narrow: when writing treemanifests, skip inspecting directories outside narrow

This provides significant speed benefits when narrow and treemanifests are in
use, see the timing numbers below. Note that like previously, differences of <5%
are considered noise.

The below timing numbers are in the same style as previously (example:
ee7ee0c516ca). 'before' is 9db85644, and does not include that example commit's
improvements.

    diff --git:
    repo  | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before
    ------+---+---+------------------------+-----------------------+------------
    m-u   |   |   | 1.327 s +-  0.051 s    | 1.296 s +-  0.009 s   |  97.7%
    m-u   |   | x | 1.310 s +-  0.020 s    | 1.295 s +-  0.015 s   |  98.9%
    m-u   | x |   | 1.295 s +-  0.018 s    | 1.296 s +-  0.007 s   | 100.1%
    m-u   | x | x | 83.5 ms +-   0.8 ms    | 84.1 ms +-   0.8 ms   | 100.7%
    l-d-r |   |   | 205.1 ms +-   3.5 ms   | 205.0 ms +-   3.8 ms  | 100.0%
    l-d-r |   | x | 194.2 ms +-   5.6 ms   | 192.3 ms +-   4.3 ms  |  99.0%
    l-d-r | x |   | 99.1 ms +-   2.2 ms    | 97.8 ms +-   0.9 ms   |  98.7%
    l-d-r | x | x | 66.2 ms +-   1.0 ms    | 67.2 ms +-   2.7 ms   | 101.5%

    diff -c . --git:
    repo  | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before
    ------+---+---+------------------------+-----------------------+------------
    m-u   |   |   | 233.9 ms +-   1.9 ms   | 235.6 ms +-   5.1 ms  | 100.7%
    m-u   |   | x | 151.4 ms +-   1.2 ms   | 152.2 ms +-   2.0 ms  | 100.5%
    m-u   | x |   | 234.8 ms +-   2.7 ms   | 235.0 ms +-   2.7 ms  | 100.1%
    m-u   | x | x | 127.8 ms +-   2.1 ms   | 126.0 ms +-   1.1 ms  |  98.6%
    l-d-r |   |   | 82.5 ms +-   1.6 ms    | 82.3 ms +-   2.0 ms   |  99.8%
    l-d-r |   | x | 3.742 s +-  0.017 s    | 3.819 s +-  0.208 s   | 102.1%
    l-d-r | x |   | 84.4 ms +-   1.5 ms    | 83.2 ms +-   1.0 ms   |  98.6%
    l-d-r | x | x | 751.2 ms +-   5.0 ms   | 755.8 ms +-  12.9 ms  | 100.6%

    rebase -r . --keep -d .^^:
    repo  | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before
    ------+---+---+------------------------+-----------------------+------------
    m-u   |   |   | 5.519 s +-  0.038 s    | 5.526 s +-  0.057 s   | 100.1%
    m-u   |   | x | 5.588 s +-  0.048 s    | 5.607 s +-  0.061 s   | 100.3%
    m-u   | x |   | 5.520 s +-  0.044 s    | 5.546 s +-  0.059 s   | 100.5%
    m-u   | x | x | 586.6 ms +-  12.8 ms   | 554.9 ms +-  21.2 ms  |  94.6% <--
    l-d-r |   |   | 629.8 ms +-   5.5 ms   | 627.4 ms +-   6.6 ms  |  99.6%
    l-d-r |   | x | 6.165 s +-  0.058 s    | 6.255 s +-  0.303 s   | 101.5%
    l-d-r | x |   | 270.2 ms +-   2.3 ms   | 271.4 ms +-   2.7 ms  | 100.4%
    l-d-r | x | x | 4.700 s +-  0.025 s    | 1.651 s +-  0.016 s   |  35.1% <--

    status --change . --copies:
    repo  | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before
    ------+---+---+------------------------+-----------------------+------------
    m-u   |   |   | 215.4 ms +-   2.3 ms   | 216.5 ms +-   4.2 ms  | 100.5%
    m-u   |   | x | 132.9 ms +-   1.2 ms   | 132.0 ms +-   1.4 ms  |  99.3%
    m-u   | x |   | 217.0 ms +-   1.9 ms   | 215.4 ms +-   1.9 ms  |  99.3%
    m-u   | x | x | 108.6 ms +-   1.0 ms   | 108.2 ms +-   1.5 ms  |  99.6%
    l-d-r |   |   | 80.0 ms +-   1.3 ms    | 80.5 ms +-   1.1 ms   | 100.6%
    l-d-r |   | x | 3.916 s +-  0.187 s    | 3.966 s +-  0.236 s   | 101.3%
    l-d-r | x |   | 84.4 ms +-   3.1 ms    | 83.9 ms +-   1.1 ms   |  99.4%
    l-d-r | x | x | 758.0 ms +-   8.2 ms   | 753.5 ms +-   5.0 ms  |  99.4%

    status --copies:
    repo  | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before
    ------+---+---+------------------------+-----------------------+------------
    m-u   |   |   | 1.905 s +-  0.025 s    | 1.910 s +-  0.044 s   | 100.3%
    m-u   |   | x | 1.892 s +-  0.009 s    | 1.895 s +-  0.012 s   | 100.2%
    m-u   | x |   | 1.891 s +-  0.012 s    | 1.902 s +-  0.018 s   | 100.6%
    m-u   | x | x | 93.3 ms +-   0.9 ms    | 93.4 ms +-   0.8 ms   | 100.1%
    l-d-r |   |   | 570.7 ms +-   7.8 ms   | 571.9 ms +-  18.5 ms  | 100.2%
    l-d-r |   | x | 561.5 ms +-   5.2 ms   | 562.9 ms +-   6.1 ms  | 100.2%
    l-d-r | x |   | 171.7 ms +-   2.6 ms   | 171.9 ms +-   1.2 ms  | 100.1%
    l-d-r | x | x | 142.7 ms +-   2.0 ms   | 140.3 ms +-   1.0 ms  |  98.3%

