manifest: add manifestctx.readdelta()

This adds an implementation of readdelta to the new manifestctx class and adds a
couple consumers of it. This currently appears to have some duplicate code, but
future patches cause this function to diverge when things like "shallow" are
introduced.

mercurial/changegroup.py

@@ -1,1027 +1,1027
 # changegroup.py - Mercurial changegroup manipulation functions
 #
 #  Copyright 2006 Matt Mackall <[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.
 
 from __future__ import absolute_import
 
 import os
 import struct
 import tempfile
 import weakref
 
 from .i18n import _
 from .node import (
     hex,
     nullrev,
     short,
 )
 
 from . import (
     branchmap,
     dagutil,
     discovery,
     error,
     mdiff,
     phases,
     util,
 )
 
 _CHANGEGROUPV1_DELTA_HEADER = "20s20s20s20s"
 _CHANGEGROUPV2_DELTA_HEADER = "20s20s20s20s20s"
 _CHANGEGROUPV3_DELTA_HEADER = ">20s20s20s20s20sH"
 
 def readexactly(stream, n):
     '''read n bytes from stream.read and abort if less was available'''
     s = stream.read(n)
     if len(s) < n:
         raise error.Abort(_("stream ended unexpectedly"
                            " (got %d bytes, expected %d)")
                           % (len(s), n))
     return s
 
 def getchunk(stream):
     """return the next chunk from stream as a string"""
     d = readexactly(stream, 4)
     l = struct.unpack(">l", d)[0]
     if l <= 4:
         if l:
             raise error.Abort(_("invalid chunk length %d") % l)
         return ""
     return readexactly(stream, l - 4)
 
 def chunkheader(length):
     """return a changegroup chunk header (string)"""
     return struct.pack(">l", length + 4)
 
 def closechunk():
     """return a changegroup chunk header (string) for a zero-length chunk"""
     return struct.pack(">l", 0)
 
 def combineresults(results):
     """logic to combine 0 or more addchangegroup results into one"""
     changedheads = 0
     result = 1
     for ret in results:
         # If any changegroup result is 0, return 0
         if ret == 0:
             result = 0
             break
         if ret < -1:
             changedheads += ret + 1
         elif ret > 1:
             changedheads += ret - 1
     if changedheads > 0:
         result = 1 + changedheads
     elif changedheads < 0:
         result = -1 + changedheads
     return result
 
 def writechunks(ui, chunks, filename, vfs=None):
     """Write chunks to a file and return its filename.
 
     The stream is assumed to be a bundle file.
     Existing files will not be overwritten.
     If no filename is specified, a temporary file is created.
     """
     fh = None
     cleanup = None
     try:
         if filename:
             if vfs:
                 fh = vfs.open(filename, "wb")
             else:
                 fh = open(filename, "wb")
         else:
             fd, filename = tempfile.mkstemp(prefix="hg-bundle-", suffix=".hg")
             fh = os.fdopen(fd, "wb")
         cleanup = filename
         for c in chunks:
             fh.write(c)
         cleanup = None
         return filename
     finally:
         if fh is not None:
             fh.close()
         if cleanup is not None:
             if filename and vfs:
                 vfs.unlink(cleanup)
             else:
                 os.unlink(cleanup)
 
 class cg1unpacker(object):
     """Unpacker for cg1 changegroup streams.
 
     A changegroup unpacker handles the framing of the revision data in
     the wire format. Most consumers will want to use the apply()
     method to add the changes from the changegroup to a repository.
 
     If you're forwarding a changegroup unmodified to another consumer,
     use getchunks(), which returns an iterator of changegroup
     chunks. This is mostly useful for cases where you need to know the
     data stream has ended by observing the end of the changegroup.
 
     deltachunk() is useful only if you're applying delta data. Most
     consumers should prefer apply() instead.
 
     A few other public methods exist. Those are used only for
     bundlerepo and some debug commands - their use is discouraged.
     """
     deltaheader = _CHANGEGROUPV1_DELTA_HEADER
     deltaheadersize = struct.calcsize(deltaheader)
     version = '01'
     _grouplistcount = 1 # One list of files after the manifests
 
     def __init__(self, fh, alg, extras=None):
         if alg == 'UN':
             alg = None # get more modern without breaking too much
         if not alg in util.decompressors:
             raise error.Abort(_('unknown stream compression type: %s')
                              % alg)
         if alg == 'BZ':
             alg = '_truncatedBZ'
         self._stream = util.decompressors[alg](fh)
         self._type = alg
         self.extras = extras or {}
         self.callback = None
 
     # These methods (compressed, read, seek, tell) all appear to only
     # be used by bundlerepo, but it's a little hard to tell.
     def compressed(self):
         return self._type is not None
     def read(self, l):
         return self._stream.read(l)
     def seek(self, pos):
         return self._stream.seek(pos)
     def tell(self):
         return self._stream.tell()
     def close(self):
         return self._stream.close()
 
     def _chunklength(self):
         d = readexactly(self._stream, 4)
         l = struct.unpack(">l", d)[0]
         if l <= 4:
             if l:
                 raise error.Abort(_("invalid chunk length %d") % l)
             return 0
         if self.callback:
             self.callback()
         return l - 4
 
     def changelogheader(self):
         """v10 does not have a changelog header chunk"""
         return {}
 
     def manifestheader(self):
         """v10 does not have a manifest header chunk"""
         return {}
 
     def filelogheader(self):
         """return the header of the filelogs chunk, v10 only has the filename"""
         l = self._chunklength()
         if not l:
             return {}
         fname = readexactly(self._stream, l)
         return {'filename': fname}
 
     def _deltaheader(self, headertuple, prevnode):
         node, p1, p2, cs = headertuple
         if prevnode is None:
             deltabase = p1
         else:
             deltabase = prevnode
         flags = 0
         return node, p1, p2, deltabase, cs, flags
 
     def deltachunk(self, prevnode):
         l = self._chunklength()
         if not l:
             return {}
         headerdata = readexactly(self._stream, self.deltaheadersize)
         header = struct.unpack(self.deltaheader, headerdata)
         delta = readexactly(self._stream, l - self.deltaheadersize)
         node, p1, p2, deltabase, cs, flags = self._deltaheader(header, prevnode)
         return {'node': node, 'p1': p1, 'p2': p2, 'cs': cs,
                 'deltabase': deltabase, 'delta': delta, 'flags': flags}
 
     def getchunks(self):
         """returns all the chunks contains in the bundle
 
         Used when you need to forward the binary stream to a file or another
         network API. To do so, it parse the changegroup data, otherwise it will
         block in case of sshrepo because it don't know the end of the stream.
         """
         # an empty chunkgroup is the end of the changegroup
         # a changegroup has at least 2 chunkgroups (changelog and manifest).
         # after that, changegroup versions 1 and 2 have a series of groups
         # with one group per file. changegroup 3 has a series of directory
         # manifests before the files.
         count = 0
         emptycount = 0
         while emptycount < self._grouplistcount:
             empty = True
             count += 1
             while True:
                 chunk = getchunk(self)
                 if not chunk:
                     if empty and count > 2:
                         emptycount += 1
                     break
                 empty = False
                 yield chunkheader(len(chunk))
                 pos = 0
                 while pos < len(chunk):
                     next = pos + 2**20
                     yield chunk[pos:next]
                     pos = next
             yield closechunk()
 
     def _unpackmanifests(self, repo, revmap, trp, prog, numchanges):
         # We know that we'll never have more manifests than we had
         # changesets.
         self.callback = prog(_('manifests'), numchanges)
         # no need to check for empty manifest group here:
         # if the result of the merge of 1 and 2 is the same in 3 and 4,
         # no new manifest will be created and the manifest group will
         # be empty during the pull
         self.manifestheader()
         repo.manifest.addgroup(self, revmap, trp)
         repo.ui.progress(_('manifests'), None)
         self.callback = None
 
     def apply(self, repo, srctype, url, emptyok=False,
               targetphase=phases.draft, expectedtotal=None):
         """Add the changegroup returned by source.read() to this repo.
         srctype is a string like 'push', 'pull', or 'unbundle'.  url is
         the URL of the repo where this changegroup is coming from.
 
         Return an integer summarizing the change to this repo:
         - nothing changed or no source: 0
         - more heads than before: 1+added heads (2..n)
         - fewer heads than before: -1-removed heads (-2..-n)
         - number of heads stays the same: 1
         """
         repo = repo.unfiltered()
         def csmap(x):
             repo.ui.debug("add changeset %s\n" % short(x))
             return len(cl)
 
         def revmap(x):
             return cl.rev(x)
 
         changesets = files = revisions = 0
 
         try:
             with repo.transaction("\n".join([srctype,
                                              util.hidepassword(url)])) as tr:
                 # The transaction could have been created before and already
                 # carries source information. In this case we use the top
                 # level data. We overwrite the argument because we need to use
                 # the top level value (if they exist) in this function.
                 srctype = tr.hookargs.setdefault('source', srctype)
                 url = tr.hookargs.setdefault('url', url)
                 repo.hook('prechangegroup', throw=True, **tr.hookargs)
 
                 # write changelog data to temp files so concurrent readers
                 # will not see an inconsistent view
                 cl = repo.changelog
                 cl.delayupdate(tr)
                 oldheads = cl.heads()
 
                 trp = weakref.proxy(tr)
                 # pull off the changeset group
                 repo.ui.status(_("adding changesets\n"))
                 clstart = len(cl)
                 class prog(object):
                     def __init__(self, step, total):
                         self._step = step
                         self._total = total
                         self._count = 1
                     def __call__(self):
                         repo.ui.progress(self._step, self._count,
                                          unit=_('chunks'), total=self._total)
                         self._count += 1
                 self.callback = prog(_('changesets'), expectedtotal)
 
                 efiles = set()
                 def onchangelog(cl, node):
                     efiles.update(cl.readfiles(node))
 
                 self.changelogheader()
                 srccontent = cl.addgroup(self, csmap, trp,
                                          addrevisioncb=onchangelog)
                 efiles = len(efiles)
 
                 if not (srccontent or emptyok):
                     raise error.Abort(_("received changelog group is empty"))
                 clend = len(cl)
                 changesets = clend - clstart
                 repo.ui.progress(_('changesets'), None)
                 self.callback = None
 
                 # pull off the manifest group
                 repo.ui.status(_("adding manifests\n"))
                 self._unpackmanifests(repo, revmap, trp, prog, changesets)
 
                 needfiles = {}
                 if repo.ui.configbool('server', 'validate', default=False):
                     # validate incoming csets have their manifests
                     for cset in xrange(clstart, clend):
                         mfnode = repo.changelog.read(
                             repo.changelog.node(cset))[0]
-                        mfest = repo.manifest.readdelta(mfnode)
+                        mfest = repo.manifestlog[mfnode].readdelta()
                         # store file nodes we must see
                         for f, n in mfest.iteritems():
                             needfiles.setdefault(f, set()).add(n)
 
                 # process the files
                 repo.ui.status(_("adding file changes\n"))
                 newrevs, newfiles = _addchangegroupfiles(
                     repo, self, revmap, trp, efiles, needfiles)
                 revisions += newrevs
                 files += newfiles
 
                 dh = 0
                 if oldheads:
                     heads = cl.heads()
                     dh = len(heads) - len(oldheads)
                     for h in heads:
                         if h not in oldheads and repo[h].closesbranch():
                             dh -= 1
                 htext = ""
                 if dh:
                     htext = _(" (%+d heads)") % dh
 
                 repo.ui.status(_("added %d changesets"
                                  " with %d changes to %d files%s\n")
                                  % (changesets, revisions, files, htext))
                 repo.invalidatevolatilesets()
 
                 if changesets > 0:
                     if 'node' not in tr.hookargs:
                         tr.hookargs['node'] = hex(cl.node(clstart))
                         tr.hookargs['node_last'] = hex(cl.node(clend - 1))
                         hookargs = dict(tr.hookargs)
                     else:
                         hookargs = dict(tr.hookargs)
                         hookargs['node'] = hex(cl.node(clstart))
                         hookargs['node_last'] = hex(cl.node(clend - 1))
                     repo.hook('pretxnchangegroup', throw=True, **hookargs)
 
                 added = [cl.node(r) for r in xrange(clstart, clend)]
                 publishing = repo.publishing()
                 if srctype in ('push', 'serve'):
                     # Old servers can not push the boundary themselves.
                     # New servers won't push the boundary if changeset already
                     # exists locally as secret
                     #
                     # We should not use added here but the list of all change in
                     # the bundle
                     if publishing:
                         phases.advanceboundary(repo, tr, phases.public,
                                                srccontent)
                     else:
                         # Those changesets have been pushed from the
                         # outside, their phases are going to be pushed
                         # alongside. Therefor `targetphase` is
                         # ignored.
                         phases.advanceboundary(repo, tr, phases.draft,
                                                srccontent)
                         phases.retractboundary(repo, tr, phases.draft, added)
                 elif srctype != 'strip':
                     # publishing only alter behavior during push
                     #
                     # strip should not touch boundary at all
                     phases.retractboundary(repo, tr, targetphase, added)
 
                 if changesets > 0:
                     if srctype != 'strip':
                         # During strip, branchcache is invalid but
                         # coming call to `destroyed` will repair it.
                         # In other case we can safely update cache on
                         # disk.
                         repo.ui.debug('updating the branch cache\n')
                         branchmap.updatecache(repo.filtered('served'))
 
                     def runhooks():
                         # These hooks run when the lock releases, not when the
                         # transaction closes. So it's possible for the changelog
                         # to have changed since we last saw it.
                         if clstart >= len(repo):
                             return
 
                         repo.hook("changegroup", **hookargs)
 
                         for n in added:
                             args = hookargs.copy()
                             args['node'] = hex(n)
                             del args['node_last']
                             repo.hook("incoming", **args)
 
                         newheads = [h for h in repo.heads()
                                     if h not in oldheads]
                         repo.ui.log("incoming",
                                     "%s incoming changes - new heads: %s\n",
                                     len(added),
                                     ', '.join([hex(c[:6]) for c in newheads]))
 
                     tr.addpostclose('changegroup-runhooks-%020i' % clstart,
                                     lambda tr: repo._afterlock(runhooks))
         finally:
             repo.ui.flush()
         # never return 0 here:
         if dh < 0:
             return dh - 1
         else:
             return dh + 1
 
 class cg2unpacker(cg1unpacker):
     """Unpacker for cg2 streams.
 
     cg2 streams add support for generaldelta, so the delta header
     format is slightly different. All other features about the data
     remain the same.
     """
     deltaheader = _CHANGEGROUPV2_DELTA_HEADER
     deltaheadersize = struct.calcsize(deltaheader)
     version = '02'
 
     def _deltaheader(self, headertuple, prevnode):
         node, p1, p2, deltabase, cs = headertuple
         flags = 0
         return node, p1, p2, deltabase, cs, flags
 
 class cg3unpacker(cg2unpacker):
     """Unpacker for cg3 streams.
 
     cg3 streams add support for exchanging treemanifests and revlog
     flags. It adds the revlog flags to the delta header and an empty chunk
     separating manifests and files.
     """
     deltaheader = _CHANGEGROUPV3_DELTA_HEADER
     deltaheadersize = struct.calcsize(deltaheader)
     version = '03'
     _grouplistcount = 2 # One list of manifests and one list of files
 
     def _deltaheader(self, headertuple, prevnode):
         node, p1, p2, deltabase, cs, flags = headertuple
         return node, p1, p2, deltabase, cs, flags
 
     def _unpackmanifests(self, repo, revmap, trp, prog, numchanges):
         super(cg3unpacker, self)._unpackmanifests(repo, revmap, trp, prog,
                                                   numchanges)
         for chunkdata in iter(self.filelogheader, {}):
             # If we get here, there are directory manifests in the changegroup
             d = chunkdata["filename"]
             repo.ui.debug("adding %s revisions\n" % d)
             dirlog = repo.manifest.dirlog(d)
             if not dirlog.addgroup(self, revmap, trp):
                 raise error.Abort(_("received dir revlog group is empty"))
 
 class headerlessfixup(object):
     def __init__(self, fh, h):
         self._h = h
         self._fh = fh
     def read(self, n):
         if self._h:
             d, self._h = self._h[:n], self._h[n:]
             if len(d) < n:
                 d += readexactly(self._fh, n - len(d))
             return d
         return readexactly(self._fh, n)
 
 class cg1packer(object):
     deltaheader = _CHANGEGROUPV1_DELTA_HEADER
     version = '01'
     def __init__(self, repo, bundlecaps=None):
         """Given a source repo, construct a bundler.
 
         bundlecaps is optional and can be used to specify the set of
         capabilities which can be used to build the bundle.
         """
         # Set of capabilities we can use to build the bundle.
         if bundlecaps is None:
             bundlecaps = set()
         self._bundlecaps = bundlecaps
         # experimental config: bundle.reorder
         reorder = repo.ui.config('bundle', 'reorder', 'auto')
         if reorder == 'auto':
             reorder = None
         else:
             reorder = util.parsebool(reorder)
         self._repo = repo
         self._reorder = reorder
         self._progress = repo.ui.progress
         if self._repo.ui.verbose and not self._repo.ui.debugflag:
             self._verbosenote = self._repo.ui.note
         else:
             self._verbosenote = lambda s: None
 
     def close(self):
         return closechunk()
 
     def fileheader(self, fname):
         return chunkheader(len(fname)) + fname
 
     # Extracted both for clarity and for overriding in extensions.
     def _sortgroup(self, revlog, nodelist, lookup):
         """Sort nodes for change group and turn them into revnums."""
         # for generaldelta revlogs, we linearize the revs; this will both be
         # much quicker and generate a much smaller bundle
         if (revlog._generaldelta and self._reorder is None) or self._reorder:
             dag = dagutil.revlogdag(revlog)
             return dag.linearize(set(revlog.rev(n) for n in nodelist))
         else:
             return sorted([revlog.rev(n) for n in nodelist])
 
     def group(self, nodelist, revlog, lookup, units=None):
         """Calculate a delta group, yielding a sequence of changegroup chunks
         (strings).
 
