util: drop the 'unpacker' helper

It is not helping anything anymore.

mercurial/obsolete.py

@@ -1,1252 +1,1252 b''
 # obsolete.py - obsolete markers handling
 #
 # Copyright 2012 Pierre-Yves David <[e-mail]>
 #                Logilab SA        <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 """Obsolete marker handling
 
 An obsolete marker maps an old changeset to a list of new
 changesets. If the list of new changesets is empty, the old changeset
 is said to be "killed". Otherwise, the old changeset is being
 "replaced" by the new changesets.
 
 Obsolete markers can be used to record and distribute changeset graph
 transformations performed by history rewrite operations, and help
 building new tools to reconcile conflicting rewrite actions. To
 facilitate conflict resolution, markers include various annotations
 besides old and news changeset identifiers, such as creation date or
 author name.
 
 The old obsoleted changeset is called a "precursor" and possible
 replacements are called "successors". Markers that used changeset X as
 a precursor are called "successor markers of X" because they hold
 information about the successors of X. Markers that use changeset Y as
 a successors are call "precursor markers of Y" because they hold
 information about the precursors of Y.
 
 Examples:
 
 - When changeset A is replaced by changeset A', one marker is stored:
 
     (A, (A',))
 
 - When changesets A and B are folded into a new changeset C, two markers are
   stored:
 
     (A, (C,)) and (B, (C,))
 
 - When changeset A is simply "pruned" from the graph, a marker is created:
 
     (A, ())
 
 - When changeset A is split into B and C, a single marker are used:
 
     (A, (C, C))
 
   We use a single marker to distinguish the "split" case from the "divergence"
   case. If two independent operations rewrite the same changeset A in to A' and
   A'', we have an error case: divergent rewriting. We can detect it because
   two markers will be created independently:
 
   (A, (B,)) and (A, (C,))
 
 Format
 ------
 
 Markers are stored in an append-only file stored in
 '.hg/store/obsstore'.
 
 The file starts with a version header:
 
 - 1 unsigned byte: version number, starting at zero.
 
 The header is followed by the markers. Marker format depend of the version. See
 comment associated with each format for details.
 
 """
 import struct
 import util, base85, node, parsers
 import phases
 from i18n import _
 
 _pack = struct.pack
 _unpack = struct.unpack
 _calcsize = struct.calcsize
 propertycache = util.propertycache
 
 # the obsolete feature is not mature enough to be enabled by default.
 # you have to rely on third party extension extension to enable this.
 _enabled = False
 
 # Options for obsolescence
 createmarkersopt = 'createmarkers'
 allowunstableopt = 'allowunstable'
 exchangeopt = 'exchange'
 
 ### obsolescence marker flag
 
 ## bumpedfix flag
 #
 # When a changeset A' succeed to a changeset A which became public, we call A'
 # "bumped" because it's a successors of a public changesets
 #
 # o    A' (bumped)
 # |`:
 # | o  A
 # |/
 # o    Z
 #
 # The way to solve this situation is to create a new changeset Ad as children
 # of A. This changeset have the same content than A'. So the diff from A to A'
 # is the same than the diff from A to Ad. Ad is marked as a successors of A'
 #
 # o   Ad
 # |`:
 # | x A'
 # |'|
 # o | A
 # |/
 # o Z
 #
 # But by transitivity Ad is also a successors of A. To avoid having Ad marked
 # as bumped too, we add the `bumpedfix` flag to the marker. <A', (Ad,)>.
 # This flag mean that the successors express the changes between the public and
 # bumped version and fix the situation, breaking the transitivity of
 # "bumped" here.
 bumpedfix = 1
 usingsha256 = 2
 
 ## Parsing and writing of version "0"
 #
 # The header is followed by the markers. Each marker is made of:
 #
 # - 1 uint8 : number of new changesets "N", can be zero.
 #
 # - 1 uint32: metadata size "M" in bytes.
 #
 # - 1 byte: a bit field. It is reserved for flags used in common
 #   obsolete marker operations, to avoid repeated decoding of metadata
 #   entries.
 #
 # - 20 bytes: obsoleted changeset identifier.
 #
 # - N*20 bytes: new changesets identifiers.
 #
 # - M bytes: metadata as a sequence of nul-terminated strings. Each
 #   string contains a key and a value, separated by a colon ':', without
 #   additional encoding. Keys cannot contain '\0' or ':' and values
 #   cannot contain '\0'.
 _fm0version = 0
 _fm0fixed   = '>BIB20s'
 _fm0node = '20s'
 _fm0fsize = _calcsize(_fm0fixed)
 _fm0fnodesize = _calcsize(_fm0node)
 
 def _fm0readmarkers(data, off):
     # Loop on markers
     l = len(data)
     while off + _fm0fsize <= l:
         # read fixed part
         cur = data[off:off + _fm0fsize]
         off += _fm0fsize
         numsuc, mdsize, flags, pre = _unpack(_fm0fixed, cur)
         # read replacement
         sucs = ()
         if numsuc:
             s = (_fm0fnodesize * numsuc)
             cur = data[off:off + s]
             sucs = _unpack(_fm0node * numsuc, cur)
             off += s
         # read metadata
         # (metadata will be decoded on demand)
         metadata = data[off:off + mdsize]
         if len(metadata) != mdsize:
             raise util.Abort(_('parsing obsolete marker: metadata is too '
                                'short, %d bytes expected, got %d')
                              % (mdsize, len(metadata)))
         off += mdsize
         metadata = _fm0decodemeta(metadata)
         try:
             when, offset = metadata.pop('date', '0 0').split(' ')
             date = float(when), int(offset)
         except ValueError:
             date = (0., 0)
         parents = None
         if 'p2' in metadata:
             parents = (metadata.pop('p1', None), metadata.pop('p2', None))
         elif 'p1' in metadata:
             parents = (metadata.pop('p1', None),)
         elif 'p0' in metadata:
             parents = ()
         if parents is not None:
             try:
                 parents = tuple(node.bin(p) for p in parents)
                 # if parent content is not a nodeid, drop the data
                 for p in parents:
                     if len(p) != 20:
                         parents = None
                         break
             except TypeError:
                 # if content cannot be translated to nodeid drop the data.
                 parents = None
 
         metadata = tuple(sorted(metadata.iteritems()))
 
         yield (pre, sucs, flags, metadata, date, parents)
 
 def _fm0encodeonemarker(marker):
     pre, sucs, flags, metadata, date, parents = marker
     if flags & usingsha256:
         raise util.Abort(_('cannot handle sha256 with old obsstore format'))
     metadata = dict(metadata)
     time, tz = date
     metadata['date'] = '%r %i' % (time, tz)
     if parents is not None:
         if not parents:
             # mark that we explicitly recorded no parents
             metadata['p0'] = ''
         for i, p in enumerate(parents):
             metadata['p%i' % (i + 1)] = node.hex(p)
     metadata = _fm0encodemeta(metadata)
     numsuc = len(sucs)
     format = _fm0fixed + (_fm0node * numsuc)
     data = [numsuc, len(metadata), flags, pre]
     data.extend(sucs)
     return _pack(format, *data) + metadata
 
 def _fm0encodemeta(meta):
     """Return encoded metadata string to string mapping.
 
     Assume no ':' in key and no '\0' in both key and value."""
     for key, value in meta.iteritems():
         if ':' in key or '\0' in key:
             raise ValueError("':' and '\0' are forbidden in metadata key'")
         if '\0' in value:
             raise ValueError("':' is forbidden in metadata value'")
     return '\0'.join(['%s:%s' % (k, meta[k]) for k in sorted(meta)])
 
 def _fm0decodemeta(data):
     """Return string to string dictionary from encoded version."""
     d = {}
     for l in data.split('\0'):
         if l:
             key, value = l.split(':')
             d[key] = value
     return d
 
 ## Parsing and writing of version "1"
 #
 # The header is followed by the markers. Each marker is made of:
 #
 # - uint32: total size of the marker (including this field)
 #
 # - float64: date in seconds since epoch
 #
 # - int16: timezone offset in minutes
 #
 # - uint16: a bit field. It is reserved for flags used in common
 #   obsolete marker operations, to avoid repeated decoding of metadata
 #   entries.
 #
 # - uint8: number of successors "N", can be zero.
 #
 # - uint8: number of parents "P", can be zero.
 #
 #     0: parents data stored but no parent,
 #     1: one parent stored,
 #     2: two parents stored,
 #     3: no parent data stored
 #
 # - uint8: number of metadata entries M
 #
 # - 20 or 32 bytes: precursor changeset identifier.
 #
 # - N*(20 or 32) bytes: successors changesets identifiers.
 #
 # - P*(20 or 32) bytes: parents of the precursors changesets.
 #
 # - M*(uint8, uint8): size of all metadata entries (key and value)
 #
 # - remaining bytes: the metadata, each (key, value) pair after the other.
 _fm1version = 1
 _fm1fixed = '>IdhHBBB20s'
 _fm1nodesha1 = '20s'
 _fm1nodesha256 = '32s'
 _fm1nodesha1size = _calcsize(_fm1nodesha1)
 _fm1nodesha256size = _calcsize(_fm1nodesha256)
 _fm1fsize = _calcsize(_fm1fixed)
 _fm1parentnone = 3
 _fm1parentshift = 14
 _fm1parentmask = (_fm1parentnone << _fm1parentshift)
 _fm1metapair = 'BB'
 _fm1metapairsize = _calcsize('BB')
 
 def _fm1purereadmarkers(data, off):
     # make some global constants local for performance
     noneflag = _fm1parentnone
     sha2flag = usingsha256
     sha1size = _fm1nodesha1size
     sha2size = _fm1nodesha256size
     sha1fmt = _fm1nodesha1
     sha2fmt = _fm1nodesha256
     metasize = _fm1metapairsize
     metafmt = _fm1metapair
     fsize = _fm1fsize
     unpack = _unpack
 
     # Loop on markers
     stop = len(data) - _fm1fsize
-    ufixed = util.unpacker(_fm1fixed)
+    ufixed = struct.Struct(_fm1fixed).unpack
 
     while off <= stop:
         # read fixed part
         o1 = off + fsize
         t, secs, tz, flags, numsuc, numpar, nummeta, prec = ufixed(data[off:o1])
 
         if flags & sha2flag:
             # FIXME: prec was read as a SHA1, needs to be amended
 
             # read 0 or more successors
             if numsuc == 1:
                 o2 = o1 + sha2size
                 sucs = (data[o1:o2],)
             else:
                 o2 = o1 + sha2size * numsuc
                 sucs = unpack(sha2fmt * numsuc, data[o1:o2])
 
             # read parents
             if numpar == noneflag:
                 o3 = o2
                 parents = None
             elif numpar == 1:
                 o3 = o2 + sha2size
                 parents = (data[o2:o3],)
             else:
                 o3 = o2 + sha2size * numpar
                 parents = unpack(sha2fmt * numpar, data[o2:o3])
         else:
             # read 0 or more successors
             if numsuc == 1:
                 o2 = o1 + sha1size
                 sucs = (data[o1:o2],)
             else:
                 o2 = o1 + sha1size * numsuc
                 sucs = unpack(sha1fmt * numsuc, data[o1:o2])
 
             # read parents
             if numpar == noneflag:
                 o3 = o2
                 parents = None
             elif numpar == 1:
                 o3 = o2 + sha1size
                 parents = (data[o2:o3],)
             else:
                 o3 = o2 + sha1size * numpar
                 parents = unpack(sha1fmt * numpar, data[o2:o3])
 
         # read metadata
         off = o3 + metasize * nummeta
         metapairsize = unpack('>' + (metafmt * nummeta), data[o3:off])
         metadata = []
         for idx in xrange(0, len(metapairsize), 2):
             o1 = off + metapairsize[idx]
             o2 = o1 + metapairsize[idx + 1]
             metadata.append((data[off:o1], data[o1:o2]))
             off = o2
 
         yield (prec, sucs, flags, tuple(metadata), (secs, tz * 60), parents)
 
 def _fm1encodeonemarker(marker):
     pre, sucs, flags, metadata, date, parents = marker
     # determine node size
     _fm1node = _fm1nodesha1
     if flags & usingsha256:
         _fm1node = _fm1nodesha256
     numsuc = len(sucs)
     numextranodes = numsuc
     if parents is None:
         numpar = _fm1parentnone
     else:
         numpar = len(parents)
         numextranodes += numpar
     formatnodes = _fm1node * numextranodes
     formatmeta = _fm1metapair * len(metadata)
     format = _fm1fixed + formatnodes + formatmeta
     # tz is stored in minutes so we divide by 60
     tz = date[1]//60
     data = [None, date[0], tz, flags, numsuc, numpar, len(metadata), pre]
     data.extend(sucs)
     if parents is not None:
         data.extend(parents)
     totalsize = _calcsize(format)
     for key, value in metadata:
         lk = len(key)
         lv = len(value)
         data.append(lk)
         data.append(lv)
         totalsize += lk + lv
     data[0] = totalsize
     data = [_pack(format, *data)]
     for key, value in metadata:
         data.append(key)
         data.append(value)
     return ''.join(data)
 
 def _fm1readmarkers(data, off):
     native = getattr(parsers, 'fm1readmarkers', None)
     if not native:
         return _fm1purereadmarkers(data, off)
     stop = len(data) - _fm1fsize
     return native(data, off, stop)
 
 # mapping to read/write various marker formats
 # <version> -> (decoder, encoder)
 formats = {_fm0version: (_fm0readmarkers, _fm0encodeonemarker),
            _fm1version: (_fm1readmarkers, _fm1encodeonemarker)}
 
