revlog: add an experimental option to mitigated delta issues (issue5480)

The general delta heuristic to select a delta do not scale with the number of
branch. The delta base is frequently too far away to be able to reuse a chain
according to the "distance" criteria. This leads to insertion of larger delta (or
even full text) that themselves push the bases for the next delta further away
leading to more large deltas and full texts. This full text and frequent
recomputation throw Mercurial performance in disarray.

For example of a slightly large repository

  280 000 files (2 150 000 versions)
  430 000 changesets (10 000 topological heads)

Number below compares repository with and without the distance criteria:

manifest size:
    with:    21.4 GB
    without:  0.3 GB

store size:
    with:    28.7 GB
    without   7.4 GB

bundle last 15 00 revisions:
    with:    800 seconds
             971 MB
    without:  50 seconds
              73 MB

unbundle time (of the last 15K revisions):
    with:    1150 seconds (~19 minutes)
    without:   35 seconds

Similar issues has been observed in other repositories.


Adding a new option or "feature" on stable is uncommon. However, given that this
issues is making Mercurial practically unusable, I'm exceptionally targeting
this patch for stable.

What is actually needed is a full rework of the delta building and reading
logic. However, that will be a longer process and churn not suitable for stable.

In the meantime, we introduces a quick and dirty mitigation of this in the
'experimental' config space. The new option introduces a way to set the maximum
amount of memory usable to store a diff in memory. This extend the ability for
Mercurial to create chains without removing all safe guard regarding memory
access. The option should be phased out when core has a more proper solution
available.

Setting the limit to '0' remove all limits, setting it to '-1' use the default
limit (textsize x 4).

mercurial/localrepo.py

@@ -1,2107 +1,2110 b''
 # localrepo.py - read/write repository 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 errno
 import hashlib
 import inspect
 import os
 import random
 import time
 import weakref
 
 from .i18n import _
 from .node import (
     hex,
     nullid,
     short,
 )
 from . import (
     bookmarks,
     branchmap,
     bundle2,
     changegroup,
     changelog,
     color,
     context,
     dirstate,
     dirstateguard,
     encoding,
     error,
     exchange,
     extensions,
     filelog,
     hook,
     lock as lockmod,
     manifest,
     match as matchmod,
     merge as mergemod,
     mergeutil,
     namespaces,
     obsolete,
     pathutil,
     peer,
     phases,
     pushkey,
     pycompat,
     repoview,
     revset,
     revsetlang,
     scmutil,
     store,
     subrepo,
     tags as tagsmod,
     transaction,
     txnutil,
     util,
     vfs as vfsmod,
 )
 
 release = lockmod.release
 urlerr = util.urlerr
 urlreq = util.urlreq
 
 class _basefilecache(scmutil.filecache):
     """All filecache usage on repo are done for logic that should be unfiltered
     """
     def __get__(self, repo, type=None):
         if repo is None:
             return self
         return super(_basefilecache, self).__get__(repo.unfiltered(), type)
     def __set__(self, repo, value):
         return super(_basefilecache, self).__set__(repo.unfiltered(), value)
     def __delete__(self, repo):
         return super(_basefilecache, self).__delete__(repo.unfiltered())
 
 class repofilecache(_basefilecache):
     """filecache for files in .hg but outside of .hg/store"""
     def join(self, obj, fname):
         return obj.vfs.join(fname)
 
 class storecache(_basefilecache):
     """filecache for files in the store"""
     def join(self, obj, fname):
         return obj.sjoin(fname)
 
 class unfilteredpropertycache(util.propertycache):
     """propertycache that apply to unfiltered repo only"""
 
     def __get__(self, repo, type=None):
         unfi = repo.unfiltered()
         if unfi is repo:
             return super(unfilteredpropertycache, self).__get__(unfi)
         return getattr(unfi, self.name)
 
 class filteredpropertycache(util.propertycache):
     """propertycache that must take filtering in account"""
 
     def cachevalue(self, obj, value):
         object.__setattr__(obj, self.name, value)
 
 
 def hasunfilteredcache(repo, name):
     """check if a repo has an unfilteredpropertycache value for <name>"""
     return name in vars(repo.unfiltered())
 
 def unfilteredmethod(orig):
     """decorate method that always need to be run on unfiltered version"""
     def wrapper(repo, *args, **kwargs):
         return orig(repo.unfiltered(), *args, **kwargs)
     return wrapper
 
 moderncaps = {'lookup', 'branchmap', 'pushkey', 'known', 'getbundle',
               'unbundle'}
 legacycaps = moderncaps.union({'changegroupsubset'})
 
 class localpeer(peer.peerrepository):
     '''peer for a local repo; reflects only the most recent API'''
 
     def __init__(self, repo, caps=None):
         if caps is None:
             caps = moderncaps.copy()
         peer.peerrepository.__init__(self)
         self._repo = repo.filtered('served')
         self.ui = repo.ui
         self._caps = repo._restrictcapabilities(caps)
         self.requirements = repo.requirements
         self.supportedformats = repo.supportedformats
 
     def close(self):
         self._repo.close()
 
     def _capabilities(self):
         return self._caps
 
     def local(self):
         return self._repo
 
     def canpush(self):
         return True
 
     def url(self):
         return self._repo.url()
 
     def lookup(self, key):
         return self._repo.lookup(key)
 
     def branchmap(self):
         return self._repo.branchmap()
 
     def heads(self):
         return self._repo.heads()
 
     def known(self, nodes):
         return self._repo.known(nodes)
 
     def getbundle(self, source, heads=None, common=None, bundlecaps=None,
                   **kwargs):
         chunks = exchange.getbundlechunks(self._repo, source, heads=heads,
                                           common=common, bundlecaps=bundlecaps,
                                           **kwargs)
         cb = util.chunkbuffer(chunks)
 
         if exchange.bundle2requested(bundlecaps):
             # When requesting a bundle2, getbundle returns a stream to make the
             # wire level function happier. We need to build a proper object
             # from it in local peer.
             return bundle2.getunbundler(self.ui, cb)
         else:
             return changegroup.getunbundler('01', cb, None)
 
     # TODO We might want to move the next two calls into legacypeer and add
     # unbundle instead.
 
     def unbundle(self, cg, heads, url):
         """apply a bundle on a repo
 
         This function handles the repo locking itself."""
         try:
             try:
                 cg = exchange.readbundle(self.ui, cg, None)
                 ret = exchange.unbundle(self._repo, cg, heads, 'push', url)
                 if util.safehasattr(ret, 'getchunks'):
                     # This is a bundle20 object, turn it into an unbundler.
                     # This little dance should be dropped eventually when the
                     # API is finally improved.
                     stream = util.chunkbuffer(ret.getchunks())
                     ret = bundle2.getunbundler(self.ui, stream)
                 return ret
             except Exception as exc:
                 # If the exception contains output salvaged from a bundle2
                 # reply, we need to make sure it is printed before continuing
                 # to fail. So we build a bundle2 with such output and consume
                 # it directly.
                 #
                 # This is not very elegant but allows a "simple" solution for
                 # issue4594
                 output = getattr(exc, '_bundle2salvagedoutput', ())
                 if output:
                     bundler = bundle2.bundle20(self._repo.ui)
                     for out in output:
                         bundler.addpart(out)
                     stream = util.chunkbuffer(bundler.getchunks())
                     b = bundle2.getunbundler(self.ui, stream)
                     bundle2.processbundle(self._repo, b)
                 raise
         except error.PushRaced as exc:
             raise error.ResponseError(_('push failed:'), str(exc))
 
     def lock(self):
         return self._repo.lock()
 
     def pushkey(self, namespace, key, old, new):
         return self._repo.pushkey(namespace, key, old, new)
 
     def listkeys(self, namespace):
         return self._repo.listkeys(namespace)
 
     def debugwireargs(self, one, two, three=None, four=None, five=None):
         '''used to test argument passing over the wire'''
         return "%s %s %s %s %s" % (one, two, three, four, five)
 
 class locallegacypeer(localpeer):
     '''peer extension which implements legacy methods too; used for tests with
     restricted capabilities'''
 
     def __init__(self, repo):
         localpeer.__init__(self, repo, caps=legacycaps)
 
     def branches(self, nodes):
         return self._repo.branches(nodes)
 
     def between(self, pairs):
         return self._repo.between(pairs)
 
     def changegroup(self, basenodes, source):
         return changegroup.changegroup(self._repo, basenodes, source)
 
     def changegroupsubset(self, bases, heads, source):
         return changegroup.changegroupsubset(self._repo, bases, heads, source)
 
 # Increment the sub-version when the revlog v2 format changes to lock out old
 # clients.
 REVLOGV2_REQUIREMENT = 'exp-revlogv2.0'
 
 class localrepository(object):
 
     supportedformats = {
         'revlogv1',
         'generaldelta',
         'treemanifest',
         'manifestv2',
         REVLOGV2_REQUIREMENT,
     }
     _basesupported = supportedformats | {
         'store',
         'fncache',
         'shared',
         'relshared',
         'dotencode',
     }
     openerreqs = {
         'revlogv1',
         'generaldelta',
         'treemanifest',
         'manifestv2',
     }
 
     # a list of (ui, featureset) functions.
     # only functions defined in module of enabled extensions are invoked
     featuresetupfuncs = set()
 
     def __init__(self, baseui, path, create=False):
         self.requirements = set()
         self.filtername = None
         # wvfs: rooted at the repository root, used to access the working copy
         self.wvfs = vfsmod.vfs(path, expandpath=True, realpath=True)
         # vfs: rooted at .hg, used to access repo files outside of .hg/store
         self.vfs = None
         # svfs: usually rooted at .hg/store, used to access repository history
         # If this is a shared repository, this vfs may point to another
         # repository's .hg/store directory.
         self.svfs = None
         self.root = self.wvfs.base
         self.path = self.wvfs.join(".hg")
         self.origroot = path
         self.auditor = pathutil.pathauditor(self.root, self._checknested)
         self.nofsauditor = pathutil.pathauditor(self.root, self._checknested,
                                                 realfs=False)
         self.vfs = vfsmod.vfs(self.path)
         self.baseui = baseui
         self.ui = baseui.copy()
         self.ui.copy = baseui.copy # prevent copying repo configuration
         # A list of callback to shape the phase if no data were found.
         # Callback are in the form: func(repo, roots) --> processed root.
         # This list it to be filled by extension during repo setup
         self._phasedefaults = []
         try:
             self.ui.readconfig(self.vfs.join("hgrc"), self.root)
             self._loadextensions()
         except IOError:
             pass
 
         if self.featuresetupfuncs:
             self.supported = set(self._basesupported) # use private copy
             extmods = set(m.__name__ for n, m
                           in extensions.extensions(self.ui))
             for setupfunc in self.featuresetupfuncs:
                 if setupfunc.__module__ in extmods:
                     setupfunc(self.ui, self.supported)
         else:
             self.supported = self._basesupported
         color.setup(self.ui)
 
         # Add compression engines.
         for name in util.compengines:
             engine = util.compengines[name]
             if engine.revlogheader():
                 self.supported.add('exp-compression-%s' % name)
 
         if not self.vfs.isdir():
             if create:
                 self.requirements = newreporequirements(self)
 
                 if not self.wvfs.exists():
                     self.wvfs.makedirs()
                 self.vfs.makedir(notindexed=True)
 
                 if 'store' in self.requirements:
                     self.vfs.mkdir("store")
 
                     # create an invalid changelog
                     self.vfs.append(
                         "00changelog.i",
                         '\0\0\0\2' # represents revlogv2
                         ' dummy changelog to prevent using the old repo layout'
                     )
             else:
                 raise error.RepoError(_("repository %s not found") % path)
         elif create:
             raise error.RepoError(_("repository %s already exists") % path)
         else:
             try:
                 self.requirements = scmutil.readrequires(
                         self.vfs, self.supported)
             except IOError as inst:
                 if inst.errno != errno.ENOENT:
                     raise
 
         self.sharedpath = self.path
         try:
             sharedpath = self.vfs.read("sharedpath").rstrip('\n')
             if 'relshared' in self.requirements:
                 sharedpath = self.vfs.join(sharedpath)
             vfs = vfsmod.vfs(sharedpath, realpath=True)
             s = vfs.base
             if not vfs.exists():
                 raise error.RepoError(
                     _('.hg/sharedpath points to nonexistent directory %s') % s)
             self.sharedpath = s
         except IOError as inst:
             if inst.errno != errno.ENOENT:
                 raise
 
         self.store = store.store(
                 self.requirements, self.sharedpath, vfsmod.vfs)
         self.spath = self.store.path
         self.svfs = self.store.vfs
         self.sjoin = self.store.join
         self.vfs.createmode = self.store.createmode
         self._applyopenerreqs()
         if create:
             self._writerequirements()
 
         self._dirstatevalidatewarned = False
 
         self._branchcaches = {}
         self._revbranchcache = None
         self.filterpats = {}
         self._datafilters = {}
         self._transref = self._lockref = self._wlockref = None
 
         # A cache for various files under .hg/ that tracks file changes,
         # (used by the filecache decorator)
         #
         # Maps a property name to its util.filecacheentry
         self._filecache = {}
 
         # hold sets of revision to be filtered
         # should be cleared when something might have changed the filter value:
         # - new changesets,
         # - phase change,
         # - new obsolescence marker,
         # - working directory parent change,
         # - bookmark changes
         self.filteredrevcache = {}
 
         # post-dirstate-status hooks
         self._postdsstatus = []
 
         # generic mapping between names and nodes
         self.names = namespaces.namespaces()
 
     def close(self):
         self._writecaches()
 
     def _loadextensions(self):
         extensions.loadall(self.ui)
 
     def _writecaches(self):
         if self._revbranchcache:
             self._revbranchcache.write()
 
     def _restrictcapabilities(self, caps):
         if self.ui.configbool('experimental', 'bundle2-advertise', True):
             caps = set(caps)
             capsblob = bundle2.encodecaps(bundle2.getrepocaps(self))
             caps.add('bundle2=' + urlreq.quote(capsblob))
         return caps
 
     def _applyopenerreqs(self):
         self.svfs.options = dict((r, 1) for r in self.requirements
                                            if r in self.openerreqs)
         # experimental config: format.chunkcachesize
         chunkcachesize = self.ui.configint('format', 'chunkcachesize')
         if chunkcachesize is not None:
             self.svfs.options['chunkcachesize'] = chunkcachesize
         # experimental config: format.maxchainlen
         maxchainlen = self.ui.configint('format', 'maxchainlen')
         if maxchainlen is not None:
             self.svfs.options['maxchainlen'] = maxchainlen
         # experimental config: format.manifestcachesize
         manifestcachesize = self.ui.configint('format', 'manifestcachesize')
         if manifestcachesize is not None:
             self.svfs.options['manifestcachesize'] = manifestcachesize
         # experimental config: format.aggressivemergedeltas
         aggressivemergedeltas = self.ui.configbool('format',
             'aggressivemergedeltas', False)
         self.svfs.options['aggressivemergedeltas'] = aggressivemergedeltas
         self.svfs.options['lazydeltabase'] = not scmutil.gddeltaconfig(self.ui)
+        chainspan = self.ui.configbytes('experimental', 'maxdeltachainspan', -1)
+        if 0 <= chainspan:
+            self.svfs.options['maxdeltachainspan'] = chainspan
 
         for r in self.requirements:
             if r.startswith('exp-compression-'):
                 self.svfs.options['compengine'] = r[len('exp-compression-'):]
 
         # TODO move "revlogv2" to openerreqs once finalized.
         if REVLOGV2_REQUIREMENT in self.requirements:
             self.svfs.options['revlogv2'] = True
 
     def _writerequirements(self):
         scmutil.writerequires(self.vfs, self.requirements)
 
     def _checknested(self, path):
         """Determine if path is a legal nested repository."""
         if not path.startswith(self.root):
             return False
         subpath = path[len(self.root) + 1:]
         normsubpath = util.pconvert(subpath)
 
         # XXX: Checking against the current working copy is wrong in
         # the sense that it can reject things like
         #
         #   $ hg cat -r 10 sub/x.txt
         #
         # if sub/ is no longer a subrepository in the working copy
         # parent revision.
         #
         # However, it can of course also allow things that would have
         # been rejected before, such as the above cat command if sub/
         # is a subrepository now, but was a normal directory before.
         # The old path auditor would have rejected by mistake since it
         # panics when it sees sub/.hg/.
         #
         # All in all, checking against the working copy seems sensible
         # since we want to prevent access to nested repositories on
         # the filesystem *now*.
         ctx = self[None]
         parts = util.splitpath(subpath)
         while parts:
             prefix = '/'.join(parts)
             if prefix in ctx.substate:
                 if prefix == normsubpath:
                     return True
                 else:
                     sub = ctx.sub(prefix)
                     return sub.checknested(subpath[len(prefix) + 1:])
             else:
                 parts.pop()
         return False
 
     def peer(self):
         return localpeer(self) # not cached to avoid reference cycle
 
     def unfiltered(self):
         """Return unfiltered version of the repository
 
         Intended to be overwritten by filtered repo."""
         return self
 
     def filtered(self, name):
         """Return a filtered version of a repository"""
         # build a new class with the mixin and the current class
         # (possibly subclass of the repo)
         class filteredrepo(repoview.repoview, self.unfiltered().__class__):
             pass
         return filteredrepo(self, name)
 
     @repofilecache('bookmarks', 'bookmarks.current')
     def _bookmarks(self):
         return bookmarks.bmstore(self)
 
     @property
     def _activebookmark(self):
         return self._bookmarks.active
 
     # _phaserevs and _phasesets depend on changelog. what we need is to
     # call _phasecache.invalidate() if '00changelog.i' was changed, but it
     # can't be easily expressed in filecache mechanism.
     @storecache('phaseroots', '00changelog.i')
     def _phasecache(self):
         return phases.phasecache(self, self._phasedefaults)
 
     @storecache('obsstore')
     def obsstore(self):
         return obsolete.makestore(self.ui, self)
 
     @storecache('00changelog.i')
     def changelog(self):
         return changelog.changelog(self.svfs,
                                    trypending=txnutil.mayhavepending(self.root))
 
     def _constructmanifest(self):
         # This is a temporary function while we migrate from manifest to
         # manifestlog. It allows bundlerepo and unionrepo to intercept the
         # manifest creation.
         return manifest.manifestrevlog(self.svfs)
 
     @storecache('00manifest.i')
     def manifestlog(self):
         return manifest.manifestlog(self.svfs, self)
 
     @repofilecache('dirstate')
     def dirstate(self):
         return dirstate.dirstate(self.vfs, self.ui, self.root,
                                  self._dirstatevalidate)
 
     def _dirstatevalidate(self, node):
         try:
             self.changelog.rev(node)
             return node
         except error.LookupError:
             if not self._dirstatevalidatewarned:
                 self._dirstatevalidatewarned = True
                 self.ui.warn(_("warning: ignoring unknown"
                                " working parent %s!\n") % short(node))
             return nullid
 
     def __getitem__(self, changeid):
         if changeid is None:
             return context.workingctx(self)
         if isinstance(changeid, slice):
             # wdirrev isn't contiguous so the slice shouldn't include it
             return [context.changectx(self, i)
                     for i in xrange(*changeid.indices(len(self)))
                     if i not in self.changelog.filteredrevs]
         try:
             return context.changectx(self, changeid)
         except error.WdirUnsupported:
             return context.workingctx(self)
 
     def __contains__(self, changeid):
         """True if the given changeid exists
 
         error.LookupError is raised if an ambiguous node specified.
         """
         try:
             self[changeid]
             return True
         except error.RepoLookupError:
             return False
 
     def __nonzero__(self):
         return True
 
     __bool__ = __nonzero__
 
     def __len__(self):
         return len(self.changelog)
 
     def __iter__(self):
         return iter(self.changelog)
 
     def revs(self, expr, *args):
         '''Find revisions matching a revset.
 
         The revset is specified as a string ``expr`` that may contain
         %-formatting to escape certain types. See ``revsetlang.formatspec``.
 
         Revset aliases from the configuration are not expanded. To expand
         user aliases, consider calling ``scmutil.revrange()`` or
         ``repo.anyrevs([expr], user=True)``.
 
         Returns a revset.abstractsmartset, which is a list-like interface
         that contains integer revisions.
         '''
         expr = revsetlang.formatspec(expr, *args)
         m = revset.match(None, expr)
         return m(self)
 
     def set(self, expr, *args):
         '''Find revisions matching a revset and emit changectx instances.
 
