X-Git-Url: https://arthur.barton.de/gitweb/?p=bup.git;a=blobdiff_plain;f=lib%2Fbup%2Fhelpers.py;h=1bae59d22e99da707daac8198bcba5c0275882b7;hp=7b5eeadafaa78e051083d9119ab687e6882d0f65;hb=41c3f3d78d79531863110f5ffd9ae8ee5a2b3986;hpb=ee285725a94584b6284871103a8ff3fe2070f7f1

diff --git a/lib/bup/helpers.py b/lib/bup/helpers.py
index 7b5eead..1bae59d 100644
--- a/lib/bup/helpers.py
+++ b/lib/bup/helpers.py
@@ -1,13 +1,39 @@
 """Helper functions and classes for bup."""
 
+from __future__ import absolute_import
+from collections import namedtuple
+from contextlib import contextmanager
+from ctypes import sizeof, c_void_p
+from os import environ
+from pipes import quote
+from subprocess import PIPE, Popen
 import sys, os, pwd, subprocess, errno, socket, select, mmap, stat, re, struct
-import heapq, operator
-from bup import _version
+import hashlib, heapq, math, operator, time, grp, tempfile
 
+from bup import _helpers
+from bup import compat
 # This function should really be in helpers, not in bup.options.  But we
 # want options.py to be standalone so people can include it in other projects.
-from bup.options import _tty_width
-tty_width = _tty_width
+from bup.options import _tty_width as tty_width
+
+
+class Nonlocal:
+    """Helper to deal with Python scoping issues"""
+    pass
+
+
+sc_page_size = os.sysconf('SC_PAGE_SIZE')
+assert(sc_page_size > 0)
+
+sc_arg_max = os.sysconf('SC_ARG_MAX')
+if sc_arg_max == -1:  # "no definite limit" - let's choose 2M
+    sc_arg_max = 2 * 1024 * 1024
+
+def last(iterable):
+    result = None
+    for result in iterable:
+        pass
+    return result
 
 
 def atoi(s):
@@ -29,6 +55,70 @@ def atof(s):
 buglvl = atoi(os.environ.get('BUP_DEBUG', 0))
 
 
+try:
+    _fdatasync = os.fdatasync
+except AttributeError:
+    _fdatasync = os.fsync
+
+if sys.platform.startswith('darwin'):
+    # Apparently os.fsync on OS X doesn't guarantee to sync all the way down
+    import fcntl
+    def fdatasync(fd):
+        try:
+            return fcntl.fcntl(fd, fcntl.F_FULLFSYNC)
+        except IOError as e:
+            # Fallback for file systems (SMB) that do not support F_FULLFSYNC
+            if e.errno == errno.ENOTSUP:
+                return _fdatasync(fd)
+            else:
+                raise
+else:
+    fdatasync = _fdatasync
+
+
+def partition(predicate, stream):
+    """Returns (leading_matches_it, rest_it), where leading_matches_it
+    must be completely exhausted before traversing rest_it.
+
+    """
+    stream = iter(stream)
+    ns = Nonlocal()
+    ns.first_nonmatch = None
+    def leading_matches():
+        for x in stream:
+            if predicate(x):
+                yield x
+            else:
+                ns.first_nonmatch = (x,)
+                break
+    def rest():
+        if ns.first_nonmatch:
+            yield ns.first_nonmatch[0]
+            for x in stream:
+                yield x
+    return (leading_matches(), rest())
+
+
+def lines_until_sentinel(f, sentinel, ex_type):
+    # sentinel must end with \n and must contain only one \n
+    while True:
+        line = f.readline()
+        if not (line and line.endswith('\n')):
+            raise ex_type('Hit EOF while reading line')
+        if line == sentinel:
+            return
+        yield line
+
+
+def stat_if_exists(path):
+    try:
+        return os.stat(path)
+    except OSError as e:
+        if e.errno != errno.ENOENT:
+            raise
+    return None
+
+
 # Write (blockingly) to sockets that may or may not be in blocking mode.
 # We need this because our stderr is sometimes eaten by subprocesses
 # (probably ssh) that sometimes make it nonblocking, if only temporarily,
@@ -40,16 +130,20 @@ def _hard_write(fd, buf):
             raise IOError('select(fd) returned without being writable')
         try:
             sz = os.write(fd, buf)
-        except OSError, e:
+        except OSError as e:
             if e.errno != errno.EAGAIN:
                 raise
         assert(sz >= 0)
         buf = buf[sz:]
 