         Given a list of changeset revs, return a set of deltas and
         metadata corresponding to nodes. The first delta is
         first parent(nodelist[0]) -> nodelist[0], the receiver is
         guaranteed to have this parent as it has all history before
         these changesets. In the case firstparent is nullrev the
         changegroup starts with a full revision.
 
         If units is not None, progress detail will be generated, units specifies
         the type of revlog that is touched (changelog, manifest, etc.).
         """
         # if we don't have any revisions touched by these changesets, bail
         if len(nodelist) == 0:
             yield self.close()
             return
 
         revs = self._sortgroup(revlog, nodelist, lookup)
 
         # add the parent of the first rev
         p = revlog.parentrevs(revs[0])[0]
         revs.insert(0, p)
 
         # build deltas
         total = len(revs) - 1
         msgbundling = _('bundling')
         for r in xrange(len(revs) - 1):
             if units is not None:
                 self._progress(msgbundling, r + 1, unit=units, total=total)
             prev, curr = revs[r], revs[r + 1]
             linknode = lookup(revlog.node(curr))
             for c in self.revchunk(revlog, curr, prev, linknode):
                 yield c
 
         if units is not None:
             self._progress(msgbundling, None)
         yield self.close()
 
     # filter any nodes that claim to be part of the known set
     def prune(self, revlog, missing, commonrevs):
         rr, rl = revlog.rev, revlog.linkrev
         return [n for n in missing if rl(rr(n)) not in commonrevs]
 
     def _packmanifests(self, dir, mfnodes, lookuplinknode):
         """Pack flat manifests into a changegroup stream."""
         assert not dir
         for chunk in self.group(mfnodes, self._repo.manifest,
                                 lookuplinknode, units=_('manifests')):
             yield chunk
 
     def _manifestsdone(self):
         return ''
 
     def generate(self, commonrevs, clnodes, fastpathlinkrev, source):
         '''yield a sequence of changegroup chunks (strings)'''
         repo = self._repo
         cl = repo.changelog
 
         clrevorder = {}
         mfs = {} # needed manifests
         fnodes = {} # needed file nodes
         changedfiles = set()
 
         # Callback for the changelog, used to collect changed files and manifest
         # nodes.
         # Returns the linkrev node (identity in the changelog case).
         def lookupcl(x):
             c = cl.read(x)
             clrevorder[x] = len(clrevorder)
             n = c[0]
             # record the first changeset introducing this manifest version
             mfs.setdefault(n, x)
             # Record a complete list of potentially-changed files in
             # this manifest.
             changedfiles.update(c[3])
             return x
 
         self._verbosenote(_('uncompressed size of bundle content:\n'))
         size = 0
         for chunk in self.group(clnodes, cl, lookupcl, units=_('changesets')):
             size += len(chunk)
             yield chunk
         self._verbosenote(_('%8.i (changelog)\n') % size)
 
         # We need to make sure that the linkrev in the changegroup refers to
         # the first changeset that introduced the manifest or file revision.
         # The fastpath is usually safer than the slowpath, because the filelogs
         # are walked in revlog order.
         #
         # When taking the slowpath with reorder=None and the manifest revlog
         # uses generaldelta, the manifest may be walked in the "wrong" order.
         # Without 'clrevorder', we would get an incorrect linkrev (see fix in
         # cc0ff93d0c0c).
         #
         # When taking the fastpath, we are only vulnerable to reordering
         # of the changelog itself. The changelog never uses generaldelta, so
         # it is only reordered when reorder=True. To handle this case, we
         # simply take the slowpath, which already has the 'clrevorder' logic.
         # This was also fixed in cc0ff93d0c0c.
         fastpathlinkrev = fastpathlinkrev and not self._reorder
         # Treemanifests don't work correctly with fastpathlinkrev
         # either, because we don't discover which directory nodes to
         # send along with files. This could probably be fixed.
         fastpathlinkrev = fastpathlinkrev and (
             'treemanifest' not in repo.requirements)
 
         for chunk in self.generatemanifests(commonrevs, clrevorder,
                 fastpathlinkrev, mfs, fnodes):
             yield chunk
         mfs.clear()
         clrevs = set(cl.rev(x) for x in clnodes)
 
         if not fastpathlinkrev:
             def linknodes(unused, fname):
                 return fnodes.get(fname, {})
         else:
             cln = cl.node
             def linknodes(filerevlog, fname):
                 llr = filerevlog.linkrev
                 fln = filerevlog.node
                 revs = ((r, llr(r)) for r in filerevlog)
                 return dict((fln(r), cln(lr)) for r, lr in revs if lr in clrevs)
 
         for chunk in self.generatefiles(changedfiles, linknodes, commonrevs,
                                         source):
             yield chunk
 
         yield self.close()
 
         if clnodes:
             repo.hook('outgoing', node=hex(clnodes[0]), source=source)
 
     def generatemanifests(self, commonrevs, clrevorder, fastpathlinkrev, mfs,
                           fnodes):
         repo = self._repo
         dirlog = repo.manifest.dirlog
         tmfnodes = {'': mfs}
 
         # Callback for the manifest, used to collect linkrevs for filelog
         # revisions.
         # Returns the linkrev node (collected in lookupcl).
         def makelookupmflinknode(dir):
             if fastpathlinkrev:
                 assert not dir
                 return mfs.__getitem__
 
             def lookupmflinknode(x):
                 """Callback for looking up the linknode for manifests.
 
                 Returns the linkrev node for the specified manifest.
 
                 SIDE EFFECT:
 
                 1) fclnodes gets populated with the list of relevant
                    file nodes if we're not using fastpathlinkrev
                 2) When treemanifests are in use, collects treemanifest nodes
                    to send
 
                 Note that this means manifests must be completely sent to
                 the client before you can trust the list of files and
                 treemanifests to send.
                 """
                 clnode = tmfnodes[dir][x]
                 mdata = dirlog(dir).readshallowfast(x)
                 for p, n, fl in mdata.iterentries():
                     if fl == 't': # subdirectory manifest
                         subdir = dir + p + '/'
                         tmfclnodes = tmfnodes.setdefault(subdir, {})
                         tmfclnode = tmfclnodes.setdefault(n, clnode)
                         if clrevorder[clnode] < clrevorder[tmfclnode]:
                             tmfclnodes[n] = clnode
                     else:
                         f = dir + p
                         fclnodes = fnodes.setdefault(f, {})
                         fclnode = fclnodes.setdefault(n, clnode)
                         if clrevorder[clnode] < clrevorder[fclnode]:
                             fclnodes[n] = clnode
                 return clnode
             return lookupmflinknode
 
         size = 0
         while tmfnodes:
             dir = min(tmfnodes)
             nodes = tmfnodes[dir]
             prunednodes = self.prune(dirlog(dir), nodes, commonrevs)
             if not dir or prunednodes:
                 for x in self._packmanifests(dir, prunednodes,
                                              makelookupmflinknode(dir)):
                     size += len(x)
                     yield x
             del tmfnodes[dir]
         self._verbosenote(_('%8.i (manifests)\n') % size)
         yield self._manifestsdone()
 
     # The 'source' parameter is useful for extensions
     def generatefiles(self, changedfiles, linknodes, commonrevs, source):
         repo = self._repo
         progress = self._progress
         msgbundling = _('bundling')
 
         total = len(changedfiles)
         # for progress output
         msgfiles = _('files')
         for i, fname in enumerate(sorted(changedfiles)):
             filerevlog = repo.file(fname)
             if not filerevlog:
                 raise error.Abort(_("empty or missing revlog for %s") % fname)
 
             linkrevnodes = linknodes(filerevlog, fname)
             # Lookup for filenodes, we collected the linkrev nodes above in the
             # fastpath case and with lookupmf in the slowpath case.
             def lookupfilelog(x):
                 return linkrevnodes[x]
 
             filenodes = self.prune(filerevlog, linkrevnodes, commonrevs)
             if filenodes:
                 progress(msgbundling, i + 1, item=fname, unit=msgfiles,
                          total=total)
                 h = self.fileheader(fname)
                 size = len(h)
                 yield h
                 for chunk in self.group(filenodes, filerevlog, lookupfilelog):
                     size += len(chunk)
                     yield chunk
                 self._verbosenote(_('%8.i  %s\n') % (size, fname))
         progress(msgbundling, None)
 
     def deltaparent(self, revlog, rev, p1, p2, prev):
         return prev
 
     def revchunk(self, revlog, rev, prev, linknode):
         node = revlog.node(rev)
         p1, p2 = revlog.parentrevs(rev)
         base = self.deltaparent(revlog, rev, p1, p2, prev)
 
         prefix = ''
         if revlog.iscensored(base) or revlog.iscensored(rev):
             try:
                 delta = revlog.revision(node)
             except error.CensoredNodeError as e:
                 delta = e.tombstone
             if base == nullrev:
                 prefix = mdiff.trivialdiffheader(len(delta))
             else:
                 baselen = revlog.rawsize(base)
                 prefix = mdiff.replacediffheader(baselen, len(delta))
         elif base == nullrev:
             delta = revlog.revision(node)
             prefix = mdiff.trivialdiffheader(len(delta))
         else:
             delta = revlog.revdiff(base, rev)
         p1n, p2n = revlog.parents(node)
         basenode = revlog.node(base)
         flags = revlog.flags(rev)
         meta = self.builddeltaheader(node, p1n, p2n, basenode, linknode, flags)
         meta += prefix
         l = len(meta) + len(delta)
         yield chunkheader(l)
         yield meta
         yield delta
     def builddeltaheader(self, node, p1n, p2n, basenode, linknode, flags):
         # do nothing with basenode, it is implicitly the previous one in HG10
         # do nothing with flags, it is implicitly 0 for cg1 and cg2
         return struct.pack(self.deltaheader, node, p1n, p2n, linknode)
 
 class cg2packer(cg1packer):
     version = '02'
     deltaheader = _CHANGEGROUPV2_DELTA_HEADER
 
     def __init__(self, repo, bundlecaps=None):
         super(cg2packer, self).__init__(repo, bundlecaps)
         if self._reorder is None:
             # Since generaldelta is directly supported by cg2, reordering
             # generally doesn't help, so we disable it by default (treating
             # bundle.reorder=auto just like bundle.reorder=False).
             self._reorder = False
 
     def deltaparent(self, revlog, rev, p1, p2, prev):
         dp = revlog.deltaparent(rev)
         # avoid storing full revisions; pick prev in those cases
         # also pick prev when we can't be sure remote has dp
         if dp == nullrev or (dp != p1 and dp != p2 and dp != prev):
             return prev
         return dp
 
     def builddeltaheader(self, node, p1n, p2n, basenode, linknode, flags):
         # Do nothing with flags, it is implicitly 0 in cg1 and cg2
         return struct.pack(self.deltaheader, node, p1n, p2n, basenode, linknode)
 
 class cg3packer(cg2packer):
     version = '03'
     deltaheader = _CHANGEGROUPV3_DELTA_HEADER
 
     def _packmanifests(self, dir, mfnodes, lookuplinknode):
         if dir:
             yield self.fileheader(dir)
         for chunk in self.group(mfnodes, self._repo.manifest.dirlog(dir),
                                 lookuplinknode, units=_('manifests')):
             yield chunk
 
     def _manifestsdone(self):
         return self.close()
 
     def builddeltaheader(self, node, p1n, p2n, basenode, linknode, flags):
         return struct.pack(
             self.deltaheader, node, p1n, p2n, basenode, linknode, flags)
 
 _packermap = {'01': (cg1packer, cg1unpacker),
              # cg2 adds support for exchanging generaldelta
              '02': (cg2packer, cg2unpacker),
              # cg3 adds support for exchanging revlog flags and treemanifests
              '03': (cg3packer, cg3unpacker),
 }
 
 def allsupportedversions(ui):
     versions = set(_packermap.keys())
     versions.discard('03')
     if (ui.configbool('experimental', 'changegroup3') or
         ui.configbool('experimental', 'treemanifest')):
         versions.add('03')
     return versions
 
 # Changegroup versions that can be applied to the repo
 def supportedincomingversions(repo):
     versions = allsupportedversions(repo.ui)
     if 'treemanifest' in repo.requirements:
         versions.add('03')
     return versions
 
 # Changegroup versions that can be created from the repo
 def supportedoutgoingversions(repo):
     versions = allsupportedversions(repo.ui)
     if 'treemanifest' in repo.requirements:
         # Versions 01 and 02 support only flat manifests and it's just too
         # expensive to convert between the flat manifest and tree manifest on
         # the fly. Since tree manifests are hashed differently, all of history
         # would have to be converted. Instead, we simply don't even pretend to
         # support versions 01 and 02.
         versions.discard('01')
         versions.discard('02')
         versions.add('03')
     return versions
 
 def safeversion(repo):
     # Finds the smallest version that it's safe to assume clients of the repo
     # will support. For example, all hg versions that support generaldelta also
     # support changegroup 02.
     versions = supportedoutgoingversions(repo)
     if 'generaldelta' in repo.requirements:
         versions.discard('01')
     assert versions
     return min(versions)
 
 def getbundler(version, repo, bundlecaps=None):
     assert version in supportedoutgoingversions(repo)
     return _packermap[version][0](repo, bundlecaps)
 
 def getunbundler(version, fh, alg, extras=None):
     return _packermap[version][1](fh, alg, extras=extras)
 
 def _changegroupinfo(repo, nodes, source):
     if repo.ui.verbose or source == 'bundle':
         repo.ui.status(_("%d changesets found\n") % len(nodes))
     if repo.ui.debugflag:
         repo.ui.debug("list of changesets:\n")
         for node in nodes:
             repo.ui.debug("%s\n" % hex(node))
 
 def getsubsetraw(repo, outgoing, bundler, source, fastpath=False):
     repo = repo.unfiltered()
     commonrevs = outgoing.common
     csets = outgoing.missing
     heads = outgoing.missingheads
     # We go through the fast path if we get told to, or if all (unfiltered
     # heads have been requested (since we then know there all linkrevs will
     # be pulled by the client).
     heads.sort()
     fastpathlinkrev = fastpath or (
             repo.filtername is None and heads == sorted(repo.heads()))
 
     repo.hook('preoutgoing', throw=True, source=source)
     _changegroupinfo(repo, csets, source)
     return bundler.generate(commonrevs, csets, fastpathlinkrev, source)
 
 def getsubset(repo, outgoing, bundler, source, fastpath=False):
     gengroup = getsubsetraw(repo, outgoing, bundler, source, fastpath)
     return getunbundler(bundler.version, util.chunkbuffer(gengroup), None,
                         {'clcount': len(outgoing.missing)})
 
 def changegroupsubset(repo, roots, heads, source, version='01'):
     """Compute a changegroup consisting of all the nodes that are
     descendants of any of the roots and ancestors of any of the heads.
     Return a chunkbuffer object whose read() method will return
     successive changegroup chunks.
 