         This is a convenience wrapper around ``revs()`` that iterates the
         result and is a generator of changectx instances.
 
         Revset aliases from the configuration are not expanded. To expand
         user aliases, consider calling ``scmutil.revrange()``.
         '''
         for r in self.revs(expr, *args):
             yield self[r]
 
     def anyrevs(self, specs, user=False):
         '''Find revisions matching one of the given revsets.
 
         Revset aliases from the configuration are not expanded by default. To
         expand user aliases, specify ``user=True``.
         '''
         if user:
             m = revset.matchany(self.ui, specs, repo=self)
         else:
             m = revset.matchany(None, specs)
         return m(self)
 
     def url(self):
         return 'file:' + self.root
 
     def hook(self, name, throw=False, **args):
         """Call a hook, passing this repo instance.
 
         This a convenience method to aid invoking hooks. Extensions likely
         won't call this unless they have registered a custom hook or are
         replacing code that is expected to call a hook.
         """
         return hook.hook(self.ui, self, name, throw, **args)
 
     @filteredpropertycache
     def _tagscache(self):
         '''Returns a tagscache object that contains various tags related
         caches.'''
 
         # This simplifies its cache management by having one decorated
         # function (this one) and the rest simply fetch things from it.
         class tagscache(object):
             def __init__(self):
                 # These two define the set of tags for this repository. tags
                 # maps tag name to node; tagtypes maps tag name to 'global' or
                 # 'local'. (Global tags are defined by .hgtags across all
                 # heads, and local tags are defined in .hg/localtags.)
                 # They constitute the in-memory cache of tags.
                 self.tags = self.tagtypes = None
 
                 self.nodetagscache = self.tagslist = None
 
         cache = tagscache()
         cache.tags, cache.tagtypes = self._findtags()
 
         return cache
 
     def tags(self):
         '''return a mapping of tag to node'''
         t = {}
         if self.changelog.filteredrevs:
             tags, tt = self._findtags()
         else:
             tags = self._tagscache.tags
         for k, v in tags.iteritems():
             try:
                 # ignore tags to unknown nodes
                 self.changelog.rev(v)
                 t[k] = v
             except (error.LookupError, ValueError):
                 pass
         return t
 
     def _findtags(self):
         '''Do the hard work of finding tags.  Return a pair of dicts
         (tags, tagtypes) where tags maps tag name to node, and tagtypes
         maps tag name to a string like \'global\' or \'local\'.
         Subclasses or extensions are free to add their own tags, but
         should be aware that the returned dicts will be retained for the
         duration of the localrepo object.'''
 
         # XXX what tagtype should subclasses/extensions use?  Currently
         # mq and bookmarks add tags, but do not set the tagtype at all.
         # Should each extension invent its own tag type?  Should there
         # be one tagtype for all such "virtual" tags?  Or is the status
         # quo fine?
 
 
         # map tag name to (node, hist)
         alltags = tagsmod.findglobaltags(self.ui, self)
         # map tag name to tag type
         tagtypes = dict((tag, 'global') for tag in alltags)
 
         tagsmod.readlocaltags(self.ui, self, alltags, tagtypes)
 
         # Build the return dicts.  Have to re-encode tag names because
         # the tags module always uses UTF-8 (in order not to lose info
         # writing to the cache), but the rest of Mercurial wants them in
         # local encoding.
         tags = {}
         for (name, (node, hist)) in alltags.iteritems():
             if node != nullid:
                 tags[encoding.tolocal(name)] = node
         tags['tip'] = self.changelog.tip()
         tagtypes = dict([(encoding.tolocal(name), value)
                          for (name, value) in tagtypes.iteritems()])
         return (tags, tagtypes)
 
     def tagtype(self, tagname):
         '''
         return the type of the given tag. result can be:
 
         'local'  : a local tag
         'global' : a global tag
         None     : tag does not exist
         '''
 
         return self._tagscache.tagtypes.get(tagname)
 
     def tagslist(self):
         '''return a list of tags ordered by revision'''
         if not self._tagscache.tagslist:
             l = []
             for t, n in self.tags().iteritems():
                 l.append((self.changelog.rev(n), t, n))
             self._tagscache.tagslist = [(t, n) for r, t, n in sorted(l)]
 
         return self._tagscache.tagslist
 
     def nodetags(self, node):
         '''return the tags associated with a node'''
         if not self._tagscache.nodetagscache:
             nodetagscache = {}
             for t, n in self._tagscache.tags.iteritems():
                 nodetagscache.setdefault(n, []).append(t)
             for tags in nodetagscache.itervalues():
                 tags.sort()
             self._tagscache.nodetagscache = nodetagscache
         return self._tagscache.nodetagscache.get(node, [])
 
     def nodebookmarks(self, node):
         """return the list of bookmarks pointing to the specified node"""
         marks = []
         for bookmark, n in self._bookmarks.iteritems():
             if n == node:
                 marks.append(bookmark)
         return sorted(marks)
 
     def branchmap(self):
         '''returns a dictionary {branch: [branchheads]} with branchheads
         ordered by increasing revision number'''
         branchmap.updatecache(self)
         return self._branchcaches[self.filtername]
 
     @unfilteredmethod
     def revbranchcache(self):
         if not self._revbranchcache:
             self._revbranchcache = branchmap.revbranchcache(self.unfiltered())
         return self._revbranchcache
 
     def branchtip(self, branch, ignoremissing=False):
         '''return the tip node for a given branch
 
         If ignoremissing is True, then this method will not raise an error.
         This is helpful for callers that only expect None for a missing branch
         (e.g. namespace).
 
         '''
         try:
             return self.branchmap().branchtip(branch)
         except KeyError:
             if not ignoremissing:
                 raise error.RepoLookupError(_("unknown branch '%s'") % branch)
             else:
                 pass
 
     def lookup(self, key):
         return self[key].node()
 
     def lookupbranch(self, key, remote=None):
         repo = remote or self
         if key in repo.branchmap():
             return key
 
         repo = (remote and remote.local()) and remote or self
         return repo[key].branch()
 
     def known(self, nodes):
         cl = self.changelog
         nm = cl.nodemap
         filtered = cl.filteredrevs
         result = []
         for n in nodes:
             r = nm.get(n)
             resp = not (r is None or r in filtered)
             result.append(resp)
         return result
 
     def local(self):
         return self
 
     def publishing(self):
         # it's safe (and desirable) to trust the publish flag unconditionally
         # so that we don't finalize changes shared between users via ssh or nfs
         return self.ui.configbool('phases', 'publish', True, untrusted=True)
 
     def cancopy(self):
         # so statichttprepo's override of local() works
         if not self.local():
             return False
         if not self.publishing():
             return True
         # if publishing we can't copy if there is filtered content
         return not self.filtered('visible').changelog.filteredrevs
 
     def shared(self):
         '''the type of shared repository (None if not shared)'''
         if self.sharedpath != self.path:
             return 'store'
         return None
 
     def wjoin(self, f, *insidef):
         return self.vfs.reljoin(self.root, f, *insidef)
 
     def file(self, f):
         if f[0] == '/':
             f = f[1:]
         return filelog.filelog(self.svfs, f)
 
     def changectx(self, changeid):
         return self[changeid]
 
     def setparents(self, p1, p2=nullid):
         with self.dirstate.parentchange():
             copies = self.dirstate.setparents(p1, p2)
             pctx = self[p1]
             if copies:
                 # Adjust copy records, the dirstate cannot do it, it
                 # requires access to parents manifests. Preserve them
                 # only for entries added to first parent.
                 for f in copies:
                     if f not in pctx and copies[f] in pctx:
                         self.dirstate.copy(copies[f], f)
             if p2 == nullid:
                 for f, s in sorted(self.dirstate.copies().items()):
                     if f not in pctx and s not in pctx:
                         self.dirstate.copy(None, f)
 
     def filectx(self, path, changeid=None, fileid=None):
         """changeid can be a changeset revision, node, or tag.
            fileid can be a file revision or node."""
         return context.filectx(self, path, changeid, fileid)
 
     def getcwd(self):
         return self.dirstate.getcwd()
 
     def pathto(self, f, cwd=None):
         return self.dirstate.pathto(f, cwd)
 
     def _loadfilter(self, filter):
         if filter not in self.filterpats:
             l = []
             for pat, cmd in self.ui.configitems(filter):
                 if cmd == '!':
                     continue
                 mf = matchmod.match(self.root, '', [pat])
                 fn = None
                 params = cmd
                 for name, filterfn in self._datafilters.iteritems():
                     if cmd.startswith(name):
                         fn = filterfn
                         params = cmd[len(name):].lstrip()
                         break
                 if not fn:
                     fn = lambda s, c, **kwargs: util.filter(s, c)
                 # Wrap old filters not supporting keyword arguments
                 if not inspect.getargspec(fn)[2]:
                     oldfn = fn
                     fn = lambda s, c, **kwargs: oldfn(s, c)
                 l.append((mf, fn, params))
             self.filterpats[filter] = l
         return self.filterpats[filter]
 
     def _filter(self, filterpats, filename, data):
         for mf, fn, cmd in filterpats:
             if mf(filename):
                 self.ui.debug("filtering %s through %s\n" % (filename, cmd))
                 data = fn(data, cmd, ui=self.ui, repo=self, filename=filename)
                 break
 
         return data
 
     @unfilteredpropertycache
     def _encodefilterpats(self):
         return self._loadfilter('encode')
 
     @unfilteredpropertycache
     def _decodefilterpats(self):
         return self._loadfilter('decode')
 
     def adddatafilter(self, name, filter):
         self._datafilters[name] = filter
 
     def wread(self, filename):
         if self.wvfs.islink(filename):
             data = self.wvfs.readlink(filename)
         else:
             data = self.wvfs.read(filename)
         return self._filter(self._encodefilterpats, filename, data)
 
     def wwrite(self, filename, data, flags, backgroundclose=False):
         """write ``data`` into ``filename`` in the working directory
 
         This returns length of written (maybe decoded) data.
         """
         data = self._filter(self._decodefilterpats, filename, data)
         if 'l' in flags:
             self.wvfs.symlink(data, filename)
         else:
             self.wvfs.write(filename, data, backgroundclose=backgroundclose)
             if 'x' in flags:
                 self.wvfs.setflags(filename, False, True)
         return len(data)
 
     def wwritedata(self, filename, data):
         return self._filter(self._decodefilterpats, filename, data)
 
     def currenttransaction(self):
         """return the current transaction or None if non exists"""
         if self._transref:
             tr = self._transref()
         else:
             tr = None
 
         if tr and tr.running():
             return tr
         return None
 
     def transaction(self, desc, report=None):
         if (self.ui.configbool('devel', 'all-warnings')
                 or self.ui.configbool('devel', 'check-locks')):
             if self._currentlock(self._lockref) is None:
                 raise error.ProgrammingError('transaction requires locking')
         tr = self.currenttransaction()
         if tr is not None:
             return tr.nest()
 
         # abort here if the journal already exists
         if self.svfs.exists("journal"):
             raise error.RepoError(
                 _("abandoned transaction found"),
                 hint=_("run 'hg recover' to clean up transaction"))
 
         idbase = "%.40f#%f" % (random.random(), time.time())
         ha = hex(hashlib.sha1(idbase).digest())
         txnid = 'TXN:' + ha
         self.hook('pretxnopen', throw=True, txnname=desc, txnid=txnid)
 
         self._writejournal(desc)
         renames = [(vfs, x, undoname(x)) for vfs, x in self._journalfiles()]
         if report:
             rp = report
         else:
             rp = self.ui.warn
         vfsmap = {'plain': self.vfs} # root of .hg/
         # we must avoid cyclic reference between repo and transaction.
         reporef = weakref.ref(self)
         # Code to track tag movement
         #
         # Since tags are all handled as file content, it is actually quite hard
         # to track these movement from a code perspective. So we fallback to a
         # tracking at the repository level. One could envision to track changes
         # to the '.hgtags' file through changegroup apply but that fails to
         # cope with case where transaction expose new heads without changegroup
         # being involved (eg: phase movement).
         #
         # For now, We gate the feature behind a flag since this likely comes
         # with performance impacts. The current code run more often than needed
         # and do not use caches as much as it could.  The current focus is on
         # the behavior of the feature so we disable it by default. The flag
         # will be removed when we are happy with the performance impact.
         #
         # Once this feature is no longer experimental move the following
         # documentation to the appropriate help section:
         #
         # The ``HG_TAG_MOVED`` variable will be set if the transaction touched
         # tags (new or changed or deleted tags). In addition the details of
         # these changes are made available in a file at:
         #     ``REPOROOT/.hg/changes/tags.changes``.
         # Make sure you check for HG_TAG_MOVED before reading that file as it
         # might exist from a previous transaction even if no tag were touched
         # in this one. Changes are recorded in a line base format::
         #
         #     <action> <hex-node> <tag-name>\n
         #
         # Actions are defined as follow:
         #   "-R": tag is removed,
         #   "+A": tag is added,
         #   "-M": tag is moved (old value),
         #   "+M": tag is moved (new value),
         tracktags = lambda x: None
         # experimental config: experimental.hook-track-tags
         shouldtracktags = self.ui.configbool('experimental', 'hook-track-tags',
                                              False)
         if desc != 'strip' and shouldtracktags:
             oldheads = self.changelog.headrevs()
             def tracktags(tr2):
                 repo = reporef()
                 oldfnodes = tagsmod.fnoderevs(repo.ui, repo, oldheads)
                 newheads = repo.changelog.headrevs()
                 newfnodes = tagsmod.fnoderevs(repo.ui, repo, newheads)
                 # notes: we compare lists here.
                 # As we do it only once buiding set would not be cheaper
                 changes = tagsmod.difftags(repo.ui, repo, oldfnodes, newfnodes)
                 if changes:
                     tr2.hookargs['tag_moved'] = '1'
                     with repo.vfs('changes/tags.changes', 'w',
                                   atomictemp=True) as changesfile:
                         # note: we do not register the file to the transaction
                         # because we needs it to still exist on the transaction
                         # is close (for txnclose hooks)
                         tagsmod.writediff(changesfile, changes)
         def validate(tr2):
             """will run pre-closing hooks"""
             # XXX the transaction API is a bit lacking here so we take a hacky
             # path for now
             #
             # We cannot add this as a "pending" hooks since the 'tr.hookargs'
             # dict is copied before these run. In addition we needs the data
             # available to in memory hooks too.
             #
             # Moreover, we also need to make sure this runs before txnclose
             # hooks and there is no "pending" mechanism that would execute
             # logic only if hooks are about to run.
             #
             # Fixing this limitation of the transaction is also needed to track
             # other families of changes (bookmarks, phases, obsolescence).
             #
             # This will have to be fixed before we remove the experimental
             # gating.
             tracktags(tr2)
             reporef().hook('pretxnclose', throw=True,
                            txnname=desc, **pycompat.strkwargs(tr.hookargs))
         def releasefn(tr, success):
             repo = reporef()
             if success:
                 # this should be explicitly invoked here, because
                 # in-memory changes aren't written out at closing
                 # transaction, if tr.addfilegenerator (via
                 # dirstate.write or so) isn't invoked while
                 # transaction running
                 repo.dirstate.write(None)
             else:
                 # discard all changes (including ones already written
                 # out) in this transaction
                 repo.dirstate.restorebackup(None, prefix='journal.')
 
                 repo.invalidate(clearfilecache=True)
 
         tr = transaction.transaction(rp, self.svfs, vfsmap,
                                      "journal",
                                      "undo",
                                      aftertrans(renames),
                                      self.store.createmode,
                                      validator=validate,
                                      releasefn=releasefn)
         tr.changes['revs'] = set()
 
         tr.hookargs['txnid'] = txnid
         # note: writing the fncache only during finalize mean that the file is
         # outdated when running hooks. As fncache is used for streaming clone,
         # this is not expected to break anything that happen during the hooks.
         tr.addfinalize('flush-fncache', self.store.write)
         def txnclosehook(tr2):
             """To be run if transaction is successful, will schedule a hook run
             """
             # Don't reference tr2 in hook() so we don't hold a reference.
             # This reduces memory consumption when there are multiple
             # transactions per lock. This can likely go away if issue5045
             # fixes the function accumulation.
             hookargs = tr2.hookargs
 
             def hook():
                 reporef().hook('txnclose', throw=False, txnname=desc,
                                **pycompat.strkwargs(hookargs))
             reporef()._afterlock(hook)
         tr.addfinalize('txnclose-hook', txnclosehook)
         tr.addpostclose('warms-cache', self._buildcacheupdater(tr))
         def txnaborthook(tr2):
             """To be run if transaction is aborted
             """
             reporef().hook('txnabort', throw=False, txnname=desc,
                            **tr2.hookargs)
         tr.addabort('txnabort-hook', txnaborthook)
         # avoid eager cache invalidation. in-memory data should be identical
         # to stored data if transaction has no error.
         tr.addpostclose('refresh-filecachestats', self._refreshfilecachestats)
         self._transref = weakref.ref(tr)
         return tr
 
     def _journalfiles(self):
         return ((self.svfs, 'journal'),
                 (self.vfs, 'journal.dirstate'),
                 (self.vfs, 'journal.branch'),
                 (self.vfs, 'journal.desc'),
                 (self.vfs, 'journal.bookmarks'),
                 (self.svfs, 'journal.phaseroots'))
 
     def undofiles(self):
         return [(vfs, undoname(x)) for vfs, x in self._journalfiles()]
 
     @unfilteredmethod
     def _writejournal(self, desc):
         self.dirstate.savebackup(None, prefix='journal.')
         self.vfs.write("journal.branch",
                           encoding.fromlocal(self.dirstate.branch()))
         self.vfs.write("journal.desc",
                           "%d\n%s\n" % (len(self), desc))
         self.vfs.write("journal.bookmarks",
                           self.vfs.tryread("bookmarks"))
         self.svfs.write("journal.phaseroots",
                            self.svfs.tryread("phaseroots"))
 
     def recover(self):
         with self.lock():
             if self.svfs.exists("journal"):
                 self.ui.status(_("rolling back interrupted transaction\n"))
                 vfsmap = {'': self.svfs,
                           'plain': self.vfs,}
                 transaction.rollback(self.svfs, vfsmap, "journal",
                                      self.ui.warn)
                 self.invalidate()
                 return True
             else:
                 self.ui.warn(_("no interrupted transaction available\n"))
                 return False
 
     def rollback(self, dryrun=False, force=False):
         wlock = lock = dsguard = None
         try:
             wlock = self.wlock()
             lock = self.lock()
             if self.svfs.exists("undo"):
                 dsguard = dirstateguard.dirstateguard(self, 'rollback')
 
                 return self._rollback(dryrun, force, dsguard)
             else:
                 self.ui.warn(_("no rollback information available\n"))
                 return 1
         finally:
             release(dsguard, lock, wlock)
 
     @unfilteredmethod # Until we get smarter cache management
     def _rollback(self, dryrun, force, dsguard):
         ui = self.ui
         try:
             args = self.vfs.read('undo.desc').splitlines()
             (oldlen, desc, detail) = (int(args[0]), args[1], None)
             if len(args) >= 3:
                 detail = args[2]
             oldtip = oldlen - 1
 
             if detail and ui.verbose:
                 msg = (_('repository tip rolled back to revision %d'
                          ' (undo %s: %s)\n')
                        % (oldtip, desc, detail))
             else:
                 msg = (_('repository tip rolled back to revision %d'
                          ' (undo %s)\n')
                        % (oldtip, desc))
         except IOError:
             msg = _('rolling back unknown transaction\n')
             desc = None
 
         if not force and self['.'] != self['tip'] and desc == 'commit':
             raise error.Abort(
                 _('rollback of last commit while not checked out '
                   'may lose data'), hint=_('use -f to force'))
 
         ui.status(msg)
         if dryrun:
             return 0
 
         parents = self.dirstate.parents()
         self.destroying()
         vfsmap = {'plain': self.vfs, '': self.svfs}
         transaction.rollback(self.svfs, vfsmap, 'undo', ui.warn)
         if self.vfs.exists('undo.bookmarks'):
             self.vfs.rename('undo.bookmarks', 'bookmarks', checkambig=True)
         if self.svfs.exists('undo.phaseroots'):
             self.svfs.rename('undo.phaseroots', 'phaseroots', checkambig=True)
         self.invalidate()
 
         parentgone = (parents[0] not in self.changelog.nodemap or
                       parents[1] not in self.changelog.nodemap)
         if parentgone:
             # prevent dirstateguard from overwriting already restored one
             dsguard.close()
 
             self.dirstate.restorebackup(None, prefix='undo.')
             try:
                 branch = self.vfs.read('undo.branch')
                 self.dirstate.setbranch(encoding.tolocal(branch))
             except IOError:
                 ui.warn(_('named branch could not be reset: '
                           'current branch is still \'%s\'\n')
                         % self.dirstate.branch())
 
             parents = tuple([p.rev() for p in self[None].parents()])
             if len(parents) > 1:
                 ui.status(_('working directory now based on '
                             'revisions %d and %d\n') % parents)
             else:
                 ui.status(_('working directory now based on '
                             'revision %d\n') % parents)
             mergemod.mergestate.clean(self, self['.'].node())
 
         # TODO: if we know which new heads may result from this rollback, pass
         # them to destroy(), which will prevent the branchhead cache from being
         # invalidated.
         self.destroyed()
         return 0
 
     def _buildcacheupdater(self, newtransaction):
         """called during transaction to build the callback updating cache
 
         Lives on the repository to help extension who might want to augment
         this logic. For this purpose, the created transaction is passed to the
         method.
         """
         # we must avoid cyclic reference between repo and transaction.
         reporef = weakref.ref(self)
         def updater(tr):
             repo = reporef()
             repo.updatecaches(tr)
         return updater
 
     @unfilteredmethod
     def updatecaches(self, tr=None):
         """warm appropriate caches
 
         If this function is called after a transaction closed. The transaction
         will be available in the 'tr' argument. This can be used to selectively
         update caches relevant to the changes in that transaction.
         """
         if tr is not None and tr.hookargs.get('source') == 'strip':
             # During strip, many caches are invalid but
             # later call to `destroyed` will refresh them.
             return
 
         if tr is None or tr.changes['revs']:
             # updating the unfiltered branchmap should refresh all the others,
             self.ui.debug('updating the branch cache\n')
             branchmap.updatecache(self.filtered('served'))
 
     def invalidatecaches(self):
 
         if '_tagscache' in vars(self):
             # can't use delattr on proxy
             del self.__dict__['_tagscache']
 
         self.unfiltered()._branchcaches.clear()
         self.invalidatevolatilesets()
 
     def invalidatevolatilesets(self):
         self.filteredrevcache.clear()
         obsolete.clearobscaches(self)
 
     def invalidatedirstate(self):
         '''Invalidates the dirstate, causing the next call to dirstate
         to check if it was modified since the last time it was read,
         rereading it if it has.
 