    update $rev^; ~/src/hg/hg{hg}/hg update $rev:
    repo  | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before
    ------+---+---+------------------------+-----------------------+------------
    m-u   |   |   | 3.126 s +-  0.016 s    | 3.128 s +-  0.015 s   | 100.1%
    m-u   |   | x | 3.014 s +-  0.068 s    | 3.008 s +-  0.031 s   |  99.8%
    m-u   | x |   | 3.143 s +-  0.037 s    | 3.184 s +-  0.086 s   | 101.3%
    m-u   | x | x | 308.0 ms +-   1.8 ms   | 308.1 ms +-   5.7 ms  | 100.0%
    l-d-r |   |   | 430.8 ms +-   4.5 ms   | 436.4 ms +-   8.7 ms  | 101.3%
    l-d-r |   | x | 9.676 s +-  0.127 s    | 9.945 s +-  0.272 s   | 102.8%
    l-d-r | x |   | 254.2 ms +-   3.3 ms   | 255.7 ms +-   3.1 ms  | 100.6%
    l-d-r | x | x | 1.571 s +-  0.030 s    | 1.555 s +-  0.014 s   |  99.0%

Differential Revision: https://phab.mercurial-scm.org/D4606

mercurial/localrepo.py

@@ -1,2530 +1,2537 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 os
 import random
 import sys
 import time
 import weakref
 
 from .i18n import _
 from .node import (
     hex,
     nullid,
     short,
 )
 from . import (
     bookmarks,
     branchmap,
     bundle2,
     changegroup,
     changelog,
     color,
     context,
     dirstate,
     dirstateguard,
     discovery,
     encoding,
     error,
     exchange,
     extensions,
     filelog,
     hook,
     lock as lockmod,
     manifest,
     match as matchmod,
     merge as mergemod,
     mergeutil,
     namespaces,
     narrowspec,
     obsolete,
     pathutil,
     phases,
     pushkey,
     pycompat,
     repository,
     repoview,
     revset,
     revsetlang,
     scmutil,
     sparse,
     store,
     subrepoutil,
     tags as tagsmod,
     transaction,
     txnutil,
     util,
     vfs as vfsmod,
 )
 from .utils import (
     interfaceutil,
     procutil,
     stringutil,
 )
 
 from .revlogutils import (
     constants as revlogconst,
 )
 
 release = lockmod.release
 urlerr = util.urlerr
 urlreq = util.urlreq
 
 # set of (path, vfs-location) tuples. vfs-location is:
 # - 'plain for vfs relative paths
 # - '' for svfs relative paths
 _cachedfiles = set()
 
 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 __init__(self, *paths):
         super(repofilecache, self).__init__(*paths)
         for path in paths:
             _cachedfiles.add((path, 'plain'))
 
     def join(self, obj, fname):
         return obj.vfs.join(fname)
 
 class storecache(_basefilecache):
     """filecache for files in the store"""
     def __init__(self, *paths):
         super(storecache, self).__init__(*paths)
         for path in paths:
             _cachedfiles.add((path, ''))
 
     def join(self, obj, fname):
         return obj.sjoin(fname)
 
 def isfilecached(repo, name):
     """check if a repo has already cached "name" filecache-ed property
 
     This returns (cachedobj-or-None, iscached) tuple.
     """
     cacheentry = repo.unfiltered()._filecache.get(name, None)
     if not cacheentry:
         return None, False
     return cacheentry.obj, True
 
 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'})
 
 @interfaceutil.implementer(repository.ipeercommandexecutor)
 class localcommandexecutor(object):
     def __init__(self, peer):
         self._peer = peer
         self._sent = False
         self._closed = False
 
     def __enter__(self):
         return self
 
     def __exit__(self, exctype, excvalue, exctb):
         self.close()
 
     def callcommand(self, command, args):
         if self._sent:
             raise error.ProgrammingError('callcommand() cannot be used after '
                                          'sendcommands()')
 
         if self._closed:
             raise error.ProgrammingError('callcommand() cannot be used after '
                                          'close()')
 
         # We don't need to support anything fancy. Just call the named
         # method on the peer and return a resolved future.
         fn = getattr(self._peer, pycompat.sysstr(command))
 
         f = pycompat.futures.Future()
 
         try:
             result = fn(**pycompat.strkwargs(args))
         except Exception:
             pycompat.future_set_exception_info(f, sys.exc_info()[1:])
         else:
             f.set_result(result)
 
         return f
 
     def sendcommands(self):
         self._sent = True
 
     def close(self):
         self._closed = True
 
 @interfaceutil.implementer(repository.ipeercommands)
 class localpeer(repository.peer):
     '''peer for a local repo; reflects only the most recent API'''
 
     def __init__(self, repo, caps=None):
         super(localpeer, self).__init__()
 
         if caps is None:
             caps = moderncaps.copy()
         self._repo = repo.filtered('served')
         self.ui = repo.ui
         self._caps = repo._restrictcapabilities(caps)
 
     # Begin of _basepeer interface.
 
     def url(self):
         return self._repo.url()
 
     def local(self):
         return self._repo
 
     def peer(self):
         return self
 
     def canpush(self):
         return True
 
     def close(self):
         self._repo.close()
 
     # End of _basepeer interface.
 
     # Begin of _basewirecommands interface.
 
     def branchmap(self):
         return self._repo.branchmap()
 
     def capabilities(self):
         return self._caps
 
     def clonebundles(self):
         return self._repo.tryread('clonebundles.manifest')
 