     It is fairly complex as determining which filenodes and which
     manifest nodes need to be included for the changeset to be complete
     is non-trivial.
 
     Another wrinkle is doing the reverse, figuring out which changeset in
     the changegroup a particular filenode or manifestnode belongs to.
     """
     outgoing = discovery.outgoing(repo, missingroots=roots, missingheads=heads)
     bundler = getbundler(version, repo)
     return getsubset(repo, outgoing, bundler, source)
 
 def getlocalchangegroupraw(repo, source, outgoing, bundlecaps=None,
                            version='01'):
     """Like getbundle, but taking a discovery.outgoing as an argument.
 
     This is only implemented for local repos and reuses potentially
     precomputed sets in outgoing. Returns a raw changegroup generator."""
     if not outgoing.missing:
         return None
     bundler = getbundler(version, repo, bundlecaps)
     return getsubsetraw(repo, outgoing, bundler, source)
 
 def getlocalchangegroup(repo, source, outgoing, bundlecaps=None,
                         version='01'):
     """Like getbundle, but taking a discovery.outgoing as an argument.
 
     This is only implemented for local repos and reuses potentially
     precomputed sets in outgoing."""
     if not outgoing.missing:
         return None
     bundler = getbundler(version, repo, bundlecaps)
     return getsubset(repo, outgoing, bundler, source)
 
 def getchangegroup(repo, source, outgoing, bundlecaps=None,
                    version='01'):
     """Like changegroupsubset, but returns the set difference between the
     ancestors of heads and the ancestors common.
 
     If heads is None, use the local heads. If common is None, use [nullid].
 
     The nodes in common might not all be known locally due to the way the
     current discovery protocol works.
     """
     return getlocalchangegroup(repo, source, outgoing, bundlecaps=bundlecaps,
                                version=version)
 
 def changegroup(repo, basenodes, source):
     # to avoid a race we use changegroupsubset() (issue1320)
     return changegroupsubset(repo, basenodes, repo.heads(), source)
 
 def _addchangegroupfiles(repo, source, revmap, trp, expectedfiles, needfiles):
     revisions = 0
     files = 0
     for chunkdata in iter(source.filelogheader, {}):
         files += 1
         f = chunkdata["filename"]
         repo.ui.debug("adding %s revisions\n" % f)
         repo.ui.progress(_('files'), files, unit=_('files'),
                          total=expectedfiles)
         fl = repo.file(f)
         o = len(fl)
         try:
             if not fl.addgroup(source, revmap, trp):
                 raise error.Abort(_("received file revlog group is empty"))
         except error.CensoredBaseError as e:
             raise error.Abort(_("received delta base is censored: %s") % e)
         revisions += len(fl) - o
         if f in needfiles:
             needs = needfiles[f]
             for new in xrange(o, len(fl)):
                 n = fl.node(new)
                 if n in needs:
                     needs.remove(n)
                 else:
                     raise error.Abort(
                         _("received spurious file revlog entry"))
             if not needs:
                 del needfiles[f]
     repo.ui.progress(_('files'), None)
 
     for f, needs in needfiles.iteritems():
         fl = repo.file(f)
         for n in needs:
             try:
                 fl.rev(n)
             except error.LookupError:
                 raise error.Abort(
                     _('missing file data for %s:%s - run hg verify') %
                     (f, hex(n)))
 
     return revisions, files

mercurial/context.py

@@ -1,1984 +1,1985
 # context.py - changeset and file context objects for mercurial
 #
 # Copyright 2006, 2007 Matt Mackall <[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.
 
 from __future__ import absolute_import
 
 import errno
 import os
 import re
 import stat
 
 from .i18n import _
 from .node import (
     bin,
     hex,
     nullid,
     nullrev,
     short,
     wdirid,
 )
 from . import (
     encoding,
     error,
     fileset,
     match as matchmod,
     mdiff,
     obsolete as obsmod,
     patch,
     phases,
     repoview,
     revlog,
     scmutil,
     subrepo,
     util,
 )
 
 propertycache = util.propertycache
 
 # Phony node value to stand-in for new files in some uses of
 # manifests. Manifests support 21-byte hashes for nodes which are
 # dirty in the working copy.
 _newnode = '!' * 21
 
 nonascii = re.compile(r'[^\x21-\x7f]').search
 
 class basectx(object):
     """A basectx object represents the common logic for its children:
     changectx: read-only context that is already present in the repo,
     workingctx: a context that represents the working directory and can
                 be committed,
     memctx: a context that represents changes in-memory and can also
             be committed."""
     def __new__(cls, repo, changeid='', *args, **kwargs):
         if isinstance(changeid, basectx):
             return changeid
 
         o = super(basectx, cls).__new__(cls)
 
         o._repo = repo
         o._rev = nullrev
         o._node = nullid
 
         return o
 
     def __str__(self):
         return short(self.node())
 
     def __int__(self):
         return self.rev()
 
     def __repr__(self):
         return "<%s %s>" % (type(self).__name__, str(self))
 
     def __eq__(self, other):
         try:
             return type(self) == type(other) and self._rev == other._rev
         except AttributeError:
             return False
 
     def __ne__(self, other):
         return not (self == other)
 
     def __contains__(self, key):
         return key in self._manifest
 
     def __getitem__(self, key):
         return self.filectx(key)
 
     def __iter__(self):
         return iter(self._manifest)
 
     def _manifestmatches(self, match, s):
         """generate a new manifest filtered by the match argument
 
         This method is for internal use only and mainly exists to provide an
         object oriented way for other contexts to customize the manifest
         generation.
         """
         return self.manifest().matches(match)
 
     def _matchstatus(self, other, match):
         """return match.always if match is none
 
         This internal method provides a way for child objects to override the
         match operator.
         """
         return match or matchmod.always(self._repo.root, self._repo.getcwd())
 
     def _buildstatus(self, other, s, match, listignored, listclean,
                      listunknown):
         """build a status with respect to another context"""
         # Load earliest manifest first for caching reasons. More specifically,
         # if you have revisions 1000 and 1001, 1001 is probably stored as a
         # delta against 1000. Thus, if you read 1000 first, we'll reconstruct
         # 1000 and cache it so that when you read 1001, we just need to apply a
         # delta to what's in the cache. So that's one full reconstruction + one
         # delta application.
         if self.rev() is not None and self.rev() < other.rev():
             self.manifest()
         mf1 = other._manifestmatches(match, s)
         mf2 = self._manifestmatches(match, s)
 
         modified, added = [], []
         removed = []
         clean = []
         deleted, unknown, ignored = s.deleted, s.unknown, s.ignored
         deletedset = set(deleted)
         d = mf1.diff(mf2, clean=listclean)
         for fn, value in d.iteritems():
             if fn in deletedset:
                 continue
             if value is None:
                 clean.append(fn)
                 continue
             (node1, flag1), (node2, flag2) = value
             if node1 is None:
                 added.append(fn)
             elif node2 is None:
                 removed.append(fn)
             elif flag1 != flag2:
                 modified.append(fn)
             elif node2 != _newnode:
                 # When comparing files between two commits, we save time by
                 # not comparing the file contents when the nodeids differ.
                 # Note that this means we incorrectly report a reverted change
                 # to a file as a modification.
                 modified.append(fn)
             elif self[fn].cmp(other[fn]):
                 modified.append(fn)
             else:
                 clean.append(fn)
 
         if removed:
             # need to filter files if they are already reported as removed
             unknown = [fn for fn in unknown if fn not in mf1]
             ignored = [fn for fn in ignored if fn not in mf1]
             # if they're deleted, don't report them as removed
             removed = [fn for fn in removed if fn not in deletedset]
 
         return scmutil.status(modified, added, removed, deleted, unknown,
                               ignored, clean)
 
     @propertycache
     def substate(self):
         return subrepo.state(self, self._repo.ui)
 
     def subrev(self, subpath):
         return self.substate[subpath][1]
 
     def rev(self):
         return self._rev
     def node(self):
         return self._node
     def hex(self):
         return hex(self.node())
     def manifest(self):
         return self._manifest
     def repo(self):
         return self._repo
     def phasestr(self):
         return phases.phasenames[self.phase()]
     def mutable(self):
         return self.phase() > phases.public
 
     def getfileset(self, expr):
         return fileset.getfileset(self, expr)
 
     def obsolete(self):
         """True if the changeset is obsolete"""
         return self.rev() in obsmod.getrevs(self._repo, 'obsolete')
 
     def extinct(self):
         """True if the changeset is extinct"""
         return self.rev() in obsmod.getrevs(self._repo, 'extinct')
 
     def unstable(self):
         """True if the changeset is not obsolete but it's ancestor are"""
         return self.rev() in obsmod.getrevs(self._repo, 'unstable')
 
     def bumped(self):
         """True if the changeset try to be a successor of a public changeset
 
         Only non-public and non-obsolete changesets may be bumped.
         """
         return self.rev() in obsmod.getrevs(self._repo, 'bumped')
 
     def divergent(self):
         """Is a successors of a changeset with multiple possible successors set
 
         Only non-public and non-obsolete changesets may be divergent.
         """
         return self.rev() in obsmod.getrevs(self._repo, 'divergent')
 
     def troubled(self):
         """True if the changeset is either unstable, bumped or divergent"""
         return self.unstable() or self.bumped() or self.divergent()
 
     def troubles(self):
         """return the list of troubles affecting this changesets.
 
         Troubles are returned as strings. possible values are:
         - unstable,
         - bumped,
         - divergent.
         """
         troubles = []
         if self.unstable():
             troubles.append('unstable')
         if self.bumped():
             troubles.append('bumped')
         if self.divergent():
             troubles.append('divergent')
         return troubles
 
     def parents(self):
         """return contexts for each parent changeset"""
         return self._parents
 
     def p1(self):
         return self._parents[0]
 
     def p2(self):
         parents = self._parents
         if len(parents) == 2:
             return parents[1]
         return changectx(self._repo, nullrev)
 
     def _fileinfo(self, path):
         if '_manifest' in self.__dict__:
             try:
                 return self._manifest[path], self._manifest.flags(path)
             except KeyError:
                 raise error.ManifestLookupError(self._node, path,
                                                 _('not found in manifest'))
         if '_manifestdelta' in self.__dict__ or path in self.files():
             if path in self._manifestdelta:
                 return (self._manifestdelta[path],
                         self._manifestdelta.flags(path))
         node, flag = self._repo.manifest.find(self._changeset.manifest, path)
         if not node:
             raise error.ManifestLookupError(self._node, path,
                                             _('not found in manifest'))
 
         return node, flag
 
     def filenode(self, path):
         return self._fileinfo(path)[0]
 
     def flags(self, path):
         try:
             return self._fileinfo(path)[1]
         except error.LookupError:
             return ''
 
     def sub(self, path, allowcreate=True):
         '''return a subrepo for the stored revision of path, never wdir()'''
         return subrepo.subrepo(self, path, allowcreate=allowcreate)
 
     def nullsub(self, path, pctx):
         return subrepo.nullsubrepo(self, path, pctx)
 
     def workingsub(self, path):
         '''return a subrepo for the stored revision, or wdir if this is a wdir
         context.
         '''
         return subrepo.subrepo(self, path, allowwdir=True)
 
     def match(self, pats=[], include=None, exclude=None, default='glob',
               listsubrepos=False, badfn=None):
         r = self._repo
         return matchmod.match(r.root, r.getcwd(), pats,
                               include, exclude, default,
                               auditor=r.nofsauditor, ctx=self,
                               listsubrepos=listsubrepos, badfn=badfn)
 
     def diff(self, ctx2=None, match=None, **opts):
         """Returns a diff generator for the given contexts and matcher"""
         if ctx2 is None:
             ctx2 = self.p1()
         if ctx2 is not None:
             ctx2 = self._repo[ctx2]
         diffopts = patch.diffopts(self._repo.ui, opts)
         return patch.diff(self._repo, ctx2, self, match=match, opts=diffopts)
 
     def dirs(self):
         return self._manifest.dirs()
 
     def hasdir(self, dir):
         return self._manifest.hasdir(dir)
 
     def dirty(self, missing=False, merge=True, branch=True):
         return False
 
     def status(self, other=None, match=None, listignored=False,
                listclean=False, listunknown=False, listsubrepos=False):
         """return status of files between two nodes or node and working
         directory.
 
         If other is None, compare this node with working directory.
 
         returns (modified, added, removed, deleted, unknown, ignored, clean)
         """
 
         ctx1 = self
         ctx2 = self._repo[other]
 
         # This next code block is, admittedly, fragile logic that tests for
         # reversing the contexts and wouldn't need to exist if it weren't for
         # the fast (and common) code path of comparing the working directory
         # with its first parent.
         #
         # What we're aiming for here is the ability to call:
         #
         # workingctx.status(parentctx)
         #
         # If we always built the manifest for each context and compared those,
         # then we'd be done. But the special case of the above call means we
         # just copy the manifest of the parent.
         reversed = False
         if (not isinstance(ctx1, changectx)
             and isinstance(ctx2, changectx)):
             reversed = True
             ctx1, ctx2 = ctx2, ctx1
 
         match = ctx2._matchstatus(ctx1, match)
         r = scmutil.status([], [], [], [], [], [], [])
         r = ctx2._buildstatus(ctx1, r, match, listignored, listclean,
                               listunknown)
 
         if reversed:
             # Reverse added and removed. Clear deleted, unknown and ignored as
             # these make no sense to reverse.
             r = scmutil.status(r.modified, r.removed, r.added, [], [], [],
                                r.clean)
 
         if listsubrepos:
             for subpath, sub in scmutil.itersubrepos(ctx1, ctx2):
                 try:
                     rev2 = ctx2.subrev(subpath)
                 except KeyError:
                     # A subrepo that existed in node1 was deleted between
                     # node1 and node2 (inclusive). Thus, ctx2's substate
                     # won't contain that subpath. The best we can do ignore it.
                     rev2 = None
                 submatch = matchmod.subdirmatcher(subpath, match)
                 s = sub.status(rev2, match=submatch, ignored=listignored,
                                clean=listclean, unknown=listunknown,
                                listsubrepos=True)
                 for rfiles, sfiles in zip(r, s):
                     rfiles.extend("%s/%s" % (subpath, f) for f in sfiles)
 
         for l in r:
             l.sort()
 
         return r
 
 
 def makememctx(repo, parents, text, user, date, branch, files, store,
                editor=None, extra=None):
     def getfilectx(repo, memctx, path):
         data, mode, copied = store.getfile(path)
         if data is None:
             return None
         islink, isexec = mode
         return memfilectx(repo, path, data, islink=islink, isexec=isexec,
                                   copied=copied, memctx=memctx)
     if extra is None:
         extra = {}
     if branch:
         extra['branch'] = encoding.fromlocal(branch)
     ctx = memctx(repo, parents, text, files, getfilectx, user,
                  date, extra, editor)
     return ctx
 
 class changectx(basectx):
     """A changecontext object makes access to data related to a particular
     changeset convenient. It represents a read-only context already present in
     the repo."""
     def __init__(self, repo, changeid=''):
         """changeid is a revision number, node, or tag"""
 
         # since basectx.__new__ already took care of copying the object, we
         # don't need to do anything in __init__, so we just exit here
         if isinstance(changeid, basectx):
             return
 
         if changeid == '':
             changeid = '.'
         self._repo = repo
 
         try:
             if isinstance(changeid, int):
                 self._node = repo.changelog.node(changeid)
                 self._rev = changeid
                 return
             if isinstance(changeid, long):
                 changeid = str(changeid)
             if changeid == 'null':
                 self._node = nullid
                 self._rev = nullrev
                 return
             if changeid == 'tip':
                 self._node = repo.changelog.tip()
                 self._rev = repo.changelog.rev(self._node)
                 return
             if changeid == '.' or changeid == repo.dirstate.p1():
                 # this is a hack to delay/avoid loading obsmarkers
                 # when we know that '.' won't be hidden
                 self._node = repo.dirstate.p1()
                 self._rev = repo.unfiltered().changelog.rev(self._node)
                 return
             if len(changeid) == 20:
                 try:
                     self._node = changeid
                     self._rev = repo.changelog.rev(changeid)
                     return
                 except error.FilteredRepoLookupError:
                     raise
                 except LookupError:
                     pass
 
             try:
                 r = int(changeid)
                 if str(r) != changeid:
                     raise ValueError
                 l = len(repo.changelog)
                 if r < 0:
                     r += l
                 if r < 0 or r >= l:
                     raise ValueError
                 self._rev = r
                 self._node = repo.changelog.node(r)
                 return
             except error.FilteredIndexError:
                 raise
             except (ValueError, OverflowError, IndexError):
                 pass
 
             if len(changeid) == 40:
                 try:
                     self._node = bin(changeid)
                     self._rev = repo.changelog.rev(self._node)
                     return
                 except error.FilteredLookupError:
                     raise
                 except (TypeError, LookupError):
                     pass
 