+
+_last_prog = 0
 def log(s):
     """Print a log message to stderr."""
+    global _last_prog
     sys.stdout.flush()
     _hard_write(sys.stderr.fileno(), s)
+    _last_prog = 0
 
 
 def debug1(s):
@@ -62,6 +156,39 @@ def debug2(s):
         log(s)
 
 
+istty1 = os.isatty(1) or (atoi(os.environ.get('BUP_FORCE_TTY')) & 1)
+istty2 = os.isatty(2) or (atoi(os.environ.get('BUP_FORCE_TTY')) & 2)
+_last_progress = ''
+def progress(s):
+    """Calls log() if stderr is a TTY.  Does nothing otherwise."""
+    global _last_progress
+    if istty2:
+        log(s)
+        _last_progress = s
+
+
+def qprogress(s):
+    """Calls progress() only if we haven't printed progress in a while.
+    
+    This avoids overloading the stderr buffer with excess junk.
+    """
+    global _last_prog
+    now = time.time()
+    if now - _last_prog > 0.1:
+        progress(s)
+        _last_prog = now
+
+
+def reprogress():
+    """Calls progress() to redisplay the most recent progress message.
+
+    Useful after you've printed some other message that wipes out the
+    progress line.
+    """
+    if _last_progress and _last_progress.endswith('\r'):
+        progress(_last_progress)
+
+
 def mkdirp(d, mode=None):
     """Recursively create directories on path 'd'.
 
@@ -73,21 +200,13 @@ def mkdirp(d, mode=None):
             os.makedirs(d, mode)
         else:
             os.makedirs(d)
-    except OSError, e:
+    except OSError as e:
         if e.errno == errno.EEXIST:
             pass
         else:
             raise
 
 
-def next(it):
-    """Get the next item from an iterator, None if we reached the end."""
-    try:
-        return it.next()
-    except StopIteration:
-        return None
-
-
 def merge_iter(iters, pfreq, pfunc, pfinal, key=None):
     if key:
         samekey = lambda e, pe: getattr(e, key) == getattr(pe, key, None)
@@ -96,7 +215,7 @@ def merge_iter(iters, pfreq, pfunc, pfinal, key=None):
     count = 0
     total = sum(len(it) for it in iters)
     iters = (iter(it) for it in iters)
-    heap = ((next(it),it) for it in iters)
+    heap = ((next(it, None),it) for it in iters)
     heap = [(e,it) for e,it in heap if e]
 
     heapq.heapify(heap)
@@ -110,7 +229,7 @@ def merge_iter(iters, pfreq, pfunc, pfinal, key=None):
             yield e
         count += 1
         try:
-            e = it.next() # Don't use next() function, it's too expensive
+            e = next(it)
         except StopIteration:
             heapq.heappop(heap) # remove current
         else:
@@ -126,21 +245,87 @@ def unlink(f):
     """
     try:
         os.unlink(f)
-    except OSError, e:
-        if e.errno == errno.ENOENT:
-            pass  # it doesn't exist, that's what you asked for
+    except OSError as e:
+        if e.errno != errno.ENOENT:
+            raise
 
 
-def readpipe(argv):
+def shstr(cmd):
+    if isinstance(cmd, compat.str_type):
+        return cmd
+    else:
+        return ' '.join(map(quote, cmd))
+
+exc = subprocess.check_call
+
+def exo(cmd,
+        input=None,
+        stdin=None,
+        stderr=None,
+        shell=False,
+        check=True,
+        preexec_fn=None):
+    if input:
+        assert stdin in (None, PIPE)
+        stdin = PIPE
+    p = Popen(cmd,
+              stdin=stdin, stdout=PIPE, stderr=stderr,
+              shell=shell,
+              preexec_fn=preexec_fn)
+    out, err = p.communicate(input)
+    if check and p.returncode != 0:
+        raise Exception('subprocess %r failed with status %d, stderr: %r'
+                        % (' '.join(map(quote, cmd)), p.returncode, err))
+    return out, err, p
+
+def readpipe(argv, preexec_fn=None, shell=False):
     """Run a subprocess and return its output."""
-    p = subprocess.Popen(argv, stdout=subprocess.PIPE)
-    r = p.stdout.read()
-    p.wait()
-    return r
+    p = subprocess.Popen(argv, stdout=subprocess.PIPE, preexec_fn=preexec_fn,
+                         shell=shell)
+    out, err = p.communicate()
+    if p.returncode != 0:
+        raise Exception('subprocess %r failed with status %d'
+                        % (' '.join(argv), p.returncode))
+    return out
 