 @util.nogc
 def _readmarkers(data):
     """Read and enumerate markers from raw data"""
     off = 0
     diskversion = _unpack('>B', data[off:off + 1])[0]
     off += 1
     if diskversion not in formats:
         raise util.Abort(_('parsing obsolete marker: unknown version %r')
                          % diskversion)
     return diskversion, formats[diskversion][0](data, off)
 
 def encodemarkers(markers, addheader=False, version=_fm0version):
     # Kept separate from flushmarkers(), it will be reused for
     # markers exchange.
     encodeone = formats[version][1]
     if addheader:
         yield _pack('>B', version)
     for marker in markers:
         yield encodeone(marker)
 
 
 class marker(object):
     """Wrap obsolete marker raw data"""
 
     def __init__(self, repo, data):
         # the repo argument will be used to create changectx in later version
         self._repo = repo
         self._data = data
         self._decodedmeta = None
 
     def __hash__(self):
         return hash(self._data)
 
     def __eq__(self, other):
         if type(other) != type(self):
             return False
         return self._data == other._data
 
     def precnode(self):
         """Precursor changeset node identifier"""
         return self._data[0]
 
     def succnodes(self):
         """List of successor changesets node identifiers"""
         return self._data[1]
 
     def parentnodes(self):
         """Parents of the precursors (None if not recorded)"""
         return self._data[5]
 
     def metadata(self):
         """Decoded metadata dictionary"""
         return dict(self._data[3])
 
     def date(self):
         """Creation date as (unixtime, offset)"""
         return self._data[4]
 
     def flags(self):
         """The flags field of the marker"""
         return self._data[2]
 
 @util.nogc
 def _addsuccessors(successors, markers):
     for mark in markers:
         successors.setdefault(mark[0], set()).add(mark)
 
 @util.nogc
 def _addprecursors(precursors, markers):
     for mark in markers:
         for suc in mark[1]:
             precursors.setdefault(suc, set()).add(mark)
 
 @util.nogc
 def _addchildren(children, markers):
     for mark in markers:
         parents = mark[5]
         if parents is not None:
             for p in parents:
                 children.setdefault(p, set()).add(mark)
 
 def _checkinvalidmarkers(markers):
     """search for marker with invalid data and raise error if needed
 
     Exist as a separated function to allow the evolve extension for a more
     subtle handling.
     """
     for mark in markers:
         if node.nullid in mark[1]:
             raise util.Abort(_('bad obsolescence marker detected: '
                                'invalid successors nullid'))
 
 class obsstore(object):
     """Store obsolete markers
 
     Markers can be accessed with two mappings:
     - precursors[x] -> set(markers on precursors edges of x)
     - successors[x] -> set(markers on successors edges of x)
     - children[x]   -> set(markers on precursors edges of children(x)
     """
 
     fields = ('prec', 'succs', 'flag', 'meta', 'date', 'parents')
     # prec:    nodeid, precursor changesets
     # succs:   tuple of nodeid, successor changesets (0-N length)
     # flag:    integer, flag field carrying modifier for the markers (see doc)
     # meta:    binary blob, encoded metadata dictionary
     # date:    (float, int) tuple, date of marker creation
     # parents: (tuple of nodeid) or None, parents of precursors
     #          None is used when no data has been recorded
 
     def __init__(self, sopener, defaultformat=_fm1version, readonly=False):
         # caches for various obsolescence related cache
         self.caches = {}
         self._all = []
         self.sopener = sopener
         data = sopener.tryread('obsstore')
         self._version = defaultformat
         self._readonly = readonly
         if data:
             self._version, markers = _readmarkers(data)
             self._addmarkers(markers)
 
     def __iter__(self):
         return iter(self._all)
 
     def __len__(self):
         return len(self._all)
 
     def __nonzero__(self):
         return bool(self._all)
 
     def create(self, transaction, prec, succs=(), flag=0, parents=None,
                date=None, metadata=None):
         """obsolete: add a new obsolete marker
 
         * ensuring it is hashable
         * check mandatory metadata
         * encode metadata
 
         If you are a human writing code creating marker you want to use the
         `createmarkers` function in this module instead.
 
         return True if a new marker have been added, False if the markers
         already existed (no op).
         """
         if metadata is None:
             metadata = {}
         if date is None:
             if 'date' in metadata:
                 # as a courtesy for out-of-tree extensions
                 date = util.parsedate(metadata.pop('date'))
             else:
                 date = util.makedate()
         if len(prec) != 20:
             raise ValueError(prec)
         for succ in succs:
             if len(succ) != 20:
                 raise ValueError(succ)
         if prec in succs:
             raise ValueError(_('in-marker cycle with %s') % node.hex(prec))
 
         metadata = tuple(sorted(metadata.iteritems()))
 
         marker = (str(prec), tuple(succs), int(flag), metadata, date, parents)
         return bool(self.add(transaction, [marker]))
 
     def add(self, transaction, markers):
         """Add new markers to the store
 
         Take care of filtering duplicate.
         Return the number of new marker."""
         if self._readonly:
             raise util.Abort('creating obsolete markers is not enabled on this '
                              'repo')
         known = set(self._all)
         new = []
         for m in markers:
             if m not in known:
                 known.add(m)
                 new.append(m)
         if new:
             f = self.sopener('obsstore', 'ab')
             try:
                 offset = f.tell()
                 transaction.add('obsstore', offset)
                 # offset == 0: new file - add the version header
                 for bytes in encodemarkers(new, offset == 0, self._version):
                     f.write(bytes)
             finally:
                 # XXX: f.close() == filecache invalidation == obsstore rebuilt.
                 # call 'filecacheentry.refresh()'  here
                 f.close()
             self._addmarkers(new)
             # new marker *may* have changed several set. invalidate the cache.
             self.caches.clear()
         # records the number of new markers for the transaction hooks
         previous = int(transaction.hookargs.get('new_obsmarkers', '0'))
         transaction.hookargs['new_obsmarkers'] = str(previous + len(new))
         return len(new)
 
     def mergemarkers(self, transaction, data):
         """merge a binary stream of markers inside the obsstore
 
         Returns the number of new markers added."""
         version, markers = _readmarkers(data)
         return self.add(transaction, markers)
 
     @propertycache
     def successors(self):
         successors = {}
         _addsuccessors(successors, self._all)
         return successors
 
     @propertycache
     def precursors(self):
         precursors = {}
         _addprecursors(precursors, self._all)
         return precursors
 
     @propertycache
     def children(self):
         children = {}
         _addchildren(children, self._all)
         return children
 
     def _cached(self, attr):
         return attr in self.__dict__
 
     def _addmarkers(self, markers):
         markers = list(markers) # to allow repeated iteration
         self._all.extend(markers)
         if self._cached('successors'):
             _addsuccessors(self.successors, markers)
         if self._cached('precursors'):
             _addprecursors(self.precursors, markers)
         if self._cached('children'):
             _addchildren(self.children, markers)
         _checkinvalidmarkers(markers)
 
     def relevantmarkers(self, nodes):
         """return a set of all obsolescence markers relevant to a set of nodes.
 
         "relevant" to a set of nodes mean:
 
         - marker that use this changeset as successor
         - prune marker of direct children on this changeset
         - recursive application of the two rules on precursors of these markers
 
         It is a set so you cannot rely on order."""
 
         pendingnodes = set(nodes)
         seenmarkers = set()
         seennodes = set(pendingnodes)
         precursorsmarkers = self.precursors
         children = self.children
         while pendingnodes:
             direct = set()
             for current in pendingnodes:
                 direct.update(precursorsmarkers.get(current, ()))
                 pruned = [m for m in children.get(current, ()) if not m[1]]
                 direct.update(pruned)
             direct -= seenmarkers
             pendingnodes = set([m[0] for m in direct])
             seenmarkers |= direct
             pendingnodes -= seennodes
             seennodes |= pendingnodes
         return seenmarkers
 
 def commonversion(versions):
     """Return the newest version listed in both versions and our local formats.
 
     Returns None if no common version exists.
     """
     versions.sort(reverse=True)
     # search for highest version known on both side
     for v in versions:
         if v in formats:
             return v
     return None
 
 # arbitrary picked to fit into 8K limit from HTTP server
 # you have to take in account:
 # - the version header
 # - the base85 encoding
 _maxpayload = 5300
 
 def _pushkeyescape(markers):
     """encode markers into a dict suitable for pushkey exchange
 
     - binary data is base85 encoded
     - split in chunks smaller than 5300 bytes"""
     keys = {}
     parts = []
     currentlen = _maxpayload * 2  # ensure we create a new part
     for marker in markers:
         nextdata = _fm0encodeonemarker(marker)
         if (len(nextdata) + currentlen > _maxpayload):
             currentpart = []
             currentlen = 0
             parts.append(currentpart)
         currentpart.append(nextdata)
         currentlen += len(nextdata)
     for idx, part in enumerate(reversed(parts)):
         data = ''.join([_pack('>B', _fm0version)] + part)
         keys['dump%i' % idx] = base85.b85encode(data)
     return keys
 
 def listmarkers(repo):
     """List markers over pushkey"""
     if not repo.obsstore:
         return {}
     return _pushkeyescape(sorted(repo.obsstore))
 
 def pushmarker(repo, key, old, new):
     """Push markers over pushkey"""
     if not key.startswith('dump'):
         repo.ui.warn(_('unknown key: %r') % key)
         return 0
     if old:
         repo.ui.warn(_('unexpected old value for %r') % key)
         return 0
     data = base85.b85decode(new)
     lock = repo.lock()
     try:
         tr = repo.transaction('pushkey: obsolete markers')
         try:
             repo.obsstore.mergemarkers(tr, data)
             tr.close()
             return 1
         finally:
             tr.release()
     finally:
         lock.release()
 
 def getmarkers(repo, nodes=None):
     """returns markers known in a repository
 
     If <nodes> is specified, only markers "relevant" to those nodes are are
     returned"""
     if nodes is None:
         rawmarkers = repo.obsstore
     else:
         rawmarkers = repo.obsstore.relevantmarkers(nodes)
 
     for markerdata in rawmarkers:
         yield marker(repo, markerdata)
 
 def relevantmarkers(repo, node):
     """all obsolete markers relevant to some revision"""
     for markerdata in repo.obsstore.relevantmarkers(node):
         yield marker(repo, markerdata)
 
 
 def precursormarkers(ctx):
     """obsolete marker marking this changeset as a successors"""
     for data in ctx.repo().obsstore.precursors.get(ctx.node(), ()):
         yield marker(ctx.repo(), data)
 
 def successormarkers(ctx):
     """obsolete marker making this changeset obsolete"""
     for data in ctx.repo().obsstore.successors.get(ctx.node(), ()):
         yield marker(ctx.repo(), data)
 
 def allsuccessors(obsstore, nodes, ignoreflags=0):
     """Yield node for every successor of <nodes>.
 
     Some successors may be unknown locally.
 
     This is a linear yield unsuited to detecting split changesets. It includes
     initial nodes too."""
     remaining = set(nodes)
     seen = set(remaining)
     while remaining:
         current = remaining.pop()
         yield current
         for mark in obsstore.successors.get(current, ()):
             # ignore marker flagged with specified flag
             if mark[2] & ignoreflags:
                 continue
             for suc in mark[1]:
                 if suc not in seen:
                     seen.add(suc)
                     remaining.add(suc)
 
 def allprecursors(obsstore, nodes, ignoreflags=0):
     """Yield node for every precursors of <nodes>.
 
     Some precursors may be unknown locally.
 
     This is a linear yield unsuited to detecting folded changesets. It includes
     initial nodes too."""
 
     remaining = set(nodes)
     seen = set(remaining)
     while remaining:
         current = remaining.pop()
         yield current
         for mark in obsstore.precursors.get(current, ()):
             # ignore marker flagged with specified flag
             if mark[2] & ignoreflags:
                 continue
             suc = mark[0]
             if suc not in seen:
                 seen.add(suc)
                 remaining.add(suc)
 
 def foreground(repo, nodes):
     """return all nodes in the "foreground" of other node
 
     The foreground of a revision is anything reachable using parent -> children
     or precursor -> successor relation. It is very similar to "descendant" but
     augmented with obsolescence information.
 
     Beware that possible obsolescence cycle may result if complex situation.
     """
     repo = repo.unfiltered()
     foreground = set(repo.set('%ln::', nodes))
     if repo.obsstore:
         # We only need this complicated logic if there is obsolescence
         # XXX will probably deserve an optimised revset.
         nm = repo.changelog.nodemap
         plen = -1
         # compute the whole set of successors or descendants
         while len(foreground) != plen:
             plen = len(foreground)
             succs = set(c.node() for c in foreground)
             mutable = [c.node() for c in foreground if c.mutable()]
             succs.update(allsuccessors(repo.obsstore, mutable))
             known = (n for n in succs if n in nm)
             foreground = set(repo.set('%ln::', known))
     return set(c.node() for c in foreground)
 
 
 def successorssets(repo, initialnode, cache=None):
     """Return all set of successors of initial nodes
 
     The successors set of a changeset A are a group of revisions that succeed
     A. It succeeds A as a consistent whole, each revision being only a partial
     replacement. The successors set contains non-obsolete changesets only.
 
     This function returns the full list of successor sets which is why it
     returns a list of tuples and not just a single tuple. Each tuple is a valid
     successors set. Not that (A,) may be a valid successors set for changeset A
     (see below).
 
     In most cases, a changeset A will have a single element (e.g. the changeset
     A is replaced by A') in its successors set. Though, it is also common for a
     changeset A to have no elements in its successor set (e.g. the changeset
     has been pruned). Therefore, the returned list of successors sets will be
     [(A',)] or [], respectively.
 