             # lookup bookmarks through the name interface
             try:
                 self._node = repo.names.singlenode(repo, changeid)
                 self._rev = repo.changelog.rev(self._node)
                 return
             except KeyError:
                 pass
             except error.FilteredRepoLookupError:
                 raise
             except error.RepoLookupError:
                 pass
 
             self._node = repo.unfiltered().changelog._partialmatch(changeid)
             if self._node is not None:
                 self._rev = repo.changelog.rev(self._node)
                 return
 
             # lookup failed
             # check if it might have come from damaged dirstate
             #
             # XXX we could avoid the unfiltered if we had a recognizable
             # exception for filtered changeset access
             if changeid in repo.unfiltered().dirstate.parents():
                 msg = _("working directory has unknown parent '%s'!")
                 raise error.Abort(msg % short(changeid))
             try:
                 if len(changeid) == 20 and nonascii(changeid):
                     changeid = hex(changeid)
             except TypeError:
                 pass
         except (error.FilteredIndexError, error.FilteredLookupError,
                 error.FilteredRepoLookupError):
             if repo.filtername.startswith('visible'):
                 msg = _("hidden revision '%s'") % changeid
                 hint = _('use --hidden to access hidden revisions')
                 raise error.FilteredRepoLookupError(msg, hint=hint)
             msg = _("filtered revision '%s' (not in '%s' subset)")
             msg %= (changeid, repo.filtername)
             raise error.FilteredRepoLookupError(msg)
         except IndexError:
             pass
         raise error.RepoLookupError(
             _("unknown revision '%s'") % changeid)
 
     def __hash__(self):
         try:
             return hash(self._rev)
         except AttributeError:
             return id(self)
 
     def __nonzero__(self):
         return self._rev != nullrev
 
     @propertycache
     def _changeset(self):
         return self._repo.changelog.changelogrevision(self.rev())
 
     @propertycache
     def _manifest(self):
         return self._repo.manifestlog[self._changeset.manifest].read()
 
     @propertycache
     def _manifestdelta(self):
-        return self._repo.manifest.readdelta(self._changeset.manifest)
+        mfnode = self._changeset.manifest
+        return self._repo.manifestlog[mfnode].readdelta()
 
     @propertycache
     def _parents(self):
         repo = self._repo
         p1, p2 = repo.changelog.parentrevs(self._rev)
         if p2 == nullrev:
             return [changectx(repo, p1)]
         return [changectx(repo, p1), changectx(repo, p2)]
 
     def changeset(self):
         c = self._changeset
         return (
             c.manifest,
             c.user,
             c.date,
             c.files,
             c.description,
             c.extra,
         )
     def manifestnode(self):
         return self._changeset.manifest
 
     def user(self):
         return self._changeset.user
     def date(self):
         return self._changeset.date
     def files(self):
         return self._changeset.files
     def description(self):
         return self._changeset.description
     def branch(self):
         return encoding.tolocal(self._changeset.extra.get("branch"))
     def closesbranch(self):
         return 'close' in self._changeset.extra
     def extra(self):
         return self._changeset.extra
     def tags(self):
         return self._repo.nodetags(self._node)
     def bookmarks(self):
         return self._repo.nodebookmarks(self._node)
     def phase(self):
         return self._repo._phasecache.phase(self._repo, self._rev)
     def hidden(self):
         return self._rev in repoview.filterrevs(self._repo, 'visible')
 
     def children(self):
         """return contexts for each child changeset"""
         c = self._repo.changelog.children(self._node)
         return [changectx(self._repo, x) for x in c]
 
     def ancestors(self):
         for a in self._repo.changelog.ancestors([self._rev]):
             yield changectx(self._repo, a)
 
     def descendants(self):
         for d in self._repo.changelog.descendants([self._rev]):
             yield changectx(self._repo, d)
 
     def filectx(self, path, fileid=None, filelog=None):
         """get a file context from this changeset"""
         if fileid is None:
             fileid = self.filenode(path)
         return filectx(self._repo, path, fileid=fileid,
                        changectx=self, filelog=filelog)
 
     def ancestor(self, c2, warn=False):
         """return the "best" ancestor context of self and c2
 
         If there are multiple candidates, it will show a message and check
         merge.preferancestor configuration before falling back to the
         revlog ancestor."""
         # deal with workingctxs
         n2 = c2._node
         if n2 is None:
             n2 = c2._parents[0]._node
         cahs = self._repo.changelog.commonancestorsheads(self._node, n2)
         if not cahs:
             anc = nullid
         elif len(cahs) == 1:
             anc = cahs[0]
         else:
             # experimental config: merge.preferancestor
             for r in self._repo.ui.configlist('merge', 'preferancestor', ['*']):
                 try:
                     ctx = changectx(self._repo, r)
                 except error.RepoLookupError:
                     continue
                 anc = ctx.node()
                 if anc in cahs:
                     break
             else:
                 anc = self._repo.changelog.ancestor(self._node, n2)
             if warn:
                 self._repo.ui.status(
                     (_("note: using %s as ancestor of %s and %s\n") %
                      (short(anc), short(self._node), short(n2))) +
                     ''.join(_("      alternatively, use --config "
                               "merge.preferancestor=%s\n") %
                             short(n) for n in sorted(cahs) if n != anc))
         return changectx(self._repo, anc)
 
     def descendant(self, other):
         """True if other is descendant of this changeset"""
         return self._repo.changelog.descendant(self._rev, other._rev)
 
     def walk(self, match):
         '''Generates matching file names.'''
 
         # Wrap match.bad method to have message with nodeid
         def bad(fn, msg):
             # The manifest doesn't know about subrepos, so don't complain about
             # paths into valid subrepos.
             if any(fn == s or fn.startswith(s + '/')
                    for s in self.substate):
                 return
             match.bad(fn, _('no such file in rev %s') % self)
 
         m = matchmod.badmatch(match, bad)
         return self._manifest.walk(m)
 
     def matches(self, match):
         return self.walk(match)
 
 class basefilectx(object):
     """A filecontext object represents the common logic for its children:
     filectx: read-only access to a filerevision that is already present
              in the repo,
     workingfilectx: a filecontext that represents files from the working
                     directory,
     memfilectx: a filecontext that represents files in-memory."""
     def __new__(cls, repo, path, *args, **kwargs):
         return super(basefilectx, cls).__new__(cls)
 
     @propertycache
     def _filelog(self):
         return self._repo.file(self._path)
 
     @propertycache
     def _changeid(self):
         if '_changeid' in self.__dict__:
             return self._changeid
         elif '_changectx' in self.__dict__:
             return self._changectx.rev()
         elif '_descendantrev' in self.__dict__:
             # this file context was created from a revision with a known
             # descendant, we can (lazily) correct for linkrev aliases
             return self._adjustlinkrev(self._path, self._filelog,
                                        self._filenode, self._descendantrev)
         else:
             return self._filelog.linkrev(self._filerev)
 
     @propertycache
     def _filenode(self):
         if '_fileid' in self.__dict__:
             return self._filelog.lookup(self._fileid)
         else:
             return self._changectx.filenode(self._path)
 
     @propertycache
     def _filerev(self):
         return self._filelog.rev(self._filenode)
 
     @propertycache
     def _repopath(self):
         return self._path
 
     def __nonzero__(self):
         try:
             self._filenode
             return True
         except error.LookupError:
             # file is missing
             return False
 
     def __str__(self):
         return "%s@%s" % (self.path(), self._changectx)
 
     def __repr__(self):
         return "<%s %s>" % (type(self).__name__, str(self))
 
     def __hash__(self):
         try:
             return hash((self._path, self._filenode))
         except AttributeError:
             return id(self)
 
     def __eq__(self, other):
         try:
             return (type(self) == type(other) and self._path == other._path
                     and self._filenode == other._filenode)
         except AttributeError:
             return False
 
     def __ne__(self, other):
         return not (self == other)
 
     def filerev(self):
         return self._filerev
     def filenode(self):
         return self._filenode
     def flags(self):
         return self._changectx.flags(self._path)
     def filelog(self):
         return self._filelog
     def rev(self):
         return self._changeid
     def linkrev(self):
         return self._filelog.linkrev(self._filerev)
     def node(self):
         return self._changectx.node()
     def hex(self):
         return self._changectx.hex()
     def user(self):
         return self._changectx.user()
     def date(self):
         return self._changectx.date()
     def files(self):
         return self._changectx.files()
     def description(self):
         return self._changectx.description()
     def branch(self):
         return self._changectx.branch()
     def extra(self):
         return self._changectx.extra()
     def phase(self):
         return self._changectx.phase()
     def phasestr(self):
         return self._changectx.phasestr()
     def manifest(self):
         return self._changectx.manifest()
     def changectx(self):
         return self._changectx
     def repo(self):
         return self._repo
 
     def path(self):
         return self._path
 
     def isbinary(self):
         try:
             return util.binary(self.data())
         except IOError:
             return False
     def isexec(self):
         return 'x' in self.flags()
     def islink(self):
         return 'l' in self.flags()
 
     def isabsent(self):
         """whether this filectx represents a file not in self._changectx
 
         This is mainly for merge code to detect change/delete conflicts. This is
         expected to be True for all subclasses of basectx."""
         return False
 
     _customcmp = False
     def cmp(self, fctx):
         """compare with other file context
 
         returns True if different than fctx.
         """
         if fctx._customcmp:
             return fctx.cmp(self)
 
         if (fctx._filenode is None
             and (self._repo._encodefilterpats
                  # if file data starts with '\1\n', empty metadata block is
                  # prepended, which adds 4 bytes to filelog.size().
                  or self.size() - 4 == fctx.size())
             or self.size() == fctx.size()):
             return self._filelog.cmp(self._filenode, fctx.data())
 
         return True
 
     def _adjustlinkrev(self, path, filelog, fnode, srcrev, inclusive=False):
         """return the first ancestor of <srcrev> introducing <fnode>
 
         If the linkrev of the file revision does not point to an ancestor of
         srcrev, we'll walk down the ancestors until we find one introducing
         this file revision.
 
         :repo: a localrepository object (used to access changelog and manifest)
         :path: the file path
         :fnode: the nodeid of the file revision
         :filelog: the filelog of this path
         :srcrev: the changeset revision we search ancestors from
         :inclusive: if true, the src revision will also be checked
         """
         repo = self._repo
         cl = repo.unfiltered().changelog
         ma = repo.manifest
         # fetch the linkrev
         fr = filelog.rev(fnode)
         lkr = filelog.linkrev(fr)
         # hack to reuse ancestor computation when searching for renames
         memberanc = getattr(self, '_ancestrycontext', None)
         iteranc = None
         if srcrev is None:
             # wctx case, used by workingfilectx during mergecopy
             revs = [p.rev() for p in self._repo[None].parents()]
             inclusive = True # we skipped the real (revless) source
         else:
             revs = [srcrev]
         if memberanc is None:
             memberanc = iteranc = cl.ancestors(revs, lkr,
                                                inclusive=inclusive)
         # check if this linkrev is an ancestor of srcrev
         if lkr not in memberanc:
             if iteranc is None:
                 iteranc = cl.ancestors(revs, lkr, inclusive=inclusive)
             for a in iteranc:
                 ac = cl.read(a) # get changeset data (we avoid object creation)
                 if path in ac[3]: # checking the 'files' field.
                     # The file has been touched, check if the content is
                     # similar to the one we search for.
                     if fnode == ma.readfast(ac[0]).get(path):
                         return a
             # In theory, we should never get out of that loop without a result.
             # But if manifest uses a buggy file revision (not children of the
             # one it replaces) we could. Such a buggy situation will likely
             # result is crash somewhere else at to some point.
         return lkr
 
     def introrev(self):
         """return the rev of the changeset which introduced this file revision
 
         This method is different from linkrev because it take into account the
         changeset the filectx was created from. It ensures the returned
         revision is one of its ancestors. This prevents bugs from
         'linkrev-shadowing' when a file revision is used by multiple
         changesets.
         """
         lkr = self.linkrev()
         attrs = vars(self)
         noctx = not ('_changeid' in attrs or '_changectx' in attrs)
         if noctx or self.rev() == lkr:
             return self.linkrev()
         return self._adjustlinkrev(self._path, self._filelog, self._filenode,
                                    self.rev(), inclusive=True)
 
     def _parentfilectx(self, path, fileid, filelog):
         """create parent filectx keeping ancestry info for _adjustlinkrev()"""
         fctx = filectx(self._repo, path, fileid=fileid, filelog=filelog)
         if '_changeid' in vars(self) or '_changectx' in vars(self):
             # If self is associated with a changeset (probably explicitly
             # fed), ensure the created filectx is associated with a
             # changeset that is an ancestor of self.changectx.
             # This lets us later use _adjustlinkrev to get a correct link.
             fctx._descendantrev = self.rev()
             fctx._ancestrycontext = getattr(self, '_ancestrycontext', None)
         elif '_descendantrev' in vars(self):
             # Otherwise propagate _descendantrev if we have one associated.
             fctx._descendantrev = self._descendantrev
             fctx._ancestrycontext = getattr(self, '_ancestrycontext', None)
         return fctx
 
     def parents(self):
         _path = self._path
         fl = self._filelog
         parents = self._filelog.parents(self._filenode)
         pl = [(_path, node, fl) for node in parents if node != nullid]
 
         r = fl.renamed(self._filenode)
         if r:
             # - In the simple rename case, both parent are nullid, pl is empty.
             # - In case of merge, only one of the parent is null id and should
             # be replaced with the rename information. This parent is -always-
             # the first one.
             #
             # As null id have always been filtered out in the previous list
             # comprehension, inserting to 0 will always result in "replacing
             # first nullid parent with rename information.
             pl.insert(0, (r[0], r[1], self._repo.file(r[0])))
 
         return [self._parentfilectx(path, fnode, l) for path, fnode, l in pl]
 
     def p1(self):
         return self.parents()[0]
 
     def p2(self):
         p = self.parents()
         if len(p) == 2:
             return p[1]
         return filectx(self._repo, self._path, fileid=-1, filelog=self._filelog)
 
     def annotate(self, follow=False, linenumber=False, diffopts=None):
         '''returns a list of tuples of ((ctx, number), line) for each line
         in the file, where ctx is the filectx of the node where
         that line was last changed; if linenumber parameter is true, number is
         the line number at the first appearance in the managed file, otherwise,
         number has a fixed value of False.
         '''
 
         def lines(text):
             if text.endswith("\n"):
                 return text.count("\n")
             return text.count("\n") + 1
 
         if linenumber:
             def decorate(text, rev):
                 return ([(rev, i) for i in xrange(1, lines(text) + 1)], text)
         else:
             def decorate(text, rev):
                 return ([(rev, False)] * lines(text), text)
 
         def pair(parent, child):
             blocks = mdiff.allblocks(parent[1], child[1], opts=diffopts,
                                      refine=True)
             for (a1, a2, b1, b2), t in blocks:
                 # Changed blocks ('!') or blocks made only of blank lines ('~')
                 # belong to the child.
                 if t == '=':
                     child[0][b1:b2] = parent[0][a1:a2]
             return child
 
         getlog = util.lrucachefunc(lambda x: self._repo.file(x))
 
         def parents(f):
             # Cut _descendantrev here to mitigate the penalty of lazy linkrev
             # adjustment. Otherwise, p._adjustlinkrev() would walk changelog
             # from the topmost introrev (= srcrev) down to p.linkrev() if it
             # isn't an ancestor of the srcrev.
             f._changeid
             pl = f.parents()
 
             # Don't return renamed parents if we aren't following.
             if not follow:
                 pl = [p for p in pl if p.path() == f.path()]
 
             # renamed filectx won't have a filelog yet, so set it
             # from the cache to save time
             for p in pl:
                 if not '_filelog' in p.__dict__:
                     p._filelog = getlog(p.path())
 
             return pl
 
         # use linkrev to find the first changeset where self appeared
         base = self
         introrev = self.introrev()
         if self.rev() != introrev:
             base = self.filectx(self.filenode(), changeid=introrev)
         if getattr(base, '_ancestrycontext', None) is None:
             cl = self._repo.changelog
             if introrev is None:
                 # wctx is not inclusive, but works because _ancestrycontext
                 # is used to test filelog revisions
                 ac = cl.ancestors([p.rev() for p in base.parents()],
                                   inclusive=True)
             else:
                 ac = cl.ancestors([introrev], inclusive=True)
             base._ancestrycontext = ac
 
         # This algorithm would prefer to be recursive, but Python is a
         # bit recursion-hostile. Instead we do an iterative
         # depth-first search.
 