 
-def realpath(p):
-    """Get the absolute path of a file.
+def _argmax_base(command):
+    base_size = 2048
+    for c in command:
+        base_size += len(command) + 1
+    for k, v in compat.items(environ):
+        base_size += len(k) + len(v) + 2 + sizeof(c_void_p)
+    return base_size
+
+
+def _argmax_args_size(args):
+    return sum(len(x) + 1 + sizeof(c_void_p) for x in args)
+
+
+def batchpipe(command, args, preexec_fn=None, arg_max=sc_arg_max):
+    """If args is not empty, yield the output produced by calling the
+command list with args as a sequence of strings (It may be necessary
+to return multiple strings in order to respect ARG_MAX)."""
+    # The optional arg_max arg is a workaround for an issue with the
+    # current wvtest behavior.
+    base_size = _argmax_base(command)
+    while args:
+        room = arg_max - base_size
+        i = 0
+        while i < len(args):
+            next_size = _argmax_args_size(args[i:i+1])
+            if room - next_size < 0:
+                break
+            room -= next_size
+            i += 1
+        sub_args = args[:i]
+        args = args[i:]
+        assert(len(sub_args))
+        yield readpipe(command + sub_args, preexec_fn=preexec_fn)
+
+
+def resolve_parent(p):
+    """Return the absolute path of a file without following any final symlink.
 
     Behaves like os.path.realpath, but doesn't follow a symlink for the last
     element. (ie. if 'p' itself is a symlink, this one won't follow it, but it
@@ -160,16 +345,97 @@ def realpath(p):
     return out
 
 
+def detect_fakeroot():
+    "Return True if we appear to be running under fakeroot."
+    return os.getenv("FAKEROOTKEY") != None
+
+
+if sys.platform.startswith('cygwin'):
+    def is_superuser():
+        # https://cygwin.com/ml/cygwin/2015-02/msg00057.html
+        groups = os.getgroups()
+        return 544 in groups or 0 in groups
+else:
+    def is_superuser():
+        return os.geteuid() == 0
+
+
+def _cache_key_value(get_value, key, cache):
+    """Return (value, was_cached).  If there is a value in the cache
+    for key, use that, otherwise, call get_value(key) which should
+    throw a KeyError if there is no value -- in which case the cached
+    and returned value will be None.
+    """
+    try: # Do we already have it (or know there wasn't one)?
+        value = cache[key]
+        return value, True
+    except KeyError:
+        pass
+    value = None
+    try:
+        cache[key] = value = get_value(key)
+    except KeyError:
+        cache[key] = None
+    return value, False
+
+
+_uid_to_pwd_cache = {}
+_name_to_pwd_cache = {}
+
+def pwd_from_uid(uid):
+    """Return password database entry for uid (may be a cached value).
+    Return None if no entry is found.
+    """
+    global _uid_to_pwd_cache, _name_to_pwd_cache
+    entry, cached = _cache_key_value(pwd.getpwuid, uid, _uid_to_pwd_cache)
+    if entry and not cached:
+        _name_to_pwd_cache[entry.pw_name] = entry
+    return entry
+
+
+def pwd_from_name(name):
+    """Return password database entry for name (may be a cached value).
+    Return None if no entry is found.
+    """
+    global _uid_to_pwd_cache, _name_to_pwd_cache
+    entry, cached = _cache_key_value(pwd.getpwnam, name, _name_to_pwd_cache)
+    if entry and not cached:
+        _uid_to_pwd_cache[entry.pw_uid] = entry
+    return entry
+
+
+_gid_to_grp_cache = {}
+_name_to_grp_cache = {}
+
+def grp_from_gid(gid):
+    """Return password database entry for gid (may be a cached value).
+    Return None if no entry is found.
+    """
+    global _gid_to_grp_cache, _name_to_grp_cache
+    entry, cached = _cache_key_value(grp.getgrgid, gid, _gid_to_grp_cache)
+    if entry and not cached:
+        _name_to_grp_cache[entry.gr_name] = entry
+    return entry
+
+
+def grp_from_name(name):
+    """Return password database entry for name (may be a cached value).
+    Return None if no entry is found.
+    """
+    global _gid_to_grp_cache, _name_to_grp_cache
+    entry, cached = _cache_key_value(grp.getgrnam, name, _name_to_grp_cache)
+    if entry and not cached:
+        _gid_to_grp_cache[entry.gr_gid] = entry
+    return entry
+
+
 _username = None
 def username():
     """Get the user's login name."""
     global _username
     if not _username:
         uid = os.getuid()
-        try:
-            _username = pwd.getpwuid(uid)[0]
-        except KeyError:
-            _username = 'user%d' % uid
+        _username = pwd_from_uid(uid)[0] or 'user%d' % uid
     return _username
 