         This is different to dirstate.invalidate() that it doesn't always
         rereads the dirstate. Use dirstate.invalidate() if you want to
         explicitly read the dirstate again (i.e. restoring it to a previous
         known good state).'''
         if hasunfilteredcache(self, 'dirstate'):
             for k in self.dirstate._filecache:
                 try:
                     delattr(self.dirstate, k)
                 except AttributeError:
                     pass
             delattr(self.unfiltered(), 'dirstate')
 
     def invalidate(self, clearfilecache=False):
         '''Invalidates both store and non-store parts other than dirstate
 
         If a transaction is running, invalidation of store is omitted,
         because discarding in-memory changes might cause inconsistency
         (e.g. incomplete fncache causes unintentional failure, but
         redundant one doesn't).
         '''
         unfiltered = self.unfiltered() # all file caches are stored unfiltered
         for k in list(self._filecache.keys()):
             # dirstate is invalidated separately in invalidatedirstate()
             if k == 'dirstate':
                 continue
 
             if clearfilecache:
                 del self._filecache[k]
             try:
                 delattr(unfiltered, k)
             except AttributeError:
                 pass
         self.invalidatecaches()
         if not self.currenttransaction():
             # TODO: Changing contents of store outside transaction
             # causes inconsistency. We should make in-memory store
             # changes detectable, and abort if changed.
             self.store.invalidatecaches()
 
     def invalidateall(self):
         '''Fully invalidates both store and non-store parts, causing the
         subsequent operation to reread any outside changes.'''
         # extension should hook this to invalidate its caches
         self.invalidate()
         self.invalidatedirstate()
 
     @unfilteredmethod
     def _refreshfilecachestats(self, tr):
         """Reload stats of cached files so that they are flagged as valid"""
         for k, ce in self._filecache.items():
             if k == 'dirstate' or k not in self.__dict__:
                 continue
             ce.refresh()
 
     def _lock(self, vfs, lockname, wait, releasefn, acquirefn, desc,
               inheritchecker=None, parentenvvar=None):
         parentlock = None
         # the contents of parentenvvar are used by the underlying lock to
         # determine whether it can be inherited
         if parentenvvar is not None:
             parentlock = encoding.environ.get(parentenvvar)
         try:
             l = lockmod.lock(vfs, lockname, 0, releasefn=releasefn,
                              acquirefn=acquirefn, desc=desc,
                              inheritchecker=inheritchecker,
                              parentlock=parentlock)
         except error.LockHeld as inst:
             if not wait:
                 raise
             # show more details for new-style locks
             if ':' in inst.locker:
                 host, pid = inst.locker.split(":", 1)
                 self.ui.warn(
                     _("waiting for lock on %s held by process %r "
                       "on host %r\n") % (desc, pid, host))
             else:
                 self.ui.warn(_("waiting for lock on %s held by %r\n") %
                              (desc, inst.locker))
             # default to 600 seconds timeout
             l = lockmod.lock(vfs, lockname,
                              int(self.ui.config("ui", "timeout", "600")),
                              releasefn=releasefn, acquirefn=acquirefn,
                              desc=desc)
             self.ui.warn(_("got lock after %s seconds\n") % l.delay)
         return l
 
     def _afterlock(self, callback):
         """add a callback to be run when the repository is fully unlocked
 
         The callback will be executed when the outermost lock is released
         (with wlock being higher level than 'lock')."""
         for ref in (self._wlockref, self._lockref):
             l = ref and ref()
             if l and l.held:
                 l.postrelease.append(callback)
                 break
         else: # no lock have been found.
             callback()
 
     def lock(self, wait=True):
         '''Lock the repository store (.hg/store) and return a weak reference
         to the lock. Use this before modifying the store (e.g. committing or
         stripping). If you are opening a transaction, get a lock as well.)
 
         If both 'lock' and 'wlock' must be acquired, ensure you always acquires
         'wlock' first to avoid a dead-lock hazard.'''
         l = self._currentlock(self._lockref)
         if l is not None:
             l.lock()
             return l
 
         l = self._lock(self.svfs, "lock", wait, None,
                        self.invalidate, _('repository %s') % self.origroot)
         self._lockref = weakref.ref(l)
         return l
 
     def _wlockchecktransaction(self):
         if self.currenttransaction() is not None:
             raise error.LockInheritanceContractViolation(
                 'wlock cannot be inherited in the middle of a transaction')
 
     def wlock(self, wait=True):
         '''Lock the non-store parts of the repository (everything under
         .hg except .hg/store) and return a weak reference to the lock.
 
         Use this before modifying files in .hg.
 
         If both 'lock' and 'wlock' must be acquired, ensure you always acquires
         'wlock' first to avoid a dead-lock hazard.'''
         l = self._wlockref and self._wlockref()
         if l is not None and l.held:
             l.lock()
             return l
 
         # We do not need to check for non-waiting lock acquisition.  Such
         # acquisition would not cause dead-lock as they would just fail.
         if wait and (self.ui.configbool('devel', 'all-warnings')
                      or self.ui.configbool('devel', 'check-locks')):
             if self._currentlock(self._lockref) is not None:
                 self.ui.develwarn('"wlock" acquired after "lock"')
 
         def unlock():
             if self.dirstate.pendingparentchange():
                 self.dirstate.invalidate()
             else:
                 self.dirstate.write(None)
 
             self._filecache['dirstate'].refresh()
 
         l = self._lock(self.vfs, "wlock", wait, unlock,
                        self.invalidatedirstate, _('working directory of %s') %
                        self.origroot,
                        inheritchecker=self._wlockchecktransaction,
                        parentenvvar='HG_WLOCK_LOCKER')
         self._wlockref = weakref.ref(l)
         return l
 
     def _currentlock(self, lockref):
         """Returns the lock if it's held, or None if it's not."""
         if lockref is None:
             return None
         l = lockref()
         if l is None or not l.held:
             return None
         return l
 
     def currentwlock(self):
         """Returns the wlock if it's held, or None if it's not."""
         return self._currentlock(self._wlockref)
 
     def _filecommit(self, fctx, manifest1, manifest2, linkrev, tr, changelist):
         """
         commit an individual file as part of a larger transaction
         """
 
         fname = fctx.path()
         fparent1 = manifest1.get(fname, nullid)
         fparent2 = manifest2.get(fname, nullid)
         if isinstance(fctx, context.filectx):
             node = fctx.filenode()
             if node in [fparent1, fparent2]:
                 self.ui.debug('reusing %s filelog entry\n' % fname)
                 if manifest1.flags(fname) != fctx.flags():
                     changelist.append(fname)
                 return node
 
         flog = self.file(fname)
         meta = {}
         copy = fctx.renamed()
         if copy and copy[0] != fname:
             # Mark the new revision of this file as a copy of another
             # file.  This copy data will effectively act as a parent
             # of this new revision.  If this is a merge, the first
             # parent will be the nullid (meaning "look up the copy data")
             # and the second one will be the other parent.  For example:
             #
             # 0 --- 1 --- 3   rev1 changes file foo
             #   \       /     rev2 renames foo to bar and changes it
             #    \- 2 -/      rev3 should have bar with all changes and
             #                      should record that bar descends from
             #                      bar in rev2 and foo in rev1
             #
             # this allows this merge to succeed:
             #
             # 0 --- 1 --- 3   rev4 reverts the content change from rev2
             #   \       /     merging rev3 and rev4 should use bar@rev2
             #    \- 2 --- 4        as the merge base
             #
 
             cfname = copy[0]
             crev = manifest1.get(cfname)
             newfparent = fparent2
 
             if manifest2: # branch merge
                 if fparent2 == nullid or crev is None: # copied on remote side
                     if cfname in manifest2:
                         crev = manifest2[cfname]
                         newfparent = fparent1
 
             # Here, we used to search backwards through history to try to find
             # where the file copy came from if the source of a copy was not in
             # the parent directory. However, this doesn't actually make sense to
             # do (what does a copy from something not in your working copy even
             # mean?) and it causes bugs (eg, issue4476). Instead, we will warn
             # the user that copy information was dropped, so if they didn't
             # expect this outcome it can be fixed, but this is the correct
             # behavior in this circumstance.
 
             if crev:
                 self.ui.debug(" %s: copy %s:%s\n" % (fname, cfname, hex(crev)))
                 meta["copy"] = cfname
                 meta["copyrev"] = hex(crev)
                 fparent1, fparent2 = nullid, newfparent
             else:
                 self.ui.warn(_("warning: can't find ancestor for '%s' "
                                "copied from '%s'!\n") % (fname, cfname))
 
         elif fparent1 == nullid:
             fparent1, fparent2 = fparent2, nullid
         elif fparent2 != nullid:
             # is one parent an ancestor of the other?
             fparentancestors = flog.commonancestorsheads(fparent1, fparent2)
             if fparent1 in fparentancestors:
                 fparent1, fparent2 = fparent2, nullid
             elif fparent2 in fparentancestors:
                 fparent2 = nullid
 
         # is the file changed?
         text = fctx.data()
         if fparent2 != nullid or flog.cmp(fparent1, text) or meta:
             changelist.append(fname)
             return flog.add(text, meta, tr, linkrev, fparent1, fparent2)
         # are just the flags changed during merge?
         elif fname in manifest1 and manifest1.flags(fname) != fctx.flags():
             changelist.append(fname)
 
         return fparent1
 
     def checkcommitpatterns(self, wctx, vdirs, match, status, fail):
         """check for commit arguments that aren't committable"""
         if match.isexact() or match.prefix():
             matched = set(status.modified + status.added + status.removed)
 
             for f in match.files():
                 f = self.dirstate.normalize(f)
                 if f == '.' or f in matched or f in wctx.substate:
                     continue
                 if f in status.deleted:
                     fail(f, _('file not found!'))
                 if f in vdirs: # visited directory
                     d = f + '/'
                     for mf in matched:
                         if mf.startswith(d):
                             break
                     else:
                         fail(f, _("no match under directory!"))
                 elif f not in self.dirstate:
                     fail(f, _("file not tracked!"))
 
     @unfilteredmethod
     def commit(self, text="", user=None, date=None, match=None, force=False,
                editor=False, extra=None):
         """Add a new revision to current repository.
 
         Revision information is gathered from the working directory,
         match can be used to filter the committed files. If editor is
         supplied, it is called to get a commit message.
         """
         if extra is None:
             extra = {}
 
         def fail(f, msg):
             raise error.Abort('%s: %s' % (f, msg))
 
         if not match:
             match = matchmod.always(self.root, '')
 
         if not force:
             vdirs = []
             match.explicitdir = vdirs.append
             match.bad = fail
 
         wlock = lock = tr = None
         try:
             wlock = self.wlock()
             lock = self.lock() # for recent changelog (see issue4368)
 
             wctx = self[None]
             merge = len(wctx.parents()) > 1
 
             if not force and merge and not match.always():
                 raise error.Abort(_('cannot partially commit a merge '
                                    '(do not specify files or patterns)'))
 
             status = self.status(match=match, clean=force)
             if force:
                 status.modified.extend(status.clean) # mq may commit clean files
 
             # check subrepos
             subs = []
             commitsubs = set()
             newstate = wctx.substate.copy()
             # only manage subrepos and .hgsubstate if .hgsub is present
             if '.hgsub' in wctx:
                 # we'll decide whether to track this ourselves, thanks
                 for c in status.modified, status.added, status.removed:
                     if '.hgsubstate' in c:
                         c.remove('.hgsubstate')
 
                 # compare current state to last committed state
                 # build new substate based on last committed state
                 oldstate = wctx.p1().substate
                 for s in sorted(newstate.keys()):
                     if not match(s):
                         # ignore working copy, use old state if present
                         if s in oldstate:
                             newstate[s] = oldstate[s]
                             continue
                         if not force:
                             raise error.Abort(
                                 _("commit with new subrepo %s excluded") % s)
                     dirtyreason = wctx.sub(s).dirtyreason(True)
                     if dirtyreason:
                         if not self.ui.configbool('ui', 'commitsubrepos'):
                             raise error.Abort(dirtyreason,
                                 hint=_("use --subrepos for recursive commit"))
                         subs.append(s)
                         commitsubs.add(s)
                     else:
                         bs = wctx.sub(s).basestate()
                         newstate[s] = (newstate[s][0], bs, newstate[s][2])
                         if oldstate.get(s, (None, None, None))[1] != bs:
                             subs.append(s)
 
                 # check for removed subrepos
                 for p in wctx.parents():
                     r = [s for s in p.substate if s not in newstate]
                     subs += [s for s in r if match(s)]
                 if subs:
                     if (not match('.hgsub') and
                         '.hgsub' in (wctx.modified() + wctx.added())):
                         raise error.Abort(
                             _("can't commit subrepos without .hgsub"))
                     status.modified.insert(0, '.hgsubstate')
 
             elif '.hgsub' in status.removed:
                 # clean up .hgsubstate when .hgsub is removed
                 if ('.hgsubstate' in wctx and
                     '.hgsubstate' not in (status.modified + status.added +
                                           status.removed)):
                     status.removed.insert(0, '.hgsubstate')
 
             # make sure all explicit patterns are matched
             if not force:
                 self.checkcommitpatterns(wctx, vdirs, match, status, fail)
 
             cctx = context.workingcommitctx(self, status,
                                             text, user, date, extra)
 
             # internal config: ui.allowemptycommit
             allowemptycommit = (wctx.branch() != wctx.p1().branch()
                                 or extra.get('close') or merge or cctx.files()
                                 or self.ui.configbool('ui', 'allowemptycommit'))
             if not allowemptycommit:
                 return None
 
             if merge and cctx.deleted():
                 raise error.Abort(_("cannot commit merge with missing files"))
 
             ms = mergemod.mergestate.read(self)
             mergeutil.checkunresolved(ms)
 
             if editor:
                 cctx._text = editor(self, cctx, subs)
             edited = (text != cctx._text)
 
             # Save commit message in case this transaction gets rolled back
             # (e.g. by a pretxncommit hook).  Leave the content alone on
             # the assumption that the user will use the same editor again.
             msgfn = self.savecommitmessage(cctx._text)
 
             # commit subs and write new state
             if subs:
                 for s in sorted(commitsubs):
                     sub = wctx.sub(s)
                     self.ui.status(_('committing subrepository %s\n') %
                         subrepo.subrelpath(sub))
                     sr = sub.commit(cctx._text, user, date)
                     newstate[s] = (newstate[s][0], sr)
                 subrepo.writestate(self, newstate)
 
             p1, p2 = self.dirstate.parents()
             hookp1, hookp2 = hex(p1), (p2 != nullid and hex(p2) or '')
             try:
                 self.hook("precommit", throw=True, parent1=hookp1,
                           parent2=hookp2)
                 tr = self.transaction('commit')
                 ret = self.commitctx(cctx, True)
             except: # re-raises
                 if edited:
                     self.ui.write(
                         _('note: commit message saved in %s\n') % msgfn)
                 raise
             # update bookmarks, dirstate and mergestate
             bookmarks.update(self, [p1, p2], ret)
             cctx.markcommitted(ret)
             ms.reset()
             tr.close()
 
         finally:
             lockmod.release(tr, lock, wlock)
 
         def commithook(node=hex(ret), parent1=hookp1, parent2=hookp2):
             # hack for command that use a temporary commit (eg: histedit)
             # temporary commit got stripped before hook release
             if self.changelog.hasnode(ret):
                 self.hook("commit", node=node, parent1=parent1,
                           parent2=parent2)
         self._afterlock(commithook)
         return ret
 
     @unfilteredmethod
     def commitctx(self, ctx, error=False):
         """Add a new revision to current repository.
         Revision information is passed via the context argument.
         """
 
         tr = None
         p1, p2 = ctx.p1(), ctx.p2()
         user = ctx.user()
 
         lock = self.lock()
         try:
             tr = self.transaction("commit")
             trp = weakref.proxy(tr)
 
             if ctx.manifestnode():
                 # reuse an existing manifest revision
                 mn = ctx.manifestnode()
                 files = ctx.files()
             elif ctx.files():
                 m1ctx = p1.manifestctx()
                 m2ctx = p2.manifestctx()
                 mctx = m1ctx.copy()
 
                 m = mctx.read()
                 m1 = m1ctx.read()
                 m2 = m2ctx.read()
 
                 # check in files
                 added = []
                 changed = []
                 removed = list(ctx.removed())
                 linkrev = len(self)
                 self.ui.note(_("committing files:\n"))
                 for f in sorted(ctx.modified() + ctx.added()):
                     self.ui.note(f + "\n")
                     try:
                         fctx = ctx[f]
                         if fctx is None:
                             removed.append(f)
                         else:
                             added.append(f)
                             m[f] = self._filecommit(fctx, m1, m2, linkrev,
                                                     trp, changed)
                             m.setflag(f, fctx.flags())
                     except OSError as inst:
                         self.ui.warn(_("trouble committing %s!\n") % f)
                         raise
                     except IOError as inst:
                         errcode = getattr(inst, 'errno', errno.ENOENT)
                         if error or errcode and errcode != errno.ENOENT:
                             self.ui.warn(_("trouble committing %s!\n") % f)
                         raise
 
                 # update manifest
                 self.ui.note(_("committing manifest\n"))
                 removed = [f for f in sorted(removed) if f in m1 or f in m2]
                 drop = [f for f in removed if f in m]
                 for f in drop:
                     del m[f]
                 mn = mctx.write(trp, linkrev,
                                 p1.manifestnode(), p2.manifestnode(),
                                 added, drop)
                 files = changed + removed
             else:
                 mn = p1.manifestnode()
                 files = []
 
             # update changelog
             self.ui.note(_("committing changelog\n"))
             self.changelog.delayupdate(tr)
             n = self.changelog.add(mn, files, ctx.description(),
                                    trp, p1.node(), p2.node(),
                                    user, ctx.date(), ctx.extra().copy())
             xp1, xp2 = p1.hex(), p2 and p2.hex() or ''
             self.hook('pretxncommit', throw=True, node=hex(n), parent1=xp1,
                       parent2=xp2)
             # set the new commit is proper phase
             targetphase = subrepo.newcommitphase(self.ui, ctx)
             if targetphase:
                 # retract boundary do not alter parent changeset.
                 # if a parent have higher the resulting phase will
                 # be compliant anyway
                 #
                 # if minimal phase was 0 we don't need to retract anything
                 phases.retractboundary(self, tr, targetphase, [n])
             tr.close()
             return n
         finally:
             if tr:
                 tr.release()
             lock.release()
 
     @unfilteredmethod
     def destroying(self):
         '''Inform the repository that nodes are about to be destroyed.
         Intended for use by strip and rollback, so there's a common
         place for anything that has to be done before destroying history.
 