     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, pycompat.bytestr(three),
                                    pycompat.bytestr(four),
                                    pycompat.bytestr(five))
 
     def getbundle(self, source, heads=None, common=None, bundlecaps=None,
                   **kwargs):
         chunks = exchange.getbundlechunks(self._repo, source, heads=heads,
                                           common=common, bundlecaps=bundlecaps,
                                           **kwargs)[1]
         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)
 
     def heads(self):
         return self._repo.heads()
 
     def known(self, nodes):
         return self._repo.known(nodes)
 
     def listkeys(self, namespace):
         return self._repo.listkeys(namespace)
 
     def lookup(self, key):
         return self._repo.lookup(key)
 
     def pushkey(self, namespace, key, old, new):
         return self._repo.pushkey(namespace, key, old, new)
 
     def stream_out(self):
         raise error.Abort(_('cannot perform stream clone against local '
                             'peer'))
 
     def unbundle(self, bundle, heads, url):
         """apply a bundle on a repo
 
         This function handles the repo locking itself."""
         try:
             try:
                 bundle = exchange.readbundle(self.ui, bundle, None)
                 ret = exchange.unbundle(self._repo, bundle, 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:'),
                                       stringutil.forcebytestr(exc))
 
     # End of _basewirecommands interface.
 
     # Begin of peer interface.
 
     def commandexecutor(self):
         return localcommandexecutor(self)
 
     # End of peer interface.
 
 @interfaceutil.implementer(repository.ipeerlegacycommands)
 class locallegacypeer(localpeer):
     '''peer extension which implements legacy methods too; used for tests with
     restricted capabilities'''
 
     def __init__(self, repo):
         super(locallegacypeer, self).__init__(repo, caps=legacycaps)
 
     # Begin of baselegacywirecommands interface.
 
     def between(self, pairs):
         return self._repo.between(pairs)
 
     def branches(self, nodes):
         return self._repo.branches(nodes)
 
     def changegroup(self, nodes, source):
         outgoing = discovery.outgoing(self._repo, missingroots=nodes,
                                       missingheads=self._repo.heads())
         return changegroup.makechangegroup(self._repo, outgoing, '01', source)
 
     def changegroupsubset(self, bases, heads, source):
         outgoing = discovery.outgoing(self._repo, missingroots=bases,
                                       missingheads=heads)
         return changegroup.makechangegroup(self._repo, outgoing, '01', source)
 
     # End of baselegacywirecommands interface.
 
 # Increment the sub-version when the revlog v2 format changes to lock out old
 # clients.
 REVLOGV2_REQUIREMENT = 'exp-revlogv2.0'
 
 # A repository with the sparserevlog feature will have delta chains that
 # can spread over a larger span. Sparse reading cuts these large spans into
 # pieces, so that each piece isn't too big.
 # Without the sparserevlog capability, reading from the repository could use
 # huge amounts of memory, because the whole span would be read at once,
 # including all the intermediate revisions that aren't pertinent for the chain.
 # This is why once a repository has enabled sparse-read, it becomes required.
 SPARSEREVLOG_REQUIREMENT = 'sparserevlog'
 
 # Functions receiving (ui, features) that extensions can register to impact
 # the ability to load repositories with custom requirements. Only
 # functions defined in loaded extensions are called.
 #
 # The function receives a set of requirement strings that the repository
 # is capable of opening. Functions will typically add elements to the
 # set to reflect that the extension knows how to handle that requirements.
 featuresetupfuncs = set()
 
 @interfaceutil.implementer(repository.completelocalrepository)
 class localrepository(object):
 
     # obsolete experimental requirements:
     #  - manifestv2: An experimental new manifest format that allowed
     #    for stem compression of long paths. Experiment ended up not
     #    being successful (repository sizes went up due to worse delta
     #    chains), and the code was deleted in 4.6.
     supportedformats = {
         'revlogv1',
         'generaldelta',
         'treemanifest',
         REVLOGV2_REQUIREMENT,
         SPARSEREVLOG_REQUIREMENT,
     }
     _basesupported = supportedformats | {
         'store',
         'fncache',
         'shared',
         'relshared',
         'dotencode',
         'exp-sparse',
         'internal-phase'
     }
     openerreqs = {
         'revlogv1',
         'generaldelta',
         'treemanifest',
     }
 
     # list of prefix for file which can be written without 'wlock'
     # Extensions should extend this list when needed
     _wlockfreeprefix = {
         # We migh consider requiring 'wlock' for the next
         # two, but pretty much all the existing code assume
         # wlock is not needed so we keep them excluded for
         # now.
         'hgrc',
         'requires',
         # XXX cache is a complicatged business someone
         # should investigate this in depth at some point
         'cache/',
         # XXX shouldn't be dirstate covered by the wlock?
         'dirstate',
         # XXX bisect was still a bit too messy at the time
         # this changeset was introduced. Someone should fix
         # the remainig bit and drop this line
         'bisect.state',
     }
 
     def __init__(self, baseui, path, intents=None):
         """Create a new local repository instance.
 
         Most callers should use ``hg.repository()`` or ``localrepo.instance()``
         for obtaining a new repository object.
         """
 
         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.baseui = baseui
         self.ui = baseui.copy()
         self.ui.copy = baseui.copy # prevent copying repo configuration
         self.vfs = vfsmod.vfs(self.path, cacheaudited=True)
         if (self.ui.configbool('devel', 'all-warnings') or
             self.ui.configbool('devel', 'check-locks')):
             self.vfs.audit = self._getvfsward(self.vfs.audit)
         # 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 featuresetupfuncs:
             self.supported = set(self._basesupported) # use private copy
             extmods = set(m.__name__ for n, m
                           in extensions.extensions(self.ui))
             for setupfunc in 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():
             try:
                 self.vfs.stat()
             except OSError as inst:
                 if inst.errno != errno.ENOENT:
                     raise
             raise error.RepoError(_("repository %s not found") % path)
         else:
             try:
                 self.requirements = scmutil.readrequires(
                         self.vfs, self.supported)
             except IOError as inst:
                 if inst.errno != errno.ENOENT:
                     raise
 
         cachepath = self.vfs.join('cache')
         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)
             cachepath = vfs.join('cache')
             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
 