         # 1st DFS pre-calculates pcache and needed
         visit = [base]
         pcache = {}
         needed = {base: 1}
         while visit:
             f = visit.pop()
             if f in pcache:
                 continue
             pl = parents(f)
             pcache[f] = pl
             for p in pl:
                 needed[p] = needed.get(p, 0) + 1
                 if p not in pcache:
                     visit.append(p)
 
         # 2nd DFS does the actual annotate
         visit[:] = [base]
         hist = {}
         while visit:
             f = visit[-1]
             if f in hist:
                 visit.pop()
                 continue
 
             ready = True
             pl = pcache[f]
             for p in pl:
                 if p not in hist:
                     ready = False
                     visit.append(p)
             if ready:
                 visit.pop()
                 curr = decorate(f.data(), f)
                 for p in pl:
                     curr = pair(hist[p], curr)
                     if needed[p] == 1:
                         del hist[p]
                         del needed[p]
                     else:
                         needed[p] -= 1
 
                 hist[f] = curr
                 del pcache[f]
 
         return zip(hist[base][0], hist[base][1].splitlines(True))
 
     def ancestors(self, followfirst=False):
         visit = {}
         c = self
         if followfirst:
             cut = 1
         else:
             cut = None
 
         while True:
             for parent in c.parents()[:cut]:
                 visit[(parent.linkrev(), parent.filenode())] = parent
             if not visit:
                 break
             c = visit.pop(max(visit))
             yield c
 
 class filectx(basefilectx):
     """A filecontext object makes access to data related to a particular
        filerevision convenient."""
     def __init__(self, repo, path, changeid=None, fileid=None,
                  filelog=None, changectx=None):
         """changeid can be a changeset revision, node, or tag.
            fileid can be a file revision or node."""
         self._repo = repo
         self._path = path
 
         assert (changeid is not None
                 or fileid is not None
                 or changectx is not None), \
                 ("bad args: changeid=%r, fileid=%r, changectx=%r"
                  % (changeid, fileid, changectx))
 
         if filelog is not None:
             self._filelog = filelog
 
         if changeid is not None:
             self._changeid = changeid
         if changectx is not None:
             self._changectx = changectx
         if fileid is not None:
             self._fileid = fileid
 
     @propertycache
     def _changectx(self):
         try:
             return changectx(self._repo, self._changeid)
         except error.FilteredRepoLookupError:
             # Linkrev may point to any revision in the repository.  When the
             # repository is filtered this may lead to `filectx` trying to build
             # `changectx` for filtered revision. In such case we fallback to
             # creating `changectx` on the unfiltered version of the reposition.
             # This fallback should not be an issue because `changectx` from
             # `filectx` are not used in complex operations that care about
             # filtering.
             #
             # This fallback is a cheap and dirty fix that prevent several
             # crashes. It does not ensure the behavior is correct. However the
             # behavior was not correct before filtering either and "incorrect
             # behavior" is seen as better as "crash"
             #
             # Linkrevs have several serious troubles with filtering that are
             # complicated to solve. Proper handling of the issue here should be
             # considered when solving linkrev issue are on the table.
             return changectx(self._repo.unfiltered(), self._changeid)
 
     def filectx(self, fileid, changeid=None):
         '''opens an arbitrary revision of the file without
         opening a new filelog'''
         return filectx(self._repo, self._path, fileid=fileid,
                        filelog=self._filelog, changeid=changeid)
 
     def data(self):
         try:
             return self._filelog.read(self._filenode)
         except error.CensoredNodeError:
             if self._repo.ui.config("censor", "policy", "abort") == "ignore":
                 return ""
             raise error.Abort(_("censored node: %s") % short(self._filenode),
                              hint=_("set censor.policy to ignore errors"))
 
     def size(self):
         return self._filelog.size(self._filerev)
 
     def renamed(self):
         """check if file was actually renamed in this changeset revision
 
         If rename logged in file revision, we report copy for changeset only
         if file revisions linkrev points back to the changeset in question
         or both changeset parents contain different file revisions.
         """
 
         renamed = self._filelog.renamed(self._filenode)
         if not renamed:
             return renamed
 
         if self.rev() == self.linkrev():
             return renamed
 
         name = self.path()
         fnode = self._filenode
         for p in self._changectx.parents():
             try:
                 if fnode == p.filenode(name):
                     return None
             except error.LookupError:
                 pass
         return renamed
 
     def children(self):
         # hard for renames
         c = self._filelog.children(self._filenode)
         return [filectx(self._repo, self._path, fileid=x,
                         filelog=self._filelog) for x in c]
 
 class committablectx(basectx):
     """A committablectx object provides common functionality for a context that
     wants the ability to commit, e.g. workingctx or memctx."""
     def __init__(self, repo, text="", user=None, date=None, extra=None,
                  changes=None):
         self._repo = repo
         self._rev = None
         self._node = None
         self._text = text
         if date:
             self._date = util.parsedate(date)
         if user:
             self._user = user
         if changes:
             self._status = changes
 
         self._extra = {}
         if extra:
             self._extra = extra.copy()
         if 'branch' not in self._extra:
             try:
                 branch = encoding.fromlocal(self._repo.dirstate.branch())
             except UnicodeDecodeError:
                 raise error.Abort(_('branch name not in UTF-8!'))
             self._extra['branch'] = branch
         if self._extra['branch'] == '':
             self._extra['branch'] = 'default'
 
     def __str__(self):
         return str(self._parents[0]) + "+"
 
     def __nonzero__(self):
         return True
 
     def _buildflagfunc(self):
         # Create a fallback function for getting file flags when the
         # filesystem doesn't support them
 
         copiesget = self._repo.dirstate.copies().get
         parents = self.parents()
         if len(parents) < 2:
             # when we have one parent, it's easy: copy from parent
             man = parents[0].manifest()
             def func(f):
                 f = copiesget(f, f)
                 return man.flags(f)
         else:
             # merges are tricky: we try to reconstruct the unstored
             # result from the merge (issue1802)
             p1, p2 = parents
             pa = p1.ancestor(p2)
             m1, m2, ma = p1.manifest(), p2.manifest(), pa.manifest()
 
             def func(f):
                 f = copiesget(f, f) # may be wrong for merges with copies
                 fl1, fl2, fla = m1.flags(f), m2.flags(f), ma.flags(f)
                 if fl1 == fl2:
                     return fl1
                 if fl1 == fla:
                     return fl2
                 if fl2 == fla:
                     return fl1
                 return '' # punt for conflicts
 
         return func
 
     @propertycache
     def _flagfunc(self):
         return self._repo.dirstate.flagfunc(self._buildflagfunc)
 
     @propertycache
     def _manifest(self):
         """generate a manifest corresponding to the values in self._status
 
         This reuse the file nodeid from parent, but we append an extra letter
         when modified. Modified files get an extra 'm' while added files get
         an extra 'a'. This is used by manifests merge to see that files
         are different and by update logic to avoid deleting newly added files.
         """
         parents = self.parents()
 
         man1 = parents[0].manifest()
         man = man1.copy()
         if len(parents) > 1:
             man2 = self.p2().manifest()
             def getman(f):
                 if f in man1:
                     return man1
                 return man2
         else:
             getman = lambda f: man1
 
         copied = self._repo.dirstate.copies()
         ff = self._flagfunc
         for i, l in (("a", self._status.added), ("m", self._status.modified)):
             for f in l:
                 orig = copied.get(f, f)
                 man[f] = getman(orig).get(orig, nullid) + i
                 try:
                     man.setflag(f, ff(f))
                 except OSError:
                     pass
 
         for f in self._status.deleted + self._status.removed:
             if f in man:
                 del man[f]
 
         return man
 
     @propertycache
     def _status(self):
         return self._repo.status()
 
     @propertycache
     def _user(self):
         return self._repo.ui.username()
 
     @propertycache
     def _date(self):
         return util.makedate()
 
     def subrev(self, subpath):
         return None
 
     def manifestnode(self):
         return None
     def user(self):
         return self._user or self._repo.ui.username()
     def date(self):
         return self._date
     def description(self):
         return self._text
     def files(self):
         return sorted(self._status.modified + self._status.added +
                       self._status.removed)
 
     def modified(self):
         return self._status.modified
     def added(self):
         return self._status.added
     def removed(self):
         return self._status.removed
     def deleted(self):
         return self._status.deleted
     def branch(self):
         return encoding.tolocal(self._extra['branch'])
     def closesbranch(self):
         return 'close' in self._extra
     def extra(self):
         return self._extra
 
     def tags(self):
         return []
 
     def bookmarks(self):
         b = []
         for p in self.parents():
             b.extend(p.bookmarks())
         return b
 
     def phase(self):
         phase = phases.draft # default phase to draft
         for p in self.parents():
             phase = max(phase, p.phase())
         return phase
 
     def hidden(self):
         return False
 
     def children(self):
         return []
 
     def flags(self, path):
         if '_manifest' in self.__dict__:
             try:
                 return self._manifest.flags(path)
             except KeyError:
                 return ''
 
         try:
             return self._flagfunc(path)
         except OSError:
             return ''
 
     def ancestor(self, c2):
         """return the "best" ancestor context of self and c2"""
         return self._parents[0].ancestor(c2) # punt on two parents for now
 
     def walk(self, match):
         '''Generates matching file names.'''
         return sorted(self._repo.dirstate.walk(match, sorted(self.substate),
                                                True, False))
 
     def matches(self, match):
         return sorted(self._repo.dirstate.matches(match))
 
     def ancestors(self):
         for p in self._parents:
             yield p
         for a in self._repo.changelog.ancestors(
             [p.rev() for p in self._parents]):
             yield changectx(self._repo, a)
 
     def markcommitted(self, node):
         """Perform post-commit cleanup necessary after committing this ctx
 
         Specifically, this updates backing stores this working context
         wraps to reflect the fact that the changes reflected by this
         workingctx have been committed.  For example, it marks
         modified and added files as normal in the dirstate.
 
         """
 
         self._repo.dirstate.beginparentchange()
         for f in self.modified() + self.added():
             self._repo.dirstate.normal(f)
         for f in self.removed():
             self._repo.dirstate.drop(f)
         self._repo.dirstate.setparents(node)
         self._repo.dirstate.endparentchange()
 
         # write changes out explicitly, because nesting wlock at
         # runtime may prevent 'wlock.release()' in 'repo.commit()'
         # from immediately doing so for subsequent changing files
         self._repo.dirstate.write(self._repo.currenttransaction())
 
 class workingctx(committablectx):
     """A workingctx object makes access to data related to
     the current working directory convenient.
     date - any valid date string or (unixtime, offset), or None.
     user - username string, or None.
     extra - a dictionary of extra values, or None.
     changes - a list of file lists as returned by localrepo.status()
                or None to use the repository status.
     """
     def __init__(self, repo, text="", user=None, date=None, extra=None,
                  changes=None):
         super(workingctx, self).__init__(repo, text, user, date, extra, changes)
 
     def __iter__(self):
         d = self._repo.dirstate
         for f in d:
             if d[f] != 'r':
                 yield f
 
     def __contains__(self, key):
         return self._repo.dirstate[key] not in "?r"
 
     def hex(self):
         return hex(wdirid)
 
     @propertycache
     def _parents(self):
         p = self._repo.dirstate.parents()
         if p[1] == nullid:
             p = p[:-1]
         return [changectx(self._repo, x) for x in p]
 
     def filectx(self, path, filelog=None):
         """get a file context from the working directory"""
         return workingfilectx(self._repo, path, workingctx=self,
                               filelog=filelog)
 
     def dirty(self, missing=False, merge=True, branch=True):
         "check whether a working directory is modified"
         # check subrepos first
         for s in sorted(self.substate):
             if self.sub(s).dirty():
                 return True
         # check current working dir
         return ((merge and self.p2()) or
                 (branch and self.branch() != self.p1().branch()) or
                 self.modified() or self.added() or self.removed() or
                 (missing and self.deleted()))
 
     def add(self, list, prefix=""):
         join = lambda f: os.path.join(prefix, f)
         with self._repo.wlock():
             ui, ds = self._repo.ui, self._repo.dirstate
             rejected = []
             lstat = self._repo.wvfs.lstat
             for f in list:
                 scmutil.checkportable(ui, join(f))
                 try:
                     st = lstat(f)
                 except OSError:
                     ui.warn(_("%s does not exist!\n") % join(f))
                     rejected.append(f)
                     continue
                 if st.st_size > 10000000:
                     ui.warn(_("%s: up to %d MB of RAM may be required "
                               "to manage this file\n"
                               "(use 'hg revert %s' to cancel the "
                               "pending addition)\n")
                               % (f, 3 * st.st_size // 1000000, join(f)))
                 if not (stat.S_ISREG(st.st_mode) or stat.S_ISLNK(st.st_mode)):
                     ui.warn(_("%s not added: only files and symlinks "
                               "supported currently\n") % join(f))
                     rejected.append(f)
                 elif ds[f] in 'amn':
                     ui.warn(_("%s already tracked!\n") % join(f))
                 elif ds[f] == 'r':
                     ds.normallookup(f)
                 else:
                     ds.add(f)
             return rejected
 
     def forget(self, files, prefix=""):
         join = lambda f: os.path.join(prefix, f)
         with self._repo.wlock():
             rejected = []
             for f in files:
                 if f not in self._repo.dirstate:
                     self._repo.ui.warn(_("%s not tracked!\n") % join(f))
                     rejected.append(f)
                 elif self._repo.dirstate[f] != 'a':
                     self._repo.dirstate.remove(f)
                 else:
                     self._repo.dirstate.drop(f)
             return rejected
 
     def undelete(self, list):
         pctxs = self.parents()
         with self._repo.wlock():
             for f in list:
                 if self._repo.dirstate[f] != 'r':
                     self._repo.ui.warn(_("%s not removed!\n") % f)
                 else:
                     fctx = f in pctxs[0] and pctxs[0][f] or pctxs[1][f]
                     t = fctx.data()
                     self._repo.wwrite(f, t, fctx.flags())
                     self._repo.dirstate.normal(f)
 
     def copy(self, source, dest):
         try:
             st = self._repo.wvfs.lstat(dest)
         except OSError as err:
             if err.errno != errno.ENOENT:
                 raise
             self._repo.ui.warn(_("%s does not exist!\n") % dest)
             return
         if not (stat.S_ISREG(st.st_mode) or stat.S_ISLNK(st.st_mode)):
             self._repo.ui.warn(_("copy failed: %s is not a file or a "
                                  "symbolic link\n") % dest)
         else:
             with self._repo.wlock():
                 if self._repo.dirstate[dest] in '?':
                     self._repo.dirstate.add(dest)
                 elif self._repo.dirstate[dest] in 'r':
                     self._repo.dirstate.normallookup(dest)
                 self._repo.dirstate.copy(source, dest)
 
     def match(self, pats=[], include=None, exclude=None, default='glob',
               listsubrepos=False, badfn=None):
         r = self._repo
 
         # Only a case insensitive filesystem needs magic to translate user input
         # to actual case in the filesystem.
         if not util.fscasesensitive(r.root):
             return matchmod.icasefsmatcher(r.root, r.getcwd(), pats, include,
                                            exclude, default, r.auditor, self,
                                            listsubrepos=listsubrepos,
                                            badfn=badfn)
         return matchmod.match(r.root, r.getcwd(), pats,
                               include, exclude, default,
                               auditor=r.auditor, ctx=self,
                               listsubrepos=listsubrepos, badfn=badfn)
 
     def _filtersuspectsymlink(self, files):
         if not files or self._repo.dirstate._checklink:
             return files
 