         This is mostly useful for saving state that is in memory and waiting
         to be flushed when the current lock is released. Because a call to
         destroyed is imminent, the repo will be invalidated causing those
         changes to stay in memory (waiting for the next unlock), or vanish
         completely.
         '''
         # When using the same lock to commit and strip, the phasecache is left
         # dirty after committing. Then when we strip, the repo is invalidated,
         # causing those changes to disappear.
         if '_phasecache' in vars(self):
             self._phasecache.write()
 
     @unfilteredmethod
     def destroyed(self):
         '''Inform the repository that nodes have been destroyed.
         Intended for use by strip and rollback, so there's a common
         place for anything that has to be done after destroying history.
         '''
         # When one tries to:
         # 1) destroy nodes thus calling this method (e.g. strip)
         # 2) use phasecache somewhere (e.g. commit)
         #
         # then 2) will fail because the phasecache contains nodes that were
         # removed. We can either remove phasecache from the filecache,
         # causing it to reload next time it is accessed, or simply filter
         # the removed nodes now and write the updated cache.
         self._phasecache.filterunknown(self)
         self._phasecache.write()
 
         # refresh all repository caches
         self.updatecaches()
 
         # Ensure the persistent tag cache is updated.  Doing it now
         # means that the tag cache only has to worry about destroyed
         # heads immediately after a strip/rollback.  That in turn
         # guarantees that "cachetip == currenttip" (comparing both rev
         # and node) always means no nodes have been added or destroyed.
 
         # XXX this is suboptimal when qrefresh'ing: we strip the current
         # head, refresh the tag cache, then immediately add a new head.
         # But I think doing it this way is necessary for the "instant
         # tag cache retrieval" case to work.
         self.invalidate()
 
     def walk(self, match, node=None):
         '''
         walk recursively through the directory tree or a given
         changeset, finding all files matched by the match
         function
         '''
         self.ui.deprecwarn('use repo[node].walk instead of repo.walk', '4.3')
         return self[node].walk(match)
 
     def status(self, node1='.', node2=None, match=None,
                ignored=False, clean=False, unknown=False,
                listsubrepos=False):
         '''a convenience method that calls node1.status(node2)'''
         return self[node1].status(node2, match, ignored, clean, unknown,
                                   listsubrepos)
 
     def addpostdsstatus(self, ps):
         """Add a callback to run within the wlock, at the point at which status
         fixups happen.
 
         On status completion, callback(wctx, status) will be called with the
         wlock held, unless the dirstate has changed from underneath or the wlock
         couldn't be grabbed.
 
         Callbacks should not capture and use a cached copy of the dirstate --
         it might change in the meanwhile. Instead, they should access the
         dirstate via wctx.repo().dirstate.
 
         This list is emptied out after each status run -- extensions should
         make sure it adds to this list each time dirstate.status is called.
         Extensions should also make sure they don't call this for statuses
         that don't involve the dirstate.
         """
 
         # The list is located here for uniqueness reasons -- it is actually
         # managed by the workingctx, but that isn't unique per-repo.
         self._postdsstatus.append(ps)
 
     def postdsstatus(self):
         """Used by workingctx to get the list of post-dirstate-status hooks."""
         return self._postdsstatus
 
     def clearpostdsstatus(self):
         """Used by workingctx to clear post-dirstate-status hooks."""
         del self._postdsstatus[:]
 
     def heads(self, start=None):
         if start is None:
             cl = self.changelog
             headrevs = reversed(cl.headrevs())
             return [cl.node(rev) for rev in headrevs]
 
         heads = self.changelog.heads(start)
         # sort the output in rev descending order
         return sorted(heads, key=self.changelog.rev, reverse=True)
 
     def branchheads(self, branch=None, start=None, closed=False):
         '''return a (possibly filtered) list of heads for the given branch
 
         Heads are returned in topological order, from newest to oldest.
         If branch is None, use the dirstate branch.
         If start is not None, return only heads reachable from start.
         If closed is True, return heads that are marked as closed as well.
         '''
         if branch is None:
             branch = self[None].branch()
         branches = self.branchmap()
         if branch not in branches:
             return []
         # the cache returns heads ordered lowest to highest
         bheads = list(reversed(branches.branchheads(branch, closed=closed)))
         if start is not None:
             # filter out the heads that cannot be reached from startrev
             fbheads = set(self.changelog.nodesbetween([start], bheads)[2])
             bheads = [h for h in bheads if h in fbheads]
         return bheads
 
     def branches(self, nodes):
         if not nodes:
             nodes = [self.changelog.tip()]
         b = []
         for n in nodes:
             t = n
             while True:
                 p = self.changelog.parents(n)
                 if p[1] != nullid or p[0] == nullid:
                     b.append((t, n, p[0], p[1]))
                     break
                 n = p[0]
         return b
 
     def between(self, pairs):
         r = []
 
         for top, bottom in pairs:
             n, l, i = top, [], 0
             f = 1
 
             while n != bottom and n != nullid:
                 p = self.changelog.parents(n)[0]
                 if i == f:
                     l.append(n)
                     f = f * 2
                 n = p
                 i += 1
 
             r.append(l)
 
         return r
 
     def checkpush(self, pushop):
         """Extensions can override this function if additional checks have
         to be performed before pushing, or call it if they override push
         command.
         """
         pass
 
     @unfilteredpropertycache
     def prepushoutgoinghooks(self):
         """Return util.hooks consists of a pushop with repo, remote, outgoing
         methods, which are called before pushing changesets.
         """
         return util.hooks()
 
     def pushkey(self, namespace, key, old, new):
         try:
             tr = self.currenttransaction()
             hookargs = {}
             if tr is not None:
                 hookargs.update(tr.hookargs)
             hookargs['namespace'] = namespace
             hookargs['key'] = key
             hookargs['old'] = old
             hookargs['new'] = new
             self.hook('prepushkey', throw=True, **hookargs)
         except error.HookAbort as exc:
             self.ui.write_err(_("pushkey-abort: %s\n") % exc)
             if exc.hint:
                 self.ui.write_err(_("(%s)\n") % exc.hint)
             return False
         self.ui.debug('pushing key for "%s:%s"\n' % (namespace, key))
         ret = pushkey.push(self, namespace, key, old, new)
         def runhook():
             self.hook('pushkey', namespace=namespace, key=key, old=old, new=new,
                       ret=ret)
         self._afterlock(runhook)
         return ret
 
     def listkeys(self, namespace):
         self.hook('prelistkeys', throw=True, namespace=namespace)
         self.ui.debug('listing keys for "%s"\n' % namespace)
         values = pushkey.list(self, namespace)
         self.hook('listkeys', namespace=namespace, values=values)
         return values
 
     def debugwireargs(self, one, two, three=None, four=None, five=None):
         '''used to test argument passing over the wire'''
         return "%s %s %s %s %s" % (one, two, three, four, five)
 
     def savecommitmessage(self, text):
         fp = self.vfs('last-message.txt', 'wb')
         try:
             fp.write(text)
         finally:
             fp.close()
         return self.pathto(fp.name[len(self.root) + 1:])
 
 # used to avoid circular references so destructors work
 def aftertrans(files):
     renamefiles = [tuple(t) for t in files]
     def a():
         for vfs, src, dest in renamefiles:
             # if src and dest refer to a same file, vfs.rename is a no-op,
             # leaving both src and dest on disk. delete dest to make sure
             # the rename couldn't be such a no-op.
             vfs.tryunlink(dest)
             try:
                 vfs.rename(src, dest)
             except OSError: # journal file does not yet exist
                 pass
     return a
 
 def undoname(fn):
     base, name = os.path.split(fn)
     assert name.startswith('journal')
     return os.path.join(base, name.replace('journal', 'undo', 1))
 
 def instance(ui, path, create):
     return localrepository(ui, util.urllocalpath(path), create)
 
 def islocal(path):
     return True
 
 def newreporequirements(repo):
     """Determine the set of requirements for a new local repository.
 
     Extensions can wrap this function to specify custom requirements for
     new repositories.
     """
     ui = repo.ui
     requirements = {'revlogv1'}
     if ui.configbool('format', 'usestore', True):
         requirements.add('store')
         if ui.configbool('format', 'usefncache', True):
             requirements.add('fncache')
             if ui.configbool('format', 'dotencode', True):
                 requirements.add('dotencode')
 
     compengine = ui.config('experimental', 'format.compression', 'zlib')
     if compengine not in util.compengines:
         raise error.Abort(_('compression engine %s defined by '
                             'experimental.format.compression not available') %
                           compengine,
                           hint=_('run "hg debuginstall" to list available '
                                  'compression engines'))
 
     # zlib is the historical default and doesn't need an explicit requirement.
     if compengine != 'zlib':
         requirements.add('exp-compression-%s' % compengine)
 
     if scmutil.gdinitconfig(ui):
         requirements.add('generaldelta')
     if ui.configbool('experimental', 'treemanifest', False):
         requirements.add('treemanifest')
     if ui.configbool('experimental', 'manifestv2', False):
         requirements.add('manifestv2')
 
     revlogv2 = ui.config('experimental', 'revlogv2')
     if revlogv2 == 'enable-unstable-format-and-corrupt-my-data':
         requirements.remove('revlogv1')
         # generaldelta is implied by revlogv2.
         requirements.discard('generaldelta')
         requirements.add(REVLOGV2_REQUIREMENT)
 
     return requirements

mercurial/revlog.py

@@ -1,2199 +1,2208 b''
 # revlog.py - storage back-end 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.
 
 """Storage back-end for Mercurial.
 
 This provides efficient delta storage with O(1) retrieve and append
 and O(changes) merge between branches.
 """
 
 from __future__ import absolute_import
 
 import binascii
 import collections
 import errno
 import hashlib
 import os
 import struct
 import zlib
 
 # import stuff from node for others to import from revlog
 from .node import (
     bin,
     hex,
     nullid,
     nullrev,
     wdirhex,
     wdirid,
     wdirrev,
 )
 from .i18n import _
 from . import (
     ancestor,
     error,
     mdiff,
     policy,
     pycompat,
     templatefilters,
     util,
 )
 
 parsers = policy.importmod(r'parsers')
 
 _pack = struct.pack
 _unpack = struct.unpack
 # Aliased for performance.
 _zlibdecompress = zlib.decompress
 
 # revlog header flags
 REVLOGV0 = 0
 REVLOGV1 = 1
 # Dummy value until file format is finalized.
 # Reminder: change the bounds check in revlog.__init__ when this is changed.
 REVLOGV2 = 0xDEAD
 FLAG_INLINE_DATA = (1 << 16)
 FLAG_GENERALDELTA = (1 << 17)
 REVLOG_DEFAULT_FLAGS = FLAG_INLINE_DATA
 REVLOG_DEFAULT_FORMAT = REVLOGV1
 REVLOG_DEFAULT_VERSION = REVLOG_DEFAULT_FORMAT | REVLOG_DEFAULT_FLAGS
 REVLOGV1_FLAGS = FLAG_INLINE_DATA | FLAG_GENERALDELTA
 REVLOGV2_FLAGS = REVLOGV1_FLAGS
 
 # revlog index flags
 REVIDX_ISCENSORED = (1 << 15) # revision has censor metadata, must be verified
 REVIDX_ELLIPSIS = (1 << 14) # revision hash does not match data (narrowhg)
 REVIDX_EXTSTORED = (1 << 13) # revision data is stored externally
 REVIDX_DEFAULT_FLAGS = 0
 # stable order in which flags need to be processed and their processors applied
 REVIDX_FLAGS_ORDER = [
     REVIDX_ISCENSORED,
     REVIDX_ELLIPSIS,
     REVIDX_EXTSTORED,
 ]
 REVIDX_KNOWN_FLAGS = util.bitsfrom(REVIDX_FLAGS_ORDER)
 
 # max size of revlog with inline data
 _maxinline = 131072
 _chunksize = 1048576
 
 RevlogError = error.RevlogError
 LookupError = error.LookupError
 CensoredNodeError = error.CensoredNodeError
 ProgrammingError = error.ProgrammingError
 
 # Store flag processors (cf. 'addflagprocessor()' to register)
 _flagprocessors = {
     REVIDX_ISCENSORED: None,
 }
 
 def addflagprocessor(flag, processor):
     """Register a flag processor on a revision data flag.
 
     Invariant:
     - Flags need to be defined in REVIDX_KNOWN_FLAGS and REVIDX_FLAGS_ORDER.
     - Only one flag processor can be registered on a specific flag.
     - flagprocessors must be 3-tuples of functions (read, write, raw) with the
       following signatures:
           - (read)  f(self, rawtext) -> text, bool
           - (write) f(self, text) -> rawtext, bool
           - (raw)   f(self, rawtext) -> bool
       "text" is presented to the user. "rawtext" is stored in revlog data, not
       directly visible to the user.
       The boolean returned by these transforms is used to determine whether
       the returned text can be used for hash integrity checking. For example,
       if "write" returns False, then "text" is used to generate hash. If
       "write" returns True, that basically means "rawtext" returned by "write"
       should be used to generate hash. Usually, "write" and "read" return
       different booleans. And "raw" returns a same boolean as "write".
 
       Note: The 'raw' transform is used for changegroup generation and in some
       debug commands. In this case the transform only indicates whether the
       contents can be used for hash integrity checks.
     """
     if not flag & REVIDX_KNOWN_FLAGS:
         msg = _("cannot register processor on unknown flag '%#x'.") % (flag)
         raise ProgrammingError(msg)
     if flag not in REVIDX_FLAGS_ORDER:
         msg = _("flag '%#x' undefined in REVIDX_FLAGS_ORDER.") % (flag)
         raise ProgrammingError(msg)
     if flag in _flagprocessors:
         msg = _("cannot register multiple processors on flag '%#x'.") % (flag)
         raise error.Abort(msg)
     _flagprocessors[flag] = processor
 
 def getoffset(q):
     return int(q >> 16)
 
 def gettype(q):
     return int(q & 0xFFFF)
 
 def offset_type(offset, type):
     if (type & ~REVIDX_KNOWN_FLAGS) != 0:
         raise ValueError('unknown revlog index flags')
     return int(int(offset) << 16 | type)
 
 _nullhash = hashlib.sha1(nullid)
 
 def hash(text, p1, p2):
     """generate a hash from the given text and its parent hashes
 
     This hash combines both the current file contents and its history
     in a manner that makes it easy to distinguish nodes with the same
     content in the revision graph.
     """
     # As of now, if one of the parent node is null, p2 is null
     if p2 == nullid:
         # deep copy of a hash is faster than creating one
         s = _nullhash.copy()
         s.update(p1)
     else:
         # none of the parent nodes are nullid
         l = [p1, p2]
         l.sort()
         s = hashlib.sha1(l[0])
         s.update(l[1])
     s.update(text)
     return s.digest()
 
 # index v0:
 #  4 bytes: offset
 #  4 bytes: compressed length
 #  4 bytes: base rev
 #  4 bytes: link rev
 # 20 bytes: parent 1 nodeid
 # 20 bytes: parent 2 nodeid
 # 20 bytes: nodeid
 indexformatv0 = ">4l20s20s20s"
 
 class revlogoldio(object):
     def __init__(self):
         self.size = struct.calcsize(indexformatv0)
 
     def parseindex(self, data, inline):
         s = self.size
         index = []
         nodemap = {nullid: nullrev}
         n = off = 0
         l = len(data)
         while off + s <= l:
             cur = data[off:off + s]
             off += s
             e = _unpack(indexformatv0, cur)
             # transform to revlogv1 format
             e2 = (offset_type(e[0], 0), e[1], -1, e[2], e[3],
                   nodemap.get(e[4], nullrev), nodemap.get(e[5], nullrev), e[6])
             index.append(e2)
             nodemap[e[6]] = n
             n += 1
 
         # add the magic null revision at -1
         index.append((0, 0, 0, -1, -1, -1, -1, nullid))
 
         return index, nodemap, None
 
     def packentry(self, entry, node, version, rev):
         if gettype(entry[0]):
             raise RevlogError(_('index entry flags need revlog version 1'))
         e2 = (getoffset(entry[0]), entry[1], entry[3], entry[4],
               node(entry[5]), node(entry[6]), entry[7])
         return _pack(indexformatv0, *e2)
 
 # index ng:
 #  6 bytes: offset
 #  2 bytes: flags
 #  4 bytes: compressed length
 #  4 bytes: uncompressed length
 #  4 bytes: base rev
 #  4 bytes: link rev
 #  4 bytes: parent 1 rev
 #  4 bytes: parent 2 rev
 # 32 bytes: nodeid
 indexformatng = ">Qiiiiii20s12x"
 versionformat = ">I"
 
 # corresponds to uncompressed length of indexformatng (2 gigs, 4-byte
 # signed integer)
 _maxentrysize = 0x7fffffff
 
 class revlogio(object):
     def __init__(self):
         self.size = struct.calcsize(indexformatng)
 
     def parseindex(self, data, inline):
         # call the C implementation to parse the index data
         index, cache = parsers.parse_index2(data, inline)
         return index, getattr(index, 'nodemap', None), cache
 
     def packentry(self, entry, node, version, rev):
         p = _pack(indexformatng, *entry)
         if rev == 0:
             p = _pack(versionformat, version) + p[4:]
         return p
 
 class revlog(object):
     """
     the underlying revision storage object
 
     A revlog consists of two parts, an index and the revision data.
 
     The index is a file with a fixed record size containing
     information on each revision, including its nodeid (hash), the
     nodeids of its parents, the position and offset of its data within
     the data file, and the revision it's based on. Finally, each entry
     contains a linkrev entry that can serve as a pointer to external
     data.
 
     The revision data itself is a linear collection of data chunks.
     Each chunk represents a revision and is usually represented as a
     delta against the previous chunk. To bound lookup time, runs of
     deltas are limited to about 2 times the length of the original
     version data. This makes retrieval of a version proportional to
     its size, or O(1) relative to the number of revisions.
 
     Both pieces of the revlog are written to in an append-only
     fashion, which means we never need to rewrite a file to insert or
     remove data, and can use some simple techniques to avoid the need
     for locking while reading.
 
