path))
1049 def stripped_path_components(path, strip_prefixes):
1050 """Strip any prefix in strip_prefixes from path and return a list
1051 of path components where each component is (name,
1052 none_or_full_fs_path_to_name). Assume path startswith('/').
1053 See thelpers.py for examples."""
1054 normalized_path = os.path.abspath(path)
1055 sorted_strip_prefixes = sorted(strip_prefixes, key=len, reverse=True)
1056 for bp in sorted_strip_prefixes:
1057 normalized_bp = os.path.abspath(bp)
1058 if normalized_bp == b'/':
1060 if normalized_path.startswith(normalized_bp):
1061 prefix = normalized_path[:len(normalized_bp)]
1063 for p in normalized_path[len(normalized_bp):].split(b'/'):
1067 result.append((p, prefix))
1070 return path_components(path)
1073 def grafted_path_components(graft_points, path):
1074 # Create a result that consists of some number of faked graft
1075 # directories before the graft point, followed by all of the real
1076 # directories from path that are after the graft point. Arrange
1077 # for the directory at the graft point in the result to correspond
1078 # to the "orig" directory in --graft orig=new. See t/thelpers.py
1079 # for some examples.
1081 # Note that given --graft orig=new, orig and new have *nothing* to
1082 # do with each other, even if some of their component names
1083 # match. i.e. --graft /foo/bar/baz=/foo/bar/bax is semantically
1084 # equivalent to --graft /foo/bar/baz=/x/y/z, or even
1087 # FIXME: This can't be the best solution...
1088 clean_path = os.path.abspath(path)
1089 for graft_point in graft_points:
1090 old_prefix, new_prefix = graft_point
1091 # Expand prefixes iff not absolute paths.
1092 old_prefix = os.path.normpath(old_prefix)
1093 new_prefix = os.path.normpath(new_prefix)
1094 if clean_path.startswith(old_prefix):
1095 escaped_prefix = re.escape(old_prefix)
1096 grafted_path = re.sub(br'^' + escaped_prefix, new_prefix, clean_path)
1097 # Handle /foo=/ (at least) -- which produces //whatever.
1098 grafted_path = b'/' + grafted_path.lstrip(b'/')
1099 clean_path_components = path_components(clean_path)
1100 # Count the components that were stripped.
1101 strip_count = 0 if old_prefix == b'/' else old_prefix.count(b'/')
1102 new_prefix_parts = new_prefix.split(b'/')
1103 result_prefix = grafted_path.split(b'/')[:new_prefix.count(b'/')]
1104 result = [(p, None) for p in result_prefix] \
1105 + clean_path_components[strip_count:]
1106 # Now set the graft point name to match the end of new_prefix.
1107 graft_point = len(result_prefix)
1108 result[graft_point] = \
1109 (new_prefix_parts[-1], clean_path_components[strip_count][1])
1110 if new_prefix == b'/': # --graft ...=/ is a special case.
1113 return path_components(clean_path)
1119 _localtime = getattr(_helpers, 'localtime', None)
1122 bup_time = namedtuple('bup_time', ['tm_year', 'tm_mon', 'tm_mday',
1123 'tm_hour', 'tm_min', 'tm_sec',
1124 'tm_wday', 'tm_yday',
1125 'tm_isdst', 'tm_gmtoff', 'tm_zone'])
1127 # Define a localtime() that returns bup_time when possible. Note:
1128 # this means that any helpers.localtime() results may need to be
1129 # passed through to_py_time() before being passed to python's time
1130 # module, which doesn't appear willing to ignore the extra items.
1132 def localtime(time):
1133 return bup_time(*_helpers.localtime(floor(time)))
1134 def utc_offset_str(t):
1135 """Return the local offset from UTC as "+hhmm" or "-hhmm" for time t.
1136 If the current UTC offset does not represent an integer number
1137 of minutes, the fractional component will be truncated."""
1138 off = localtime(t).tm_gmtoff
1139 # Note: // doesn't truncate like C for negative values, it rounds down.
1140 offmin = abs(off) // 60
1142 h = (offmin - m) // 60
1143 return b'%+03d%02d' % (-h if off < 0 else h, m)
1145 if isinstance(x, time.struct_time):
1147 return time.struct_time(x[:9])
1149 localtime = time.localtime
1150 def utc_offset_str(t):
1151 return time.strftime(b'%z', localtime(t))
1156 _some_invalid_save_parts_rx = re.compile(br'[\[ ~^:?*\\]|\.\.|//|@{')
1158 def valid_save_name(name):
1159 # Enforce a superset of the restrictions in git-check-ref-format(1)
1161 or name.startswith(b'/') or name.endswith(b'/') \
1162 or name.endswith(b'.'):
1164 if _some_invalid_save_parts_rx.search(name):
1167 if byte_int(c) < 0x20 or byte_int(c) == 0x7f:
1169 for part in name.split(b'/'):
1170 if part.startswith(b'.') or part.endswith(b'.lock'):
1175 _period_rx = re.compile(br'^([0-9]+)(s|min|h|d|w|m|y)$')
1177 def period_as_secs(s):
1180 match = _period_rx.match(s)
1183 mag = int(match.group(1))
1184 scale = match.group(2)
1185 return mag * {b's': 1,
1189 b'w': 60 * 60 * 24 * 7,
1190 b'm': 60 * 60 * 24 * 31,
1191 b'y': 60 * 60 * 24 * 366}[scale]