     When a changeset A is split into A' and B', however, it will result in a
     successors set containing more than a single element, i.e. [(A',B')].
     Divergent changesets will result in multiple successors sets, i.e. [(A',),
     (A'')].
 
     If a changeset A is not obsolete, then it will conceptually have no
     successors set. To distinguish this from a pruned changeset, the successor
     set will only contain itself, i.e. [(A,)].
 
     Finally, successors unknown locally are considered to be pruned (obsoleted
     without any successors).
 
     The optional `cache` parameter is a dictionary that may contain precomputed
     successors sets. It is meant to reuse the computation of a previous call to
     `successorssets` when multiple calls are made at the same time. The cache
     dictionary is updated in place. The caller is responsible for its live
     spawn. Code that makes multiple calls to `successorssets` *must* use this
     cache mechanism or suffer terrible performances.
 
     """
 
     succmarkers = repo.obsstore.successors
 
     # Stack of nodes we search successors sets for
     toproceed = [initialnode]
     # set version of above list for fast loop detection
     # element added to "toproceed" must be added here
     stackedset = set(toproceed)
     if cache is None:
         cache = {}
 
     # This while loop is the flattened version of a recursive search for
     # successors sets
     #
     # def successorssets(x):
     #    successors = directsuccessors(x)
     #    ss = [[]]
     #    for succ in directsuccessors(x):
     #        # product as in itertools cartesian product
     #        ss = product(ss, successorssets(succ))
     #    return ss
     #
     # But we can not use plain recursive calls here:
     # - that would blow the python call stack
     # - obsolescence markers may have cycles, we need to handle them.
     #
     # The `toproceed` list act as our call stack. Every node we search
     # successors set for are stacked there.
     #
     # The `stackedset` is set version of this stack used to check if a node is
     # already stacked. This check is used to detect cycles and prevent infinite
     # loop.
     #
     # successors set of all nodes are stored in the `cache` dictionary.
     #
     # After this while loop ends we use the cache to return the successors sets
     # for the node requested by the caller.
     while toproceed:
         # Every iteration tries to compute the successors sets of the topmost
         # node of the stack: CURRENT.
         #
         # There are four possible outcomes:
         #
         # 1) We already know the successors sets of CURRENT:
         #    -> mission accomplished, pop it from the stack.
         # 2) Node is not obsolete:
         #    -> the node is its own successors sets. Add it to the cache.
         # 3) We do not know successors set of direct successors of CURRENT:
         #    -> We add those successors to the stack.
         # 4) We know successors sets of all direct successors of CURRENT:
         #    -> We can compute CURRENT successors set and add it to the
         #       cache.
         #
         current = toproceed[-1]
         if current in cache:
             # case (1): We already know the successors sets
             stackedset.remove(toproceed.pop())
         elif current not in succmarkers:
             # case (2): The node is not obsolete.
             if current in repo:
                 # We have a valid last successors.
                 cache[current] = [(current,)]
             else:
                 # Final obsolete version is unknown locally.
                 # Do not count that as a valid successors
                 cache[current] = []
         else:
             # cases (3) and (4)
             #
             # We proceed in two phases. Phase 1 aims to distinguish case (3)
             # from case (4):
             #
             #     For each direct successors of CURRENT, we check whether its
             #     successors sets are known. If they are not, we stack the
             #     unknown node and proceed to the next iteration of the while
             #     loop. (case 3)
             #
             #     During this step, we may detect obsolescence cycles: a node
             #     with unknown successors sets but already in the call stack.
             #     In such a situation, we arbitrary set the successors sets of
             #     the node to nothing (node pruned) to break the cycle.
             #
             #     If no break was encountered we proceed to phase 2.
             #
             # Phase 2 computes successors sets of CURRENT (case 4); see details
             # in phase 2 itself.
             #
             # Note the two levels of iteration in each phase.
             # - The first one handles obsolescence markers using CURRENT as
             #   precursor (successors markers of CURRENT).
             #
             #   Having multiple entry here means divergence.
             #
             # - The second one handles successors defined in each marker.
             #
             #   Having none means pruned node, multiple successors means split,
             #   single successors are standard replacement.
             #
             for mark in sorted(succmarkers[current]):
                 for suc in mark[1]:
                     if suc not in cache:
                         if suc in stackedset:
                             # cycle breaking
                             cache[suc] = []
                         else:
                             # case (3) If we have not computed successors sets
                             # of one of those successors we add it to the
                             # `toproceed` stack and stop all work for this
                             # iteration.
                             toproceed.append(suc)
                             stackedset.add(suc)
                             break
                 else:
                     continue
                 break
             else:
                 # case (4): we know all successors sets of all direct
                 # successors
                 #
                 # Successors set contributed by each marker depends on the
                 # successors sets of all its "successors" node.
                 #
                 # Each different marker is a divergence in the obsolescence
                 # history. It contributes successors sets distinct from other
                 # markers.
                 #
                 # Within a marker, a successor may have divergent successors
                 # sets. In such a case, the marker will contribute multiple
                 # divergent successors sets. If multiple successors have
                 # divergent successors sets, a Cartesian product is used.
                 #
                 # At the end we post-process successors sets to remove
                 # duplicated entry and successors set that are strict subset of
                 # another one.
                 succssets = []
                 for mark in sorted(succmarkers[current]):
                     # successors sets contributed by this marker
                     markss = [[]]
                     for suc in mark[1]:
                         # cardinal product with previous successors
                         productresult = []
                         for prefix in markss:
                             for suffix in cache[suc]:
                                 newss = list(prefix)
                                 for part in suffix:
                                     # do not duplicated entry in successors set
                                     # first entry wins.
                                     if part not in newss:
                                         newss.append(part)
                                 productresult.append(newss)
                         markss = productresult
                     succssets.extend(markss)
                 # remove duplicated and subset
                 seen = []
                 final = []
                 candidate = sorted(((set(s), s) for s in succssets if s),
                                    key=lambda x: len(x[1]), reverse=True)
                 for setversion, listversion in candidate:
                     for seenset in seen:
                         if setversion.issubset(seenset):
                             break
                     else:
                         final.append(listversion)
                         seen.append(setversion)
                 final.reverse() # put small successors set first
                 cache[current] = final
     return cache[initialnode]
 
 def _knownrevs(repo, nodes):
     """yield revision numbers of known nodes passed in parameters
 
     Unknown revisions are silently ignored."""
     torev = repo.changelog.nodemap.get
     for n in nodes:
         rev = torev(n)
         if rev is not None:
             yield rev
 
 # mapping of 'set-name' -> <function to compute this set>
 cachefuncs = {}
 def cachefor(name):
     """Decorator to register a function as computing the cache for a set"""
     def decorator(func):
         assert name not in cachefuncs
         cachefuncs[name] = func
         return func
     return decorator
 
 def getrevs(repo, name):
     """Return the set of revision that belong to the <name> set
 
     Such access may compute the set and cache it for future use"""
     repo = repo.unfiltered()
     if not repo.obsstore:
         return frozenset()
     if name not in repo.obsstore.caches:
         repo.obsstore.caches[name] = cachefuncs[name](repo)
     return repo.obsstore.caches[name]
 
 # To be simple we need to invalidate obsolescence cache when:
 #
 # - new changeset is added:
 # - public phase is changed
 # - obsolescence marker are added
 # - strip is used a repo
 def clearobscaches(repo):
     """Remove all obsolescence related cache from a repo
 
     This remove all cache in obsstore is the obsstore already exist on the
     repo.
 
     (We could be smarter here given the exact event that trigger the cache
     clearing)"""
     # only clear cache is there is obsstore data in this repo
     if 'obsstore' in repo._filecache:
         repo.obsstore.caches.clear()
 
 @cachefor('obsolete')
 def _computeobsoleteset(repo):
     """the set of obsolete revisions"""
     obs = set()
     getrev = repo.changelog.nodemap.get
     getphase = repo._phasecache.phase
     for n in repo.obsstore.successors:
         rev = getrev(n)
         if rev is not None and getphase(repo, rev):
             obs.add(rev)
     return obs
 
 @cachefor('unstable')
 def _computeunstableset(repo):
     """the set of non obsolete revisions with obsolete parents"""
     revs = [(ctx.rev(), ctx) for ctx in
             repo.set('(not public()) and (not obsolete())')]
     revs.sort(key=lambda x:x[0])
     unstable = set()
     for rev, ctx in revs:
         # A rev is unstable if one of its parent is obsolete or unstable
         # this works since we traverse following growing rev order
         if any((x.obsolete() or (x.rev() in unstable))
                 for x in ctx.parents()):
             unstable.add(rev)
     return unstable
 
 @cachefor('suspended')
 def _computesuspendedset(repo):
     """the set of obsolete parents with non obsolete descendants"""
     suspended = repo.changelog.ancestors(getrevs(repo, 'unstable'))
     return set(r for r in getrevs(repo, 'obsolete') if r in suspended)
 
 @cachefor('extinct')
 def _computeextinctset(repo):
     """the set of obsolete parents without non obsolete descendants"""
     return getrevs(repo, 'obsolete') - getrevs(repo, 'suspended')
 
 
 @cachefor('bumped')
 def _computebumpedset(repo):
     """the set of revs trying to obsolete public revisions"""
     bumped = set()
     # util function (avoid attribute lookup in the loop)
     phase = repo._phasecache.phase # would be faster to grab the full list
     public = phases.public
     cl = repo.changelog
     torev = cl.nodemap.get
     for ctx in repo.set('(not public()) and (not obsolete())'):
         rev = ctx.rev()
         # We only evaluate mutable, non-obsolete revision
         node = ctx.node()
         # (future) A cache of precursors may worth if split is very common
         for pnode in allprecursors(repo.obsstore, [node],
                                    ignoreflags=bumpedfix):
             prev = torev(pnode) # unfiltered! but so is phasecache
             if (prev is not None) and (phase(repo, prev) <= public):
                 # we have a public precursors
                 bumped.add(rev)
                 break # Next draft!
     return bumped
 
 @cachefor('divergent')
 def _computedivergentset(repo):
     """the set of rev that compete to be the final successors of some revision.
     """
     divergent = set()
     obsstore = repo.obsstore
     newermap = {}
     for ctx in repo.set('(not public()) - obsolete()'):
         mark = obsstore.precursors.get(ctx.node(), ())
         toprocess = set(mark)
         seen = set()
         while toprocess:
             prec = toprocess.pop()[0]
             if prec in seen:
                 continue # emergency cycle hanging prevention
             seen.add(prec)
             if prec not in newermap:
                 successorssets(repo, prec, newermap)
             newer = [n for n in newermap[prec] if n]
             if len(newer) > 1:
                 divergent.add(ctx.rev())
                 break
             toprocess.update(obsstore.precursors.get(prec, ()))
     return divergent
 
 
 def createmarkers(repo, relations, flag=0, date=None, metadata=None):
     """Add obsolete markers between changesets in a repo
 
     <relations> must be an iterable of (<old>, (<new>, ...)[,{metadata}])
     tuple. `old` and `news` are changectx. metadata is an optional dictionary
     containing metadata for this marker only. It is merged with the global
     metadata specified through the `metadata` argument of this function,
 
     Trying to obsolete a public changeset will raise an exception.
 
     Current user and date are used except if specified otherwise in the
     metadata attribute.
 
     This function operates within a transaction of its own, but does
     not take any lock on the repo.
     """
     # prepare metadata
     if metadata is None:
         metadata = {}
     if 'user' not in metadata:
         metadata['user'] = repo.ui.username()
     tr = repo.transaction('add-obsolescence-marker')
     try:
         for rel in relations:
             prec = rel[0]
             sucs = rel[1]
             localmetadata = metadata.copy()
             if 2 < len(rel):
                 localmetadata.update(rel[2])
 
             if not prec.mutable():
                 raise util.Abort("cannot obsolete immutable changeset: %s"
                                  % prec)
             nprec = prec.node()
             nsucs = tuple(s.node() for s in sucs)
             npare = None
             if not nsucs:
                 npare = tuple(p.node() for p in prec.parents())
             if nprec in nsucs:
                 raise util.Abort("changeset %s cannot obsolete itself" % prec)
             repo.obsstore.create(tr, nprec, nsucs, flag, parents=npare,
                                  date=date, metadata=localmetadata)
             repo.filteredrevcache.clear()
         tr.close()
     finally:
         tr.release()
 
 def isenabled(repo, option):
     """Returns True if the given repository has the given obsolete option
     enabled.
     """
     result = set(repo.ui.configlist('experimental', 'evolution'))
     if 'all' in result:
         return True
 
     # For migration purposes, temporarily return true if the config hasn't been
     # set but _enabled is true.
     if len(result) == 0 and _enabled:
         return True
 
     # createmarkers must be enabled if other options are enabled
     if ((allowunstableopt in result or exchangeopt in result) and
         not createmarkersopt in result):
         raise util.Abort(_("'createmarkers' obsolete option must be enabled "
                            "if other obsolete options are enabled"))
 
     return option in result

mercurial/util.py

@@ -1,2249 +1,2245 b''
 # util.py - Mercurial utility functions and platform specific implementations
 #
 #  Copyright 2005 K. Thananchayan <[e-mail]>
 #  Copyright 2005-2007 Matt Mackall <[e-mail]>
 #  Copyright 2006 Vadim Gelfer <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 """Mercurial utility functions and platform specific implementations.
 