 
@@ -179,9 +445,10 @@ def userfullname():
     global _userfullname
     if not _userfullname:
         uid = os.getuid()
-        try:
-            _userfullname = pwd.getpwuid(uid)[4].split(',')[0]
-        except KeyError:
+        entry = pwd_from_uid(uid)
+        if entry:
+            _userfullname = entry[4].split(',')[0] or entry[0]
+        if not _userfullname:
             _userfullname = 'user%d' % uid
     return _userfullname
 
@@ -202,6 +469,15 @@ def resource_path(subdir=''):
         _resource_path = os.environ.get('BUP_RESOURCE_PATH') or '.'
     return os.path.join(_resource_path, subdir)
 
+def format_filesize(size):
+    unit = 1024.0
+    size = float(size)
+    if size < unit:
+        return "%d" % (size)
+    exponent = int(math.log(size) / math.log(unit))
+    size_prefix = "KMGTPE"[exponent - 1]
+    return "%.1f%s" % (size / math.pow(unit, exponent), size_prefix)
+
 
 class NotOk(Exception):
     pass
@@ -369,7 +645,7 @@ class DemuxConn(BaseConn):
                 if not self._next_packet(timeout):
                     return False
             try:
-                self.buf = self.reader.next()
+                self.buf = next(self.reader)
                 return True
             except StopIteration:
                 self.reader = None
@@ -442,6 +718,42 @@ def chunkyreader(f, count = None):
             yield b
 
 
+@contextmanager
+def atomically_replaced_file(name, mode='w', buffering=-1):
+    """Yield a file that will be atomically renamed name when leaving the block.
+
+    This contextmanager yields an open file object that is backed by a
+    temporary file which will be renamed (atomically) to the target
+    name if everything succeeds.
+
+    The mode and buffering arguments are handled exactly as with open,
+    and the yielded file will have very restrictive permissions, as
+    per mkstemp.
+
+    E.g.::
+
+        with atomically_replaced_file('foo.txt', 'w') as f:
+            f.write('hello jack.')
+
+    """
+
+    (ffd, tempname) = tempfile.mkstemp(dir=os.path.dirname(name),
+                                       text=('b' not in mode))
+    try:
+        try:
+            f = os.fdopen(ffd, mode, buffering)
+        except:
+            os.close(ffd)
+            raise
+        try:
+            yield f
+        finally:
+            f.close()
+        os.rename(tempname, name)
+    finally:
+        unlink(tempname)  # nonexistant file is ignored
+
+
 def slashappend(s):
     """Append "/" to 's' if it doesn't aleady end in "/"."""
     if s and not s.endswith('/'):
@@ -467,7 +779,6 @@ def _mmap_do(f, sz, flags, prot, close):
 
 def mmap_read(f, sz = 0, close=True):
     """Create a read-only memory mapped region on file 'f'.
-
     If sz is 0, the region will cover the entire file.
     """
     return _mmap_do(f, sz, mmap.MAP_PRIVATE, mmap.PROT_READ, close)
@@ -475,13 +786,90 @@ def mmap_read(f, sz = 0, close=True):
 
 def mmap_readwrite(f, sz = 0, close=True):
     """Create a read-write memory mapped region on file 'f'.
-
     If sz is 0, the region will cover the entire file.
     """
     return _mmap_do(f, sz, mmap.MAP_SHARED, mmap.PROT_READ|mmap.PROT_WRITE,
                     close)
 