     If checkambig, indexfile is opened with checkambig=True at
     writing, to avoid file stat ambiguity.
     """
     def __init__(self, opener, indexfile, datafile=None, checkambig=False):
         """
         create a revlog object
 
         opener is a function that abstracts the file opening operation
         and can be used to implement COW semantics or the like.
         """
         self.indexfile = indexfile
         self.datafile = datafile or (indexfile[:-2] + ".d")
         self.opener = opener
         #  When True, indexfile is opened with checkambig=True at writing, to
         #  avoid file stat ambiguity.
         self._checkambig = checkambig
         # 3-tuple of (node, rev, text) for a raw revision.
         self._cache = None
         # Maps rev to chain base rev.
         self._chainbasecache = util.lrucachedict(100)
         # 2-tuple of (offset, data) of raw data from the revlog at an offset.
         self._chunkcache = (0, '')
         # How much data to read and cache into the raw revlog data cache.
         self._chunkcachesize = 65536
         self._maxchainlen = None
         self._aggressivemergedeltas = False
         self.index = []
         # Mapping of partial identifiers to full nodes.
         self._pcache = {}
         # Mapping of revision integer to full node.
         self._nodecache = {nullid: nullrev}
         self._nodepos = None
         self._compengine = 'zlib'
+        self._maxdeltachainspan = -1
 
         v = REVLOG_DEFAULT_VERSION
         opts = getattr(opener, 'options', None)
         if opts is not None:
             if 'revlogv2' in opts:
                 # version 2 revlogs always use generaldelta.
                 v = REVLOGV2 | FLAG_GENERALDELTA | FLAG_INLINE_DATA
             elif 'revlogv1' in opts:
                 if 'generaldelta' in opts:
                     v |= FLAG_GENERALDELTA
             else:
                 v = 0
             if 'chunkcachesize' in opts:
                 self._chunkcachesize = opts['chunkcachesize']
             if 'maxchainlen' in opts:
                 self._maxchainlen = opts['maxchainlen']
             if 'aggressivemergedeltas' in opts:
                 self._aggressivemergedeltas = opts['aggressivemergedeltas']
             self._lazydeltabase = bool(opts.get('lazydeltabase', False))
             if 'compengine' in opts:
                 self._compengine = opts['compengine']
+            if 'maxdeltachainspan' in opts:
+                self._maxdeltachainspan = opts['maxdeltachainspan']
 
         if self._chunkcachesize <= 0:
             raise RevlogError(_('revlog chunk cache size %r is not greater '
                                 'than 0') % self._chunkcachesize)
         elif self._chunkcachesize & (self._chunkcachesize - 1):
             raise RevlogError(_('revlog chunk cache size %r is not a power '
                                 'of 2') % self._chunkcachesize)
 
         indexdata = ''
         self._initempty = True
         try:
             f = self.opener(self.indexfile)
             indexdata = f.read()
             f.close()
             if len(indexdata) > 0:
                 v = struct.unpack(versionformat, indexdata[:4])[0]
                 self._initempty = False
         except IOError as inst:
             if inst.errno != errno.ENOENT:
                 raise
 
         self.version = v
         self._inline = v & FLAG_INLINE_DATA
         self._generaldelta = v & FLAG_GENERALDELTA
         flags = v & ~0xFFFF
         fmt = v & 0xFFFF
         if fmt == REVLOGV0:
             if flags:
                 raise RevlogError(_('unknown flags (%#04x) in version %d '
                                     'revlog %s') %
                                   (flags >> 16, fmt, self.indexfile))
         elif fmt == REVLOGV1:
             if flags & ~REVLOGV1_FLAGS:
                 raise RevlogError(_('unknown flags (%#04x) in version %d '
                                     'revlog %s') %
                                   (flags >> 16, fmt, self.indexfile))
         elif fmt == REVLOGV2:
             if flags & ~REVLOGV2_FLAGS:
                 raise RevlogError(_('unknown flags (%#04x) in version %d '
                                     'revlog %s') %
                                   (flags >> 16, fmt, self.indexfile))
         else:
             raise RevlogError(_('unknown version (%d) in revlog %s') %
                               (fmt, self.indexfile))
 
         self.storedeltachains = True
 
         self._io = revlogio()
         if self.version == REVLOGV0:
             self._io = revlogoldio()
         try:
             d = self._io.parseindex(indexdata, self._inline)
         except (ValueError, IndexError):
             raise RevlogError(_("index %s is corrupted") % (self.indexfile))
         self.index, nodemap, self._chunkcache = d
         if nodemap is not None:
             self.nodemap = self._nodecache = nodemap
         if not self._chunkcache:
             self._chunkclear()
         # revnum -> (chain-length, sum-delta-length)
         self._chaininfocache = {}
         # revlog header -> revlog compressor
         self._decompressors = {}
 
     @util.propertycache
     def _compressor(self):
         return util.compengines[self._compengine].revlogcompressor()
 
     def tip(self):
         return self.node(len(self.index) - 2)
     def __contains__(self, rev):
         return 0 <= rev < len(self)
     def __len__(self):
         return len(self.index) - 1
     def __iter__(self):
         return iter(xrange(len(self)))
     def revs(self, start=0, stop=None):
         """iterate over all rev in this revlog (from start to stop)"""
         step = 1
         if stop is not None:
             if start > stop:
                 step = -1
             stop += step
         else:
             stop = len(self)
         return xrange(start, stop, step)
 
     @util.propertycache
     def nodemap(self):
         self.rev(self.node(0))
         return self._nodecache
 
     def hasnode(self, node):
         try:
             self.rev(node)
             return True
         except KeyError:
             return False
 
     def clearcaches(self):
         self._cache = None
         self._chainbasecache.clear()
         self._chunkcache = (0, '')
         self._pcache = {}
 
         try:
             self._nodecache.clearcaches()
         except AttributeError:
             self._nodecache = {nullid: nullrev}
             self._nodepos = None
 
     def rev(self, node):
         try:
             return self._nodecache[node]
         except TypeError:
             raise
         except RevlogError:
             # parsers.c radix tree lookup failed
             if node == wdirid:
                 raise error.WdirUnsupported
             raise LookupError(node, self.indexfile, _('no node'))
         except KeyError:
             # pure python cache lookup failed
             n = self._nodecache
             i = self.index
             p = self._nodepos
             if p is None:
                 p = len(i) - 2
             for r in xrange(p, -1, -1):
                 v = i[r][7]
                 n[v] = r
                 if v == node:
                     self._nodepos = r - 1
                     return r
             if node == wdirid:
                 raise error.WdirUnsupported
             raise LookupError(node, self.indexfile, _('no node'))
 
     # Accessors for index entries.
 
     # First tuple entry is 8 bytes. First 6 bytes are offset. Last 2 bytes
     # are flags.
     def start(self, rev):
         return int(self.index[rev][0] >> 16)
 
     def flags(self, rev):
         return self.index[rev][0] & 0xFFFF
 
     def length(self, rev):
         return self.index[rev][1]
 
     def rawsize(self, rev):
         """return the length of the uncompressed text for a given revision"""
         l = self.index[rev][2]
         if l >= 0:
             return l
 
         t = self.revision(rev, raw=True)
         return len(t)
 
     def size(self, rev):
         """length of non-raw text (processed by a "read" flag processor)"""
         # fast path: if no "read" flag processor could change the content,
         # size is rawsize. note: ELLIPSIS is known to not change the content.
         flags = self.flags(rev)
         if flags & (REVIDX_KNOWN_FLAGS ^ REVIDX_ELLIPSIS) == 0:
             return self.rawsize(rev)
 
         return len(self.revision(rev, raw=False))
 
     def chainbase(self, rev):
         base = self._chainbasecache.get(rev)
         if base is not None:
             return base
 
         index = self.index
         base = index[rev][3]
         while base != rev:
             rev = base
             base = index[rev][3]
 
         self._chainbasecache[rev] = base
         return base
 
     def linkrev(self, rev):
         return self.index[rev][4]
 
     def parentrevs(self, rev):
         try:
             return self.index[rev][5:7]
         except IndexError:
             if rev == wdirrev:
                 raise error.WdirUnsupported
             raise
 
     def node(self, rev):
         try:
             return self.index[rev][7]
         except IndexError:
             if rev == wdirrev:
                 raise error.WdirUnsupported
             raise
 
     # Derived from index values.
 
     def end(self, rev):
         return self.start(rev) + self.length(rev)
 
     def parents(self, node):
         i = self.index
         d = i[self.rev(node)]
         return i[d[5]][7], i[d[6]][7] # map revisions to nodes inline
 
     def chainlen(self, rev):
         return self._chaininfo(rev)[0]
 
     def _chaininfo(self, rev):
         chaininfocache = self._chaininfocache
         if rev in chaininfocache:
             return chaininfocache[rev]
         index = self.index
         generaldelta = self._generaldelta
         iterrev = rev
         e = index[iterrev]
         clen = 0
         compresseddeltalen = 0
         while iterrev != e[3]:
             clen += 1
             compresseddeltalen += e[1]
             if generaldelta:
                 iterrev = e[3]
             else:
                 iterrev -= 1
             if iterrev in chaininfocache:
                 t = chaininfocache[iterrev]
                 clen += t[0]
                 compresseddeltalen += t[1]
                 break
             e = index[iterrev]
         else:
             # Add text length of base since decompressing that also takes
             # work. For cache hits the length is already included.
             compresseddeltalen += e[1]
         r = (clen, compresseddeltalen)
         chaininfocache[rev] = r
         return r
 
     def _deltachain(self, rev, stoprev=None):
         """Obtain the delta chain for a revision.
 
         ``stoprev`` specifies a revision to stop at. If not specified, we
         stop at the base of the chain.
 
         Returns a 2-tuple of (chain, stopped) where ``chain`` is a list of
         revs in ascending order and ``stopped`` is a bool indicating whether
         ``stoprev`` was hit.
         """
         # Try C implementation.
         try:
             return self.index.deltachain(rev, stoprev, self._generaldelta)
         except AttributeError:
             pass
 
         chain = []
 
         # Alias to prevent attribute lookup in tight loop.
         index = self.index
         generaldelta = self._generaldelta
 
         iterrev = rev
         e = index[iterrev]
         while iterrev != e[3] and iterrev != stoprev:
             chain.append(iterrev)
             if generaldelta:
                 iterrev = e[3]
             else:
                 iterrev -= 1
             e = index[iterrev]
 
         if iterrev == stoprev:
             stopped = True
         else:
             chain.append(iterrev)
             stopped = False
 
         chain.reverse()
         return chain, stopped
 
     def ancestors(self, revs, stoprev=0, inclusive=False):
         """Generate the ancestors of 'revs' in reverse topological order.
         Does not generate revs lower than stoprev.
 
         See the documentation for ancestor.lazyancestors for more details."""
 
         return ancestor.lazyancestors(self.parentrevs, revs, stoprev=stoprev,
                                       inclusive=inclusive)
 
     def descendants(self, revs):
         """Generate the descendants of 'revs' in revision order.
 
         Yield a sequence of revision numbers starting with a child of
         some rev in revs, i.e., each revision is *not* considered a
         descendant of itself.  Results are ordered by revision number (a
         topological sort)."""
         first = min(revs)
         if first == nullrev:
             for i in self:
                 yield i
             return
 
         seen = set(revs)
         for i in self.revs(start=first + 1):
             for x in self.parentrevs(i):
                 if x != nullrev and x in seen:
                     seen.add(i)
                     yield i
                     break
 
     def findcommonmissing(self, common=None, heads=None):
         """Return a tuple of the ancestors of common and the ancestors of heads
         that are not ancestors of common. In revset terminology, we return the
         tuple:
 
           ::common, (::heads) - (::common)
 
         The list is sorted by revision number, meaning it is
         topologically sorted.
 
         'heads' and 'common' are both lists of node IDs.  If heads is
         not supplied, uses all of the revlog's heads.  If common is not
         supplied, uses nullid."""
         if common is None:
             common = [nullid]
         if heads is None:
             heads = self.heads()
 
         common = [self.rev(n) for n in common]
         heads = [self.rev(n) for n in heads]
 
         # we want the ancestors, but inclusive
         class lazyset(object):
             def __init__(self, lazyvalues):
                 self.addedvalues = set()
                 self.lazyvalues = lazyvalues
 
             def __contains__(self, value):
                 return value in self.addedvalues or value in self.lazyvalues
 
             def __iter__(self):
                 added = self.addedvalues
                 for r in added:
                     yield r
                 for r in self.lazyvalues:
                     if not r in added:
                         yield r
 
             def add(self, value):
                 self.addedvalues.add(value)
 
             def update(self, values):
                 self.addedvalues.update(values)
 
         has = lazyset(self.ancestors(common))
         has.add(nullrev)
         has.update(common)
 
         # take all ancestors from heads that aren't in has
         missing = set()
         visit = collections.deque(r for r in heads if r not in has)
         while visit:
             r = visit.popleft()
             if r in missing:
                 continue
             else:
                 missing.add(r)
                 for p in self.parentrevs(r):
                     if p not in has:
                         visit.append(p)
         missing = list(missing)
         missing.sort()
         return has, [self.node(miss) for miss in missing]
 
     def incrementalmissingrevs(self, common=None):
         """Return an object that can be used to incrementally compute the
         revision numbers of the ancestors of arbitrary sets that are not
         ancestors of common. This is an ancestor.incrementalmissingancestors
         object.
 
         'common' is a list of revision numbers. If common is not supplied, uses
         nullrev.
         """
         if common is None:
             common = [nullrev]
 
         return ancestor.incrementalmissingancestors(self.parentrevs, common)
 
     def findmissingrevs(self, common=None, heads=None):
         """Return the revision numbers of the ancestors of heads that
         are not ancestors of common.
 
         More specifically, return a list of revision numbers corresponding to
         nodes N such that every N satisfies the following constraints:
 
           1. N is an ancestor of some node in 'heads'
           2. N is not an ancestor of any node in 'common'
 
         The list is sorted by revision number, meaning it is
         topologically sorted.
 
         'heads' and 'common' are both lists of revision numbers.  If heads is
         not supplied, uses all of the revlog's heads.  If common is not
         supplied, uses nullid."""
         if common is None:
             common = [nullrev]
         if heads is None:
             heads = self.headrevs()
 
         inc = self.incrementalmissingrevs(common=common)
         return inc.missingancestors(heads)
 
     def findmissing(self, common=None, heads=None):
         """Return the ancestors of heads that are not ancestors of common.
 
         More specifically, return a list of nodes N such that every N
         satisfies the following constraints:
 
           1. N is an ancestor of some node in 'heads'
           2. N is not an ancestor of any node in 'common'
 
         The list is sorted by revision number, meaning it is
         topologically sorted.
 
         'heads' and 'common' are both lists of node IDs.  If heads is
         not supplied, uses all of the revlog's heads.  If common is not
         supplied, uses nullid."""
         if common is None:
             common = [nullid]
         if heads is None:
             heads = self.heads()
 
         common = [self.rev(n) for n in common]
         heads = [self.rev(n) for n in heads]
 
         inc = self.incrementalmissingrevs(common=common)
         return [self.node(r) for r in inc.missingancestors(heads)]
 
     def nodesbetween(self, roots=None, heads=None):
         """Return a topological path from 'roots' to 'heads'.
 
         Return a tuple (nodes, outroots, outheads) where 'nodes' is a
         topologically sorted list of all nodes N that satisfy both of
         these constraints:
 
           1. N is a descendant of some node in 'roots'
           2. N is an ancestor of some node in 'heads'
 
         Every node is considered to be both a descendant and an ancestor
         of itself, so every reachable node in 'roots' and 'heads' will be
         included in 'nodes'.
 
         'outroots' is the list of reachable nodes in 'roots', i.e., the
         subset of 'roots' that is returned in 'nodes'.  Likewise,
         'outheads' is the subset of 'heads' that is also in 'nodes'.
 
         'roots' and 'heads' are both lists of node IDs.  If 'roots' is
         unspecified, uses nullid as the only root.  If 'heads' is
         unspecified, uses list of all of the revlog's heads."""
         nonodes = ([], [], [])
         if roots is not None:
             roots = list(roots)
             if not roots:
                 return nonodes
             lowestrev = min([self.rev(n) for n in roots])
         else:
             roots = [nullid] # Everybody's a descendant of nullid
             lowestrev = nullrev
         if (lowestrev == nullrev) and (heads is None):
             # We want _all_ the nodes!
             return ([self.node(r) for r in self], [nullid], list(self.heads()))
         if heads is None:
             # All nodes are ancestors, so the latest ancestor is the last
             # node.
             highestrev = len(self) - 1
             # Set ancestors to None to signal that every node is an ancestor.
             ancestors = None
             # Set heads to an empty dictionary for later discovery of heads
             heads = {}
         else:
             heads = list(heads)
             if not heads:
                 return nonodes
             ancestors = set()
             # Turn heads into a dictionary so we can remove 'fake' heads.
             # Also, later we will be using it to filter out the heads we can't
             # find from roots.
             heads = dict.fromkeys(heads, False)
             # Start at the top and keep marking parents until we're done.
             nodestotag = set(heads)
             # Remember where the top was so we can use it as a limit later.
             highestrev = max([self.rev(n) for n in nodestotag])
             while nodestotag:
                 # grab a node to tag
                 n = nodestotag.pop()
                 # Never tag nullid
                 if n == nullid:
                     continue
                 # A node's revision number represents its place in a
                 # topologically sorted list of nodes.
                 r = self.rev(n)
                 if r >= lowestrev:
                     if n not in ancestors:
                         # If we are possibly a descendant of one of the roots
                         # and we haven't already been marked as an ancestor
                         ancestors.add(n) # Mark as ancestor
                         # Add non-nullid parents to list of nodes to tag.
                         nodestotag.update([p for p in self.parents(n) if
                                            p != nullid])
                     elif n in heads: # We've seen it before, is it a fake head?
                         # So it is, real heads should not be the ancestors of
                         # any other heads.
                         heads.pop(n)
             if not ancestors:
                 return nonodes
             # Now that we have our set of ancestors, we want to remove any
             # roots that are not ancestors.
 
             # If one of the roots was nullid, everything is included anyway.
             if lowestrev > nullrev:
                 # But, since we weren't, let's recompute the lowest rev to not
                 # include roots that aren't ancestors.
 
                 # Filter out roots that aren't ancestors of heads
                 roots = [root for root in roots if root in ancestors]
                 # Recompute the lowest revision
                 if roots:
                     lowestrev = min([self.rev(root) for root in roots])
                 else:
                     # No more roots?  Return empty list
                     return nonodes
             else:
                 # We are descending from nullid, and don't need to care about
                 # any other roots.
                 lowestrev = nullrev
                 roots = [nullid]
         # Transform our roots list into a set.
         descendants = set(roots)
         # Also, keep the original roots so we can filter out roots that aren't
         # 'real' roots (i.e. are descended from other roots).
         roots = descendants.copy()
         # Our topologically sorted list of output nodes.
         orderedout = []
         # Don't start at nullid since we don't want nullid in our output list,
         # and if nullid shows up in descendants, empty parents will look like
         # they're descendants.
         for r in self.revs(start=max(lowestrev, 0), stop=highestrev + 1):
             n = self.node(r)
             isdescendant = False
             if lowestrev == nullrev:  # Everybody is a descendant of nullid
                 isdescendant = True
             elif n in descendants:
                 # n is already a descendant
                 isdescendant = True
                 # This check only needs to be done here because all the roots
                 # will start being marked is descendants before the loop.
                 if n in roots:
                     # If n was a root, check if it's a 'real' root.
                     p = tuple(self.parents(n))
                     # If any of its parents are descendants, it's not a root.
                     if (p[0] in descendants) or (p[1] in descendants):
                         roots.remove(n)
             else:
                 p = tuple(self.parents(n))
                 # A node is a descendant if either of its parents are
                 # descendants.  (We seeded the dependents list with the roots
                 # up there, remember?)
                 if (p[0] in descendants) or (p[1] in descendants):
                     descendants.add(n)
                     isdescendant = True
             if isdescendant and ((ancestors is None) or (n in ancestors)):
                 # Only include nodes that are both descendants and ancestors.
                 orderedout.append(n)
                 if (ancestors is not None) and (n in heads):
                     # We're trying to figure out which heads are reachable
                     # from roots.
                     # Mark this head as having been reached
                     heads[n] = True
                 elif ancestors is None:
                     # Otherwise, we're trying to discover the heads.
                     # Assume this is a head because if it isn't, the next step
                     # will eventually remove it.
                     heads[n] = True
                     # But, obviously its parents aren't.
                     for p in self.parents(n):
                         heads.pop(p, None)
         heads = [head for head, flag in heads.iteritems() if flag]
         roots = list(roots)
         assert orderedout
         assert roots
         assert heads
         return (orderedout, roots, heads)
 
     def headrevs(self):
         try:
             return self.index.headrevs()
         except AttributeError:
             return self._headrevs()
 
     def computephases(self, roots):
         return self.index.computephasesmapsets(roots)
 
     def _headrevs(self):
         count = len(self)
         if not count:
             return [nullrev]
         # we won't iter over filtered rev so nobody is a head at start
         ishead = [0] * (count + 1)
         index = self.index
         for r in self:
             ishead[r] = 1  # I may be an head
             e = index[r]
             ishead[e[5]] = ishead[e[6]] = 0  # my parent are not
         return [r for r, val in enumerate(ishead) if val]
 
     def heads(self, start=None, stop=None):
         """return the list of all nodes that have no children
 
         if start is specified, only heads that are descendants of
         start will be returned
         if stop is specified, it will consider all the revs from stop
         as if they had no children
         """
         if start is None and stop is None:
             if not len(self):
                 return [nullid]
             return [self.node(r) for r in self.headrevs()]
 
         if start is None:
             start = nullid
         if stop is None:
             stop = []
         stoprevs = set([self.rev(n) for n in stop])
         startrev = self.rev(start)
         reachable = {startrev}
         heads = {startrev}
 
         parentrevs = self.parentrevs
         for r in self.revs(start=startrev + 1):
             for p in parentrevs(r):
                 if p in reachable:
                     if r not in stoprevs:
                         reachable.add(r)
                     heads.add(r)
                 if p in heads and p not in stoprevs:
                     heads.remove(p)
 
         return [self.node(r) for r in heads]
 
     def children(self, node):
         """find the children of a given node"""
         c = []
         p = self.rev(node)
         for r in self.revs(start=p + 1):
             prevs = [pr for pr in self.parentrevs(r) if pr != nullrev]
             if prevs:
                 for pr in prevs:
                     if pr == p:
                         c.append(self.node(r))
             elif p == nullrev:
                 c.append(self.node(r))
         return c
 
     def descendant(self, start, end):
         if start == nullrev:
             return True
         for i in self.descendants([start]):
             if i == end:
                 return True
             elif i > end:
                 break
         return False
 
     def commonancestorsheads(self, a, b):
         """calculate all the heads of the common ancestors of nodes a and b"""
         a, b = self.rev(a), self.rev(b)
         try:
             ancs = self.index.commonancestorsheads(a, b)
         except (AttributeError, OverflowError): # C implementation failed
             ancs = ancestor.commonancestorsheads(self.parentrevs, a, b)
         return pycompat.maplist(self.node, ancs)
 
     def isancestor(self, a, b):
         """return True if node a is an ancestor of node b
 