         if 'exp-sparse' in self.requirements and not sparse.enabled:
             raise error.RepoError(_('repository is using sparse feature but '
                                     'sparse is not enabled; enable the '
                                     '"sparse" extensions to access'))
 
         self.store = store.store(
             self.requirements, self.sharedpath,
             lambda base: vfsmod.vfs(base, cacheaudited=True))
         self.spath = self.store.path
         self.svfs = self.store.vfs
         self.sjoin = self.store.join
         self.vfs.createmode = self.store.createmode
         self.cachevfs = vfsmod.vfs(cachepath, cacheaudited=True)
         self.cachevfs.createmode = self.store.createmode
         if (self.ui.configbool('devel', 'all-warnings') or
             self.ui.configbool('devel', 'check-locks')):
             if util.safehasattr(self.svfs, 'vfs'): # this is filtervfs
                 self.svfs.vfs.audit = self._getsvfsward(self.svfs.vfs.audit)
             else: # standard vfs
                 self.svfs.audit = self._getsvfsward(self.svfs.audit)
         self._applyopenerreqs()
 
         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()
 
         # Key to signature value.
         self._sparsesignaturecache = {}
         # Signature to cached matcher instance.
         self._sparsematchercache = {}
 
     def _getvfsward(self, origfunc):
         """build a ward for self.vfs"""
         rref = weakref.ref(self)
         def checkvfs(path, mode=None):
             ret = origfunc(path, mode=mode)
             repo = rref()
             if (repo is None
                 or not util.safehasattr(repo, '_wlockref')
                 or not util.safehasattr(repo, '_lockref')):
                 return
             if mode in (None, 'r', 'rb'):
                 return
             if path.startswith(repo.path):
                 # truncate name relative to the repository (.hg)
                 path = path[len(repo.path) + 1:]
             if path.startswith('cache/'):
                 msg = 'accessing cache with vfs instead of cachevfs: "%s"'
                 repo.ui.develwarn(msg % path, stacklevel=2, config="cache-vfs")
             if path.startswith('journal.'):
                 # journal is covered by 'lock'
                 if repo._currentlock(repo._lockref) is None:
                     repo.ui.develwarn('write with no lock: "%s"' % path,
                                       stacklevel=2, config='check-locks')
             elif repo._currentlock(repo._wlockref) is None:
                 # rest of vfs files are covered by 'wlock'
                 #
                 # exclude special files
                 for prefix in self._wlockfreeprefix:
                     if path.startswith(prefix):
                         return
                 repo.ui.develwarn('write with no wlock: "%s"' % path,
                                   stacklevel=2, config='check-locks')
             return ret
         return checkvfs
 
     def _getsvfsward(self, origfunc):
         """build a ward for self.svfs"""
         rref = weakref.ref(self)
         def checksvfs(path, mode=None):
             ret = origfunc(path, mode=mode)
             repo = rref()
             if repo is None or not util.safehasattr(repo, '_lockref'):
                 return
             if mode in (None, 'r', 'rb'):
                 return
             if path.startswith(repo.sharedpath):
                 # truncate name relative to the repository (.hg)
                 path = path[len(repo.sharedpath) + 1:]
             if repo._currentlock(repo._lockref) is None:
                 repo.ui.develwarn('write with no lock: "%s"' % path,
                                   stacklevel=3)
             return ret
         return checksvfs
 
     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'):
             caps = set(caps)
             capsblob = bundle2.encodecaps(bundle2.getrepocaps(self,
                                                               role='client'))
             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.manifestcachesize
         manifestcachesize = self.ui.configint('format', 'manifestcachesize')
         if manifestcachesize is not None:
             self.svfs.options['manifestcachesize'] = manifestcachesize
         deltabothparents = self.ui.configbool('storage',
             'revlog.optimize-delta-parent-choice')
         self.svfs.options['deltabothparents'] = deltabothparents
         self.svfs.options['lazydeltabase'] = not scmutil.gddeltaconfig(self.ui)
         chainspan = self.ui.configbytes('experimental', 'maxdeltachainspan')
         if 0 <= chainspan:
             self.svfs.options['maxdeltachainspan'] = chainspan
         mmapindexthreshold = self.ui.configbytes('experimental',
                                                  'mmapindexthreshold')
         if mmapindexthreshold is not None:
             self.svfs.options['mmapindexthreshold'] = mmapindexthreshold
         withsparseread = self.ui.configbool('experimental', 'sparse-read')
         srdensitythres = float(self.ui.config('experimental',
                                               'sparse-read.density-threshold'))
         srmingapsize = self.ui.configbytes('experimental',
                                            'sparse-read.min-gap-size')
         self.svfs.options['with-sparse-read'] = withsparseread
         self.svfs.options['sparse-read-density-threshold'] = srdensitythres
         self.svfs.options['sparse-read-min-gap-size'] = srmingapsize
         sparserevlog = SPARSEREVLOG_REQUIREMENT in self.requirements
         self.svfs.options['sparse-revlog'] = sparserevlog
         if sparserevlog:
             self.svfs.options['generaldelta'] = True
         maxchainlen = None
         if sparserevlog:
             maxchainlen = revlogconst.SPARSE_REVLOG_MAX_CHAIN_LENGTH
         # experimental config: format.maxchainlen
         maxchainlen = self.ui.configint('format', 'maxchainlen', maxchainlen)
         if maxchainlen is not None:
             self.svfs.options['maxchainlen'] = maxchainlen
 
         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)
 
     # Don't cache auditor/nofsauditor, or you'll end up with reference cycle:
     # self -> auditor -> self._checknested -> self
 
     @property
     def auditor(self):
         # This is only used by context.workingctx.match in order to
         # detect files in subrepos.
         return pathutil.pathauditor(self.root, callback=self._checknested)
 
     @property
     def nofsauditor(self):
         # This is only used by context.basectx.match in order to detect
         # files in subrepos.
         return pathutil.pathauditor(self.root, callback=self._checknested,
                                     realfs=False, cached=True)
 