         # Symlink placeholders may get non-symlink-like contents
         # via user error or dereferencing by NFS or Samba servers,
         # so we filter out any placeholders that don't look like a
         # symlink
         sane = []
         for f in files:
             if self.flags(f) == 'l':
                 d = self[f].data()
                 if d == '' or len(d) >= 1024 or '\n' in d or util.binary(d):
                     self._repo.ui.debug('ignoring suspect symlink placeholder'
                                         ' "%s"\n' % f)
                     continue
             sane.append(f)
         return sane
 
     def _checklookup(self, files):
         # check for any possibly clean files
         if not files:
             return [], []
 
         modified = []
         fixup = []
         pctx = self._parents[0]
         # do a full compare of any files that might have changed
         for f in sorted(files):
             if (f not in pctx or self.flags(f) != pctx.flags(f)
                 or pctx[f].cmp(self[f])):
                 modified.append(f)
             else:
                 fixup.append(f)
 
         # update dirstate for files that are actually clean
         if fixup:
             try:
                 # updating the dirstate is optional
                 # so we don't wait on the lock
                 # wlock can invalidate the dirstate, so cache normal _after_
                 # taking the lock
                 with self._repo.wlock(False):
                     normal = self._repo.dirstate.normal
                     for f in fixup:
                         normal(f)
                     # write changes out explicitly, because nesting
                     # wlock at runtime may prevent 'wlock.release()'
                     # after this block from doing so for subsequent
                     # changing files
                     self._repo.dirstate.write(self._repo.currenttransaction())
             except error.LockError:
                 pass
         return modified, fixup
 
     def _manifestmatches(self, match, s):
         """Slow path for workingctx
 
         The fast path is when we compare the working directory to its parent
         which means this function is comparing with a non-parent; therefore we
         need to build a manifest and return what matches.
         """
         mf = self._repo['.']._manifestmatches(match, s)
         for f in s.modified + s.added:
             mf[f] = _newnode
             mf.setflag(f, self.flags(f))
         for f in s.removed:
             if f in mf:
                 del mf[f]
         return mf
 
     def _dirstatestatus(self, match=None, ignored=False, clean=False,
                         unknown=False):
         '''Gets the status from the dirstate -- internal use only.'''
         listignored, listclean, listunknown = ignored, clean, unknown
         match = match or matchmod.always(self._repo.root, self._repo.getcwd())
         subrepos = []
         if '.hgsub' in self:
             subrepos = sorted(self.substate)
         cmp, s = self._repo.dirstate.status(match, subrepos, listignored,
                                             listclean, listunknown)
 
         # check for any possibly clean files
         if cmp:
             modified2, fixup = self._checklookup(cmp)
             s.modified.extend(modified2)
 
             # update dirstate for files that are actually clean
             if fixup and listclean:
                 s.clean.extend(fixup)
 
         if match.always():
             # cache for performance
             if s.unknown or s.ignored or s.clean:
                 # "_status" is cached with list*=False in the normal route
                 self._status = scmutil.status(s.modified, s.added, s.removed,
                                               s.deleted, [], [], [])
             else:
                 self._status = s
 
         return s
 
     def _buildstatus(self, other, s, match, listignored, listclean,
                      listunknown):
         """build a status with respect to another context
 
         This includes logic for maintaining the fast path of status when
         comparing the working directory against its parent, which is to skip
         building a new manifest if self (working directory) is not comparing
         against its parent (repo['.']).
         """
         s = self._dirstatestatus(match, listignored, listclean, listunknown)
         # Filter out symlinks that, in the case of FAT32 and NTFS filesystems,
         # might have accidentally ended up with the entire contents of the file
         # they are supposed to be linking to.
         s.modified[:] = self._filtersuspectsymlink(s.modified)
         if other != self._repo['.']:
             s = super(workingctx, self)._buildstatus(other, s, match,
                                                      listignored, listclean,
                                                      listunknown)
         return s
 
     def _matchstatus(self, other, match):
         """override the match method with a filter for directory patterns
 
         We use inheritance to customize the match.bad method only in cases of
         workingctx since it belongs only to the working directory when
         comparing against the parent changeset.
 
         If we aren't comparing against the working directory's parent, then we
         just use the default match object sent to us.
         """
         superself = super(workingctx, self)
         match = superself._matchstatus(other, match)
         if other != self._repo['.']:
             def bad(f, msg):
                 # 'f' may be a directory pattern from 'match.files()',
                 # so 'f not in ctx1' is not enough
                 if f not in other and not other.hasdir(f):
                     self._repo.ui.warn('%s: %s\n' %
                                        (self._repo.dirstate.pathto(f), msg))
             match.bad = bad
         return match
 
 class committablefilectx(basefilectx):
     """A committablefilectx provides common functionality for a file context
     that wants the ability to commit, e.g. workingfilectx or memfilectx."""
     def __init__(self, repo, path, filelog=None, ctx=None):
         self._repo = repo
         self._path = path
         self._changeid = None
         self._filerev = self._filenode = None
 
         if filelog is not None:
             self._filelog = filelog
         if ctx:
             self._changectx = ctx
 
     def __nonzero__(self):
         return True
 
     def linkrev(self):
         # linked to self._changectx no matter if file is modified or not
         return self.rev()
 
     def parents(self):
         '''return parent filectxs, following copies if necessary'''
         def filenode(ctx, path):
             return ctx._manifest.get(path, nullid)
 
         path = self._path
         fl = self._filelog
         pcl = self._changectx._parents
         renamed = self.renamed()
 
         if renamed:
             pl = [renamed + (None,)]
         else:
             pl = [(path, filenode(pcl[0], path), fl)]
 
         for pc in pcl[1:]:
             pl.append((path, filenode(pc, path), fl))
 
         return [self._parentfilectx(p, fileid=n, filelog=l)
                 for p, n, l in pl if n != nullid]
 
     def children(self):
         return []
 
 class workingfilectx(committablefilectx):
     """A workingfilectx object makes access to data related to a particular
        file in the working directory convenient."""
     def __init__(self, repo, path, filelog=None, workingctx=None):
         super(workingfilectx, self).__init__(repo, path, filelog, workingctx)
 
     @propertycache
     def _changectx(self):
         return workingctx(self._repo)
 
     def data(self):
         return self._repo.wread(self._path)
     def renamed(self):
         rp = self._repo.dirstate.copied(self._path)
         if not rp:
             return None
         return rp, self._changectx._parents[0]._manifest.get(rp, nullid)
 
     def size(self):
         return self._repo.wvfs.lstat(self._path).st_size
     def date(self):
         t, tz = self._changectx.date()
         try:
             return (self._repo.wvfs.lstat(self._path).st_mtime, tz)
         except OSError as err:
             if err.errno != errno.ENOENT:
                 raise
             return (t, tz)
 
     def cmp(self, fctx):
         """compare with other file context
 
         returns True if different than fctx.
         """
         # fctx should be a filectx (not a workingfilectx)
         # invert comparison to reuse the same code path
         return fctx.cmp(self)
 
     def remove(self, ignoremissing=False):
         """wraps unlink for a repo's working directory"""
         util.unlinkpath(self._repo.wjoin(self._path), ignoremissing)
 
     def write(self, data, flags):
         """wraps repo.wwrite"""
         self._repo.wwrite(self._path, data, flags)
 
 class workingcommitctx(workingctx):
     """A workingcommitctx object makes access to data related to
     the revision being committed convenient.
 
     This hides changes in the working directory, if they aren't
     committed in this context.
     """
     def __init__(self, repo, changes,
                  text="", user=None, date=None, extra=None):
         super(workingctx, self).__init__(repo, text, user, date, extra,
                                          changes)
 
     def _dirstatestatus(self, match=None, ignored=False, clean=False,
                         unknown=False):
         """Return matched files only in ``self._status``
 
         Uncommitted files appear "clean" via this context, even if
         they aren't actually so in the working directory.
         """
         match = match or matchmod.always(self._repo.root, self._repo.getcwd())
         if clean:
             clean = [f for f in self._manifest if f not in self._changedset]
         else:
             clean = []
         return scmutil.status([f for f in self._status.modified if match(f)],
                               [f for f in self._status.added if match(f)],
                               [f for f in self._status.removed if match(f)],
                               [], [], [], clean)
 
     @propertycache
     def _changedset(self):
         """Return the set of files changed in this context
         """
         changed = set(self._status.modified)
         changed.update(self._status.added)
         changed.update(self._status.removed)
         return changed
 
 def makecachingfilectxfn(func):
     """Create a filectxfn that caches based on the path.
 
     We can't use util.cachefunc because it uses all arguments as the cache
     key and this creates a cycle since the arguments include the repo and
     memctx.
     """
     cache = {}
 
     def getfilectx(repo, memctx, path):
         if path not in cache:
             cache[path] = func(repo, memctx, path)
         return cache[path]
 
     return getfilectx
 
 class memctx(committablectx):
     """Use memctx to perform in-memory commits via localrepo.commitctx().
 
     Revision information is supplied at initialization time while
     related files data and is made available through a callback
     mechanism.  'repo' is the current localrepo, 'parents' is a
     sequence of two parent revisions identifiers (pass None for every
     missing parent), 'text' is the commit message and 'files' lists
     names of files touched by the revision (normalized and relative to
     repository root).
 
     filectxfn(repo, memctx, path) is a callable receiving the
     repository, the current memctx object and the normalized path of
     requested file, relative to repository root. It is fired by the
     commit function for every file in 'files', but calls order is
     undefined. If the file is available in the revision being
     committed (updated or added), filectxfn returns a memfilectx
     object. If the file was removed, filectxfn raises an
     IOError. Moved files are represented by marking the source file
     removed and the new file added with copy information (see
     memfilectx).
 
     user receives the committer name and defaults to current
     repository username, date is the commit date in any format
     supported by util.parsedate() and defaults to current date, extra
     is a dictionary of metadata or is left empty.
     """
 
     # Mercurial <= 3.1 expects the filectxfn to raise IOError for missing files.
     # Extensions that need to retain compatibility across Mercurial 3.1 can use
     # this field to determine what to do in filectxfn.
     _returnnoneformissingfiles = True
 
     def __init__(self, repo, parents, text, files, filectxfn, user=None,
                  date=None, extra=None, editor=False):
         super(memctx, self).__init__(repo, text, user, date, extra)
         self._rev = None
         self._node = None
         parents = [(p or nullid) for p in parents]
         p1, p2 = parents
         self._parents = [changectx(self._repo, p) for p in (p1, p2)]
         files = sorted(set(files))
         self._files = files
         self.substate = {}
 
         # if store is not callable, wrap it in a function
         if not callable(filectxfn):
             def getfilectx(repo, memctx, path):
                 fctx = filectxfn[path]
                 # this is weird but apparently we only keep track of one parent
                 # (why not only store that instead of a tuple?)
                 copied = fctx.renamed()
                 if copied:
                     copied = copied[0]
                 return memfilectx(repo, path, fctx.data(),
                                   islink=fctx.islink(), isexec=fctx.isexec(),
                                   copied=copied, memctx=memctx)
             self._filectxfn = getfilectx
         else:
             # memoizing increases performance for e.g. vcs convert scenarios.
             self._filectxfn = makecachingfilectxfn(filectxfn)
 
         if extra:
             self._extra = extra.copy()
         else:
             self._extra = {}
 
         if self._extra.get('branch', '') == '':
             self._extra['branch'] = 'default'
 
         if editor:
             self._text = editor(self._repo, self, [])
             self._repo.savecommitmessage(self._text)
 
     def filectx(self, path, filelog=None):
         """get a file context from the working directory
 
         Returns None if file doesn't exist and should be removed."""
         return self._filectxfn(self._repo, self, path)
 
     def commit(self):
         """commit context to the repo"""
         return self._repo.commitctx(self)
 
     @propertycache
     def _manifest(self):
         """generate a manifest based on the return values of filectxfn"""
 
         # keep this simple for now; just worry about p1
         pctx = self._parents[0]
         man = pctx.manifest().copy()
 
         for f in self._status.modified:
             p1node = nullid
             p2node = nullid
             p = pctx[f].parents() # if file isn't in pctx, check p2?
             if len(p) > 0:
                 p1node = p[0].filenode()
                 if len(p) > 1:
                     p2node = p[1].filenode()
             man[f] = revlog.hash(self[f].data(), p1node, p2node)
 
         for f in self._status.added:
             man[f] = revlog.hash(self[f].data(), nullid, nullid)
 
         for f in self._status.removed:
             if f in man:
                 del man[f]
 
         return man
 
     @propertycache
     def _status(self):
         """Calculate exact status from ``files`` specified at construction
         """
         man1 = self.p1().manifest()
         p2 = self._parents[1]
         # "1 < len(self._parents)" can't be used for checking
         # existence of the 2nd parent, because "memctx._parents" is
         # explicitly initialized by the list, of which length is 2.
         if p2.node() != nullid:
             man2 = p2.manifest()
             managing = lambda f: f in man1 or f in man2
         else:
             managing = lambda f: f in man1
 
         modified, added, removed = [], [], []
         for f in self._files:
             if not managing(f):
                 added.append(f)
             elif self[f]:
                 modified.append(f)
             else:
                 removed.append(f)
 
         return scmutil.status(modified, added, removed, [], [], [], [])
 
 class memfilectx(committablefilectx):
     """memfilectx represents an in-memory file to commit.
 
     See memctx and committablefilectx for more details.
     """
     def __init__(self, repo, path, data, islink=False,
                  isexec=False, copied=None, memctx=None):
         """
         path is the normalized file path relative to repository root.
         data is the file content as a string.
         islink is True if the file is a symbolic link.
         isexec is True if the file is executable.
         copied is the source file path if current file was copied in the
         revision being committed, or None."""
         super(memfilectx, self).__init__(repo, path, None, memctx)
         self._data = data
         self._flags = (islink and 'l' or '') + (isexec and 'x' or '')
         self._copied = None
         if copied:
             self._copied = (copied, nullid)
 
     def data(self):
         return self._data
     def size(self):
         return len(self.data())
     def flags(self):
         return self._flags
     def renamed(self):
         return self._copied
 
     def remove(self, ignoremissing=False):
         """wraps unlink for a repo's working directory"""
         # need to figure out what to do here
         del self._changectx[self._path]
 
     def write(self, data, flags):
         """wraps repo.wwrite"""
         self._data = data

mercurial/manifest.py

@@ -1,1245 +1,1278
 # manifest.py - manifest revision class for mercurial
 #
 # Copyright 2005-2007 Matt Mackall <[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.
 
 from __future__ import absolute_import
 
 import array
 import heapq
 import os
 import struct
 
 from .i18n import _
 from . import (
     error,
     mdiff,
     parsers,
     revlog,
     util,
 )
 
 propertycache = util.propertycache
 
 def _parsev1(data):
     # This method does a little bit of excessive-looking
     # precondition checking. This is so that the behavior of this
     # class exactly matches its C counterpart to try and help
     # prevent surprise breakage for anyone that develops against
     # the pure version.
     if data and data[-1] != '\n':
         raise ValueError('Manifest did not end in a newline.')
     prev = None
     for l in data.splitlines():
         if prev is not None and prev > l:
             raise ValueError('Manifest lines not in sorted order.')
         prev = l
         f, n = l.split('\0')
         if len(n) > 40:
             yield f, revlog.bin(n[:40]), n[40:]
         else:
             yield f, revlog.bin(n), ''
 
 def _parsev2(data):
     metadataend = data.find('\n')
     # Just ignore metadata for now
     pos = metadataend + 1
     prevf = ''
     while pos < len(data):
         end = data.find('\n', pos + 1) # +1 to skip stem length byte
         if end == -1:
             raise ValueError('Manifest ended with incomplete file entry.')
         stemlen = ord(data[pos])
         items = data[pos + 1:end].split('\0')
         f = prevf[:stemlen] + items[0]
         if prevf > f:
             raise ValueError('Manifest entries not in sorted order.')
         fl = items[1]
         # Just ignore metadata (items[2:] for now)
         n = data[end + 1:end + 21]
         yield f, n, fl
         pos = end + 22
         prevf = f
 
 def _parse(data):
     """Generates (path, node, flags) tuples from a manifest text"""
     if data.startswith('\0'):
         return iter(_parsev2(data))
     else:
         return iter(_parsev1(data))
 
 def _text(it, usemanifestv2):
     """Given an iterator over (path, node, flags) tuples, returns a manifest
     text"""
     if usemanifestv2:
         return _textv2(it)
     else:
         return _textv1(it)
 
 def _textv1(it):
     files = []
     lines = []
     _hex = revlog.hex
     for f, n, fl in it:
         files.append(f)
         # if this is changed to support newlines in filenames,
         # be sure to check the templates/ dir again (especially *-raw.tmpl)
         lines.append("%s\0%s%s\n" % (f, _hex(n), fl))
 
     _checkforbidden(files)
     return ''.join(lines)
 
 def _textv2(it):
     files = []
     lines = ['\0\n']
     prevf = ''
     for f, n, fl in it:
         files.append(f)
         stem = os.path.commonprefix([prevf, f])
         stemlen = min(len(stem), 255)
         lines.append("%c%s\0%s\n%s\n" % (stemlen, f[stemlen:], fl, n))
         prevf = f
     _checkforbidden(files)
     return ''.join(lines)
 
 class _lazymanifest(dict):
     """This is the pure implementation of lazymanifest.
 