         The implementation of this is trivial but the use of
         commonancestorsheads is not."""
         return a in self.commonancestorsheads(a, b)
 
     def ancestor(self, a, b):
         """calculate the "best" common ancestor of nodes a and b"""
 
         a, b = self.rev(a), self.rev(b)
         try:
             ancs = self.index.ancestors(a, b)
         except (AttributeError, OverflowError):
             ancs = ancestor.ancestors(self.parentrevs, a, b)
         if ancs:
             # choose a consistent winner when there's a tie
             return min(map(self.node, ancs))
         return nullid
 
     def _match(self, id):
         if isinstance(id, int):
             # rev
             return self.node(id)
         if len(id) == 20:
             # possibly a binary node
             # odds of a binary node being all hex in ASCII are 1 in 10**25
             try:
                 node = id
                 self.rev(node) # quick search the index
                 return node
             except LookupError:
                 pass # may be partial hex id
         try:
             # str(rev)
             rev = int(id)
             if str(rev) != id:
                 raise ValueError
             if rev < 0:
                 rev = len(self) + rev
             if rev < 0 or rev >= len(self):
                 raise ValueError
             return self.node(rev)
         except (ValueError, OverflowError):
             pass
         if len(id) == 40:
             try:
                 # a full hex nodeid?
                 node = bin(id)
                 self.rev(node)
                 return node
             except (TypeError, LookupError):
                 pass
 
     def _partialmatch(self, id):
         maybewdir = wdirhex.startswith(id)
         try:
             partial = self.index.partialmatch(id)
             if partial and self.hasnode(partial):
                 if maybewdir:
                     # single 'ff...' match in radix tree, ambiguous with wdir
                     raise RevlogError
                 return partial
             if maybewdir:
                 # no 'ff...' match in radix tree, wdir identified
                 raise error.WdirUnsupported
             return None
         except RevlogError:
             # parsers.c radix tree lookup gave multiple matches
             # fast path: for unfiltered changelog, radix tree is accurate
             if not getattr(self, 'filteredrevs', None):
                 raise LookupError(id, self.indexfile,
                                   _('ambiguous identifier'))
             # fall through to slow path that filters hidden revisions
         except (AttributeError, ValueError):
             # we are pure python, or key was too short to search radix tree
             pass
 
         if id in self._pcache:
             return self._pcache[id]
 
         if len(id) < 40:
             try:
                 # hex(node)[:...]
                 l = len(id) // 2  # grab an even number of digits
                 prefix = bin(id[:l * 2])
                 nl = [e[7] for e in self.index if e[7].startswith(prefix)]
                 nl = [n for n in nl if hex(n).startswith(id) and
                       self.hasnode(n)]
                 if len(nl) > 0:
                     if len(nl) == 1 and not maybewdir:
                         self._pcache[id] = nl[0]
                         return nl[0]
                     raise LookupError(id, self.indexfile,
                                       _('ambiguous identifier'))
                 if maybewdir:
                     raise error.WdirUnsupported
                 return None
             except (TypeError, binascii.Error):
                 pass
 
     def lookup(self, id):
         """locate a node based on:
             - revision number or str(revision number)
             - nodeid or subset of hex nodeid
         """
         n = self._match(id)
         if n is not None:
             return n
         n = self._partialmatch(id)
         if n:
             return n
 
         raise LookupError(id, self.indexfile, _('no match found'))
 
     def cmp(self, node, text):
         """compare text with a given file revision
 
         returns True if text is different than what is stored.
         """
         p1, p2 = self.parents(node)
         return hash(text, p1, p2) != node
 
     def _cachesegment(self, offset, data):
         """Add a segment to the revlog cache.
 
         Accepts an absolute offset and the data that is at that location.
         """
         o, d = self._chunkcache
         # try to add to existing cache
         if o + len(d) == offset and len(d) + len(data) < _chunksize:
             self._chunkcache = o, d + data
         else:
             self._chunkcache = offset, data
 
     def _readsegment(self, offset, length, df=None):
         """Load a segment of raw data from the revlog.
 
         Accepts an absolute offset, length to read, and an optional existing
         file handle to read from.
 
         If an existing file handle is passed, it will be seeked and the
         original seek position will NOT be restored.
 
         Returns a str or buffer of raw byte data.
         """
         if df is not None:
             closehandle = False
         else:
             if self._inline:
                 df = self.opener(self.indexfile)
             else:
                 df = self.opener(self.datafile)
             closehandle = True
 
         # Cache data both forward and backward around the requested
         # data, in a fixed size window. This helps speed up operations
         # involving reading the revlog backwards.
         cachesize = self._chunkcachesize
         realoffset = offset & ~(cachesize - 1)
         reallength = (((offset + length + cachesize) & ~(cachesize - 1))
                       - realoffset)
         df.seek(realoffset)
         d = df.read(reallength)
         if closehandle:
             df.close()
         self._cachesegment(realoffset, d)
         if offset != realoffset or reallength != length:
             return util.buffer(d, offset - realoffset, length)
         return d
 
     def _getsegment(self, offset, length, df=None):
         """Obtain a segment of raw data from the revlog.
 
         Accepts an absolute offset, length of bytes to obtain, and an
         optional file handle to the already-opened revlog. If the file
         handle is used, it's original seek position will not be preserved.
 
         Requests for data may be returned from a cache.
 
         Returns a str or a buffer instance of raw byte data.
         """
         o, d = self._chunkcache
         l = len(d)
 
         # is it in the cache?
         cachestart = offset - o
         cacheend = cachestart + length
         if cachestart >= 0 and cacheend <= l:
             if cachestart == 0 and cacheend == l:
                 return d # avoid a copy
             return util.buffer(d, cachestart, cacheend - cachestart)
 
         return self._readsegment(offset, length, df=df)
 
     def _getsegmentforrevs(self, startrev, endrev, df=None):
         """Obtain a segment of raw data corresponding to a range of revisions.
 
         Accepts the start and end revisions and an optional already-open
         file handle to be used for reading. If the file handle is read, its
         seek position will not be preserved.
 
         Requests for data may be satisfied by a cache.
 
         Returns a 2-tuple of (offset, data) for the requested range of
         revisions. Offset is the integer offset from the beginning of the
         revlog and data is a str or buffer of the raw byte data.
 
         Callers will need to call ``self.start(rev)`` and ``self.length(rev)``
         to determine where each revision's data begins and ends.
         """
         # Inlined self.start(startrev) & self.end(endrev) for perf reasons
         # (functions are expensive).
         index = self.index
         istart = index[startrev]
         start = int(istart[0] >> 16)
         if startrev == endrev:
             end = start + istart[1]
         else:
             iend = index[endrev]
             end = int(iend[0] >> 16) + iend[1]
 
         if self._inline:
             start += (startrev + 1) * self._io.size
             end += (endrev + 1) * self._io.size
         length = end - start
 
         return start, self._getsegment(start, length, df=df)
 
     def _chunk(self, rev, df=None):
         """Obtain a single decompressed chunk for a revision.
 
         Accepts an integer revision and an optional already-open file handle
         to be used for reading. If used, the seek position of the file will not
         be preserved.
 
         Returns a str holding uncompressed data for the requested revision.
         """
         return self.decompress(self._getsegmentforrevs(rev, rev, df=df)[1])
 
     def _chunks(self, revs, df=None):
         """Obtain decompressed chunks for the specified revisions.
 
         Accepts an iterable of numeric revisions that are assumed to be in
         ascending order. Also accepts an optional already-open file handle
         to be used for reading. If used, the seek position of the file will
         not be preserved.
 
         This function is similar to calling ``self._chunk()`` multiple times,
         but is faster.
 
         Returns a list with decompressed data for each requested revision.
         """
         if not revs:
             return []
         start = self.start
         length = self.length
         inline = self._inline
         iosize = self._io.size
         buffer = util.buffer
 
         l = []
         ladd = l.append
 
         try:
             offset, data = self._getsegmentforrevs(revs[0], revs[-1], df=df)
         except OverflowError:
             # issue4215 - we can't cache a run of chunks greater than
             # 2G on Windows
             return [self._chunk(rev, df=df) for rev in revs]
 
         decomp = self.decompress
         for rev in revs:
             chunkstart = start(rev)
             if inline:
                 chunkstart += (rev + 1) * iosize
             chunklength = length(rev)
             ladd(decomp(buffer(data, chunkstart - offset, chunklength)))
 
         return l
 
     def _chunkclear(self):
         """Clear the raw chunk cache."""
         self._chunkcache = (0, '')
 
     def deltaparent(self, rev):
         """return deltaparent of the given revision"""
         base = self.index[rev][3]
         if base == rev:
             return nullrev
         elif self._generaldelta:
             return base
         else:
             return rev - 1
 
     def revdiff(self, rev1, rev2):
         """return or calculate a delta between two revisions
 
         The delta calculated is in binary form and is intended to be written to
         revlog data directly. So this function needs raw revision data.
         """
         if rev1 != nullrev and self.deltaparent(rev2) == rev1:
             return bytes(self._chunk(rev2))
 
         return mdiff.textdiff(self.revision(rev1, raw=True),
                               self.revision(rev2, raw=True))
 
     def revision(self, nodeorrev, _df=None, raw=False):
         """return an uncompressed revision of a given node or revision
         number.
 
         _df - an existing file handle to read from. (internal-only)
         raw - an optional argument specifying if the revision data is to be
         treated as raw data when applying flag transforms. 'raw' should be set
         to True when generating changegroups or in debug commands.
         """
         if isinstance(nodeorrev, int):
             rev = nodeorrev
             node = self.node(rev)
         else:
             node = nodeorrev
             rev = None
 
         cachedrev = None
         flags = None
         rawtext = None
         if node == nullid:
             return ""
         if self._cache:
             if self._cache[0] == node:
                 # _cache only stores rawtext
                 if raw:
                     return self._cache[2]
                 # duplicated, but good for perf
                 if rev is None:
                     rev = self.rev(node)
                 if flags is None:
                     flags = self.flags(rev)
                 # no extra flags set, no flag processor runs, text = rawtext
                 if flags == REVIDX_DEFAULT_FLAGS:
                     return self._cache[2]
                 # rawtext is reusable. need to run flag processor
                 rawtext = self._cache[2]
 
             cachedrev = self._cache[1]
 
         # look up what we need to read
         if rawtext is None:
             if rev is None:
                 rev = self.rev(node)
 
             chain, stopped = self._deltachain(rev, stoprev=cachedrev)
             if stopped:
                 rawtext = self._cache[2]
 
             # drop cache to save memory
             self._cache = None
 
             bins = self._chunks(chain, df=_df)
             if rawtext is None:
                 rawtext = bytes(bins[0])
                 bins = bins[1:]
 
             rawtext = mdiff.patches(rawtext, bins)
             self._cache = (node, rev, rawtext)
 
         if flags is None:
             if rev is None:
                 rev = self.rev(node)
             flags = self.flags(rev)
 
         text, validatehash = self._processflags(rawtext, flags, 'read', raw=raw)
         if validatehash:
             self.checkhash(text, node, rev=rev)
 
         return text
 
     def hash(self, text, p1, p2):
         """Compute a node hash.
 
         Available as a function so that subclasses can replace the hash
         as needed.
         """
         return hash(text, p1, p2)
 
     def _processflags(self, text, flags, operation, raw=False):
         """Inspect revision data flags and applies transforms defined by
         registered flag processors.
 
         ``text`` - the revision data to process
         ``flags`` - the revision flags
         ``operation`` - the operation being performed (read or write)
         ``raw`` - an optional argument describing if the raw transform should be
         applied.
 
         This method processes the flags in the order (or reverse order if
         ``operation`` is 'write') defined by REVIDX_FLAGS_ORDER, applying the
         flag processors registered for present flags. The order of flags defined
         in REVIDX_FLAGS_ORDER needs to be stable to allow non-commutativity.
 
         Returns a 2-tuple of ``(text, validatehash)`` where ``text`` is the
         processed text and ``validatehash`` is a bool indicating whether the
         returned text should be checked for hash integrity.
 
         Note: If the ``raw`` argument is set, it has precedence over the
         operation and will only update the value of ``validatehash``.
         """
         # fast path: no flag processors will run
         if flags == 0:
             return text, True
         if not operation in ('read', 'write'):
             raise ProgrammingError(_("invalid '%s' operation ") % (operation))
         # Check all flags are known.
         if flags & ~REVIDX_KNOWN_FLAGS:
             raise RevlogError(_("incompatible revision flag '%#x'") %
                               (flags & ~REVIDX_KNOWN_FLAGS))
         validatehash = True
         # Depending on the operation (read or write), the order might be
         # reversed due to non-commutative transforms.
         orderedflags = REVIDX_FLAGS_ORDER
         if operation == 'write':
             orderedflags = reversed(orderedflags)
 
         for flag in orderedflags:
             # If a flagprocessor has been registered for a known flag, apply the
             # related operation transform and update result tuple.
             if flag & flags:
                 vhash = True
 
                 if flag not in _flagprocessors:
                     message = _("missing processor for flag '%#x'") % (flag)
                     raise RevlogError(message)
 
                 processor = _flagprocessors[flag]
                 if processor is not None:
                     readtransform, writetransform, rawtransform = processor
 
                     if raw:
                         vhash = rawtransform(self, text)
                     elif operation == 'read':
                         text, vhash = readtransform(self, text)
                     else: # write operation
                         text, vhash = writetransform(self, text)
                 validatehash = validatehash and vhash
 
         return text, validatehash
 
     def checkhash(self, text, node, p1=None, p2=None, rev=None):
         """Check node hash integrity.
 
         Available as a function so that subclasses can extend hash mismatch
         behaviors as needed.
         """
         if p1 is None and p2 is None:
             p1, p2 = self.parents(node)
         if node != self.hash(text, p1, p2):
             revornode = rev
             if revornode is None:
                 revornode = templatefilters.short(hex(node))
             raise RevlogError(_("integrity check failed on %s:%s")
                 % (self.indexfile, revornode))
 
     def checkinlinesize(self, tr, fp=None):
         """Check if the revlog is too big for inline and convert if so.
 
         This should be called after revisions are added to the revlog. If the
         revlog has grown too large to be an inline revlog, it will convert it
         to use multiple index and data files.
         """
         if not self._inline or (self.start(-2) + self.length(-2)) < _maxinline:
             return
 
         trinfo = tr.find(self.indexfile)
         if trinfo is None:
             raise RevlogError(_("%s not found in the transaction")
                               % self.indexfile)
 
         trindex = trinfo[2]
         if trindex is not None:
             dataoff = self.start(trindex)
         else:
             # revlog was stripped at start of transaction, use all leftover data
             trindex = len(self) - 1
             dataoff = self.end(-2)
 
         tr.add(self.datafile, dataoff)
 
         if fp:
             fp.flush()
             fp.close()
 
         df = self.opener(self.datafile, 'w')
         try:
             for r in self:
                 df.write(self._getsegmentforrevs(r, r)[1])
         finally:
             df.close()
 
         fp = self.opener(self.indexfile, 'w', atomictemp=True,
                          checkambig=self._checkambig)
         self.version &= ~FLAG_INLINE_DATA
         self._inline = False
         for i in self:
             e = self._io.packentry(self.index[i], self.node, self.version, i)
             fp.write(e)
 
         # if we don't call close, the temp file will never replace the
         # real index
         fp.close()
 
         tr.replace(self.indexfile, trindex * self._io.size)
         self._chunkclear()
 
     def addrevision(self, text, transaction, link, p1, p2, cachedelta=None,
                     node=None, flags=REVIDX_DEFAULT_FLAGS):
         """add a revision to the log
 
         text - the revision data to add
         transaction - the transaction object used for rollback
         link - the linkrev data to add
         p1, p2 - the parent nodeids of the revision
         cachedelta - an optional precomputed delta
         node - nodeid of revision; typically node is not specified, and it is
             computed by default as hash(text, p1, p2), however subclasses might
             use different hashing method (and override checkhash() in such case)
         flags - the known flags to set on the revision
         """
         if link == nullrev:
             raise RevlogError(_("attempted to add linkrev -1 to %s")
                               % self.indexfile)
 
         if flags:
             node = node or self.hash(text, p1, p2)
 
         rawtext, validatehash = self._processflags(text, flags, 'write')
 
         # If the flag processor modifies the revision data, ignore any provided
         # cachedelta.
         if rawtext != text:
             cachedelta = None
 
         if len(rawtext) > _maxentrysize:
             raise RevlogError(
                 _("%s: size of %d bytes exceeds maximum revlog storage of 2GiB")
                 % (self.indexfile, len(rawtext)))
 
         node = node or self.hash(rawtext, p1, p2)
         if node in self.nodemap:
             return node
 
         if validatehash:
             self.checkhash(rawtext, node, p1=p1, p2=p2)
 
         return self.addrawrevision(rawtext, transaction, link, p1, p2, node,
                                    flags, cachedelta=cachedelta)
 
     def addrawrevision(self, rawtext, transaction, link, p1, p2, node, flags,
                        cachedelta=None):
         """add a raw revision with known flags, node and parents
         useful when reusing a revision not stored in this revlog (ex: received
         over wire, or read from an external bundle).
         """
         dfh = None
         if not self._inline:
             dfh = self.opener(self.datafile, "a+")
         ifh = self.opener(self.indexfile, "a+", checkambig=self._checkambig)
         try:
             return self._addrevision(node, rawtext, transaction, link, p1, p2,
                                      flags, cachedelta, ifh, dfh)
         finally:
             if dfh:
                 dfh.close()
             ifh.close()
 
     def compress(self, data):
         """Generate a possibly-compressed representation of data."""
         if not data:
             return '', data
 
         compressed = self._compressor.compress(data)
 
         if compressed:
             # The revlog compressor added the header in the returned data.
             return '', compressed
 
         if data[0:1] == '\0':
             return '', data
         return 'u', data
 
     def decompress(self, data):
         """Decompress a revlog chunk.
 