     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, visibilityexceptions=None):
         """Return a filtered version of a repository"""
         cls = repoview.newtype(self.unfiltered().__class__)
         return cls(self, name, visibilityexceptions)
 
     @repofilecache('bookmarks', 'bookmarks.current')
     def _bookmarks(self):
         return bookmarks.bmstore(self)
 
     @property
     def _activebookmark(self):
         return self._bookmarks.active
 
     # _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 self._makedirstate()
 
     def _makedirstate(self):
         """Extension point for wrapping the dirstate per-repo."""
         sparsematchfn = lambda: sparse.matcher(self)
 
         return dirstate.dirstate(self.vfs, self.ui, self.root,
                                  self._dirstatevalidate, sparsematchfn)
 
     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
 
     @storecache(narrowspec.FILENAME)
     def narrowpats(self):
         """matcher patterns for this repository's narrowspec
 
         A tuple of (includes, excludes).
         """
         source = self
         if self.shared():
             from . import hg
             source = hg.sharedreposource(self)
         return narrowspec.load(source)
 
     @storecache(narrowspec.FILENAME)
     def _narrowmatch(self):
         if repository.NARROW_REQUIREMENT not in self.requirements:
             return matchmod.always(self.root, '')
         include, exclude = self.narrowpats
         return narrowspec.match(self.root, include=include, exclude=exclude)
 
     # TODO(martinvonz): make this property-like instead?
     def narrowmatch(self):
         return self._narrowmatch
 
     def setnarrowpats(self, newincludes, newexcludes):
         narrowspec.save(self, newincludes, newexcludes)
         self.invalidate(clearfilecache=True)
 
     def __getitem__(self, changeid):
         if changeid is None:
             return context.workingctx(self)
         if isinstance(changeid, context.basectx):
             return changeid
         if isinstance(changeid, slice):
             # wdirrev isn't contiguous so the slice shouldn't include it
             return [context.changectx(self, i)
                     for i in pycompat.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.AmbiguousPrefixLookupError 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):
         # no need to pay the cost of repoview.changelog
         unfi = self.unfiltered()
         return len(unfi.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, localalias=None):
         '''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``. To provide some local
         definitions overriding user aliases, set ``localalias`` to
         ``{name: definitionstring}``.
         '''
         if user:
             m = revset.matchany(self.ui, specs,
                                 lookup=revset.lookupfn(self),
                                 localalias=localalias)
         else:
             m = revset.matchany(None, specs, localalias=localalias)
         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"""
         return self._bookmarks.names(node)
 
     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 scmutil.revsymbol(self, key).node()
 
     def lookupbranch(self, key):
         if key in self.branchmap():
             return key
 
         return scmutil.revsymbol(self, 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', 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 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, changectx=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,
                                changectx=changectx)
 
     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: procutil.filter(s, c)
                 # Wrap old filters not supporting keyword arguments
                 if not pycompat.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, **kwargs):
         """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,
                             **kwargs)
             if 'x' in flags:
                 self.wvfs.setflags(filename, False, True)
             else:
                 self.wvfs.setflags(filename, False, False)
         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(name=desc)
 
         # 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')
         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)
             repo = reporef()
             if repo.ui.configbool('experimental', 'single-head-per-branch'):
                 scmutil.enforcesinglehead(repo, tr2, desc)
             if hook.hashook(repo.ui, 'pretxnclose-bookmark'):
                 for name, (old, new) in sorted(tr.changes['bookmarks'].items()):
                     args = tr.hookargs.copy()
                     args.update(bookmarks.preparehookargs(name, old, new))
                     repo.hook('pretxnclose-bookmark', throw=True,
                               txnname=desc,
                               **pycompat.strkwargs(args))
             if hook.hashook(repo.ui, 'pretxnclose-phase'):
                 cl = repo.unfiltered().changelog
                 for rev, (old, new) in tr.changes['phases'].items():
                     args = tr.hookargs.copy()
                     node = hex(cl.node(rev))
                     args.update(phases.preparehookargs(node, old, new))
                     repo.hook('pretxnclose-phase', throw=True, txnname=desc,
                               **pycompat.strkwargs(args))
 
             repo.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
                 narrowspec.restorebackup(self, 'journal.narrowspec')
                 repo.dirstate.restorebackup(None, 'journal.dirstate')
 
                 repo.invalidate(clearfilecache=True)
 
         tr = transaction.transaction(rp, self.svfs, vfsmap,
                                      "journal",
                                      "undo",
                                      aftertrans(renames),
                                      self.store.createmode,
                                      validator=validate,
                                      releasefn=releasefn,
                                      checkambigfiles=_cachedfiles,
                                      name=desc)
         tr.changes['origrepolen'] = len(self)
         tr.changes['obsmarkers'] = set()
         tr.changes['phases'] = {}
         tr.changes['bookmarks'] = {}
 
         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 hookfunc():
                 repo = reporef()
                 if hook.hashook(repo.ui, 'txnclose-bookmark'):
                     bmchanges = sorted(tr.changes['bookmarks'].items())
                     for name, (old, new) in bmchanges:
                         args = tr.hookargs.copy()
                         args.update(bookmarks.preparehookargs(name, old, new))
                         repo.hook('txnclose-bookmark', throw=False,
                                   txnname=desc, **pycompat.strkwargs(args))
 
                 if hook.hashook(repo.ui, 'txnclose-phase'):
                     cl = repo.unfiltered().changelog
                     phasemv = sorted(tr.changes['phases'].items())
                     for rev, (old, new) in phasemv:
                         args = tr.hookargs.copy()
                         node = hex(cl.node(rev))
                         args.update(phases.preparehookargs(node, old, new))
                         repo.hook('txnclose-phase', throw=False, txnname=desc,
                                   **pycompat.strkwargs(args))
 