 
+def mmap_readwrite_private(f, sz = 0, close=True):
+    """Create a read-write memory mapped region on file 'f'.
+    If sz is 0, the region will cover the entire file.
+    The map is private, which means the changes are never flushed back to the
+    file.
+    """
+    return _mmap_do(f, sz, mmap.MAP_PRIVATE, mmap.PROT_READ|mmap.PROT_WRITE,
+                    close)
+
+
+_mincore = getattr(_helpers, 'mincore', None)
+if _mincore:
+    # ./configure ensures that we're on Linux if MINCORE_INCORE isn't defined.
+    MINCORE_INCORE = getattr(_helpers, 'MINCORE_INCORE', 1)
+
+    _fmincore_chunk_size = None
+    def _set_fmincore_chunk_size():
+        global _fmincore_chunk_size
+        pref_chunk_size = 64 * 1024 * 1024
+        chunk_size = sc_page_size
+        if (sc_page_size < pref_chunk_size):
+            chunk_size = sc_page_size * (pref_chunk_size / sc_page_size)
+        _fmincore_chunk_size = chunk_size
+
+    def fmincore(fd):
+        """Return the mincore() data for fd as a bytearray whose values can be
+        tested via MINCORE_INCORE, or None if fd does not fully
+        support the operation."""
+        st = os.fstat(fd)
+        if (st.st_size == 0):
+            return bytearray(0)
+        if not _fmincore_chunk_size:
+            _set_fmincore_chunk_size()
+        pages_per_chunk = _fmincore_chunk_size / sc_page_size;
+        page_count = (st.st_size + sc_page_size - 1) / sc_page_size;
+        chunk_count = page_count / _fmincore_chunk_size
+        if chunk_count < 1:
+            chunk_count = 1
+        result = bytearray(page_count)
+        for ci in xrange(chunk_count):
+            pos = _fmincore_chunk_size * ci;
+            msize = min(_fmincore_chunk_size, st.st_size - pos)
+            try:
+                m = mmap.mmap(fd, msize, mmap.MAP_PRIVATE, 0, 0, pos)
+            except mmap.error as ex:
+                if ex.errno == errno.EINVAL or ex.errno == errno.ENODEV:
+                    # Perhaps the file was a pipe, i.e. "... | bup split ..."
+                    return None
+                raise ex
+            try:
+                _mincore(m, msize, 0, result, ci * pages_per_chunk)
+            except OSError as ex:
+                if ex.errno == errno.ENOSYS:
+                    return None
+                raise
+        return result
+
+
+def parse_timestamp(epoch_str):
+    """Return the number of nanoseconds since the epoch that are described
+by epoch_str (100ms, 100ns, ...); when epoch_str cannot be parsed,
+throw a ValueError that may contain additional information."""
+    ns_per = {'s' :  1000000000,
+              'ms' : 1000000,
+              'us' : 1000,
+              'ns' : 1}
+    match = re.match(r'^((?:[-+]?[0-9]+)?)(s|ms|us|ns)$', epoch_str)
+    if not match:
+        if re.match(r'^([-+]?[0-9]+)$', epoch_str):
+            raise ValueError('must include units, i.e. 100ns, 100ms, ...')
+        raise ValueError()
+    (n, units) = match.group(1, 2)
+    if not n:
+        n = 1
+    n = int(n)
+    return n * ns_per[units]
+
+
 def parse_num(s):
     """Parse data size information into a float number.
 
@@ -527,11 +915,18 @@ def add_error(e):
     log('%-70s\n' % e)
 
 
-istty = os.isatty(2) or atoi(os.environ.get('BUP_FORCE_TTY'))
-def progress(s):
-    """Calls log(s) if stderr is a TTY.  Does nothing otherwise."""
-    if istty:
-        log(s)
+def clear_errors():
+    global saved_errors
+    saved_errors = []
+
+
+def die_if_errors(msg=None, status=1):
+    global saved_errors
+    if saved_errors:
+        if not msg:
+            msg = 'warning: %d errors encountered\n' % len(saved_errors)
+        log(msg)
+        sys.exit(status)
 