         The chunk is expected to begin with a header identifying the
         format type so it can be routed to an appropriate decompressor.
         """
         if not data:
             return data
 
         # Revlogs are read much more frequently than they are written and many
         # chunks only take microseconds to decompress, so performance is
         # important here.
         #
         # We can make a few assumptions about revlogs:
         #
         # 1) the majority of chunks will be compressed (as opposed to inline
         #    raw data).
         # 2) decompressing *any* data will likely by at least 10x slower than
         #    returning raw inline data.
         # 3) we want to prioritize common and officially supported compression
         #    engines
         #
         # It follows that we want to optimize for "decompress compressed data
         # when encoded with common and officially supported compression engines"
         # case over "raw data" and "data encoded by less common or non-official
         # compression engines." That is why we have the inline lookup first
         # followed by the compengines lookup.
         #
         # According to `hg perfrevlogchunks`, this is ~0.5% faster for zlib
         # compressed chunks. And this matters for changelog and manifest reads.
         t = data[0:1]
 
         if t == 'x':
             try:
                 return _zlibdecompress(data)
             except zlib.error as e:
                 raise RevlogError(_('revlog decompress error: %s') % str(e))
         # '\0' is more common than 'u' so it goes first.
         elif t == '\0':
             return data
         elif t == 'u':
             return util.buffer(data, 1)
 
         try:
             compressor = self._decompressors[t]
         except KeyError:
             try:
                 engine = util.compengines.forrevlogheader(t)
                 compressor = engine.revlogcompressor()
                 self._decompressors[t] = compressor
             except KeyError:
                 raise RevlogError(_('unknown compression type %r') % t)
 
         return compressor.decompress(data)
 
     def _isgooddelta(self, d, textlen):
         """Returns True if the given delta is good. Good means that it is within
         the disk span, disk size, and chain length bounds that we know to be
         performant."""
         if d is None:
             return False
 
         # - 'dist' is the distance from the base revision -- bounding it limits
         #   the amount of I/O we need to do.
         # - 'compresseddeltalen' is the sum of the total size of deltas we need
         #   to apply -- bounding it limits the amount of CPU we consume.
         dist, l, data, base, chainbase, chainlen, compresseddeltalen = d
-        if (dist > textlen * 4 or l > textlen or
+
+        defaultmax = textlen * 4
+        maxdist = self._maxdeltachainspan
+        if not maxdist:
+            maxdist = dist # ensure the conditional pass
+        maxdist = max(maxdist, defaultmax)
+        if (dist > maxdist or l > textlen or
             compresseddeltalen > textlen * 2 or
             (self._maxchainlen and chainlen > self._maxchainlen)):
             return False
 
         return True
 
     def _addrevision(self, node, rawtext, transaction, link, p1, p2, flags,
                      cachedelta, ifh, dfh, alwayscache=False):
         """internal function to add revisions to the log
 
         see addrevision for argument descriptions.
 
         note: "addrevision" takes non-raw text, "_addrevision" takes raw text.
 
         invariants:
         - rawtext is optional (can be None); if not set, cachedelta must be set.
           if both are set, they must correspond to each other.
         """
         btext = [rawtext]
         def buildtext():
             if btext[0] is not None:
                 return btext[0]
             baserev = cachedelta[0]
             delta = cachedelta[1]
             # special case deltas which replace entire base; no need to decode
             # base revision. this neatly avoids censored bases, which throw when
             # they're decoded.
             hlen = struct.calcsize(">lll")
             if delta[:hlen] == mdiff.replacediffheader(self.rawsize(baserev),
                                                        len(delta) - hlen):
                 btext[0] = delta[hlen:]
             else:
                 if self._inline:
                     fh = ifh
                 else:
                     fh = dfh
                 basetext = self.revision(baserev, _df=fh, raw=True)
                 btext[0] = mdiff.patch(basetext, delta)
 
             try:
                 res = self._processflags(btext[0], flags, 'read', raw=True)
                 btext[0], validatehash = res
                 if validatehash:
                     self.checkhash(btext[0], node, p1=p1, p2=p2)
                 if flags & REVIDX_ISCENSORED:
                     raise RevlogError(_('node %s is not censored') % node)
             except CensoredNodeError:
                 # must pass the censored index flag to add censored revisions
                 if not flags & REVIDX_ISCENSORED:
                     raise
             return btext[0]
 
         def builddelta(rev):
             # can we use the cached delta?
             if cachedelta and cachedelta[0] == rev:
                 delta = cachedelta[1]
             else:
                 t = buildtext()
                 if self.iscensored(rev):
                     # deltas based on a censored revision must replace the
                     # full content in one patch, so delta works everywhere
                     header = mdiff.replacediffheader(self.rawsize(rev), len(t))
                     delta = header + t
                 else:
                     if self._inline:
                         fh = ifh
                     else:
                         fh = dfh
                     ptext = self.revision(rev, _df=fh, raw=True)
                     delta = mdiff.textdiff(ptext, t)
             header, data = self.compress(delta)
             deltalen = len(header) + len(data)
             chainbase = self.chainbase(rev)
             dist = deltalen + offset - self.start(chainbase)
             if self._generaldelta:
                 base = rev
             else:
                 base = chainbase
             chainlen, compresseddeltalen = self._chaininfo(rev)
             chainlen += 1
             compresseddeltalen += deltalen
             return (dist, deltalen, (header, data), base,
                     chainbase, chainlen, compresseddeltalen)
 
         curr = len(self)
         prev = curr - 1
         offset = self.end(prev)
         delta = None
         p1r, p2r = self.rev(p1), self.rev(p2)
 
         # full versions are inserted when the needed deltas
         # become comparable to the uncompressed text
         if rawtext is None:
             textlen = mdiff.patchedsize(self.rawsize(cachedelta[0]),
                                         cachedelta[1])
         else:
             textlen = len(rawtext)
 
         # should we try to build a delta?
         if prev != nullrev and self.storedeltachains:
             tested = set()
             # This condition is true most of the time when processing
             # changegroup data into a generaldelta repo. The only time it
             # isn't true is if this is the first revision in a delta chain
             # or if ``format.generaldelta=true`` disabled ``lazydeltabase``.
             if cachedelta and self._generaldelta and self._lazydeltabase:
                 # Assume what we received from the server is a good choice
                 # build delta will reuse the cache
                 candidatedelta = builddelta(cachedelta[0])
                 tested.add(cachedelta[0])
                 if self._isgooddelta(candidatedelta, textlen):
                     delta = candidatedelta
             if delta is None and self._generaldelta:
                 # exclude already lazy tested base if any
                 parents = [p for p in (p1r, p2r)
                            if p != nullrev and p not in tested]
                 if parents and not self._aggressivemergedeltas:
                     # Pick whichever parent is closer to us (to minimize the
                     # chance of having to build a fulltext).
                     parents = [max(parents)]
                 tested.update(parents)
                 pdeltas = []
                 for p in parents:
                     pd = builddelta(p)
                     if self._isgooddelta(pd, textlen):
                         pdeltas.append(pd)
                 if pdeltas:
                     delta = min(pdeltas, key=lambda x: x[1])
             if delta is None and prev not in tested:
                 # other approach failed try against prev to hopefully save us a
                 # fulltext.
                 candidatedelta = builddelta(prev)
                 if self._isgooddelta(candidatedelta, textlen):
                     delta = candidatedelta
         if delta is not None:
             dist, l, data, base, chainbase, chainlen, compresseddeltalen = delta
         else:
             rawtext = buildtext()
             data = self.compress(rawtext)
             l = len(data[1]) + len(data[0])
             base = chainbase = curr
 
         e = (offset_type(offset, flags), l, textlen,
              base, link, p1r, p2r, node)
         self.index.insert(-1, e)
         self.nodemap[node] = curr
 
         entry = self._io.packentry(e, self.node, self.version, curr)
         self._writeentry(transaction, ifh, dfh, entry, data, link, offset)
 
         if alwayscache and rawtext is None:
             rawtext = buildtext()
 
         if type(rawtext) == str: # only accept immutable objects
             self._cache = (node, curr, rawtext)
         self._chainbasecache[curr] = chainbase
         return node
 
     def _writeentry(self, transaction, ifh, dfh, entry, data, link, offset):
         # Files opened in a+ mode have inconsistent behavior on various
         # platforms. Windows requires that a file positioning call be made
         # when the file handle transitions between reads and writes. See
         # 3686fa2b8eee and the mixedfilemodewrapper in windows.py. On other
         # platforms, Python or the platform itself can be buggy. Some versions
         # of Solaris have been observed to not append at the end of the file
         # if the file was seeked to before the end. See issue4943 for more.
         #
         # We work around this issue by inserting a seek() before writing.
         # Note: This is likely not necessary on Python 3.
         ifh.seek(0, os.SEEK_END)
         if dfh:
             dfh.seek(0, os.SEEK_END)
 
         curr = len(self) - 1
         if not self._inline:
             transaction.add(self.datafile, offset)
             transaction.add(self.indexfile, curr * len(entry))
             if data[0]:
                 dfh.write(data[0])
             dfh.write(data[1])
             ifh.write(entry)
         else:
             offset += curr * self._io.size
             transaction.add(self.indexfile, offset, curr)
             ifh.write(entry)
             ifh.write(data[0])
             ifh.write(data[1])
             self.checkinlinesize(transaction, ifh)
 
     def addgroup(self, cg, linkmapper, transaction, addrevisioncb=None):
         """
         add a delta group
 
         given a set of deltas, add them to the revision log. the
         first delta is against its parent, which should be in our
         log, the rest are against the previous delta.
 
         If ``addrevisioncb`` is defined, it will be called with arguments of
         this revlog and the node that was added.
         """
 
         nodes = []
 
         r = len(self)
         end = 0
         if r:
             end = self.end(r - 1)
         ifh = self.opener(self.indexfile, "a+", checkambig=self._checkambig)
         isize = r * self._io.size
         if self._inline:
             transaction.add(self.indexfile, end + isize, r)
             dfh = None
         else:
             transaction.add(self.indexfile, isize, r)
             transaction.add(self.datafile, end)
             dfh = self.opener(self.datafile, "a+")
         def flush():
             if dfh:
                 dfh.flush()
             ifh.flush()
         try:
             # loop through our set of deltas
             chain = None
             for chunkdata in iter(lambda: cg.deltachunk(chain), {}):
                 node = chunkdata['node']
                 p1 = chunkdata['p1']
                 p2 = chunkdata['p2']
                 cs = chunkdata['cs']
                 deltabase = chunkdata['deltabase']
                 delta = chunkdata['delta']
                 flags = chunkdata['flags'] or REVIDX_DEFAULT_FLAGS
 
                 nodes.append(node)
 
                 link = linkmapper(cs)
                 if node in self.nodemap:
                     # this can happen if two branches make the same change
                     chain = node
                     continue
 
                 for p in (p1, p2):
                     if p not in self.nodemap:
                         raise LookupError(p, self.indexfile,
                                           _('unknown parent'))
 
                 if deltabase not in self.nodemap:
                     raise LookupError(deltabase, self.indexfile,
                                       _('unknown delta base'))
 
                 baserev = self.rev(deltabase)
 
                 if baserev != nullrev and self.iscensored(baserev):
                     # if base is censored, delta must be full replacement in a
                     # single patch operation
                     hlen = struct.calcsize(">lll")
                     oldlen = self.rawsize(baserev)
                     newlen = len(delta) - hlen
                     if delta[:hlen] != mdiff.replacediffheader(oldlen, newlen):
                         raise error.CensoredBaseError(self.indexfile,
                                                       self.node(baserev))
 
                 if not flags and self._peek_iscensored(baserev, delta, flush):
                     flags |= REVIDX_ISCENSORED
 
                 # We assume consumers of addrevisioncb will want to retrieve
                 # the added revision, which will require a call to
                 # revision(). revision() will fast path if there is a cache
                 # hit. So, we tell _addrevision() to always cache in this case.
                 # We're only using addgroup() in the context of changegroup
                 # generation so the revision data can always be handled as raw
                 # by the flagprocessor.
                 chain = self._addrevision(node, None, transaction, link,
                                           p1, p2, flags, (baserev, delta),
                                           ifh, dfh,
                                           alwayscache=bool(addrevisioncb))
 
                 if addrevisioncb:
                     addrevisioncb(self, chain)
 
                 if not dfh and not self._inline:
                     # addrevision switched from inline to conventional
                     # reopen the index
                     ifh.close()
                     dfh = self.opener(self.datafile, "a+")
                     ifh = self.opener(self.indexfile, "a+",
                                       checkambig=self._checkambig)
         finally:
             if dfh:
                 dfh.close()
             ifh.close()
 
         return nodes
 
     def iscensored(self, rev):
         """Check if a file revision is censored."""
         return False
 
     def _peek_iscensored(self, baserev, delta, flush):
         """Quickly check if a delta produces a censored revision."""
         return False
 
     def getstrippoint(self, minlink):
         """find the minimum rev that must be stripped to strip the linkrev
 
         Returns a tuple containing the minimum rev and a set of all revs that
         have linkrevs that will be broken by this strip.
         """
         brokenrevs = set()
         strippoint = len(self)
 
         heads = {}
         futurelargelinkrevs = set()
         for head in self.headrevs():
             headlinkrev = self.linkrev(head)
             heads[head] = headlinkrev
             if headlinkrev >= minlink:
                 futurelargelinkrevs.add(headlinkrev)
 
         # This algorithm involves walking down the rev graph, starting at the
         # heads. Since the revs are topologically sorted according to linkrev,
         # once all head linkrevs are below the minlink, we know there are
         # no more revs that could have a linkrev greater than minlink.
         # So we can stop walking.
         while futurelargelinkrevs:
             strippoint -= 1
             linkrev = heads.pop(strippoint)
 
             if linkrev < minlink:
                 brokenrevs.add(strippoint)
             else:
                 futurelargelinkrevs.remove(linkrev)
 
             for p in self.parentrevs(strippoint):
                 if p != nullrev:
                     plinkrev = self.linkrev(p)
                     heads[p] = plinkrev
                     if plinkrev >= minlink:
                         futurelargelinkrevs.add(plinkrev)
 
         return strippoint, brokenrevs
 
     def strip(self, minlink, transaction):
         """truncate the revlog on the first revision with a linkrev >= minlink
 
         This function is called when we're stripping revision minlink and
         its descendants from the repository.
 
         We have to remove all revisions with linkrev >= minlink, because
         the equivalent changelog revisions will be renumbered after the
         strip.
 
         So we truncate the revlog on the first of these revisions, and
         trust that the caller has saved the revisions that shouldn't be
         removed and that it'll re-add them after this truncation.
         """
         if len(self) == 0:
             return
 
         rev, _ = self.getstrippoint(minlink)
         if rev == len(self):
             return
 
         # first truncate the files on disk
         end = self.start(rev)
         if not self._inline:
             transaction.add(self.datafile, end)
             end = rev * self._io.size
         else:
             end += rev * self._io.size
 
         transaction.add(self.indexfile, end)
 
         # then reset internal state in memory to forget those revisions
         self._cache = None
         self._chaininfocache = {}
         self._chunkclear()
         for x in xrange(rev, len(self)):
             del self.nodemap[self.node(x)]
 
         del self.index[rev:-1]
 
     def checksize(self):
         expected = 0
         if len(self):
             expected = max(0, self.end(len(self) - 1))
 
         try:
             f = self.opener(self.datafile)
             f.seek(0, 2)
             actual = f.tell()
             f.close()
             dd = actual - expected
         except IOError as inst:
             if inst.errno != errno.ENOENT:
                 raise
             dd = 0
 
         try:
             f = self.opener(self.indexfile)
             f.seek(0, 2)
             actual = f.tell()
             f.close()
             s = self._io.size
             i = max(0, actual // s)
             di = actual - (i * s)
             if self._inline:
                 databytes = 0
                 for r in self:
                     databytes += max(0, self.length(r))
                 dd = 0
                 di = actual - len(self) * s - databytes
         except IOError as inst:
             if inst.errno != errno.ENOENT:
                 raise
             di = 0
 
         return (dd, di)
 
     def files(self):
         res = [self.indexfile]
         if not self._inline:
             res.append(self.datafile)
         return res
 
     DELTAREUSEALWAYS = 'always'
     DELTAREUSESAMEREVS = 'samerevs'
     DELTAREUSENEVER = 'never'
 
     DELTAREUSEALL = {'always', 'samerevs', 'never'}
 
     def clone(self, tr, destrevlog, addrevisioncb=None,
               deltareuse=DELTAREUSESAMEREVS, aggressivemergedeltas=None):
         """Copy this revlog to another, possibly with format changes.
 
         The destination revlog will contain the same revisions and nodes.
         However, it may not be bit-for-bit identical due to e.g. delta encoding
         differences.
 
         The ``deltareuse`` argument control how deltas from the existing revlog
         are preserved in the destination revlog. The argument can have the
         following values:
 
         DELTAREUSEALWAYS
            Deltas will always be reused (if possible), even if the destination
            revlog would not select the same revisions for the delta. This is the
            fastest mode of operation.
         DELTAREUSESAMEREVS
            Deltas will be reused if the destination revlog would pick the same
            revisions for the delta. This mode strikes a balance between speed
            and optimization.
         DELTAREUSENEVER
            Deltas will never be reused. This is the slowest mode of execution.
            This mode can be used to recompute deltas (e.g. if the diff/delta
            algorithm changes).
 
         Delta computation can be slow, so the choice of delta reuse policy can
         significantly affect run time.
 
         The default policy (``DELTAREUSESAMEREVS``) strikes a balance between
         two extremes. Deltas will be reused if they are appropriate. But if the
         delta could choose a better revision, it will do so. This means if you
         are converting a non-generaldelta revlog to a generaldelta revlog,
         deltas will be recomputed if the delta's parent isn't a parent of the
         revision.
 
         In addition to the delta policy, the ``aggressivemergedeltas`` argument
         controls whether to compute deltas against both parents for merges.
         By default, the current default is used.
         """
         if deltareuse not in self.DELTAREUSEALL:
             raise ValueError(_('value for deltareuse invalid: %s') % deltareuse)
 
         if len(destrevlog):
             raise ValueError(_('destination revlog is not empty'))
 
         if getattr(self, 'filteredrevs', None):
             raise ValueError(_('source revlog has filtered revisions'))
         if getattr(destrevlog, 'filteredrevs', None):
             raise ValueError(_('destination revlog has filtered revisions'))
 
         # lazydeltabase controls whether to reuse a cached delta, if possible.
         oldlazydeltabase = destrevlog._lazydeltabase
         oldamd = destrevlog._aggressivemergedeltas
 
         try:
             if deltareuse == self.DELTAREUSEALWAYS:
                 destrevlog._lazydeltabase = True
             elif deltareuse == self.DELTAREUSESAMEREVS:
                 destrevlog._lazydeltabase = False
 
             destrevlog._aggressivemergedeltas = aggressivemergedeltas or oldamd
 
             populatecachedelta = deltareuse in (self.DELTAREUSEALWAYS,
                                                 self.DELTAREUSESAMEREVS)
 
             index = self.index
             for rev in self:
                 entry = index[rev]
 
                 # Some classes override linkrev to take filtered revs into
                 # account. Use raw entry from index.
                 flags = entry[0] & 0xffff
                 linkrev = entry[4]
                 p1 = index[entry[5]][7]
                 p2 = index[entry[6]][7]
                 node = entry[7]
 
                 # (Possibly) reuse the delta from the revlog if allowed and
                 # the revlog chunk is a delta.
                 cachedelta = None
                 rawtext = None
                 if populatecachedelta:
                     dp = self.deltaparent(rev)
                     if dp != nullrev:
                         cachedelta = (dp, str(self._chunk(rev)))
 
                 if not cachedelta:
                     rawtext = self.revision(rev, raw=True)
 
                 ifh = destrevlog.opener(destrevlog.indexfile, 'a+',
                                         checkambig=False)
                 dfh = None
                 if not destrevlog._inline:
                     dfh = destrevlog.opener(destrevlog.datafile, 'a+')
                 try:
                     destrevlog._addrevision(node, rawtext, tr, linkrev, p1, p2,
                                             flags, cachedelta, ifh, dfh)
                 finally:
                     if dfh:
                         dfh.close()
                     ifh.close()
 
                 if addrevisioncb:
                     addrevisioncb(self, rev, node)
         finally:
             destrevlog._lazydeltabase = oldlazydeltabase
             destrevlog._aggressivemergedeltas = oldamd

tests/test-generaldelta.t

@@ -1,163 +1,351 b''
 Check whether size of generaldelta revlog is not bigger than its
 regular equivalent. Test would fail if generaldelta was naive
 implementation of parentdelta: third manifest revision would be fully
 inserted due to big distance from its paren revision (zero).
 
   $ hg init repo --config format.generaldelta=no --config format.usegeneraldelta=no
   $ cd repo
   $ echo foo > foo
   $ echo bar > bar
   $ echo baz > baz
   $ hg commit -q -Am boo
   $ hg clone --pull . ../gdrepo -q --config format.generaldelta=yes
   $ for r in 1 2 3; do
   >   echo $r > foo
   >   hg commit -q -m $r
   >   hg up -q -r 0
   >   hg pull . -q -r $r -R ../gdrepo
   > done
 
   $ cd ..
   >>> from __future__ import print_function
   >>> import os
   >>> regsize = os.stat("repo/.hg/store/00manifest.i").st_size
   >>> gdsize = os.stat("gdrepo/.hg/store/00manifest.i").st_size
   >>> if regsize < gdsize:
   ...     print('generaldata increased size of manifest')
 
 Verify rev reordering doesnt create invalid bundles (issue4462)
 This requires a commit tree that when pulled will reorder manifest revs such
 that the second manifest to create a file rev will be ordered before the first
 manifest to create that file rev. We also need to do a partial pull to ensure
 reordering happens. At the end we verify the linkrev points at the earliest
 commit.
 