                 repo.hook('txnclose', throw=False, txnname=desc,
                           **pycompat.strkwargs(hookargs))
             reporef()._afterlock(hookfunc)
         tr.addfinalize('txnclose-hook', txnclosehook)
         # Include a leading "-" to make it happen before the transaction summary
         # reports registered via scmutil.registersummarycallback() whose names
         # are 00-txnreport etc. That way, the caches will be warm when the
         # callbacks run.
         tr.addpostclose('-warm-cache', self._buildcacheupdater(tr))
         def txnaborthook(tr2):
             """To be run if transaction is aborted
             """
             reporef().hook('txnabort', throw=False, txnname=desc,
                            **pycompat.strkwargs(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)
         scmutil.registersummarycallback(self, tr, desc)
         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, 'journal.dirstate')
         narrowspec.savebackup(self, 'journal.narrowspec')
         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,
                                      checkambigfiles=_cachedfiles)
                 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,
                              checkambigfiles=_cachedfiles)
         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()
 
             narrowspec.restorebackup(self, 'undo.narrowspec')
             self.dirstate.restorebackup(None, 'undo.dirstate')
             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, full=False):
         """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 'full' is set, make sure all caches the function knows about have
         up-to-date data. Even the ones usually loaded more lazily.
         """
         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['origrepolen'] < len(self):
             # updating the unfiltered branchmap should refresh all the others,
             self.ui.debug('updating the branch cache\n')
             branchmap.updatecache(self.filtered('served'))
 
         if full:
             rbc = self.revbranchcache()
             for r in self.changelog:
                 rbc.branchinfo(r)
             rbc.write()
 
             # ensure the working copy parents are in the manifestfulltextcache
             for ctx in self['.'].parents():
                 ctx.manifest()  # accessing the manifest is enough
 
     def invalidatecaches(self):
 
         if '_tagscache' in vars(self):
             # can't use delattr on proxy
             del self.__dict__['_tagscache']
 
         self.unfiltered()._branchcaches.clear()
         self.invalidatevolatilesets()
         self._sparsesignaturecache.clear()
 
     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 (k == 'changelog' and
                 self.currenttransaction() and
                 self.changelog._delayed):
                 # The changelog object may store unwritten revisions. We don't
                 # want to lose them.
                 # TODO: Solve the problem instead of working around it.
                 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():
             k = pycompat.sysstr(k)
             if k == r'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)
 
         timeout = 0
         warntimeout = 0
         if wait:
             timeout = self.ui.configint("ui", "timeout")
             warntimeout = self.ui.configint("ui", "timeout.warn")
         # internal config: ui.signal-safe-lock
         signalsafe = self.ui.configbool('ui', 'signal-safe-lock')
 
         l = lockmod.trylock(self.ui, vfs, lockname, timeout, warntimeout,
                             releasefn=releasefn,
                             acquirefn=acquirefn, desc=desc,
                             inheritchecker=inheritchecker,
                             parentlock=parentlock,
                             signalsafe=signalsafe)
         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, newstate = subrepoutil.precommit(
                 self.ui, wctx, status, match, force=force)
 
             # 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') %
                                    subrepoutil.subrelpath(sub))
                     sr = sub.commit(cctx._text, user, date)
                     newstate[s] = (newstate[s][0], sr)
                 subrepoutil.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.
 
         ctx.files() should list all files involved in this commit, i.e.
         modified/added/removed files. On merge, it may be wider than the
         ctx.files() to be committed, since any file nodes derived directly
         from p1 or p2 are excluded from the committed ctx.files().
         """
 
         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
                 self.ui.debug('reusing known manifest\n')
                 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
                 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]
                 files = changed + removed
                 md = None
                 if not files:
                     # if no "files" actually changed in terms of the changelog,
                     # try hard to detect unmodified manifest entry so that the
                     # exact same commit can be reproduced later on convert.
                     md = m1.diff(m, scmutil.matchfiles(self, ctx.files()))
                 if not files and md:
                     self.ui.debug('not reusing manifest (no file change in '
                                   'changelog, but manifest differs)\n')
                 if files or md:
                     self.ui.note(_("committing manifest\n"))
+                    # we're using narrowmatch here since it's already applied at
+                    # other stages (such as dirstate.walk), so we're already
+                    # ignoring things outside of narrowspec in most cases. The
+                    # one case where we might have files outside the narrowspec
+                    # at this point is merges, and we already error out in the
+                    # case where the merge has files outside of the narrowspec,
+                    # so this is safe.
                     mn = mctx.write(trp, linkrev,
                                     p1.manifestnode(), p2.manifestnode(),
-                                    added, drop)
+                                    added, drop, match=self.narrowmatch())
                 else:
                     self.ui.debug('reusing manifest form p1 (listed files '
                                   'actually unchanged)\n')
                     mn = p1.manifestnode()
             else:
                 self.ui.debug('reusing manifest from p1 (no file change)\n')
                 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 = subrepoutil.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.registernew(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 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.
         """
 
     @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 = pycompat.strkwargs(hookargs)
             hookargs[r'namespace'] = namespace
             hookargs[r'key'] = key
             hookargs[r'old'] = old
             hookargs[r'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, pycompat.bytestr(three),
                                    pycompat.bytestr(four),
                                    pycompat.bytestr(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, intents=None, createopts=None):
     localpath = util.urllocalpath(path)
     if create:
         createrepository(ui, localpath, createopts=createopts)
 
     return localrepository(ui, localpath, intents=intents)
 
 def islocal(path):
     return True
 
 def newreporequirements(ui, createopts=None):
     """Determine the set of requirements for a new local repository.
 