 
 def handle_ctrl_c():
@@ -543,7 +938,7 @@ def handle_ctrl_c():
     oldhook = sys.excepthook
     def newhook(exctype, value, traceback):
         if exctype == KeyboardInterrupt:
-            log('Interrupted.\n')
+            log('\nInterrupted.\n')
         else:
             return oldhook(exctype, value, traceback)
     sys.excepthook = newhook
@@ -559,14 +954,14 @@ def columnate(l, prefix):
         return ""
     l = l[:]
     clen = max(len(s) for s in l)
-    ncols = (tty_width() - len(prefix)) / (clen + 2)
+    ncols = (tty_width() - len(prefix)) // (clen + 2)
     if ncols <= 1:
         ncols = 1
         clen = 0
     cols = []
     while len(l) % ncols:
         l.append('')
-    rows = len(l)/ncols
+    rows = len(l) // ncols
     for s in range(0, len(l), rows):
         cols.append(l[s:s+rows])
     out = ''
@@ -579,95 +974,239 @@ def parse_date_or_fatal(str, fatal):
     """Parses the given date or calls Option.fatal().
     For now we expect a string that contains a float."""
     try:
-        date = atof(str)
-    except ValueError, e:
+        date = float(str)
+    except ValueError as e:
         raise fatal('invalid date format (should be a float): %r' % e)
     else:
         return date
 
 
-def strip_path(prefix, path):
-    """Strips a given prefix from a path.
-
-    First both paths are normalized.
-
-    Raises an Exception if no prefix is given.
-    """
-    if prefix == None:
-        raise Exception('no path given')
-
-    normalized_prefix = os.path.realpath(prefix)
-    debug2("normalized_prefix: %s\n" % normalized_prefix)
-    normalized_path = os.path.realpath(path)
-    debug2("normalized_path: %s\n" % normalized_path)
-    if normalized_path.startswith(normalized_prefix):
-        return normalized_path[len(normalized_prefix):]
-    else:
-        return path
-
-
-def strip_base_path(path, base_paths):
-    """Strips the base path from a given path.
-
-
-    Determines the base path for the given string and then strips it
-    using strip_path().
-    Iterates over all base_paths from long to short, to prevent that
-    a too short base_path is removed.
-    """
-    normalized_path = os.path.realpath(path)
-    sorted_base_paths = sorted(base_paths, key=len, reverse=True)
-    for bp in sorted_base_paths:
-        if normalized_path.startswith(os.path.realpath(bp)):
-            return strip_path(bp, normalized_path)
-    return path
-
+def parse_excludes(options, fatal):
+    """Traverse the options and extract all excludes, or call Option.fatal()."""
+    excluded_paths = []
 