 This contains helper routines that are independent of the SCM core and
 hide platform-specific details from the core.
 """
 
 import i18n
 _ = i18n._
 import error, osutil, encoding, parsers
 import errno, shutil, sys, tempfile, traceback
 import re as remod
 import os, time, datetime, calendar, textwrap, signal, collections
-import imp, socket, urllib, struct
+import imp, socket, urllib
 import gc
 
 if os.name == 'nt':
     import windows as platform
 else:
     import posix as platform
 
 cachestat = platform.cachestat
 checkexec = platform.checkexec
 checklink = platform.checklink
 copymode = platform.copymode
 executablepath = platform.executablepath
 expandglobs = platform.expandglobs
 explainexit = platform.explainexit
 findexe = platform.findexe
 gethgcmd = platform.gethgcmd
 getuser = platform.getuser
 groupmembers = platform.groupmembers
 groupname = platform.groupname
 hidewindow = platform.hidewindow
 isexec = platform.isexec
 isowner = platform.isowner
 localpath = platform.localpath
 lookupreg = platform.lookupreg
 makedir = platform.makedir
 nlinks = platform.nlinks
 normpath = platform.normpath
 normcase = platform.normcase
 normcasespec = platform.normcasespec
 normcasefallback = platform.normcasefallback
 openhardlinks = platform.openhardlinks
 oslink = platform.oslink
 parsepatchoutput = platform.parsepatchoutput
 pconvert = platform.pconvert
 popen = platform.popen
 posixfile = platform.posixfile
 quotecommand = platform.quotecommand
 readpipe = platform.readpipe
 rename = platform.rename
 removedirs = platform.removedirs
 samedevice = platform.samedevice
 samefile = platform.samefile
 samestat = platform.samestat
 setbinary = platform.setbinary
 setflags = platform.setflags
 setsignalhandler = platform.setsignalhandler
 shellquote = platform.shellquote
 spawndetached = platform.spawndetached
 split = platform.split
 sshargs = platform.sshargs
 statfiles = getattr(osutil, 'statfiles', platform.statfiles)
 statisexec = platform.statisexec
 statislink = platform.statislink
 termwidth = platform.termwidth
 testpid = platform.testpid
 umask = platform.umask
 unlink = platform.unlink
 unlinkpath = platform.unlinkpath
 username = platform.username
 
 # Python compatibility
 
 _notset = object()
 
 def safehasattr(thing, attr):
     return getattr(thing, attr, _notset) is not _notset
 
 def sha1(s=''):
     '''
     Low-overhead wrapper around Python's SHA support
 
     >>> f = _fastsha1
     >>> a = sha1()
     >>> a = f()
     >>> a.hexdigest()
     'da39a3ee5e6b4b0d3255bfef95601890afd80709'
     '''
 
     return _fastsha1(s)
 
 def _fastsha1(s=''):
     # This function will import sha1 from hashlib or sha (whichever is
     # available) and overwrite itself with it on the first call.
     # Subsequent calls will go directly to the imported function.
     if sys.version_info >= (2, 5):
         from hashlib import sha1 as _sha1
     else:
         from sha import sha as _sha1
     global _fastsha1, sha1
     _fastsha1 = sha1 = _sha1
     return _sha1(s)
 
 def md5(s=''):
     try:
         from hashlib import md5 as _md5
     except ImportError:
         from md5 import md5 as _md5
     global md5
     md5 = _md5
     return _md5(s)
 
 DIGESTS = {
     'md5': md5,
     'sha1': sha1,
 }
 # List of digest types from strongest to weakest
 DIGESTS_BY_STRENGTH = ['sha1', 'md5']
 
 try:
     import hashlib
     DIGESTS.update({
         'sha512': hashlib.sha512,
     })
     DIGESTS_BY_STRENGTH.insert(0, 'sha512')
 except ImportError:
     pass
 
 for k in DIGESTS_BY_STRENGTH:
     assert k in DIGESTS
 
 class digester(object):
     """helper to compute digests.
 
     This helper can be used to compute one or more digests given their name.
 
     >>> d = digester(['md5', 'sha1'])
     >>> d.update('foo')
     >>> [k for k in sorted(d)]
     ['md5', 'sha1']
     >>> d['md5']
     'acbd18db4cc2f85cedef654fccc4a4d8'
     >>> d['sha1']
     '0beec7b5ea3f0fdbc95d0dd47f3c5bc275da8a33'
     >>> digester.preferred(['md5', 'sha1'])
     'sha1'
     """
 
     def __init__(self, digests, s=''):
         self._hashes = {}
         for k in digests:
             if k not in DIGESTS:
                 raise Abort(_('unknown digest type: %s') % k)
             self._hashes[k] = DIGESTS[k]()
         if s:
             self.update(s)
 
     def update(self, data):
         for h in self._hashes.values():
             h.update(data)
 
     def __getitem__(self, key):
         if key not in DIGESTS:
             raise Abort(_('unknown digest type: %s') % k)
         return self._hashes[key].hexdigest()
 
     def __iter__(self):
         return iter(self._hashes)
 
     @staticmethod
     def preferred(supported):
         """returns the strongest digest type in both supported and DIGESTS."""
 
         for k in DIGESTS_BY_STRENGTH:
             if k in supported:
                 return k
         return None
 
 class digestchecker(object):
     """file handle wrapper that additionally checks content against a given
     size and digests.
 
         d = digestchecker(fh, size, {'md5': '...'})
 
     When multiple digests are given, all of them are validated.
     """
 
     def __init__(self, fh, size, digests):
         self._fh = fh
         self._size = size
         self._got = 0
         self._digests = dict(digests)
         self._digester = digester(self._digests.keys())
 
     def read(self, length=-1):
         content = self._fh.read(length)
         self._digester.update(content)
         self._got += len(content)
         return content
 
     def validate(self):
         if self._size != self._got:
             raise Abort(_('size mismatch: expected %d, got %d') %
                 (self._size, self._got))
         for k, v in self._digests.items():
             if v != self._digester[k]:
                 # i18n: first parameter is a digest name
                 raise Abort(_('%s mismatch: expected %s, got %s') %
                     (k, v, self._digester[k]))
 
 try:
     buffer = buffer
 except NameError:
     if sys.version_info[0] < 3:
         def buffer(sliceable, offset=0):
             return sliceable[offset:]
     else:
         def buffer(sliceable, offset=0):
             return memoryview(sliceable)[offset:]
 
 import subprocess
 closefds = os.name == 'posix'
 
-def unpacker(fmt):
-    """create a struct unpacker for the specified format"""
-    return struct.Struct(fmt).unpack
-
 def popen2(cmd, env=None, newlines=False):
     # Setting bufsize to -1 lets the system decide the buffer size.
     # The default for bufsize is 0, meaning unbuffered. This leads to
     # poor performance on Mac OS X: http://bugs.python.org/issue4194
     p = subprocess.Popen(cmd, shell=True, bufsize=-1,
                          close_fds=closefds,
                          stdin=subprocess.PIPE, stdout=subprocess.PIPE,
                          universal_newlines=newlines,
                          env=env)
     return p.stdin, p.stdout
 
 def popen3(cmd, env=None, newlines=False):
     stdin, stdout, stderr, p = popen4(cmd, env, newlines)
     return stdin, stdout, stderr
 
 def popen4(cmd, env=None, newlines=False):
     p = subprocess.Popen(cmd, shell=True, bufsize=-1,
                          close_fds=closefds,
                          stdin=subprocess.PIPE, stdout=subprocess.PIPE,
                          stderr=subprocess.PIPE,
                          universal_newlines=newlines,
                          env=env)
     return p.stdin, p.stdout, p.stderr, p
 
 def version():
     """Return version information if available."""
     try:
         import __version__
         return __version__.version
     except ImportError:
         return 'unknown'
 
 # used by parsedate
 defaultdateformats = (
     '%Y-%m-%d %H:%M:%S',
     '%Y-%m-%d %I:%M:%S%p',
     '%Y-%m-%d %H:%M',
     '%Y-%m-%d %I:%M%p',
     '%Y-%m-%d',
     '%m-%d',
     '%m/%d',
     '%m/%d/%y',
     '%m/%d/%Y',
     '%a %b %d %H:%M:%S %Y',
     '%a %b %d %I:%M:%S%p %Y',
     '%a, %d %b %Y %H:%M:%S',        #  GNU coreutils "/bin/date --rfc-2822"
     '%b %d %H:%M:%S %Y',
     '%b %d %I:%M:%S%p %Y',
     '%b %d %H:%M:%S',
     '%b %d %I:%M:%S%p',
     '%b %d %H:%M',
     '%b %d %I:%M%p',
     '%b %d %Y',
     '%b %d',
     '%H:%M:%S',
     '%I:%M:%S%p',
     '%H:%M',
     '%I:%M%p',
 )
 
 extendeddateformats = defaultdateformats + (
     "%Y",
     "%Y-%m",
     "%b",
     "%b %Y",
     )
 
 def cachefunc(func):
     '''cache the result of function calls'''
     # XXX doesn't handle keywords args
     if func.func_code.co_argcount == 0:
         cache = []
         def f():
             if len(cache) == 0:
                 cache.append(func())
             return cache[0]
         return f
     cache = {}
     if func.func_code.co_argcount == 1:
         # we gain a small amount of time because
         # we don't need to pack/unpack the list
         def f(arg):
             if arg not in cache:
                 cache[arg] = func(arg)
             return cache[arg]
     else:
         def f(*args):
             if args not in cache:
                 cache[args] = func(*args)
             return cache[args]
 
     return f
 
 class sortdict(dict):
     '''a simple sorted dictionary'''
     def __init__(self, data=None):
         self._list = []
         if data:
             self.update(data)
     def copy(self):
         return sortdict(self)
     def __setitem__(self, key, val):
         if key in self:
             self._list.remove(key)
         self._list.append(key)
         dict.__setitem__(self, key, val)
     def __iter__(self):
         return self._list.__iter__()
     def update(self, src):
         if isinstance(src, dict):
             src = src.iteritems()
         for k, v in src:
             self[k] = v
     def clear(self):
         dict.clear(self)
         self._list = []
     def items(self):
         return [(k, self[k]) for k in self._list]
     def __delitem__(self, key):
         dict.__delitem__(self, key)
         self._list.remove(key)
     def pop(self, key, *args, **kwargs):
         dict.pop(self, key, *args, **kwargs)
         try:
             self._list.remove(key)
         except ValueError:
             pass
     def keys(self):
         return self._list
     def iterkeys(self):
         return self._list.__iter__()
     def iteritems(self):
         for k in self._list:
             yield k, self[k]
     def insert(self, index, key, val):
         self._list.insert(index, key)
         dict.__setitem__(self, key, val)
 
 class lrucachedict(object):
     '''cache most recent gets from or sets to this dictionary'''
     def __init__(self, maxsize):
         self._cache = {}
         self._maxsize = maxsize
         self._order = collections.deque()
 
     def __getitem__(self, key):
         value = self._cache[key]
         self._order.remove(key)
         self._order.append(key)
         return value
 
     def __setitem__(self, key, value):
         if key not in self._cache:
             if len(self._cache) >= self._maxsize:
                 del self._cache[self._order.popleft()]
         else:
             self._order.remove(key)
         self._cache[key] = value
         self._order.append(key)
 
     def __contains__(self, key):
         return key in self._cache
 
     def clear(self):
         self._cache.clear()
         self._order = collections.deque()
 
 def lrucachefunc(func):
     '''cache most recent results of function calls'''
     cache = {}
     order = collections.deque()
     if func.func_code.co_argcount == 1:
         def f(arg):
             if arg not in cache:
                 if len(cache) > 20:
                     del cache[order.popleft()]
                 cache[arg] = func(arg)
             else:
                 order.remove(arg)
             order.append(arg)
             return cache[arg]
     else:
         def f(*args):
             if args not in cache:
                 if len(cache) > 20:
                     del cache[order.popleft()]
                 cache[args] = func(*args)
             else:
                 order.remove(args)
             order.append(args)
             return cache[args]
 
     return f
 
 class propertycache(object):
     def __init__(self, func):
         self.func = func
         self.name = func.__name__
     def __get__(self, obj, type=None):
         result = self.func(obj)
         self.cachevalue(obj, result)
         return result
 
     def cachevalue(self, obj, value):
         # __dict__ assignment required to bypass __setattr__ (eg: repoview)
         obj.__dict__[self.name] = value
 
 def pipefilter(s, cmd):
     '''filter string S through command CMD, returning its output'''
     p = subprocess.Popen(cmd, shell=True, close_fds=closefds,
                          stdin=subprocess.PIPE, stdout=subprocess.PIPE)
     pout, perr = p.communicate(s)
     return pout
 
 def tempfilter(s, cmd):
     '''filter string S through a pair of temporary files with CMD.
     CMD is used as a template to create the real command to be run,
     with the strings INFILE and OUTFILE replaced by the real names of
     the temporary files generated.'''
     inname, outname = None, None
     try:
         infd, inname = tempfile.mkstemp(prefix='hg-filter-in-')
         fp = os.fdopen(infd, 'wb')
         fp.write(s)
         fp.close()
         outfd, outname = tempfile.mkstemp(prefix='hg-filter-out-')
         os.close(outfd)
         cmd = cmd.replace('INFILE', inname)
         cmd = cmd.replace('OUTFILE', outname)
         code = os.system(cmd)
         if sys.platform == 'OpenVMS' and code & 1:
             code = 0
         if code:
             raise Abort(_("command '%s' failed: %s") %
                         (cmd, explainexit(code)))
         fp = open(outname, 'rb')
         r = fp.read()
         fp.close()
         return r
     finally:
         try:
             if inname:
                 os.unlink(inname)
         except OSError:
             pass
         try:
             if outname:
                 os.unlink(outname)
         except OSError:
             pass
 
 filtertable = {
     'tempfile:': tempfilter,
     'pipe:': pipefilter,
     }
 
 def filter(s, cmd):
     "filter a string through a command that transforms its input to its output"
     for name, fn in filtertable.iteritems():
         if cmd.startswith(name):
             return fn(s, cmd[len(name):].lstrip())
     return pipefilter(s, cmd)
 
 def binary(s):
     """return true if a string is binary data"""
     return bool(s and '\0' in s)
 
 def increasingchunks(source, min=1024, max=65536):
     '''return no less than min bytes per chunk while data remains,
     doubling min after each chunk until it reaches max'''
     def log2(x):
         if not x:
             return 0
         i = 0
         while x:
             x >>= 1
             i += 1
         return i - 1
 
     buf = []
     blen = 0
     for chunk in source:
         buf.append(chunk)
         blen += len(chunk)
         if blen >= min:
             if min < max:
                 min = min << 1
                 nmin = 1 << log2(blen)
                 if nmin > min:
                     min = nmin
                 if min > max:
                     min = max
             yield ''.join(buf)
             blen = 0
             buf = []
     if buf:
         yield ''.join(buf)
 
 Abort = error.Abort
 
 def always(fn):
     return True
 
 def never(fn):
     return False
 
 def nogc(func):
     """disable garbage collector
 
     Python's garbage collector triggers a GC each time a certain number of
     container objects (the number being defined by gc.get_threshold()) are
     allocated even when marked not to be tracked by the collector. Tracking has
     no effect on when GCs are triggered, only on what objects the GC looks
     into. As a workaround, disable GC while building complex (huge)
     containers.
 