     Extensions can wrap this function to specify custom requirements for
     new repositories.
     """
     createopts = createopts or {}
 
     requirements = {'revlogv1'}
     if ui.configbool('format', 'usestore'):
         requirements.add('store')
         if ui.configbool('format', 'usefncache'):
             requirements.add('fncache')
             if ui.configbool('format', 'dotencode'):
                 requirements.add('dotencode')
 
     compengine = ui.config('experimental', 'format.compression')
     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'):
         requirements.add('treemanifest')
     # experimental config: format.sparse-revlog
     if ui.configbool('format', 'sparse-revlog'):
         requirements.add(SPARSEREVLOG_REQUIREMENT)
 
     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)
     # experimental config: format.internal-phase
     if ui.configbool('format', 'internal-phase'):
         requirements.add('internal-phase')
 
     if createopts.get('narrowfiles'):
         requirements.add(repository.NARROW_REQUIREMENT)
 
     return requirements
 
 def filterknowncreateopts(ui, createopts):
     """Filters a dict of repo creation options against options that are known.
 
     Receives a dict of repo creation options and returns a dict of those
     options that we don't know how to handle.
 
     This function is called as part of repository creation. If the
     returned dict contains any items, repository creation will not
     be allowed, as it means there was a request to create a repository
     with options not recognized by loaded code.
 
     Extensions can wrap this function to filter out creation options
     they know how to handle.
     """
     known = {'narrowfiles'}
 
     return {k: v for k, v in createopts.items() if k not in known}
 
 def createrepository(ui, path, createopts=None):
     """Create a new repository in a vfs.
 
     ``path`` path to the new repo's working directory.
     ``createopts`` options for the new repository.
     """
     createopts = createopts or {}
 
     unknownopts = filterknowncreateopts(ui, createopts)
 
     if not isinstance(unknownopts, dict):
         raise error.ProgrammingError('filterknowncreateopts() did not return '
                                      'a dict')
 
     if unknownopts:
         raise error.Abort(_('unable to create repository because of unknown '
                             'creation option: %s') %
                           ', '.sorted(unknownopts),
                           hint=_('is a required extension not loaded?'))
 
     requirements = newreporequirements(ui, createopts=createopts)
 
     wdirvfs = vfsmod.vfs(path, expandpath=True, realpath=True)
     if not wdirvfs.exists():
         wdirvfs.makedirs()
 
     hgvfs = vfsmod.vfs(wdirvfs.join(b'.hg'))
     if hgvfs.exists():
         raise error.RepoError(_('repository %s already exists') % path)
 
     hgvfs.makedir(notindexed=True)
 
     if b'store' in requirements:
         hgvfs.mkdir(b'store')
 
         # We create an invalid changelog outside the store so very old
         # Mercurial versions (which didn't know about the requirements
         # file) encounter an error on reading the changelog. This
         # effectively locks out old clients and prevents them from
         # mucking with a repo in an unknown format.
         #
         # The revlog header has version 2, which won't be recognized by
         # such old clients.
         hgvfs.append(b'00changelog.i',
                      b'\0\0\0\2 dummy changelog to prevent using the old repo '
                      b'layout')
 
     scmutil.writerequires(hgvfs, requirements)
 
 def poisonrepository(repo):
     """Poison a repository instance so it can no longer be used."""
     # Perform any cleanup on the instance.
     repo.close()
 
     # Our strategy is to replace the type of the object with one that
     # has all attribute lookups result in error.
     #
     # But we have to allow the close() method because some constructors
     # of repos call close() on repo references.
     class poisonedrepository(object):
         def __getattribute__(self, item):
             if item == r'close':
                 return object.__getattribute__(self, item)
 
             raise error.ProgrammingError('repo instances should not be used '
                                          'after unshare')
 
         def close(self):
             pass
 
     # We may have a repoview, which intercepts __setattr__. So be sure
     # we operate at the lowest level possible.
     object.__setattr__(repo, r'__class__', poisonedrepository)

mercurial/manifest.py

@@ -1,2004 +1,2017 b''
 # manifest.py - manifest revision class for mercurial
 #
 # Copyright 2005-2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import heapq
 import itertools
 import struct
 import weakref
 
 from .i18n import _
 from .node import (
     bin,
     hex,
     nullid,
     nullrev,
 )
 from . import (
     error,
     mdiff,
     policy,
     pycompat,
     repository,
     revlog,
     util,
 )
 from .utils import (
     interfaceutil,
 )
 
 parsers = policy.importmod(r'parsers')
 propertycache = util.propertycache
 
 def _parse(data):
     # This method does a little bit of excessive-looking
     # precondition checking. This is so that the behavior of this
     # class exactly matches its C counterpart to try and help
     # prevent surprise breakage for anyone that develops against
     # the pure version.
     if data and data[-1:] != '\n':
         raise ValueError('Manifest did not end in a newline.')
     prev = None
     for l in data.splitlines():
         if prev is not None and prev > l:
             raise ValueError('Manifest lines not in sorted order.')
         prev = l
         f, n = l.split('\0')
         if len(n) > 40:
             yield f, bin(n[:40]), n[40:]
         else:
             yield f, bin(n), ''
 
 def _text(it):
     files = []
     lines = []
     for f, n, fl in it:
         files.append(f)
         # if this is changed to support newlines in filenames,
         # be sure to check the templates/ dir again (especially *-raw.tmpl)
         lines.append("%s\0%s%s\n" % (f, hex(n), fl))
 
     _checkforbidden(files)
     return ''.join(lines)
 
 class lazymanifestiter(object):
     def __init__(self, lm):
         self.pos = 0
         self.lm = lm
 
     def __iter__(self):
         return self
 
     def next(self):
         try:
             data, pos = self.lm._get(self.pos)
         except IndexError:
             raise StopIteration
         if pos == -1:
             self.pos += 1
             return data[0]
         self.pos += 1
         zeropos = data.find('\x00', pos)
         return data[pos:zeropos]
 
     __next__ = next
 
 class lazymanifestiterentries(object):
     def __init__(self, lm):
         self.lm = lm
         self.pos = 0
 
     def __iter__(self):
         return self
 
     def next(self):
         try:
             data, pos = self.lm._get(self.pos)
         except IndexError:
             raise StopIteration
         if pos == -1:
             self.pos += 1
             return data
         zeropos = data.find('\x00', pos)
         hashval = unhexlify(data, self.lm.extrainfo[self.pos],
                             zeropos + 1, 40)
         flags = self.lm._getflags(data, self.pos, zeropos)
         self.pos += 1
         return (data[pos:zeropos], hashval, flags)
 