     It has not been optimized *at all* and is not lazy.
     """
 
     def __init__(self, data):
         dict.__init__(self)
         for f, n, fl in _parse(data):
             self[f] = n, fl
 
     def __setitem__(self, k, v):
         node, flag = v
         assert node is not None
         if len(node) > 21:
             node = node[:21] # match c implementation behavior
         dict.__setitem__(self, k, (node, flag))
 
     def __iter__(self):
         return iter(sorted(dict.keys(self)))
 
     def iterkeys(self):
         return iter(sorted(dict.keys(self)))
 
     def iterentries(self):
         return ((f, e[0], e[1]) for f, e in sorted(self.iteritems()))
 
     def copy(self):
         c = _lazymanifest('')
         c.update(self)
         return c
 
     def diff(self, m2, clean=False):
         '''Finds changes between the current manifest and m2.'''
         diff = {}
 
         for fn, e1 in self.iteritems():
             if fn not in m2:
                 diff[fn] = e1, (None, '')
             else:
                 e2 = m2[fn]
                 if e1 != e2:
                     diff[fn] = e1, e2
                 elif clean:
                     diff[fn] = None
 
         for fn, e2 in m2.iteritems():
             if fn not in self:
                 diff[fn] = (None, ''), e2
 
         return diff
 
     def filtercopy(self, filterfn):
         c = _lazymanifest('')
         for f, n, fl in self.iterentries():
             if filterfn(f):
                 c[f] = n, fl
         return c
 
     def text(self):
         """Get the full data of this manifest as a bytestring."""
         return _textv1(self.iterentries())
 
 try:
     _lazymanifest = parsers.lazymanifest
 except AttributeError:
     pass
 
 class manifestdict(object):
     def __init__(self, data=''):
         if data.startswith('\0'):
             #_lazymanifest can not parse v2
             self._lm = _lazymanifest('')
             for f, n, fl in _parsev2(data):
                 self._lm[f] = n, fl
         else:
             self._lm = _lazymanifest(data)
 
     def __getitem__(self, key):
         return self._lm[key][0]
 
     def find(self, key):
         return self._lm[key]
 
     def __len__(self):
         return len(self._lm)
 
     def __setitem__(self, key, node):
         self._lm[key] = node, self.flags(key, '')
 
     def __contains__(self, key):
         return key in self._lm
 
     def __delitem__(self, key):
         del self._lm[key]
 
     def __iter__(self):
         return self._lm.__iter__()
 
     def iterkeys(self):
         return self._lm.iterkeys()
 
     def keys(self):
         return list(self.iterkeys())
 
     def filesnotin(self, m2):
         '''Set of files in this manifest that are not in the other'''
         diff = self.diff(m2)
         files = set(filepath
                     for filepath, hashflags in diff.iteritems()
                     if hashflags[1][0] is None)
         return files
 
     @propertycache
     def _dirs(self):
         return util.dirs(self)
 
     def dirs(self):
         return self._dirs
 
     def hasdir(self, dir):
         return dir in self._dirs
 
     def _filesfastpath(self, match):
         '''Checks whether we can correctly and quickly iterate over matcher
         files instead of over manifest files.'''
         files = match.files()
         return (len(files) < 100 and (match.isexact() or
             (match.prefix() and all(fn in self for fn in files))))
 
     def walk(self, match):
         '''Generates matching file names.
 
         Equivalent to manifest.matches(match).iterkeys(), but without creating
         an entirely new manifest.
 
         It also reports nonexistent files by marking them bad with match.bad().
         '''
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         # avoid the entire walk if we're only looking for specific files
         if self._filesfastpath(match):
             for fn in sorted(fset):
                 yield fn
             return
 
         for fn in self:
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             if match(fn):
                 yield fn
 
         # for dirstate.walk, files=['.'] means "walk the whole tree".
         # follow that here, too
         fset.discard('.')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def matches(self, match):
         '''generate a new manifest filtered by the match argument'''
         if match.always():
             return self.copy()
 
         if self._filesfastpath(match):
             m = manifestdict()
             lm = self._lm
             for fn in match.files():
                 if fn in lm:
                     m._lm[fn] = lm[fn]
             return m
 
         m = manifestdict()
         m._lm = self._lm.filtercopy(match)
         return m
 
     def diff(self, m2, clean=False):
         '''Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         '''
         return self._lm.diff(m2._lm, clean)
 
     def setflag(self, key, flag):
         self._lm[key] = self[key], flag
 
     def get(self, key, default=None):
         try:
             return self._lm[key][0]
         except KeyError:
             return default
 
     def flags(self, key, default=''):
         try:
             return self._lm[key][1]
         except KeyError:
             return default
 
     def copy(self):
         c = manifestdict()
         c._lm = self._lm.copy()
         return c
 
     def iteritems(self):
         return (x[:2] for x in self._lm.iterentries())
 
     def iterentries(self):
         return self._lm.iterentries()
 
     def text(self, usemanifestv2=False):
         if usemanifestv2:
             return _textv2(self._lm.iterentries())
         else:
             # use (probably) native version for v1
             return self._lm.text()
 
     def fastdelta(self, base, changes):
         """Given a base manifest text as an array.array and a list of changes
         relative to that text, compute a delta that can be used by revlog.
         """
         delta = []
         dstart = None
         dend = None
         dline = [""]
         start = 0
         # zero copy representation of base as a buffer
         addbuf = util.buffer(base)
 
         changes = list(changes)
         if len(changes) < 1000:
             # start with a readonly loop that finds the offset of
             # each line and creates the deltas
             for f, todelete in changes:
                 # bs will either be the index of the item or the insert point
                 start, end = _msearch(addbuf, f, start)
                 if not todelete:
                     h, fl = self._lm[f]
                     l = "%s\0%s%s\n" % (f, revlog.hex(h), fl)
                 else:
                     if start == end:
                         # item we want to delete was not found, error out
                         raise AssertionError(
                                 _("failed to remove %s from manifest") % f)
                     l = ""
                 if dstart is not None and dstart <= start and dend >= start:
                     if dend < end:
                         dend = end
                     if l:
                         dline.append(l)
                 else:
                     if dstart is not None:
                         delta.append([dstart, dend, "".join(dline)])
                     dstart = start
                     dend = end
                     dline = [l]
 
             if dstart is not None:
                 delta.append([dstart, dend, "".join(dline)])
             # apply the delta to the base, and get a delta for addrevision
             deltatext, arraytext = _addlistdelta(base, delta)
         else:
             # For large changes, it's much cheaper to just build the text and
             # diff it.
             arraytext = array.array('c', self.text())
             deltatext = mdiff.textdiff(base, arraytext)
 
         return arraytext, deltatext
 
 def _msearch(m, s, lo=0, hi=None):
     '''return a tuple (start, end) that says where to find s within m.
 
     If the string is found m[start:end] are the line containing
     that string.  If start == end the string was not found and
     they indicate the proper sorted insertion point.
 
     m should be a buffer or a string
     s is a string'''
     def advance(i, c):
         while i < lenm and m[i] != c:
             i += 1
         return i
     if not s:
         return (lo, lo)
     lenm = len(m)
     if not hi:
         hi = lenm
     while lo < hi:
         mid = (lo + hi) // 2
         start = mid
         while start > 0 and m[start - 1] != '\n':
             start -= 1
         end = advance(start, '\0')
         if m[start:end] < s:
             # we know that after the null there are 40 bytes of sha1
             # this translates to the bisect lo = mid + 1
             lo = advance(end + 40, '\n') + 1
         else:
             # this translates to the bisect hi = mid
             hi = start
     end = advance(lo, '\0')
     found = m[lo:end]
     if s == found:
         # we know that after the null there are 40 bytes of sha1
         end = advance(end + 40, '\n')
         return (lo, end + 1)
     else:
         return (lo, lo)
 
 def _checkforbidden(l):
     """Check filenames for illegal characters."""
     for f in l:
         if '\n' in f or '\r' in f:
             raise error.RevlogError(
                 _("'\\n' and '\\r' disallowed in filenames: %r") % f)
 
 
 # apply the changes collected during the bisect loop to our addlist
 # return a delta suitable for addrevision
 def _addlistdelta(addlist, x):
     # for large addlist arrays, building a new array is cheaper
     # than repeatedly modifying the existing one
     currentposition = 0
     newaddlist = array.array('c')
 
     for start, end, content in x:
         newaddlist += addlist[currentposition:start]
         if content:
             newaddlist += array.array('c', content)
 
         currentposition = end
 
     newaddlist += addlist[currentposition:]
 
     deltatext = "".join(struct.pack(">lll", start, end, len(content))
                    + content for start, end, content in x)
     return deltatext, newaddlist
 
 def _splittopdir(f):
     if '/' in f:
         dir, subpath = f.split('/', 1)
         return dir + '/', subpath
     else:
         return '', f
 
 _noop = lambda s: None
 
 class treemanifest(object):
     def __init__(self, dir='', text=''):
         self._dir = dir
         self._node = revlog.nullid
         self._loadfunc = _noop
         self._copyfunc = _noop
         self._dirty = False
         self._dirs = {}
         # Using _lazymanifest here is a little slower than plain old dicts
         self._files = {}
         self._flags = {}
         if text:
             def readsubtree(subdir, subm):
                 raise AssertionError('treemanifest constructor only accepts '
                                      'flat manifests')
             self.parse(text, readsubtree)
             self._dirty = True # Mark flat manifest dirty after parsing
 
     def _subpath(self, path):
         return self._dir + path
 
     def __len__(self):
         self._load()
         size = len(self._files)
         for m in self._dirs.values():
             size += m.__len__()
         return size
 
     def _isempty(self):
         self._load() # for consistency; already loaded by all callers
         return (not self._files and (not self._dirs or
                 all(m._isempty() for m in self._dirs.values())))
 
     def __repr__(self):
         return ('<treemanifest dir=%s, node=%s, loaded=%s, dirty=%s at 0x%x>' %
                 (self._dir, revlog.hex(self._node),
                  bool(self._loadfunc is _noop),
                  self._dirty, id(self)))
 
     def dir(self):
         '''The directory that this tree manifest represents, including a
         trailing '/'. Empty string for the repo root directory.'''
         return self._dir
 
     def node(self):
         '''This node of this instance. nullid for unsaved instances. Should
         be updated when the instance is read or written from a revlog.
         '''
         assert not self._dirty
         return self._node
 
     def setnode(self, node):
         self._node = node
         self._dirty = False
 
     def iterentries(self):
         self._load()
         for p, n in sorted(self._dirs.items() + self._files.items()):
             if p in self._files:
                 yield self._subpath(p), n, self._flags.get(p, '')
             else:
                 for x in n.iterentries():
                     yield x
 
     def iteritems(self):
         self._load()
         for p, n in sorted(self._dirs.items() + self._files.items()):
             if p in self._files:
                 yield self._subpath(p), n
             else:
                 for f, sn in n.iteritems():
                     yield f, sn
 
     def iterkeys(self):
         self._load()
         for p in sorted(self._dirs.keys() + self._files.keys()):
             if p in self._files:
                 yield self._subpath(p)
             else:
                 for f in self._dirs[p].iterkeys():
                     yield f
 
     def keys(self):
         return list(self.iterkeys())
 
     def __iter__(self):
         return self.iterkeys()
 
     def __contains__(self, f):
         if f is None:
             return False
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 return False
             return self._dirs[dir].__contains__(subpath)
         else:
             return f in self._files
 
     def get(self, f, default=None):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 return default
             return self._dirs[dir].get(subpath, default)
         else:
             return self._files.get(f, default)
 
     def __getitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             return self._dirs[dir].__getitem__(subpath)
         else:
             return self._files[f]
 
     def flags(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 return ''
             return self._dirs[dir].flags(subpath)
         else:
             if f in self._dirs:
                 return ''
             return self._flags.get(f, '')
 
     def find(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             return self._dirs[dir].find(subpath)
         else:
             return self._files[f], self._flags.get(f, '')
 
     def __delitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._dirs[dir].__delitem__(subpath)
             # If the directory is now empty, remove it
             if self._dirs[dir]._isempty():
                 del self._dirs[dir]
         else:
             del self._files[f]
             if f in self._flags:
                 del self._flags[f]
         self._dirty = True
 
     def __setitem__(self, f, n):
         assert n is not None
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].__setitem__(subpath, n)
         else:
             self._files[f] = n[:21] # to match manifestdict's behavior
         self._dirty = True
 
     def _load(self):
         if self._loadfunc is not _noop:
             lf, self._loadfunc = self._loadfunc, _noop
             lf(self)
         elif self._copyfunc is not _noop:
             cf, self._copyfunc = self._copyfunc, _noop
             cf(self)
 
     def setflag(self, f, flags):
         """Set the flags (symlink, executable) for path f."""
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].setflag(subpath, flags)
         else:
             self._flags[f] = flags
         self._dirty = True
 
     def copy(self):
         copy = treemanifest(self._dir)
         copy._node = self._node
         copy._dirty = self._dirty
         if self._copyfunc is _noop:
             def _copyfunc(s):
                 self._load()
                 for d in self._dirs:
                     s._dirs[d] = self._dirs[d].copy()
                 s._files = dict.copy(self._files)
                 s._flags = dict.copy(self._flags)
             if self._loadfunc is _noop:
                 _copyfunc(copy)
             else:
                 copy._copyfunc = _copyfunc
         else:
             copy._copyfunc = self._copyfunc
         return copy
 
     def filesnotin(self, m2):
         '''Set of files in this manifest that are not in the other'''
         files = set()
         def _filesnotin(t1, t2):
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             for d, m1 in t1._dirs.iteritems():
                 if d in t2._dirs:
                     m2 = t2._dirs[d]
                     _filesnotin(m1, m2)
                 else:
                     files.update(m1.iterkeys())
 
             for fn in t1._files.iterkeys():
                 if fn not in t2._files:
                     files.add(t1._subpath(fn))
 
         _filesnotin(self, m2)
         return files
 
     @propertycache
     def _alldirs(self):
         return util.dirs(self)
 
     def dirs(self):
         return self._alldirs
 
     def hasdir(self, dir):
         self._load()
         topdir, subdir = _splittopdir(dir)
         if topdir:
             if topdir in self._dirs:
                 return self._dirs[topdir].hasdir(subdir)
             return False
         return (dir + '/') in self._dirs
 
     def walk(self, match):
         '''Generates matching file names.
 
         Equivalent to manifest.matches(match).iterkeys(), but without creating
         an entirely new manifest.
 