     This garbage collector issue have been fixed in 2.7.
     """
     def wrapper(*args, **kwargs):
         gcenabled = gc.isenabled()
         gc.disable()
         try:
             return func(*args, **kwargs)
         finally:
             if gcenabled:
                 gc.enable()
     return wrapper
 
 def pathto(root, n1, n2):
     '''return the relative path from one place to another.
     root should use os.sep to separate directories
     n1 should use os.sep to separate directories
     n2 should use "/" to separate directories
     returns an os.sep-separated path.
 
     If n1 is a relative path, it's assumed it's
     relative to root.
     n2 should always be relative to root.
     '''
     if not n1:
         return localpath(n2)
     if os.path.isabs(n1):
         if os.path.splitdrive(root)[0] != os.path.splitdrive(n1)[0]:
             return os.path.join(root, localpath(n2))
         n2 = '/'.join((pconvert(root), n2))
     a, b = splitpath(n1), n2.split('/')
     a.reverse()
     b.reverse()
     while a and b and a[-1] == b[-1]:
         a.pop()
         b.pop()
     b.reverse()
     return os.sep.join((['..'] * len(a)) + b) or '.'
 
 def mainfrozen():
     """return True if we are a frozen executable.
 
     The code supports py2exe (most common, Windows only) and tools/freeze
     (portable, not much used).
     """
     return (safehasattr(sys, "frozen") or # new py2exe
             safehasattr(sys, "importers") or # old py2exe
             imp.is_frozen("__main__")) # tools/freeze
 
 # the location of data files matching the source code
 if mainfrozen():
     # executable version (py2exe) doesn't support __file__
     datapath = os.path.dirname(sys.executable)
 else:
     datapath = os.path.dirname(__file__)
 
 i18n.setdatapath(datapath)
 
 _hgexecutable = None
 
 def hgexecutable():
     """return location of the 'hg' executable.
 
     Defaults to $HG or 'hg' in the search path.
     """
     if _hgexecutable is None:
         hg = os.environ.get('HG')
         mainmod = sys.modules['__main__']
         if hg:
             _sethgexecutable(hg)
         elif mainfrozen():
             _sethgexecutable(sys.executable)
         elif os.path.basename(getattr(mainmod, '__file__', '')) == 'hg':
             _sethgexecutable(mainmod.__file__)
         else:
             exe = findexe('hg') or os.path.basename(sys.argv[0])
             _sethgexecutable(exe)
     return _hgexecutable
 
 def _sethgexecutable(path):
     """set location of the 'hg' executable"""
     global _hgexecutable
     _hgexecutable = path
 
 def system(cmd, environ={}, cwd=None, onerr=None, errprefix=None, out=None):
     '''enhanced shell command execution.
     run with environment maybe modified, maybe in different dir.
 
     if command fails and onerr is None, return status, else raise onerr
     object as exception.
 
     if out is specified, it is assumed to be a file-like object that has a
     write() method. stdout and stderr will be redirected to out.'''
     try:
         sys.stdout.flush()
     except Exception:
         pass
     def py2shell(val):
         'convert python object into string that is useful to shell'
         if val is None or val is False:
             return '0'
         if val is True:
             return '1'
         return str(val)
     origcmd = cmd
     cmd = quotecommand(cmd)
     if sys.platform == 'plan9' and (sys.version_info[0] == 2
                                     and sys.version_info[1] < 7):
         # subprocess kludge to work around issues in half-baked Python
         # ports, notably bichued/python:
         if not cwd is None:
             os.chdir(cwd)
         rc = os.system(cmd)
     else:
         env = dict(os.environ)
         env.update((k, py2shell(v)) for k, v in environ.iteritems())
         env['HG'] = hgexecutable()
         if out is None or out == sys.__stdout__:
             rc = subprocess.call(cmd, shell=True, close_fds=closefds,
                                  env=env, cwd=cwd)
         else:
             proc = subprocess.Popen(cmd, shell=True, close_fds=closefds,
                                     env=env, cwd=cwd, stdout=subprocess.PIPE,
                                     stderr=subprocess.STDOUT)
             while True:
                 line = proc.stdout.readline()
                 if not line:
                     break
                 out.write(line)
             proc.wait()
             rc = proc.returncode
         if sys.platform == 'OpenVMS' and rc & 1:
             rc = 0
     if rc and onerr:
         errmsg = '%s %s' % (os.path.basename(origcmd.split(None, 1)[0]),
                             explainexit(rc)[0])
         if errprefix:
             errmsg = '%s: %s' % (errprefix, errmsg)
         raise onerr(errmsg)
     return rc
 
 def checksignature(func):
     '''wrap a function with code to check for calling errors'''
     def check(*args, **kwargs):
         try:
             return func(*args, **kwargs)
         except TypeError:
             if len(traceback.extract_tb(sys.exc_info()[2])) == 1:
                 raise error.SignatureError
             raise
 
     return check
 
 def copyfile(src, dest, hardlink=False):
     "copy a file, preserving mode and atime/mtime"
     if os.path.lexists(dest):
         unlink(dest)
     # hardlinks are problematic on CIFS, quietly ignore this flag
     # until we find a way to work around it cleanly (issue4546)
     if False and hardlink:
         try:
             oslink(src, dest)
             return
         except (IOError, OSError):
             pass # fall back to normal copy
     if os.path.islink(src):
         os.symlink(os.readlink(src), dest)
     else:
         try:
             shutil.copyfile(src, dest)
             shutil.copymode(src, dest)
         except shutil.Error, inst:
             raise Abort(str(inst))
 
 def copyfiles(src, dst, hardlink=None, progress=lambda t, pos: None):
     """Copy a directory tree using hardlinks if possible."""
     num = 0
 
     if hardlink is None:
         hardlink = (os.stat(src).st_dev ==
                     os.stat(os.path.dirname(dst)).st_dev)
     if hardlink:
         topic = _('linking')
     else:
         topic = _('copying')
 
     if os.path.isdir(src):
         os.mkdir(dst)
         for name, kind in osutil.listdir(src):
             srcname = os.path.join(src, name)
             dstname = os.path.join(dst, name)
             def nprog(t, pos):
                 if pos is not None:
                     return progress(t, pos + num)
             hardlink, n = copyfiles(srcname, dstname, hardlink, progress=nprog)
             num += n
     else:
         if hardlink:
             try:
                 oslink(src, dst)
             except (IOError, OSError):
                 hardlink = False
                 shutil.copy(src, dst)
         else:
             shutil.copy(src, dst)
         num += 1
         progress(topic, num)
     progress(topic, None)
 
     return hardlink, num
 
 _winreservednames = '''con prn aux nul
     com1 com2 com3 com4 com5 com6 com7 com8 com9
     lpt1 lpt2 lpt3 lpt4 lpt5 lpt6 lpt7 lpt8 lpt9'''.split()
 _winreservedchars = ':*?"<>|'
 def checkwinfilename(path):
     r'''Check that the base-relative path is a valid filename on Windows.
     Returns None if the path is ok, or a UI string describing the problem.
 
     >>> checkwinfilename("just/a/normal/path")
     >>> checkwinfilename("foo/bar/con.xml")
     "filename contains 'con', which is reserved on Windows"
     >>> checkwinfilename("foo/con.xml/bar")
     "filename contains 'con', which is reserved on Windows"
     >>> checkwinfilename("foo/bar/xml.con")
     >>> checkwinfilename("foo/bar/AUX/bla.txt")
     "filename contains 'AUX', which is reserved on Windows"
     >>> checkwinfilename("foo/bar/bla:.txt")
     "filename contains ':', which is reserved on Windows"
     >>> checkwinfilename("foo/bar/b\07la.txt")
     "filename contains '\\x07', which is invalid on Windows"
     >>> checkwinfilename("foo/bar/bla ")
     "filename ends with ' ', which is not allowed on Windows"
     >>> checkwinfilename("../bar")
     >>> checkwinfilename("foo\\")
     "filename ends with '\\', which is invalid on Windows"
     >>> checkwinfilename("foo\\/bar")
     "directory name ends with '\\', which is invalid on Windows"
     '''
     if path.endswith('\\'):
         return _("filename ends with '\\', which is invalid on Windows")
     if '\\/' in path:
         return _("directory name ends with '\\', which is invalid on Windows")
     for n in path.replace('\\', '/').split('/'):
         if not n:
             continue
         for c in n:
             if c in _winreservedchars:
                 return _("filename contains '%s', which is reserved "
                          "on Windows") % c
             if ord(c) <= 31:
                 return _("filename contains %r, which is invalid "
                          "on Windows") % c
         base = n.split('.')[0]
         if base and base.lower() in _winreservednames:
             return _("filename contains '%s', which is reserved "
                      "on Windows") % base
         t = n[-1]
         if t in '. ' and n not in '..':
             return _("filename ends with '%s', which is not allowed "
                      "on Windows") % t
 
 if os.name == 'nt':
     checkosfilename = checkwinfilename
 else:
     checkosfilename = platform.checkosfilename
 
 def makelock(info, pathname):
     try:
         return os.symlink(info, pathname)
     except OSError, why:
         if why.errno == errno.EEXIST:
             raise
     except AttributeError: # no symlink in os
         pass
 
     ld = os.open(pathname, os.O_CREAT | os.O_WRONLY | os.O_EXCL)
     os.write(ld, info)
     os.close(ld)
 
 def readlock(pathname):
     try:
         return os.readlink(pathname)
     except OSError, why:
         if why.errno not in (errno.EINVAL, errno.ENOSYS):
             raise
     except AttributeError: # no symlink in os
         pass
     fp = posixfile(pathname)
     r = fp.read()
     fp.close()
     return r
 
 def fstat(fp):
     '''stat file object that may not have fileno method.'''
     try:
         return os.fstat(fp.fileno())
     except AttributeError:
         return os.stat(fp.name)
 
 # File system features
 
 def checkcase(path):
     """
     Return true if the given path is on a case-sensitive filesystem
 
     Requires a path (like /foo/.hg) ending with a foldable final
     directory component.
     """
     s1 = os.lstat(path)
     d, b = os.path.split(path)
     b2 = b.upper()
     if b == b2:
         b2 = b.lower()
         if b == b2:
             return True # no evidence against case sensitivity
     p2 = os.path.join(d, b2)
     try:
         s2 = os.lstat(p2)
         if s2 == s1:
             return False
         return True
     except OSError:
         return True
 
 try:
     import re2
     _re2 = None
 except ImportError:
     _re2 = False
 
 class _re(object):
     def _checkre2(self):
         global _re2
         try:
             # check if match works, see issue3964
             _re2 = bool(re2.match(r'\[([^\[]+)\]', '[ui]'))
         except ImportError:
             _re2 = False
 
     def compile(self, pat, flags=0):
         '''Compile a regular expression, using re2 if possible
 
         For best performance, use only re2-compatible regexp features. The
         only flags from the re module that are re2-compatible are
         IGNORECASE and MULTILINE.'''
         if _re2 is None:
             self._checkre2()
         if _re2 and (flags & ~(remod.IGNORECASE | remod.MULTILINE)) == 0:
             if flags & remod.IGNORECASE:
                 pat = '(?i)' + pat
             if flags & remod.MULTILINE:
                 pat = '(?m)' + pat
             try:
                 return re2.compile(pat)
             except re2.error:
                 pass
         return remod.compile(pat, flags)
 
     @propertycache
     def escape(self):
         '''Return the version of escape corresponding to self.compile.
 
         This is imperfect because whether re2 or re is used for a particular
         function depends on the flags, etc, but it's the best we can do.
         '''
         global _re2
         if _re2 is None:
             self._checkre2()
         if _re2:
             return re2.escape
         else:
             return remod.escape
 
 re = _re()
 
 _fspathcache = {}
 def fspath(name, root):
     '''Get name in the case stored in the filesystem
 
     The name should be relative to root, and be normcase-ed for efficiency.
 
     Note that this function is unnecessary, and should not be
     called, for case-sensitive filesystems (simply because it's expensive).
 