-def graft_path(graft_points, path):
-    normalized_path = os.path.realpath(path)
+    for flag in options:
+        (option, parameter) = flag
+        if option == '--exclude':
+            excluded_paths.append(resolve_parent(parameter))
+        elif option == '--exclude-from':
+            try:
+                f = open(resolve_parent(parameter))
+            except IOError as e:
+                raise fatal("couldn't read %s" % parameter)
+            for exclude_path in f.readlines():
+                # FIXME: perhaps this should be rstrip('\n')
+                exclude_path = resolve_parent(exclude_path.strip())
+                if exclude_path:
+                    excluded_paths.append(exclude_path)
+    return sorted(frozenset(excluded_paths))
+
+
+def parse_rx_excludes(options, fatal):
+    """Traverse the options and extract all rx excludes, or call
+    Option.fatal()."""
+    excluded_patterns = []
+
+    for flag in options:
+        (option, parameter) = flag
+        if option == '--exclude-rx':
+            try:
+                excluded_patterns.append(re.compile(parameter))
+            except re.error as ex:
+                fatal('invalid --exclude-rx pattern (%s): %s' % (parameter, ex))
+        elif option == '--exclude-rx-from':
+            try:
+                f = open(resolve_parent(parameter))
+            except IOError as e:
+                raise fatal("couldn't read %s" % parameter)
+            for pattern in f.readlines():
+                spattern = pattern.rstrip('\n')
+                if not spattern:
+                    continue
+                try:
+                    excluded_patterns.append(re.compile(spattern))
+                except re.error as ex:
+                    fatal('invalid --exclude-rx pattern (%s): %s' % (spattern, ex))
+    return excluded_patterns
+
+
+def should_rx_exclude_path(path, exclude_rxs):
+    """Return True if path matches a regular expression in exclude_rxs."""
+    for rx in exclude_rxs:
+        if rx.search(path):
+            debug1('Skipping %r: excluded by rx pattern %r.\n'
+                   % (path, rx.pattern))
+            return True
+    return False
+
+
+# FIXME: Carefully consider the use of functions (os.path.*, etc.)
+# that resolve against the current filesystem in the strip/graft
+# functions for example, but elsewhere as well.  I suspect bup's not
+# always being careful about that.  For some cases, the contents of
+# the current filesystem should be irrelevant, and consulting it might
+# produce the wrong result, perhaps via unintended symlink resolution,
+# for example.
+
+def path_components(path):
+    """Break path into a list of pairs of the form (name,
+    full_path_to_name).  Path must start with '/'.
+    Example:
+      '/home/foo' -> [('', '/'), ('home', '/home'), ('foo', '/home/foo')]"""
+    if not path.startswith('/'):
+        raise Exception('path must start with "/": %s' % path)
+    # Since we assume path startswith('/'), we can skip the first element.
+    result = [('', '/')]
+    norm_path = os.path.abspath(path)
+    if norm_path == '/':
+        return result
+    full_path = ''
+    for p in norm_path.split('/')[1:]:
+        full_path += '/' + p
+        result.append((p, full_path))
+    return result
+
+
+def stripped_path_components(path, strip_prefixes):
+    """Strip any prefix in strip_prefixes from path and return a list
+    of path components where each component is (name,
+    none_or_full_fs_path_to_name).  Assume path startswith('/').
+    See thelpers.py for examples."""
+    normalized_path = os.path.abspath(path)
+    sorted_strip_prefixes = sorted(strip_prefixes, key=len, reverse=True)
+    for bp in sorted_strip_prefixes:
+        normalized_bp = os.path.abspath(bp)
+        if normalized_bp == '/':
+            continue
+        if normalized_path.startswith(normalized_bp):
+            prefix = normalized_path[:len(normalized_bp)]
+            result = []
+            for p in normalized_path[len(normalized_bp):].split('/'):
+                if p: # not root
+                    prefix += '/'
+                prefix += p
+                result.append((p, prefix))
+            return result
+    # Nothing to strip.
+    return path_components(path)
+
+
+def grafted_path_components(graft_points, path):
+    # Create a result that consists of some number of faked graft
+    # directories before the graft point, followed by all of the real
+    # directories from path that are after the graft point.  Arrange
+    # for the directory at the graft point in the result to correspond
+    # to the "orig" directory in --graft orig=new.  See t/thelpers.py
+    # for some examples.
+
+    # Note that given --graft orig=new, orig and new have *nothing* to
+    # do with each other, even if some of their component names
+    # match. i.e. --graft /foo/bar/baz=/foo/bar/bax is semantically
+    # equivalent to --graft /foo/bar/baz=/x/y/z, or even
+    # /foo/bar/baz=/x.
+
+    # FIXME: This can't be the best solution...
+    clean_path = os.path.abspath(path)
     for graft_point in graft_points:
         old_prefix, new_prefix = graft_point
-        if normalized_path.startswith(old_prefix):
-            return re.sub(r'^' + old_prefix, new_prefix, normalized_path)
-    return normalized_path
-
-
-# hashlib is only available in python 2.5 or higher, but the 'sha' module
-# produces a DeprecationWarning in python 2.6 or higher.  We want to support
-# python 2.4 and above without any stupid warnings, so let's try using hashlib
-# first, and downgrade if it fails.
-try:
-    import hashlib
-except ImportError:
-    import sha
-    Sha1 = sha.sha
+        # Expand prefixes iff not absolute paths.
+        old_prefix = os.path.normpath(old_prefix)
+        new_prefix = os.path.normpath(new_prefix)
+        if clean_path.startswith(old_prefix):
+            escaped_prefix = re.escape(old_prefix)
+            grafted_path = re.sub(r'^' + escaped_prefix, new_prefix, clean_path)
+            # Handle /foo=/ (at least) -- which produces //whatever.
+            grafted_path = '/' + grafted_path.lstrip('/')
+            clean_path_components = path_components(clean_path)
+            # Count the components that were stripped.
+            strip_count = 0 if old_prefix == '/' else old_prefix.count('/')
+            new_prefix_parts = new_prefix.split('/')
+            result_prefix = grafted_path.split('/')[:new_prefix.count('/')]
+            result = [(p, None) for p in result_prefix] \
+                + clean_path_components[strip_count:]
+            # Now set the graft point name to match the end of new_prefix.
+            graft_point = len(result_prefix)
+            result[graft_point] = \
+                (new_prefix_parts[-1], clean_path_components[strip_count][1])
+            if new_prefix == '/': # --graft ...=/ is a special case.
+                return result[1:]
+            return result
+    return path_components(clean_path)
+
+
+Sha1 = hashlib.sha1
+
+
+_localtime = getattr(_helpers, 'localtime', None)
+
+if _localtime:
+    bup_time = namedtuple('bup_time', ['tm_year', 'tm_mon', 'tm_mday',
+                                       'tm_hour', 'tm_min', 'tm_sec',
+                                       'tm_wday', 'tm_yday',
+                                       'tm_isdst', 'tm_gmtoff', 'tm_zone'])
+
+# Define a localtime() that returns bup_time when possible.  Note:
+# this means that any helpers.localtime() results may need to be
+# passed through to_py_time() before being passed to python's time
+# module, which doesn't appear willing to ignore the extra items.
+if _localtime:
+    def localtime(time):
+        return bup_time(*_helpers.localtime(time))
+    def utc_offset_str(t):
+        """Return the local offset from UTC as "+hhmm" or "-hhmm" for time t.
+        If the current UTC offset does not represent an integer number
+        of minutes, the fractional component will be truncated."""
+        off = localtime(t).tm_gmtoff
+        # Note: // doesn't truncate like C for negative values, it rounds down.
+        offmin = abs(off) // 60
+        m = offmin % 60
+        h = (offmin - m) // 60
+        return "%+03d%02d" % (-h if off < 0 else h, m)
+    def to_py_time(x):
+        if isinstance(x, time.struct_time):
+            return x
+        return time.struct_time(x[:9])
 else:
-    Sha1 = hashlib.sha1
+    localtime = time.localtime
+    def utc_offset_str(t):
+        return time.strftime('%z', localtime(t))
+    def to_py_time(x):
+        return x
 