         It also reports nonexistent files by marking them bad with match.bad().
         '''
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         for fn in self._walk(match):
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             yield fn
 
         # for dirstate.walk, files=['.'] means "walk the whole tree".
         # follow that here, too
         fset.discard('.')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def _walk(self, match):
         '''Recursively generates matching file names for walk().'''
         if not match.visitdir(self._dir[:-1] or '.'):
             return
 
         # yield this dir's files and walk its submanifests
         self._load()
         for p in sorted(self._dirs.keys() + self._files.keys()):
             if p in self._files:
                 fullp = self._subpath(p)
                 if match(fullp):
                     yield fullp
             else:
                 for f in self._dirs[p]._walk(match):
                     yield f
 
     def matches(self, match):
         '''generate a new manifest filtered by the match argument'''
         if match.always():
             return self.copy()
 
         return self._matches(match)
 
     def _matches(self, match):
         '''recursively generate a new manifest filtered by the match argument.
         '''
 
         visit = match.visitdir(self._dir[:-1] or '.')
         if visit == 'all':
             return self.copy()
         ret = treemanifest(self._dir)
         if not visit:
             return ret
 
         self._load()
         for fn in self._files:
             fullp = self._subpath(fn)
             if not match(fullp):
                 continue
             ret._files[fn] = self._files[fn]
             if fn in self._flags:
                 ret._flags[fn] = self._flags[fn]
 
         for dir, subm in self._dirs.iteritems():
             m = subm._matches(match)
             if not m._isempty():
                 ret._dirs[dir] = m
 
         if not ret._isempty():
             ret._dirty = True
         return ret
 
     def diff(self, m2, clean=False):
         '''Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         '''
         result = {}
         emptytree = treemanifest()
         def _diff(t1, t2):
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             for d, m1 in t1._dirs.iteritems():
                 m2 = t2._dirs.get(d, emptytree)
                 _diff(m1, m2)
 
             for d, m2 in t2._dirs.iteritems():
                 if d not in t1._dirs:
                     _diff(emptytree, m2)
 
             for fn, n1 in t1._files.iteritems():
                 fl1 = t1._flags.get(fn, '')
                 n2 = t2._files.get(fn, None)
                 fl2 = t2._flags.get(fn, '')
                 if n1 != n2 or fl1 != fl2:
                     result[t1._subpath(fn)] = ((n1, fl1), (n2, fl2))
                 elif clean:
                     result[t1._subpath(fn)] = None
 
             for fn, n2 in t2._files.iteritems():
                 if fn not in t1._files:
                     fl2 = t2._flags.get(fn, '')
                     result[t2._subpath(fn)] = ((None, ''), (n2, fl2))
 
         _diff(self, m2)
         return result
 
     def unmodifiedsince(self, m2):
         return not self._dirty and not m2._dirty and self._node == m2._node
 
     def parse(self, text, readsubtree):
         for f, n, fl in _parse(text):
             if fl == 't':
                 f = f + '/'
                 self._dirs[f] = readsubtree(self._subpath(f), n)
             elif '/' in f:
                 # This is a flat manifest, so use __setitem__ and setflag rather
                 # than assigning directly to _files and _flags, so we can
                 # assign a path in a subdirectory, and to mark dirty (compared
                 # to nullid).
                 self[f] = n
                 if fl:
                     self.setflag(f, fl)
             else:
                 # Assigning to _files and _flags avoids marking as dirty,
                 # and should be a little faster.
                 self._files[f] = n
                 if fl:
                     self._flags[f] = fl
 
     def text(self, usemanifestv2=False):
         """Get the full data of this manifest as a bytestring."""
         self._load()
         return _text(self.iterentries(), usemanifestv2)
 
     def dirtext(self, usemanifestv2=False):
         """Get the full data of this directory as a bytestring. Make sure that
         any submanifests have been written first, so their nodeids are correct.
         """
         self._load()
         flags = self.flags
         dirs = [(d[:-1], self._dirs[d]._node, 't') for d in self._dirs]
         files = [(f, self._files[f], flags(f)) for f in self._files]
         return _text(sorted(dirs + files), usemanifestv2)
 
     def read(self, gettext, readsubtree):
         def _load_for_read(s):
             s.parse(gettext(), readsubtree)
             s._dirty = False
         self._loadfunc = _load_for_read
 
     def writesubtrees(self, m1, m2, writesubtree):
         self._load() # for consistency; should never have any effect here
         m1._load()
         m2._load()
         emptytree = treemanifest()
         for d, subm in self._dirs.iteritems():
             subp1 = m1._dirs.get(d, emptytree)._node
             subp2 = m2._dirs.get(d, emptytree)._node
             if subp1 == revlog.nullid:
                 subp1, subp2 = subp2, subp1
             writesubtree(subm, subp1, subp2)
 
 class manifestrevlog(revlog.revlog):
     '''A revlog that stores manifest texts. This is responsible for caching the
     full-text manifest contents.
     '''
     def __init__(self, opener, indexfile):
         super(manifestrevlog, self).__init__(opener, indexfile)
 
         # During normal operations, we expect to deal with not more than four
         # revs at a time (such as during commit --amend). When rebasing large
         # stacks of commits, the number can go up, hence the config knob below.
         cachesize = 4
         opts = getattr(opener, 'options', None)
         if opts is not None:
             cachesize = opts.get('manifestcachesize', cachesize)
         self._fulltextcache = util.lrucachedict(cachesize)
 
     @property
     def fulltextcache(self):
         return self._fulltextcache
 
     def clearcaches(self):
         super(manifestrevlog, self).clearcaches()
         self._fulltextcache.clear()
 
 class manifestlog(object):
     """A collection class representing the collection of manifest snapshots
     referenced by commits in the repository.
 
     In this situation, 'manifest' refers to the abstract concept of a snapshot
     of the list of files in the given commit. Consumers of the output of this
     class do not care about the implementation details of the actual manifests
     they receive (i.e. tree or flat or lazily loaded, etc)."""
     def __init__(self, opener, repo):
         self._repo = repo
 
         # We'll separate this into it's own cache once oldmanifest is no longer
         # used
         self._mancache = repo.manifest._mancache
 
     @property
     def _revlog(self):
         return self._repo.manifest
 
     @property
     def _oldmanifest(self):
         # _revlog is the same as _oldmanifest right now, but we eventually want
         # to delete _oldmanifest while still allowing manifestlog to access the
         # revlog specific apis.
         return self._repo.manifest
 
     def __getitem__(self, node):
         """Retrieves the manifest instance for the given node. Throws a KeyError
         if not found.
         """
         if node in self._mancache:
             cachemf = self._mancache[node]
             # The old manifest may put non-ctx manifests in the cache, so skip
             # those since they don't implement the full api.
             if (isinstance(cachemf, manifestctx) or
                 isinstance(cachemf, treemanifestctx)):
                 return cachemf
 
         if self._oldmanifest._treeinmem:
             m = treemanifestctx(self._revlog, '', node)
         else:
             m = manifestctx(self._revlog, node)
         if node != revlog.nullid:
             self._mancache[node] = m
         return m
 
 class manifestctx(object):
     """A class representing a single revision of a manifest, including its
     contents, its parent revs, and its linkrev.
     """
     def __init__(self, revlog, node):
         self._revlog = revlog
         self._data = None
 
         self._node = node
 
         # TODO: We eventually want p1, p2, and linkrev exposed on this class,
         # but let's add it later when something needs it and we can load it
         # lazily.
         #self.p1, self.p2 = revlog.parents(node)
         #rev = revlog.rev(node)
         #self.linkrev = revlog.linkrev(rev)
 
     def node(self):
         return self._node
 
     def read(self):
         if not self._data:
             if self._node == revlog.nullid:
                 self._data = manifestdict()
             else:
                 text = self._revlog.revision(self._node)
                 arraytext = array.array('c', text)
                 self._revlog._fulltextcache[self._node] = arraytext
                 self._data = manifestdict(text)
         return self._data
 
+    def readdelta(self):
+        revlog = self._revlog
+        if revlog._usemanifestv2:
+            # Need to perform a slow delta
+            r0 = revlog.deltaparent(revlog.rev(self._node))
+            m0 = manifestctx(revlog, revlog.node(r0)).read()
+            m1 = self.read()
+            md = manifestdict()
+            for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).iteritems():
+                if n1:
+                    md[f] = n1
+                    if fl1:
+                        md.setflag(f, fl1)
+            return md
+
+        r = revlog.rev(self._node)
+        d = mdiff.patchtext(revlog.revdiff(revlog.deltaparent(r), r))
+        return manifestdict(d)
+
 class treemanifestctx(object):
     def __init__(self, revlog, dir, node):
         revlog = revlog.dirlog(dir)
         self._revlog = revlog
         self._dir = dir
         self._data = None
 
         self._node = node
 
         # TODO: Load p1/p2/linkrev lazily. They need to be lazily loaded so that
         # we can instantiate treemanifestctx objects for directories we don't
         # have on disk.
         #self.p1, self.p2 = revlog.parents(node)
         #rev = revlog.rev(node)
         #self.linkrev = revlog.linkrev(rev)
 
     def read(self):
         if not self._data:
             if self._node == revlog.nullid:
                 self._data = treemanifest()
             elif self._revlog._treeondisk:
                 m = treemanifest(dir=self._dir)
                 def gettext():
                     return self._revlog.revision(self._node)
                 def readsubtree(dir, subm):
                     return treemanifestctx(self._revlog, dir, subm).read()
                 m.read(gettext, readsubtree)
                 m.setnode(self._node)
                 self._data = m
             else:
                 text = self._revlog.revision(self._node)
                 arraytext = array.array('c', text)
                 self._revlog.fulltextcache[self._node] = arraytext
                 self._data = treemanifest(dir=self._dir, text=text)
 
         return self._data
 
     def node(self):
         return self._node
 
+    def readdelta(self):
+        # Need to perform a slow delta
+        revlog = self._revlog
+        r0 = revlog.deltaparent(revlog.rev(self._node))
+        m0 = treemanifestctx(revlog, revlog.node(r0), dir=self._dir).read()
+        m1 = self.read()
+        md = treemanifest(dir=self._dir)
+        for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).iteritems():
+            if n1:
+                md[f] = n1
+                if fl1:
+                    md.setflag(f, fl1)
+        return md
+
 class manifest(manifestrevlog):
     def __init__(self, opener, dir='', dirlogcache=None):
         '''The 'dir' and 'dirlogcache' arguments are for internal use by
         manifest.manifest only. External users should create a root manifest
         log with manifest.manifest(opener) and call dirlog() on it.
         '''
         # During normal operations, we expect to deal with not more than four
         # revs at a time (such as during commit --amend). When rebasing large
         # stacks of commits, the number can go up, hence the config knob below.
         cachesize = 4
         usetreemanifest = False
         usemanifestv2 = False
         opts = getattr(opener, 'options', None)
         if opts is not None:
             cachesize = opts.get('manifestcachesize', cachesize)
             usetreemanifest = opts.get('treemanifest', usetreemanifest)
             usemanifestv2 = opts.get('manifestv2', usemanifestv2)
         self._mancache = util.lrucachedict(cachesize)
         self._treeinmem = usetreemanifest
         self._treeondisk = usetreemanifest
         self._usemanifestv2 = usemanifestv2
         indexfile = "00manifest.i"
         if dir:
             assert self._treeondisk, 'opts is %r' % opts
             if not dir.endswith('/'):
                 dir = dir + '/'
             indexfile = "meta/" + dir + "00manifest.i"
         super(manifest, self).__init__(opener, indexfile)
         self._dir = dir
         # The dirlogcache is kept on the root manifest log
         if dir:
             self._dirlogcache = dirlogcache
         else:
             self._dirlogcache = {'': self}
 
     def _newmanifest(self, data=''):
         if self._treeinmem:
             return treemanifest(self._dir, data)
         return manifestdict(data)
 
     def dirlog(self, dir):
         if dir:
             assert self._treeondisk
         if dir not in self._dirlogcache:
             self._dirlogcache[dir] = manifest(self.opener, dir,
                                               self._dirlogcache)
         return self._dirlogcache[dir]
 
     def _slowreaddelta(self, node):
         r0 = self.deltaparent(self.rev(node))
         m0 = self.read(self.node(r0))
         m1 = self.read(node)
         md = self._newmanifest()
         for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).iteritems():
             if n1:
                 md[f] = n1
                 if fl1:
                     md.setflag(f, fl1)
         return md
 
     def readdelta(self, node):
         if self._usemanifestv2 or self._treeondisk:
             return self._slowreaddelta(node)
         r = self.rev(node)
         d = mdiff.patchtext(self.revdiff(self.deltaparent(r), r))
         return self._newmanifest(d)
 
     def readshallowdelta(self, node):
         '''For flat manifests, this is the same as readdelta(). For
         treemanifests, this will read the delta for this revlog's directory,
         without recursively reading subdirectory manifests. Instead, any
         subdirectory entry will be reported as it appears in the manifests, i.e.
         the subdirectory will be reported among files and distinguished only by
         its 't' flag.'''
         if not self._treeondisk:
             return self.readdelta(node)
         if self._usemanifestv2:
             raise error.Abort(
                 _("readshallowdelta() not implemented for manifestv2"))
         r = self.rev(node)
         d = mdiff.patchtext(self.revdiff(self.deltaparent(r), r))
         return manifestdict(d)
 
     def readfast(self, node):
         '''use the faster of readdelta or read
 
         This will return a manifest which is either only the files
         added/modified relative to p1, or all files in the
         manifest. Which one is returned depends on the codepath used
         to retrieve the data.
         '''
         r = self.rev(node)
         deltaparent = self.deltaparent(r)
         if deltaparent != revlog.nullrev and deltaparent in self.parentrevs(r):
             return self.readdelta(node)
         return self.read(node)
 
     def readshallowfast(self, node):
         '''like readfast(), but calls readshallowdelta() instead of readdelta()
         '''
         r = self.rev(node)
         deltaparent = self.deltaparent(r)
         if deltaparent != revlog.nullrev and deltaparent in self.parentrevs(r):
             return self.readshallowdelta(node)
         return self.readshallow(node)
 
     def read(self, node):
         if node == revlog.nullid:
             return self._newmanifest() # don't upset local cache
         if node in self._mancache:
             cached = self._mancache[node]
             if (isinstance(cached, manifestctx) or
                 isinstance(cached, treemanifestctx)):
                 cached = cached.read()
             return cached
         if self._treeondisk:
             def gettext():
                 return self.revision(node)
             def readsubtree(dir, subm):
                 return self.dirlog(dir).read(subm)
             m = self._newmanifest()
             m.read(gettext, readsubtree)
             m.setnode(node)
             arraytext = None
         else:
             text = self.revision(node)
             m = self._newmanifest(text)
             arraytext = array.array('c', text)
         self._mancache[node] = m
         self.fulltextcache[node] = arraytext
         return m
 
     def readshallow(self, node):
         '''Reads the manifest in this directory. When using flat manifests,
         this manifest will generally have files in subdirectories in it. Does
         not cache the manifest as the callers generally do not read the same
         version twice.'''
         return manifestdict(self.revision(node))
 
     def find(self, node, f):
         '''look up entry for a single file efficiently.
         return (node, flags) pair if found, (None, None) if not.'''
         m = self.read(node)
         try:
             return m.find(f)
         except KeyError:
             return None, None
 
     def add(self, m, transaction, link, p1, p2, added, removed):
         if (p1 in self.fulltextcache and not self._treeinmem
             and not self._usemanifestv2):
             # If our first parent is in the manifest cache, we can
             # compute a delta here using properties we know about the
             # manifest up-front, which may save time later for the
             # revlog layer.
 
             _checkforbidden(added)
             # combine the changed lists into one sorted iterator
             work = heapq.merge([(x, False) for x in added],
                                [(x, True) for x in removed])
 
             arraytext, deltatext = m.fastdelta(self.fulltextcache[p1], work)
             cachedelta = self.rev(p1), deltatext
             text = util.buffer(arraytext)
             n = self.addrevision(text, transaction, link, p1, p2, cachedelta)
         else:
             # The first parent manifest isn't already loaded, so we'll
             # just encode a fulltext of the manifest and pass that
             # through to the revlog layer, and let it handle the delta
             # process.
             if self._treeondisk:
                 m1 = self.read(p1)
                 m2 = self.read(p2)
                 n = self._addtree(m, transaction, link, m1, m2)
                 arraytext = None
             else:
                 text = m.text(self._usemanifestv2)
                 n = self.addrevision(text, transaction, link, p1, p2)
                 arraytext = array.array('c', text)
 
         self._mancache[n] = m
         self.fulltextcache[n] = arraytext
 
         return n
 
     def _addtree(self, m, transaction, link, m1, m2):
         # If the manifest is unchanged compared to one parent,
         # don't write a new revision
         if m.unmodifiedsince(m1) or m.unmodifiedsince(m2):
             return m.node()
         def writesubtree(subm, subp1, subp2):
             sublog = self.dirlog(subm.dir())
             sublog.add(subm, transaction, link, subp1, subp2, None, None)
         m.writesubtrees(m1, m2, writesubtree)
         text = m.dirtext(self._usemanifestv2)
         # Double-check whether contents are unchanged to one parent
         if text == m1.dirtext(self._usemanifestv2):
             n = m1.node()
         elif text == m2.dirtext(self._usemanifestv2):
             n = m2.node()
         else:
             n = self.addrevision(text, transaction, link, m1.node(), m2.node())
         # Save nodeid so parent manifest can calculate its nodeid
         m.setnode(n)
         return n
 
     def clearcaches(self):
         super(manifest, self).clearcaches()
         self._mancache.clear()
         self._dirlogcache = {'': self}