     The root should be normcase-ed, too.
     '''
     def _makefspathcacheentry(dir):
         return dict((normcase(n), n) for n in os.listdir(dir))
 
     seps = os.sep
     if os.altsep:
         seps = seps + os.altsep
     # Protect backslashes. This gets silly very quickly.
     seps.replace('\\','\\\\')
     pattern = remod.compile(r'([^%s]+)|([%s]+)' % (seps, seps))
     dir = os.path.normpath(root)
     result = []
     for part, sep in pattern.findall(name):
         if sep:
             result.append(sep)
             continue
 
         if dir not in _fspathcache:
             _fspathcache[dir] = _makefspathcacheentry(dir)
         contents = _fspathcache[dir]
 
         found = contents.get(part)
         if not found:
             # retry "once per directory" per "dirstate.walk" which
             # may take place for each patches of "hg qpush", for example
             _fspathcache[dir] = contents = _makefspathcacheentry(dir)
             found = contents.get(part)
 
         result.append(found or part)
         dir = os.path.join(dir, part)
 
     return ''.join(result)
 
 def checknlink(testfile):
     '''check whether hardlink count reporting works properly'''
 
     # testfile may be open, so we need a separate file for checking to
     # work around issue2543 (or testfile may get lost on Samba shares)
     f1 = testfile + ".hgtmp1"
     if os.path.lexists(f1):
         return False
     try:
         posixfile(f1, 'w').close()
     except IOError:
         return False
 
     f2 = testfile + ".hgtmp2"
     fd = None
     try:
         oslink(f1, f2)
         # nlinks() may behave differently for files on Windows shares if
         # the file is open.
         fd = posixfile(f2)
         return nlinks(f2) > 1
     except OSError:
         return False
     finally:
         if fd is not None:
             fd.close()
         for f in (f1, f2):
             try:
                 os.unlink(f)
             except OSError:
                 pass
 
 def endswithsep(path):
     '''Check path ends with os.sep or os.altsep.'''
     return path.endswith(os.sep) or os.altsep and path.endswith(os.altsep)
 
 def splitpath(path):
     '''Split path by os.sep.
     Note that this function does not use os.altsep because this is
     an alternative of simple "xxx.split(os.sep)".
     It is recommended to use os.path.normpath() before using this
     function if need.'''
     return path.split(os.sep)
 
 def gui():
     '''Are we running in a GUI?'''
     if sys.platform == 'darwin':
         if 'SSH_CONNECTION' in os.environ:
             # handle SSH access to a box where the user is logged in
             return False
         elif getattr(osutil, 'isgui', None):
             # check if a CoreGraphics session is available
             return osutil.isgui()
         else:
             # pure build; use a safe default
             return True
     else:
         return os.name == "nt" or os.environ.get("DISPLAY")
 
 def mktempcopy(name, emptyok=False, createmode=None):
     """Create a temporary file with the same contents from name
 
     The permission bits are copied from the original file.
 
     If the temporary file is going to be truncated immediately, you
     can use emptyok=True as an optimization.
 
     Returns the name of the temporary file.
     """
     d, fn = os.path.split(name)
     fd, temp = tempfile.mkstemp(prefix='.%s-' % fn, dir=d)
     os.close(fd)
     # Temporary files are created with mode 0600, which is usually not
     # what we want.  If the original file already exists, just copy
     # its mode.  Otherwise, manually obey umask.
     copymode(name, temp, createmode)
     if emptyok:
         return temp
     try:
         try:
             ifp = posixfile(name, "rb")
         except IOError, inst:
             if inst.errno == errno.ENOENT:
                 return temp
             if not getattr(inst, 'filename', None):
                 inst.filename = name
             raise
         ofp = posixfile(temp, "wb")
         for chunk in filechunkiter(ifp):
             ofp.write(chunk)
         ifp.close()
         ofp.close()
     except: # re-raises
         try: os.unlink(temp)
         except OSError: pass
         raise
     return temp
 
 class atomictempfile(object):
     '''writable file object that atomically updates a file
 
     All writes will go to a temporary copy of the original file. Call
     close() when you are done writing, and atomictempfile will rename
     the temporary copy to the original name, making the changes
     visible. If the object is destroyed without being closed, all your
     writes are discarded.
     '''
     def __init__(self, name, mode='w+b', createmode=None):
         self.__name = name      # permanent name
         self._tempname = mktempcopy(name, emptyok=('w' in mode),
                                     createmode=createmode)
         self._fp = posixfile(self._tempname, mode)
 
         # delegated methods
         self.write = self._fp.write
         self.seek = self._fp.seek
         self.tell = self._fp.tell
         self.fileno = self._fp.fileno
 
     def close(self):
         if not self._fp.closed:
             self._fp.close()
             rename(self._tempname, localpath(self.__name))
 
     def discard(self):
         if not self._fp.closed:
             try:
                 os.unlink(self._tempname)
             except OSError:
                 pass
             self._fp.close()
 
     def __del__(self):
         if safehasattr(self, '_fp'): # constructor actually did something
             self.discard()
 
 def makedirs(name, mode=None, notindexed=False):
     """recursive directory creation with parent mode inheritance"""
     try:
         makedir(name, notindexed)
     except OSError, err:
         if err.errno == errno.EEXIST:
             return
         if err.errno != errno.ENOENT or not name:
             raise
         parent = os.path.dirname(os.path.abspath(name))
         if parent == name:
             raise
         makedirs(parent, mode, notindexed)
         makedir(name, notindexed)
     if mode is not None:
         os.chmod(name, mode)
 
 def ensuredirs(name, mode=None, notindexed=False):
     """race-safe recursive directory creation
 
     Newly created directories are marked as "not to be indexed by
     the content indexing service", if ``notindexed`` is specified
     for "write" mode access.
     """
     if os.path.isdir(name):
         return
     parent = os.path.dirname(os.path.abspath(name))
     if parent != name:
         ensuredirs(parent, mode, notindexed)
     try:
         makedir(name, notindexed)
     except OSError, err:
         if err.errno == errno.EEXIST and os.path.isdir(name):
             # someone else seems to have won a directory creation race
             return
         raise
     if mode is not None:
         os.chmod(name, mode)
 
 def readfile(path):
     fp = open(path, 'rb')
     try:
         return fp.read()
     finally:
         fp.close()
 
 def writefile(path, text):
     fp = open(path, 'wb')
     try:
         fp.write(text)
     finally:
         fp.close()
 
 def appendfile(path, text):
     fp = open(path, 'ab')
     try:
         fp.write(text)
     finally:
         fp.close()
 
 class chunkbuffer(object):
     """Allow arbitrary sized chunks of data to be efficiently read from an
     iterator over chunks of arbitrary size."""
 
     def __init__(self, in_iter):
         """in_iter is the iterator that's iterating over the input chunks.
         targetsize is how big a buffer to try to maintain."""
         def splitbig(chunks):
             for chunk in chunks:
                 if len(chunk) > 2**20:
                     pos = 0
                     while pos < len(chunk):
                         end = pos + 2 ** 18
                         yield chunk[pos:end]
                         pos = end
                 else:
                     yield chunk
         self.iter = splitbig(in_iter)
         self._queue = collections.deque()
 
     def read(self, l=None):
         """Read L bytes of data from the iterator of chunks of data.
         Returns less than L bytes if the iterator runs dry.
 
         If size parameter is omitted, read everything"""
         left = l
         buf = []
         queue = self._queue
         while left is None or left > 0:
             # refill the queue
             if not queue:
                 target = 2**18
                 for chunk in self.iter:
                     queue.append(chunk)
                     target -= len(chunk)
                     if target <= 0:
                         break
                 if not queue:
                     break
 
             chunk = queue.popleft()
             if left is not None:
                 left -= len(chunk)
             if left is not None and left < 0:
                 queue.appendleft(chunk[left:])
                 buf.append(chunk[:left])
             else:
                 buf.append(chunk)
 
         return ''.join(buf)
 
 def filechunkiter(f, size=65536, limit=None):
     """Create a generator that produces the data in the file size
     (default 65536) bytes at a time, up to optional limit (default is
     to read all data).  Chunks may be less than size bytes if the
     chunk is the last chunk in the file, or the file is a socket or
     some other type of file that sometimes reads less data than is
     requested."""
     assert size >= 0
     assert limit is None or limit >= 0
     while True:
         if limit is None:
             nbytes = size
         else:
             nbytes = min(limit, size)
         s = nbytes and f.read(nbytes)
         if not s:
             break
         if limit:
             limit -= len(s)
         yield s
 
 def makedate(timestamp=None):
     '''Return a unix timestamp (or the current time) as a (unixtime,
     offset) tuple based off the local timezone.'''
     if timestamp is None:
         timestamp = time.time()
     if timestamp < 0:
         hint = _("check your clock")
         raise Abort(_("negative timestamp: %d") % timestamp, hint=hint)
     delta = (datetime.datetime.utcfromtimestamp(timestamp) -
              datetime.datetime.fromtimestamp(timestamp))
     tz = delta.days * 86400 + delta.seconds
     return timestamp, tz
 
 def datestr(date=None, format='%a %b %d %H:%M:%S %Y %1%2'):
     """represent a (unixtime, offset) tuple as a localized time.
     unixtime is seconds since the epoch, and offset is the time zone's
     number of seconds away from UTC. if timezone is false, do not
     append time zone to string."""
     t, tz = date or makedate()
     if t < 0:
         t = 0   # time.gmtime(lt) fails on Windows for lt < -43200
         tz = 0
     if "%1" in format or "%2" in format or "%z" in format:
         sign = (tz > 0) and "-" or "+"
         minutes = abs(tz) // 60
         format = format.replace("%z", "%1%2")
         format = format.replace("%1", "%c%02d" % (sign, minutes // 60))
         format = format.replace("%2", "%02d" % (minutes % 60))
     try:
         t = time.gmtime(float(t) - tz)
     except ValueError:
         # time was out of range
         t = time.gmtime(sys.maxint)
     s = time.strftime(format, t)
     return s
 
 def shortdate(date=None):
     """turn (timestamp, tzoff) tuple into iso 8631 date."""
     return datestr(date, format='%Y-%m-%d')
 
 def strdate(string, format, defaults=[]):
     """parse a localized time string and return a (unixtime, offset) tuple.
     if the string cannot be parsed, ValueError is raised."""
     def timezone(string):
         tz = string.split()[-1]
         if tz[0] in "+-" and len(tz) == 5 and tz[1:].isdigit():
             sign = (tz[0] == "+") and 1 or -1
             hours = int(tz[1:3])
             minutes = int(tz[3:5])
             return -sign * (hours * 60 + minutes) * 60
         if tz == "GMT" or tz == "UTC":
             return 0
         return None
 
     # NOTE: unixtime = localunixtime + offset
     offset, date = timezone(string), string
     if offset is not None:
         date = " ".join(string.split()[:-1])
 
     # add missing elements from defaults
     usenow = False # default to using biased defaults
     for part in ("S", "M", "HI", "d", "mb", "yY"): # decreasing specificity
         found = [True for p in part if ("%"+p) in format]
         if not found:
             date += "@" + defaults[part][usenow]
             format += "@%" + part[0]
         else:
             # We've found a specific time element, less specific time
             # elements are relative to today
             usenow = True
 
     timetuple = time.strptime(date, format)
     localunixtime = int(calendar.timegm(timetuple))
     if offset is None:
         # local timezone
         unixtime = int(time.mktime(timetuple))
         offset = unixtime - localunixtime
     else:
         unixtime = localunixtime + offset
     return unixtime, offset
 
 def parsedate(date, formats=None, bias={}):
     """parse a localized date/time and return a (unixtime, offset) tuple.
 
     The date may be a "unixtime offset" string or in one of the specified
     formats. If the date already is a (unixtime, offset) tuple, it is returned.
 