   $ hg init server --config format.generaldelta=True
   $ cd server
   $ touch a
   $ hg commit -Aqm a
   $ echo x > x
   $ echo y > y
   $ hg commit -Aqm xy
   $ hg up -q '.^'
   $ echo x > x
   $ echo z > z
   $ hg commit -Aqm xz
   $ hg up -q 1
   $ echo b > b
   $ hg commit -Aqm b
   $ hg merge -q 2
   $ hg commit -Aqm merge
   $ echo c > c
   $ hg commit -Aqm c
   $ hg log -G -T '{rev} {shortest(node)} {desc}'
   @  5 ebb8 c
   |
   o    4 baf7 merge
   |\
   | o  3 a129 b
   | |
   o |  2 958c xz
   | |
   | o  1 f00c xy
   |/
   o  0 3903 a
   
   $ cd ..
   $ hg init client --config format.generaldelta=false --config format.usegeneraldelta=false
   $ cd client
   $ hg pull -q ../server -r 4
   $ hg debugindex x
      rev    offset  length   base linkrev nodeid       p1           p2
        0         0       3      0       1 1406e7411862 000000000000 000000000000
 
   $ cd ..
 
 Test "usegeneraldelta" config
 (repo are general delta, but incoming bundle are not re-deltafied)
 
 delta coming from the server base delta server are not recompressed.
 (also include the aggressive version for comparison)
 
   $ hg clone repo --pull --config format.usegeneraldelta=1 usegd
   requesting all changes
   adding changesets
   adding manifests
   adding file changes
   added 4 changesets with 6 changes to 3 files (+2 heads)
   updating to branch default
   3 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg clone repo --pull --config format.generaldelta=1 full
   requesting all changes
   adding changesets
   adding manifests
   adding file changes
   added 4 changesets with 6 changes to 3 files (+2 heads)
   updating to branch default
   3 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg -R repo debugindex -m
      rev    offset  length   base linkrev nodeid       p1           p2
        0         0     104      0       0 cef96823c800 000000000000 000000000000
        1       104      57      0       1 58ab9a8d541d cef96823c800 000000000000
        2       161      57      0       2 134fdc6fd680 cef96823c800 000000000000
        3       218     104      3       3 723508934dad cef96823c800 000000000000
   $ hg -R usegd debugindex -m
      rev    offset  length  delta linkrev nodeid       p1           p2
        0         0     104     -1       0 cef96823c800 000000000000 000000000000
        1       104      57      0       1 58ab9a8d541d cef96823c800 000000000000
        2       161      57      1       2 134fdc6fd680 cef96823c800 000000000000
        3       218      57      0       3 723508934dad cef96823c800 000000000000
   $ hg -R full debugindex -m
      rev    offset  length  delta linkrev nodeid       p1           p2
        0         0     104     -1       0 cef96823c800 000000000000 000000000000
        1       104      57      0       1 58ab9a8d541d cef96823c800 000000000000
        2       161      57      0       2 134fdc6fd680 cef96823c800 000000000000
        3       218      57      0       3 723508934dad cef96823c800 000000000000
 
 Test format.aggressivemergedeltas
 
   $ hg init --config format.generaldelta=1 aggressive
   $ cd aggressive
   $ cat << EOF >> .hg/hgrc
   > [format]
   > generaldelta = 1
   > EOF
   $ touch a b c d e
   $ hg commit -Aqm side1
   $ hg up -q null
   $ touch x y
   $ hg commit -Aqm side2
 
 - Verify non-aggressive merge uses p1 (commit 1) as delta parent
   $ hg merge -q 0
   $ hg commit -q -m merge
   $ hg debugindex -m
      rev    offset  length  delta linkrev nodeid       p1           p2
        0         0      59     -1       0 8dde941edb6e 000000000000 000000000000
        1        59      61      0       1 315c023f341d 000000000000 000000000000
        2       120      65      1       2 2ab389a983eb 315c023f341d 8dde941edb6e
 
   $ hg strip -q -r . --config extensions.strip=
 
 - Verify aggressive merge uses p2 (commit 0) as delta parent
   $ hg up -q -C 1
   $ hg merge -q 0
   $ hg commit -q -m merge --config format.aggressivemergedeltas=True
   $ hg debugindex -m
      rev    offset  length  delta linkrev nodeid       p1           p2
        0         0      59     -1       0 8dde941edb6e 000000000000 000000000000
        1        59      61      0       1 315c023f341d 000000000000 000000000000
        2       120      62      0       2 2ab389a983eb 315c023f341d 8dde941edb6e
 
 Test that strip bundle use bundle2
   $ hg --config extensions.strip= strip .
   0 files updated, 0 files merged, 5 files removed, 0 files unresolved
   saved backup bundle to $TESTTMP/aggressive/.hg/strip-backup/1c5d4dc9a8b8-6c68e60c-backup.hg (glob)
   $ hg debugbundle .hg/strip-backup/*
   Stream params: sortdict([('Compression', 'BZ')])
   changegroup -- "sortdict([('version', '02'), ('nbchanges', '1')])"
       1c5d4dc9a8b8d6e1750966d343e94db665e7a1e9
   phase-heads -- 'sortdict()'
       1c5d4dc9a8b8d6e1750966d343e94db665e7a1e9 draft
 
   $ cd ..
+
+test maxdeltachainspan
+
+  $ hg init source-repo
+  $ cd source-repo
+  $ hg debugbuilddag --new-file '.+5:brancha$.+11:branchb$.+30:branchc<brancha+2<branchb+2'
+  $ cd ..
+  $ hg -R source-repo debugindex -m
+     rev    offset  length  delta linkrev nodeid       p1           p2
+       0         0      46     -1       0 19deeef41503 000000000000 000000000000
+       1        46      57      0       1 fffc37b38c40 19deeef41503 000000000000
+       2       103      57      1       2 5822d75c83d9 fffc37b38c40 000000000000
+       3       160      57      2       3 19cf2273e601 5822d75c83d9 000000000000
+       4       217      57      3       4 d45ead487afe 19cf2273e601 000000000000
+       5       274      57      4       5 96e0c2ce55ed d45ead487afe 000000000000
+       6       331      46     -1       6 0c2ea5222c74 000000000000 000000000000
+       7       377      57      6       7 4ca08a89134d 0c2ea5222c74 000000000000
+       8       434      57      7       8 c973dbfd30ac 4ca08a89134d 000000000000
+       9       491      57      8       9 d81d878ff2cd c973dbfd30ac 000000000000
+      10       548      58      9      10 dbee7f0dd760 d81d878ff2cd 000000000000
+      11       606      58     10      11 474be9f1fd4e dbee7f0dd760 000000000000
+      12       664      58     11      12 594a27502c85 474be9f1fd4e 000000000000
+      13       722      58     12      13 a7d25307d6a9 594a27502c85 000000000000
+      14       780      58     13      14 3eb53082272e a7d25307d6a9 000000000000
+      15       838      58     14      15 d1e94c85caf6 3eb53082272e 000000000000
+      16       896      58     15      16 8933d9629788 d1e94c85caf6 000000000000
+      17       954      58     16      17 a33416e52d91 8933d9629788 000000000000
+      18      1012      47     -1      18 4ccbf31021ed 000000000000 000000000000
+      19      1059      58     18      19 dcad7a25656c 4ccbf31021ed 000000000000
+      20      1117      58     19      20 617c4f8be75f dcad7a25656c 000000000000
+      21      1175      58     20      21 975b9c1d75bb 617c4f8be75f 000000000000
+      22      1233      58     21      22 74f09cd33b70 975b9c1d75bb 000000000000
+      23      1291      58     22      23 54e79bfa7ef1 74f09cd33b70 000000000000
+      24      1349      58     23      24 c556e7ff90af 54e79bfa7ef1 000000000000
+      25      1407      58     24      25 42daedfe9c6b c556e7ff90af 000000000000
+      26      1465      58     25      26 f302566947c7 42daedfe9c6b 000000000000
+      27      1523      58     26      27 2346959851cb f302566947c7 000000000000
+      28      1581      58     27      28 ca8d867106b4 2346959851cb 000000000000
+      29      1639      58     28      29 fd9152decab2 ca8d867106b4 000000000000
+      30      1697      58     29      30 3fe34080a79b fd9152decab2 000000000000
+      31      1755      58     30      31 bce61a95078e 3fe34080a79b 000000000000
+      32      1813      58     31      32 1dd9ba54ba15 bce61a95078e 000000000000
+      33      1871      58     32      33 3cd9b90a9972 1dd9ba54ba15 000000000000
+      34      1929      58     33      34 5db8c9754ef5 3cd9b90a9972 000000000000
+      35      1987      58     34      35 ee4a240cc16c 5db8c9754ef5 000000000000
+      36      2045      58     35      36 9e1d38725343 ee4a240cc16c 000000000000
+      37      2103      58     36      37 3463f73086a8 9e1d38725343 000000000000
+      38      2161      58     37      38 88af72fab449 3463f73086a8 000000000000
+      39      2219      58     38      39 472f5ce73785 88af72fab449 000000000000
+      40      2277      58     39      40 c91b8351e5b8 472f5ce73785 000000000000
+      41      2335      58     40      41 9c8289c5c5c0 c91b8351e5b8 000000000000
+      42      2393      58     41      42 a13fd4a09d76 9c8289c5c5c0 000000000000
+      43      2451      58     42      43 2ec2c81cafe0 a13fd4a09d76 000000000000
+      44      2509      58     43      44 f27fdd174392 2ec2c81cafe0 000000000000
+      45      2567      58     44      45 a539ec59fe41 f27fdd174392 000000000000
+      46      2625      58     45      46 5e98b9ecb738 a539ec59fe41 000000000000
+      47      2683      58     46      47 31e6b47899d0 5e98b9ecb738 000000000000
+      48      2741      58     47      48 2cf25d6636bd 31e6b47899d0 000000000000
+      49      2799     197     -1      49 9fff62ea0624 96e0c2ce55ed 000000000000
+      50      2996      58     49      50 467f8e30a066 9fff62ea0624 000000000000
+      51      3054     356     50      51 346db97283df a33416e52d91 000000000000
+      52      3410      58     51      52 4e003fd4d5cd 346db97283df 000000000000
+  $ hg clone --pull source-repo --config experimental.maxdeltachainspan=2800 relax-chain --config format.generaldelta=yes
+  requesting all changes
+  adding changesets
+  adding manifests
+  adding file changes
+  added 53 changesets with 53 changes to 53 files (+2 heads)
+  updating to branch default
+  14 files updated, 0 files merged, 0 files removed, 0 files unresolved
+  $ hg -R relax-chain debugindex -m
+     rev    offset  length  delta linkrev nodeid       p1           p2
+       0         0      46     -1       0 19deeef41503 000000000000 000000000000
+       1        46      57      0       1 fffc37b38c40 19deeef41503 000000000000
+       2       103      57      1       2 5822d75c83d9 fffc37b38c40 000000000000
+       3       160      57      2       3 19cf2273e601 5822d75c83d9 000000000000
+       4       217      57      3       4 d45ead487afe 19cf2273e601 000000000000
+       5       274      57      4       5 96e0c2ce55ed d45ead487afe 000000000000
+       6       331      46     -1       6 0c2ea5222c74 000000000000 000000000000
+       7       377      57      6       7 4ca08a89134d 0c2ea5222c74 000000000000
+       8       434      57      7       8 c973dbfd30ac 4ca08a89134d 000000000000
+       9       491      57      8       9 d81d878ff2cd c973dbfd30ac 000000000000
+      10       548      58      9      10 dbee7f0dd760 d81d878ff2cd 000000000000
+      11       606      58     10      11 474be9f1fd4e dbee7f0dd760 000000000000
+      12       664      58     11      12 594a27502c85 474be9f1fd4e 000000000000
+      13       722      58     12      13 a7d25307d6a9 594a27502c85 000000000000
+      14       780      58     13      14 3eb53082272e a7d25307d6a9 000000000000
+      15       838      58     14      15 d1e94c85caf6 3eb53082272e 000000000000
+      16       896      58     15      16 8933d9629788 d1e94c85caf6 000000000000
+      17       954      58     16      17 a33416e52d91 8933d9629788 000000000000
+      18      1012      47     -1      18 4ccbf31021ed 000000000000 000000000000
+      19      1059      58     18      19 dcad7a25656c 4ccbf31021ed 000000000000
+      20      1117      58     19      20 617c4f8be75f dcad7a25656c 000000000000
+      21      1175      58     20      21 975b9c1d75bb 617c4f8be75f 000000000000
+      22      1233      58     21      22 74f09cd33b70 975b9c1d75bb 000000000000
+      23      1291      58     22      23 54e79bfa7ef1 74f09cd33b70 000000000000
+      24      1349      58     23      24 c556e7ff90af 54e79bfa7ef1 000000000000
+      25      1407      58     24      25 42daedfe9c6b c556e7ff90af 000000000000
+      26      1465      58     25      26 f302566947c7 42daedfe9c6b 000000000000
+      27      1523      58     26      27 2346959851cb f302566947c7 000000000000
+      28      1581      58     27      28 ca8d867106b4 2346959851cb 000000000000
+      29      1639      58     28      29 fd9152decab2 ca8d867106b4 000000000000
+      30      1697      58     29      30 3fe34080a79b fd9152decab2 000000000000
+      31      1755      58     30      31 bce61a95078e 3fe34080a79b 000000000000
+      32      1813      58     31      32 1dd9ba54ba15 bce61a95078e 000000000000
+      33      1871      58     32      33 3cd9b90a9972 1dd9ba54ba15 000000000000
+      34      1929      58     33      34 5db8c9754ef5 3cd9b90a9972 000000000000
+      35      1987      58     34      35 ee4a240cc16c 5db8c9754ef5 000000000000
+      36      2045      58     35      36 9e1d38725343 ee4a240cc16c 000000000000
+      37      2103      58     36      37 3463f73086a8 9e1d38725343 000000000000
+      38      2161      58     37      38 88af72fab449 3463f73086a8 000000000000
+      39      2219      58     38      39 472f5ce73785 88af72fab449 000000000000
+      40      2277      58     39      40 c91b8351e5b8 472f5ce73785 000000000000
+      41      2335      58     40      41 9c8289c5c5c0 c91b8351e5b8 000000000000
+      42      2393      58     41      42 a13fd4a09d76 9c8289c5c5c0 000000000000
+      43      2451      58     42      43 2ec2c81cafe0 a13fd4a09d76 000000000000
+      44      2509      58     43      44 f27fdd174392 2ec2c81cafe0 000000000000
+      45      2567      58     44      45 a539ec59fe41 f27fdd174392 000000000000
+      46      2625      58     45      46 5e98b9ecb738 a539ec59fe41 000000000000
+      47      2683      58     46      47 31e6b47899d0 5e98b9ecb738 000000000000
+      48      2741      58     47      48 2cf25d6636bd 31e6b47899d0 000000000000
+      49      2799     197     -1      49 9fff62ea0624 96e0c2ce55ed 000000000000
+      50      2996      58     49      50 467f8e30a066 9fff62ea0624 000000000000
+      51      3054      58     17      51 346db97283df a33416e52d91 000000000000
+      52      3112     369     -1      52 4e003fd4d5cd 346db97283df 000000000000
+  $ hg clone --pull source-repo --config experimental.maxdeltachainspan=0 noconst-chain --config format.generaldelta=yes
+  requesting all changes
+  adding changesets
+  adding manifests
+  adding file changes
+  added 53 changesets with 53 changes to 53 files (+2 heads)
+  updating to branch default
+  14 files updated, 0 files merged, 0 files removed, 0 files unresolved
+  $ hg -R noconst-chain debugindex -m
+     rev    offset  length  delta linkrev nodeid       p1           p2
+       0         0      46     -1       0 19deeef41503 000000000000 000000000000
+       1        46      57      0       1 fffc37b38c40 19deeef41503 000000000000
+       2       103      57      1       2 5822d75c83d9 fffc37b38c40 000000000000
+       3       160      57      2       3 19cf2273e601 5822d75c83d9 000000000000
+       4       217      57      3       4 d45ead487afe 19cf2273e601 000000000000
+       5       274      57      4       5 96e0c2ce55ed d45ead487afe 000000000000
+       6       331      46     -1       6 0c2ea5222c74 000000000000 000000000000
+       7       377      57      6       7 4ca08a89134d 0c2ea5222c74 000000000000
+       8       434      57      7       8 c973dbfd30ac 4ca08a89134d 000000000000
+       9       491      57      8       9 d81d878ff2cd c973dbfd30ac 000000000000
+      10       548      58      9      10 dbee7f0dd760 d81d878ff2cd 000000000000
+      11       606      58     10      11 474be9f1fd4e dbee7f0dd760 000000000000
+      12       664      58     11      12 594a27502c85 474be9f1fd4e 000000000000
+      13       722      58     12      13 a7d25307d6a9 594a27502c85 000000000000
+      14       780      58     13      14 3eb53082272e a7d25307d6a9 000000000000
+      15       838      58     14      15 d1e94c85caf6 3eb53082272e 000000000000
+      16       896      58     15      16 8933d9629788 d1e94c85caf6 000000000000
+      17       954      58     16      17 a33416e52d91 8933d9629788 000000000000
+      18      1012      47     -1      18 4ccbf31021ed 000000000000 000000000000
+      19      1059      58     18      19 dcad7a25656c 4ccbf31021ed 000000000000
+      20      1117      58     19      20 617c4f8be75f dcad7a25656c 000000000000
+      21      1175      58     20      21 975b9c1d75bb 617c4f8be75f 000000000000
+      22      1233      58     21      22 74f09cd33b70 975b9c1d75bb 000000000000
+      23      1291      58     22      23 54e79bfa7ef1 74f09cd33b70 000000000000
+      24      1349      58     23      24 c556e7ff90af 54e79bfa7ef1 000000000000
+      25      1407      58     24      25 42daedfe9c6b c556e7ff90af 000000000000
+      26      1465      58     25      26 f302566947c7 42daedfe9c6b 000000000000
+      27      1523      58     26      27 2346959851cb f302566947c7 000000000000
+      28      1581      58     27      28 ca8d867106b4 2346959851cb 000000000000
+      29      1639      58     28      29 fd9152decab2 ca8d867106b4 000000000000
+      30      1697      58     29      30 3fe34080a79b fd9152decab2 000000000000
+      31      1755      58     30      31 bce61a95078e 3fe34080a79b 000000000000
+      32      1813      58     31      32 1dd9ba54ba15 bce61a95078e 000000000000
+      33      1871      58     32      33 3cd9b90a9972 1dd9ba54ba15 000000000000
+      34      1929      58     33      34 5db8c9754ef5 3cd9b90a9972 000000000000
+      35      1987      58     34      35 ee4a240cc16c 5db8c9754ef5 000000000000
+      36      2045      58     35      36 9e1d38725343 ee4a240cc16c 000000000000
+      37      2103      58     36      37 3463f73086a8 9e1d38725343 000000000000
+      38      2161      58     37      38 88af72fab449 3463f73086a8 000000000000
+      39      2219      58     38      39 472f5ce73785 88af72fab449 000000000000
+      40      2277      58     39      40 c91b8351e5b8 472f5ce73785 000000000000
+      41      2335      58     40      41 9c8289c5c5c0 c91b8351e5b8 000000000000
+      42      2393      58     41      42 a13fd4a09d76 9c8289c5c5c0 000000000000
+      43      2451      58     42      43 2ec2c81cafe0 a13fd4a09d76 000000000000
+      44      2509      58     43      44 f27fdd174392 2ec2c81cafe0 000000000000
+      45      2567      58     44      45 a539ec59fe41 f27fdd174392 000000000000
+      46      2625      58     45      46 5e98b9ecb738 a539ec59fe41 000000000000
+      47      2683      58     46      47 31e6b47899d0 5e98b9ecb738 000000000000
+      48      2741      58     47      48 2cf25d6636bd 31e6b47899d0 000000000000
+      49      2799      58      5      49 9fff62ea0624 96e0c2ce55ed 000000000000
+      50      2857      58     49      50 467f8e30a066 9fff62ea0624 000000000000
+      51      2915      58     17      51 346db97283df a33416e52d91 000000000000
+      52      2973      58     51      52 4e003fd4d5cd 346db97283df 000000000000