 
-def version_date():
-    """Format bup's version date string for output."""
-    return _version.DATE.split(' ')[0]
+_some_invalid_save_parts_rx = re.compile(r'[[ ~^:?*\\]|\.\.|//|@{')
 
-
-def version_commit():
-    """Get the commit hash of bup's current version."""
-    return _version.COMMIT
-
-
-def version_tag():
-    """Format bup's version tag (the official version number).
-
-    When generated from a commit other than one pointed to with a tag, the
-    returned string will be "unknown-" followed by the first seven positions of
-    the commit hash.
-    """
-    names = _version.NAMES.strip()
-    assert(names[0] == '(')
-    assert(names[-1] == ')')
-    names = names[1:-1]
-    l = [n.strip() for n in names.split(',')]
-    for n in l:
-        if n.startswith('tag: bup-'):
-            return n[9:]
-    return 'unknown-%s' % _version.COMMIT[:7]
+def valid_save_name(name):
+    # Enforce a superset of the restrictions in git-check-ref-format(1)
+    if name == '@' \
+       or name.startswith('/') or name.endswith('/') \
+       or name.endswith('.'):
+        return False
+    if _some_invalid_save_parts_rx.search(name):
+        return False
+    for c in name:
+        if ord(c) < 0x20 or ord(c) == 0x7f:
+            return False
+    for part in name.split('/'):
+        if part.startswith('.') or part.endswith('.lock'):
+            return False
+    return True
+
+
+_period_rx = re.compile(r'^([0-9]+)(s|min|h|d|w|m|y)$')
+
+def period_as_secs(s):
+    if s == 'forever':
+        return float('inf')
+    match = _period_rx.match(s)
+    if not match:
+        return None
+    mag = int(match.group(1))
+    scale = match.group(2)
+    return mag * {'s': 1,
+                  'min': 60,
+                  'h': 60 * 60,
+                  'd': 60 * 60 * 24,
+                  'w': 60 * 60 * 24 * 7,
+                  'm': 60 * 60 * 24 * 31,
+                  'y': 60 * 60 * 24 * 366}[scale]