     >>> parsedate(' today ') == parsedate(\
                                   datetime.date.today().strftime('%b %d'))
     True
     >>> parsedate( 'yesterday ') == parsedate((datetime.date.today() -\
                                                datetime.timedelta(days=1)\
                                               ).strftime('%b %d'))
     True
     >>> now, tz = makedate()
     >>> strnow, strtz = parsedate('now')
     >>> (strnow - now) < 1
     True
     >>> tz == strtz
     True
     """
     if not date:
         return 0, 0
     if isinstance(date, tuple) and len(date) == 2:
         return date
     if not formats:
         formats = defaultdateformats
     date = date.strip()
 
     if date == 'now' or date == _('now'):
         return makedate()
     if date == 'today' or date == _('today'):
         date = datetime.date.today().strftime('%b %d')
     elif date == 'yesterday' or date == _('yesterday'):
         date = (datetime.date.today() -
                 datetime.timedelta(days=1)).strftime('%b %d')
 
     try:
         when, offset = map(int, date.split(' '))
     except ValueError:
         # fill out defaults
         now = makedate()
         defaults = {}
         for part in ("d", "mb", "yY", "HI", "M", "S"):
             # this piece is for rounding the specific end of unknowns
             b = bias.get(part)
             if b is None:
                 if part[0] in "HMS":
                     b = "00"
                 else:
                     b = "0"
 
             # this piece is for matching the generic end to today's date
             n = datestr(now, "%" + part[0])
 
             defaults[part] = (b, n)
 
         for format in formats:
             try:
                 when, offset = strdate(date, format, defaults)
             except (ValueError, OverflowError):
                 pass
             else:
                 break
         else:
             raise Abort(_('invalid date: %r') % date)
     # validate explicit (probably user-specified) date and
     # time zone offset. values must fit in signed 32 bits for
     # current 32-bit linux runtimes. timezones go from UTC-12
     # to UTC+14
     if abs(when) > 0x7fffffff:
         raise Abort(_('date exceeds 32 bits: %d') % when)
     if when < 0:
         raise Abort(_('negative date value: %d') % when)
     if offset < -50400 or offset > 43200:
         raise Abort(_('impossible time zone offset: %d') % offset)
     return when, offset
 
 def matchdate(date):
     """Return a function that matches a given date match specifier
 
     Formats include:
 
     '{date}' match a given date to the accuracy provided
 
     '<{date}' on or before a given date
 
     '>{date}' on or after a given date
 
     >>> p1 = parsedate("10:29:59")
     >>> p2 = parsedate("10:30:00")
     >>> p3 = parsedate("10:30:59")
     >>> p4 = parsedate("10:31:00")
     >>> p5 = parsedate("Sep 15 10:30:00 1999")
     >>> f = matchdate("10:30")
     >>> f(p1[0])
     False
     >>> f(p2[0])
     True
     >>> f(p3[0])
     True
     >>> f(p4[0])
     False
     >>> f(p5[0])
     False
     """
 
     def lower(date):
         d = {'mb': "1", 'd': "1"}
         return parsedate(date, extendeddateformats, d)[0]
 
     def upper(date):
         d = {'mb': "12", 'HI': "23", 'M': "59", 'S': "59"}
         for days in ("31", "30", "29"):
             try:
                 d["d"] = days
                 return parsedate(date, extendeddateformats, d)[0]
             except Abort:
                 pass
         d["d"] = "28"
         return parsedate(date, extendeddateformats, d)[0]
 
     date = date.strip()
 
     if not date:
         raise Abort(_("dates cannot consist entirely of whitespace"))
     elif date[0] == "<":
         if not date[1:]:
             raise Abort(_("invalid day spec, use '<DATE'"))
         when = upper(date[1:])
         return lambda x: x <= when
     elif date[0] == ">":
         if not date[1:]:
             raise Abort(_("invalid day spec, use '>DATE'"))
         when = lower(date[1:])
         return lambda x: x >= when
     elif date[0] == "-":
         try:
             days = int(date[1:])
         except ValueError:
             raise Abort(_("invalid day spec: %s") % date[1:])
         if days < 0:
             raise Abort(_('%s must be nonnegative (see "hg help dates")')
                 % date[1:])
         when = makedate()[0] - days * 3600 * 24
         return lambda x: x >= when
     elif " to " in date:
         a, b = date.split(" to ")
         start, stop = lower(a), upper(b)
         return lambda x: x >= start and x <= stop
     else:
         start, stop = lower(date), upper(date)
         return lambda x: x >= start and x <= stop
 
 def shortuser(user):
     """Return a short representation of a user name or email address."""
     f = user.find('@')
     if f >= 0:
         user = user[:f]
     f = user.find('<')
     if f >= 0:
         user = user[f + 1:]
     f = user.find(' ')
     if f >= 0:
         user = user[:f]
     f = user.find('.')
     if f >= 0:
         user = user[:f]
     return user
 
 def emailuser(user):
     """Return the user portion of an email address."""
     f = user.find('@')
     if f >= 0:
         user = user[:f]
     f = user.find('<')
     if f >= 0:
         user = user[f + 1:]
     return user
 
 def email(author):
     '''get email of author.'''
     r = author.find('>')
     if r == -1:
         r = None
     return author[author.find('<') + 1:r]
 
 def ellipsis(text, maxlength=400):
     """Trim string to at most maxlength (default: 400) columns in display."""
     return encoding.trim(text, maxlength, ellipsis='...')
 
 def unitcountfn(*unittable):
     '''return a function that renders a readable count of some quantity'''
 
     def go(count):
         for multiplier, divisor, format in unittable:
             if count >= divisor * multiplier:
                 return format % (count / float(divisor))
         return unittable[-1][2] % count
 
     return go
 
 bytecount = unitcountfn(
     (100, 1 << 30, _('%.0f GB')),
     (10, 1 << 30, _('%.1f GB')),
     (1, 1 << 30, _('%.2f GB')),
     (100, 1 << 20, _('%.0f MB')),
     (10, 1 << 20, _('%.1f MB')),
     (1, 1 << 20, _('%.2f MB')),
     (100, 1 << 10, _('%.0f KB')),
     (10, 1 << 10, _('%.1f KB')),
     (1, 1 << 10, _('%.2f KB')),
     (1, 1, _('%.0f bytes')),
     )
 
 def uirepr(s):
     # Avoid double backslash in Windows path repr()
     return repr(s).replace('\\\\', '\\')
 
 # delay import of textwrap
 def MBTextWrapper(**kwargs):
     class tw(textwrap.TextWrapper):
         """
         Extend TextWrapper for width-awareness.
 
         Neither number of 'bytes' in any encoding nor 'characters' is
         appropriate to calculate terminal columns for specified string.
 
         Original TextWrapper implementation uses built-in 'len()' directly,
         so overriding is needed to use width information of each characters.
 
         In addition, characters classified into 'ambiguous' width are
         treated as wide in East Asian area, but as narrow in other.
 
         This requires use decision to determine width of such characters.
         """
         def _cutdown(self, ucstr, space_left):
             l = 0
             colwidth = encoding.ucolwidth
             for i in xrange(len(ucstr)):
                 l += colwidth(ucstr[i])
                 if space_left < l:
                     return (ucstr[:i], ucstr[i:])
             return ucstr, ''
 
         # overriding of base class
         def _handle_long_word(self, reversed_chunks, cur_line, cur_len, width):
             space_left = max(width - cur_len, 1)
 
             if self.break_long_words:
                 cut, res = self._cutdown(reversed_chunks[-1], space_left)
                 cur_line.append(cut)
                 reversed_chunks[-1] = res
             elif not cur_line:
                 cur_line.append(reversed_chunks.pop())
 
         # this overriding code is imported from TextWrapper of python 2.6
         # to calculate columns of string by 'encoding.ucolwidth()'
         def _wrap_chunks(self, chunks):
             colwidth = encoding.ucolwidth
 
             lines = []
             if self.width <= 0:
                 raise ValueError("invalid width %r (must be > 0)" % self.width)
 
             # Arrange in reverse order so items can be efficiently popped
             # from a stack of chucks.
             chunks.reverse()
 
             while chunks:
 
                 # Start the list of chunks that will make up the current line.
                 # cur_len is just the length of all the chunks in cur_line.
                 cur_line = []
                 cur_len = 0
 
                 # Figure out which static string will prefix this line.
                 if lines:
                     indent = self.subsequent_indent
                 else:
                     indent = self.initial_indent
 
                 # Maximum width for this line.
                 width = self.width - len(indent)
 
                 # First chunk on line is whitespace -- drop it, unless this
                 # is the very beginning of the text (i.e. no lines started yet).
                 if self.drop_whitespace and chunks[-1].strip() == '' and lines:
                     del chunks[-1]
 
                 while chunks:
                     l = colwidth(chunks[-1])
 
                     # Can at least squeeze this chunk onto the current line.
                     if cur_len + l <= width:
                         cur_line.append(chunks.pop())
                         cur_len += l
 
                     # Nope, this line is full.
                     else:
                         break
 
                 # The current line is full, and the next chunk is too big to
                 # fit on *any* line (not just this one).
                 if chunks and colwidth(chunks[-1]) > width:
                     self._handle_long_word(chunks, cur_line, cur_len, width)
 
                 # If the last chunk on this line is all whitespace, drop it.
                 if (self.drop_whitespace and
                     cur_line and cur_line[-1].strip() == ''):
                     del cur_line[-1]
 
                 # Convert current line back to a string and store it in list
                 # of all lines (return value).
                 if cur_line:
                     lines.append(indent + ''.join(cur_line))
 
             return lines
 
     global MBTextWrapper
     MBTextWrapper = tw
     return tw(**kwargs)
 
 def wrap(line, width, initindent='', hangindent=''):
     maxindent = max(len(hangindent), len(initindent))
     if width <= maxindent:
         # adjust for weird terminal size
         width = max(78, maxindent + 1)
     line = line.decode(encoding.encoding, encoding.encodingmode)
     initindent = initindent.decode(encoding.encoding, encoding.encodingmode)
     hangindent = hangindent.decode(encoding.encoding, encoding.encodingmode)
     wrapper = MBTextWrapper(width=width,
                             initial_indent=initindent,
                             subsequent_indent=hangindent)
     return wrapper.fill(line).encode(encoding.encoding)
 
 def iterlines(iterator):
     for chunk in iterator:
         for line in chunk.splitlines():
             yield line
 
 def expandpath(path):
     return os.path.expanduser(os.path.expandvars(path))
 
 def hgcmd():
     """Return the command used to execute current hg
 
     This is different from hgexecutable() because on Windows we want
     to avoid things opening new shell windows like batch files, so we
     get either the python call or current executable.
     """
     if mainfrozen():
         return [sys.executable]
     return gethgcmd()
 
 def rundetached(args, condfn):
     """Execute the argument list in a detached process.
 
     condfn is a callable which is called repeatedly and should return
     True once the child process is known to have started successfully.
     At this point, the child process PID is returned. If the child
     process fails to start or finishes before condfn() evaluates to
     True, return -1.
     """
     # Windows case is easier because the child process is either
     # successfully starting and validating the condition or exiting
     # on failure. We just poll on its PID. On Unix, if the child
     # process fails to start, it will be left in a zombie state until
     # the parent wait on it, which we cannot do since we expect a long
     # running process on success. Instead we listen for SIGCHLD telling
     # us our child process terminated.
     terminated = set()
     def handler(signum, frame):
         terminated.add(os.wait())
     prevhandler = None
     SIGCHLD = getattr(signal, 'SIGCHLD', None)
     if SIGCHLD is not None:
         prevhandler = signal.signal(SIGCHLD, handler)
     try:
         pid = spawndetached(args)
         while not condfn():
             if ((pid in terminated or not testpid(pid))
                 and not condfn()):
                 return -1
             time.sleep(0.1)
         return pid
     finally:
         if prevhandler is not None:
             signal.signal(signal.SIGCHLD, prevhandler)
 
 def interpolate(prefix, mapping, s, fn=None, escape_prefix=False):
     """Return the result of interpolating items in the mapping into string s.
 
     prefix is a single character string, or a two character string with
     a backslash as the first character if the prefix needs to be escaped in
     a regular expression.
 
     fn is an optional function that will be applied to the replacement text
     just before replacement.
 
     escape_prefix is an optional flag that allows using doubled prefix for
     its escaping.
     """
     fn = fn or (lambda s: s)
     patterns = '|'.join(mapping.keys())
     if escape_prefix:
         patterns += '|' + prefix
         if len(prefix) > 1:
             prefix_char = prefix[1:]
         else:
             prefix_char = prefix
         mapping[prefix_char] = prefix_char
     r = remod.compile(r'%s(%s)' % (prefix, patterns))
     return r.sub(lambda x: fn(mapping[x.group()[1:]]), s)
 
 def getport(port):
     """Return the port for a given network service.
 
     If port is an integer, it's returned as is. If it's a string, it's
     looked up using socket.getservbyname(). If there's no matching
     service, util.Abort is raised.
     """
     try:
         return int(port)
     except ValueError:
         pass
 
     try:
         return socket.getservbyname(port)
     except socket.error:
         raise Abort(_("no port number associated with service '%s'") % port)
 
 _booleans = {'1': True, 'yes': True, 'true': True, 'on': True, 'always': True,
              '0': False, 'no': False, 'false': False, 'off': False,
              'never': False}
 
 def parsebool(s):
     """Parse s into a boolean.
 
     If s is not a valid boolean, returns None.
     """
     return _booleans.get(s.lower(), None)
 
 _hexdig = '0123456789ABCDEFabcdef'
 _hextochr = dict((a + b, chr(int(a + b, 16)))
                  for a in _hexdig for b in _hexdig)
 
 def _urlunquote(s):
     """Decode HTTP/HTML % encoding.
 
     >>> _urlunquote('abc%20def')
     'abc def'
     """
     res = s.split('%')
     # fastpath
     if len(res) == 1:
         return s
     s = res[0]
     for item in res[1:]:
         try:
             s += _hextochr[item[:2]] + item[2:]
         except KeyError:
             s += '%' + item
         except UnicodeDecodeError:
             s += unichr(int(item[:2], 16)) + item[2:]
     return s
 
 class url(object):
     r"""Reliable URL parser.
 
     This parses URLs and provides attributes for the following
     components:
 
     <scheme>://<user>:<passwd>@<host>:<port>/<path>?<query>#<fragment>
 
     Missing components are set to None. The only exception is
     fragment, which is set to '' if present but empty.
 
     If parsefragment is False, fragment is included in query. If
     parsequery is False, query is included in path. If both are
     False, both fragment and query are included in path.
 
     See http://www.ietf.org/rfc/rfc2396.txt for more information.
 
     Note that for backward compatibility reasons, bundle URLs do not
     take host names. That means 'bundle://../' has a path of '../'.
 
     Examples:
 
     >>> url('http://www.ietf.org/rfc/rfc2396.txt')
     <url scheme: 'http', host: 'www.ietf.org', path: 'rfc/rfc2396.txt'>
     >>> url('ssh://[::1]:2200//home/joe/repo')
     <url scheme: 'ssh', host: '[::1]', port: '2200', path: '/home/joe/repo'>
     >>> url('file:///home/joe/repo')
     <url scheme: 'file', path: '/home/joe/repo'>
     >>> url('file:///c:/temp/foo/')
     <url scheme: 'file', path: 'c:/temp/foo/'>
     >>> url('bundle:foo')
     <url scheme: 'bundle', path: 'foo'>
     >>> url('bundle://../foo')
     <url scheme: 'bundle', path: '../foo'>
     >>> url(r'c:\foo\bar')
     <url path: 'c:\\foo\\bar'>
     >>> url(r'\\blah\blah\blah')
     <url path: '\\\\blah\\blah\\blah'>
     >>> url(r'\\blah\blah\blah#baz')
     <url path: '\\\\blah\\blah\\blah', fragment: 'baz'>
     >>> url(r'file:///C:\users\me')
     <url scheme: 'file', path: 'C:\\users\\me'>
 
     Authentication credentials:
 
     >>> url('ssh://joe:xyz@x/repo')
     <url scheme: 'ssh', user: 'joe', passwd: 'xyz', host: 'x', path: 'repo'>
     >>> url('ssh://joe@x/repo')
     <url scheme: 'ssh', user: 'joe', host: 'x', path: 'repo'>
 
     Query strings and fragments:
 
     >>> url('http://host/a?b#c')
     <url scheme: 'http', host: 'host', path: 'a', query: 'b', fragment: 'c'>
     >>> url('http://host/a?b#c', parsequery=False, parsefragment=False)
     <url scheme: 'http', host: 'host', path: 'a?b#c'>
     """
 
     _safechars = "!~*'()+"
     _safepchars = "/!~*'()+:\\"
     _matchscheme = remod.compile(r'^[a-zA-Z0-9+.\-]+:').match
 
     def __init__(self, path, parsequery=True, parsefragment=True):
         # We slowly chomp away at path until we have only the path left
         self.scheme = self.user = self.passwd = self.host = None
         self.port = self.path = self.query = self.fragment = None
         self._localpath = True
         self._hostport = ''
         self._origpath = path
 
         if parsefragment and '#' in path:
             path, self.fragment = path.split('#', 1)
             if not path:
                 path = None
 
         # special case for Windows drive letters and UNC paths
         if hasdriveletter(path) or path.startswith(r'\\'):
             self.path = path
             return
 
         # For compatibility reasons, we can't handle bundle paths as
         # normal URLS
         if path.startswith('bundle:'):
             self.scheme = 'bundle'
             path = path[7:]
             if path.startswith('//'):
                 path = path[2:]
             self.path = path
             return
 
         if self._matchscheme(path):
             parts = path.split(':', 1)
             if parts[0]:
                 self.scheme, path = parts
                 self._localpath = False
 
         if not path:
             path = None
             if self._localpath:
                 self.path = ''
                 return
         else:
             if self._localpath:
                 self.path = path
                 return
 
             if parsequery and '?' in path:
                 path, self.query = path.split('?', 1)
                 if not path:
                     path = None
                 if not self.query:
                     self.query = None
 
             # // is required to specify a host/authority
             if path and path.startswith('//'):
                 parts = path[2:].split('/', 1)
                 if len(parts) > 1:
                     self.host, path = parts
                 else:
                     self.host = parts[0]
                     path = None
                 if not self.host:
                     self.host = None
                     # path of file:///d is /d
                     # path of file:///d:/ is d:/, not /d:/
                     if path and not hasdriveletter(path):
                         path = '/' + path
 
             if self.host and '@' in self.host:
                 self.user, self.host = self.host.rsplit('@', 1)
                 if ':' in self.user:
                     self.user, self.passwd = self.user.split(':', 1)
                 if not self.host:
                     self.host = None
 
             # Don't split on colons in IPv6 addresses without ports
             if (self.host and ':' in self.host and
                 not (self.host.startswith('[') and self.host.endswith(']'))):
                 self._hostport = self.host
                 self.host, self.port = self.host.rsplit(':', 1)
                 if not self.host:
                     self.host = None
 
             if (self.host and self.scheme == 'file' and
                 self.host not in ('localhost', '127.0.0.1', '[::1]')):
                 raise Abort(_('file:// URLs can only refer to localhost'))
 
         self.path = path
 
         # leave the query string escaped
         for a in ('user', 'passwd', 'host', 'port',
                   'path', 'fragment'):
             v = getattr(self, a)
             if v is not None:
                 setattr(self, a, _urlunquote(v))
 
     def __repr__(self):
         attrs = []
         for a in ('scheme', 'user', 'passwd', 'host', 'port', 'path',
                   'query', 'fragment'):
             v = getattr(self, a)
             if v is not None:
                 attrs.append('%s: %r' % (a, v))
         return '<url %s>' % ', '.join(attrs)
 
     def __str__(self):
         r"""Join the URL's components back into a URL string.
 
         Examples:
 
         >>> str(url('http://user:pw@host:80/c:/bob?fo:oo#ba:ar'))
         'http://user:pw@host:80/c:/bob?fo:oo#ba:ar'
         >>> str(url('http://user:pw@host:80/?foo=bar&baz=42'))
         'http://user:pw@host:80/?foo=bar&baz=42'
         >>> str(url('http://user:pw@host:80/?foo=bar%3dbaz'))
         'http://user:pw@host:80/?foo=bar%3dbaz'
         >>> str(url('ssh://user:pw@[::1]:2200//home/joe#'))
         'ssh://user:pw@[::1]:2200//home/joe#'
         >>> str(url('http://localhost:80//'))
         'http://localhost:80//'
         >>> str(url('http://localhost:80/'))
         'http://localhost:80/'
         >>> str(url('http://localhost:80'))
         'http://localhost:80/'
         >>> str(url('bundle:foo'))
         'bundle:foo'
         >>> str(url('bundle://../foo'))
         'bundle:../foo'
         >>> str(url('path'))
         'path'
         >>> str(url('file:///tmp/foo/bar'))
         'file:///tmp/foo/bar'
         >>> str(url('file:///c:/tmp/foo/bar'))
         'file:///c:/tmp/foo/bar'
         >>> print url(r'bundle:foo\bar')
         bundle:foo\bar
         >>> print url(r'file:///D:\data\hg')
         file:///D:\data\hg
         """
         if self._localpath:
             s = self.path
             if self.scheme == 'bundle':
                 s = 'bundle:' + s
             if self.fragment:
                 s += '#' + self.fragment
             return s
 
         s = self.scheme + ':'
         if self.user or self.passwd or self.host:
             s += '//'
         elif self.scheme and (not self.path or self.path.startswith('/')
                               or hasdriveletter(self.path)):
             s += '//'
             if hasdriveletter(self.path):
                 s += '/'
         if self.user:
             s += urllib.quote(self.user, safe=self._safechars)
         if self.passwd:
             s += ':' + urllib.quote(self.passwd, safe=self._safechars)
         if self.user or self.passwd:
             s += '@'
         if self.host:
             if not (self.host.startswith('[') and self.host.endswith(']')):
                 s += urllib.quote(self.host)
             else:
                 s += self.host
         if self.port:
             s += ':' + urllib.quote(self.port)
         if self.host:
             s += '/'
         if self.path:
             # TODO: similar to the query string, we should not unescape the
             # path when we store it, the path might contain '%2f' = '/',
             # which we should *not* escape.
             s += urllib.quote(self.path, safe=self._safepchars)
         if self.query:
             # we store the query in escaped form.
             s += '?' + self.query
         if self.fragment is not None:
             s += '#' + urllib.quote(self.fragment, safe=self._safepchars)
         return s
 
     def authinfo(self):
         user, passwd = self.user, self.passwd
         try:
             self.user, self.passwd = None, None
             s = str(self)
         finally:
             self.user, self.passwd = user, passwd
         if not self.user:
             return (s, None)
         # authinfo[1] is passed to urllib2 password manager, and its
         # URIs must not contain credentials. The host is passed in the
         # URIs list because Python < 2.4.3 uses only that to search for
         # a password.
         return (s, (None, (s, self.host),
                     self.user, self.passwd or ''))
 
     def isabs(self):
         if self.scheme and self.scheme != 'file':
             return True # remote URL
         if hasdriveletter(self.path):
             return True # absolute for our purposes - can't be joined()
         if self.path.startswith(r'\\'):
             return True # Windows UNC path
         if self.path.startswith('/'):
             return True # POSIX-style
         return False
 
     def localpath(self):
         if self.scheme == 'file' or self.scheme == 'bundle':
             path = self.path or '/'
             # For Windows, we need to promote hosts containing drive
             # letters to paths with drive letters.
             if hasdriveletter(self._hostport):
                 path = self._hostport + '/' + self.path
             elif (self.host is not None and self.path
                   and not hasdriveletter(path)):
                 path = '/' + path
             return path
         return self._origpath
 
     def islocal(self):
         '''whether localpath will return something that posixfile can open'''
         return (not self.scheme or self.scheme == 'file'
                 or self.scheme == 'bundle')
 
 def hasscheme(path):
     return bool(url(path).scheme)
 
 def hasdriveletter(path):
     return path and path[1:2] == ':' and path[0:1].isalpha()
 
 def urllocalpath(path):
     return url(path, parsequery=False, parsefragment=False).localpath()
 
 def hidepassword(u):
     '''hide user credential in a url string'''
     u = url(u)
     if u.passwd:
         u.passwd = '***'
     return str(u)
 
 def removeauth(u):
     '''remove all authentication information from a url string'''
     u = url(u)
     u.user = u.passwd = None
     return str(u)
 
 def isatty(fd):
     try:
         return fd.isatty()
     except AttributeError:
         return False
 
 timecount = unitcountfn(
     (1, 1e3, _('%.0f s')),
     (100, 1, _('%.1f s')),
     (10, 1, _('%.2f s')),
     (1, 1, _('%.3f s')),
     (100, 0.001, _('%.1f ms')),
     (10, 0.001, _('%.2f ms')),
     (1, 0.001, _('%.3f ms')),
     (100, 0.000001, _('%.1f us')),
     (10, 0.000001, _('%.2f us')),
     (1, 0.000001, _('%.3f us')),
     (100, 0.000000001, _('%.1f ns')),
     (10, 0.000000001, _('%.2f ns')),
     (1, 0.000000001, _('%.3f ns')),
     )
 
 _timenesting = [0]
 
 def timed(func):
     '''Report the execution time of a function call to stderr.
 
     During development, use as a decorator when you need to measure
     the cost of a function, e.g. as follows:
 
     @util.timed
     def foo(a, b, c):
         pass
     '''
 
     def wrapper(*args, **kwargs):
         start = time.time()
         indent = 2
         _timenesting[0] += indent
         try:
             return func(*args, **kwargs)
         finally:
             elapsed = time.time() - start
             _timenesting[0] -= indent
             sys.stderr.write('%s%s: %s\n' %
                              (' ' * _timenesting[0], func.__name__,
                               timecount(elapsed)))
     return wrapper
 
 _sizeunits = (('m', 2**20), ('k', 2**10), ('g', 2**30),
               ('kb', 2**10), ('mb', 2**20), ('gb', 2**30), ('b', 1))
 
 def sizetoint(s):
     '''Convert a space specifier to a byte count.
 
     >>> sizetoint('30')
     30
     >>> sizetoint('2.2kb')
     2252
     >>> sizetoint('6M')
     6291456
     '''
     t = s.strip().lower()
     try:
         for k, u in _sizeunits:
             if t.endswith(k):
                 return int(float(t[:-len(k)]) * u)
         return int(t)
     except ValueError:
         raise error.ParseError(_("couldn't parse size: %s") % s)
 
 class hooks(object):
     '''A collection of hook functions that can be used to extend a
     function's behaviour. Hooks are called in lexicographic order,
     based on the names of their sources.'''
 
     def __init__(self):
         self._hooks = []
 
     def add(self, source, hook):
         self._hooks.append((source, hook))
 
     def __call__(self, *args):
         self._hooks.sort(key=lambda x: x[0])
         results = []
         for source, hook in self._hooks:
             results.append(hook(*args))
         return results
 
 def debugstacktrace(msg='stacktrace', skip=0, f=sys.stderr, otherf=sys.stdout):
     '''Writes a message to f (stderr) with a nicely formatted stacktrace.
     Skips the 'skip' last entries. By default it will flush stdout first.
     It can be used everywhere and do intentionally not require an ui object.
     Not be used in production code but very convenient while developing.
     '''
     if otherf:
         otherf.flush()
     f.write('%s at:\n' % msg)
     entries = [('%s:%s' % (fn, ln), func)
         for fn, ln, func, _text in traceback.extract_stack()[:-skip - 1]]
     if entries:
         fnmax = max(len(entry[0]) for entry in entries)
         for fnln, func in entries:
             f.write(' %-*s in %s\n' % (fnmax, fnln, func))
     f.flush()
 
 class dirs(object):
     '''a multiset of directory names from a dirstate or manifest'''
 
     def __init__(self, map, skip=None):
         self._dirs = {}
         addpath = self.addpath
         if safehasattr(map, 'iteritems') and skip is not None:
             for f, s in map.iteritems():
                 if s[0] != skip:
                     addpath(f)
         else:
             for f in map:
                 addpath(f)
 
     def addpath(self, path):
         dirs = self._dirs
         for base in finddirs(path):
             if base in dirs:
                 dirs[base] += 1
                 return
             dirs[base] = 1
 
     def delpath(self, path):
         dirs = self._dirs
         for base in finddirs(path):
             if dirs[base] > 1:
                 dirs[base] -= 1
                 return
             del dirs[base]
 
     def __iter__(self):
         return self._dirs.iterkeys()
 
     def __contains__(self, d):
         return d in self._dirs
 
 if safehasattr(parsers, 'dirs'):
     dirs = parsers.dirs
 
 def finddirs(path):
     pos = path.rfind('/')
     while pos != -1:
         yield path[:pos]
         pos = path.rfind('/', 0, pos)
 
 # convenient shortcut
 dst = debugstacktrace