     __next__ = next
 
 def unhexlify(data, extra, pos, length):
     s = bin(data[pos:pos + length])
     if extra:
         s += chr(extra & 0xff)
     return s
 
 def _cmp(a, b):
     return (a > b) - (a < b)
 
 class _lazymanifest(object):
     def __init__(self, data, positions=None, extrainfo=None, extradata=None):
         if positions is None:
             self.positions = self.findlines(data)
             self.extrainfo = [0] * len(self.positions)
             self.data = data
             self.extradata = []
         else:
             self.positions = positions[:]
             self.extrainfo = extrainfo[:]
             self.extradata = extradata[:]
             self.data = data
 
     def findlines(self, data):
         if not data:
             return []
         pos = data.find("\n")
         if pos == -1 or data[-1:] != '\n':
             raise ValueError("Manifest did not end in a newline.")
         positions = [0]
         prev = data[:data.find('\x00')]
         while pos < len(data) - 1 and pos != -1:
             positions.append(pos + 1)
             nexts = data[pos + 1:data.find('\x00', pos + 1)]
             if nexts < prev:
                 raise ValueError("Manifest lines not in sorted order.")
             prev = nexts
             pos = data.find("\n", pos + 1)
         return positions
 
     def _get(self, index):
         # get the position encoded in pos:
         #   positive number is an index in 'data'
         #   negative number is in extrapieces
         pos = self.positions[index]
         if pos >= 0:
             return self.data, pos
         return self.extradata[-pos - 1], -1
 
     def _getkey(self, pos):
         if pos >= 0:
             return self.data[pos:self.data.find('\x00', pos + 1)]
         return self.extradata[-pos - 1][0]
 
     def bsearch(self, key):
         first = 0
         last = len(self.positions) - 1
 
         while first <= last:
             midpoint = (first + last)//2
             nextpos = self.positions[midpoint]
             candidate = self._getkey(nextpos)
             r = _cmp(key, candidate)
             if r == 0:
                 return midpoint
             else:
                 if r < 0:
                     last = midpoint - 1
                 else:
                     first = midpoint + 1
         return -1
 
     def bsearch2(self, key):
         # same as the above, but will always return the position
         # done for performance reasons
         first = 0
         last = len(self.positions) - 1
 
         while first <= last:
             midpoint = (first + last)//2
             nextpos = self.positions[midpoint]
             candidate = self._getkey(nextpos)
             r = _cmp(key, candidate)
             if r == 0:
                 return (midpoint, True)
             else:
                 if r < 0:
                     last = midpoint - 1
                 else:
                     first = midpoint + 1
         return (first, False)
 
     def __contains__(self, key):
         return self.bsearch(key) != -1
 
     def _getflags(self, data, needle, pos):
         start = pos + 41
         end = data.find("\n", start)
         if end == -1:
             end = len(data) - 1
         if start == end:
             return ''
         return self.data[start:end]
 
     def __getitem__(self, key):
         if not isinstance(key, bytes):
             raise TypeError("getitem: manifest keys must be a bytes.")
         needle = self.bsearch(key)
         if needle == -1:
             raise KeyError
         data, pos = self._get(needle)
         if pos == -1:
             return (data[1], data[2])
         zeropos = data.find('\x00', pos)
         assert 0 <= needle <= len(self.positions)
         assert len(self.extrainfo) == len(self.positions)
         hashval = unhexlify(data, self.extrainfo[needle], zeropos + 1, 40)
         flags = self._getflags(data, needle, zeropos)
         return (hashval, flags)
 
     def __delitem__(self, key):
         needle, found = self.bsearch2(key)
         if not found:
             raise KeyError
         cur = self.positions[needle]
         self.positions = self.positions[:needle] + self.positions[needle + 1:]
         self.extrainfo = self.extrainfo[:needle] + self.extrainfo[needle + 1:]
         if cur >= 0:
             self.data = self.data[:cur] + '\x00' + self.data[cur + 1:]
 
     def __setitem__(self, key, value):
         if not isinstance(key, bytes):
             raise TypeError("setitem: manifest keys must be a byte string.")
         if not isinstance(value, tuple) or len(value) != 2:
             raise TypeError("Manifest values must be a tuple of (node, flags).")
         hashval = value[0]
         if not isinstance(hashval, bytes) or not 20 <= len(hashval) <= 22:
             raise TypeError("node must be a 20-byte byte string")
         flags = value[1]
         if len(hashval) == 22:
             hashval = hashval[:-1]
         if not isinstance(flags, bytes) or len(flags) > 1:
             raise TypeError("flags must a 0 or 1 byte string, got %r", flags)
         needle, found = self.bsearch2(key)
         if found:
             # put the item
             pos = self.positions[needle]
             if pos < 0:
                 self.extradata[-pos - 1] = (key, hashval, value[1])
             else:
                 # just don't bother
                 self.extradata.append((key, hashval, value[1]))
                 self.positions[needle] = -len(self.extradata)
         else:
             # not found, put it in with extra positions
             self.extradata.append((key, hashval, value[1]))
             self.positions = (self.positions[:needle] + [-len(self.extradata)]
                               + self.positions[needle:])
             self.extrainfo = (self.extrainfo[:needle] + [0] +
                               self.extrainfo[needle:])
 
     def copy(self):
         # XXX call _compact like in C?
         return _lazymanifest(self.data, self.positions, self.extrainfo,
             self.extradata)
 
     def _compact(self):
         # hopefully not called TOO often
         if len(self.extradata) == 0:
             return
         l = []
         last_cut = 0
         i = 0
         offset = 0
         self.extrainfo = [0] * len(self.positions)
         while i < len(self.positions):
             if self.positions[i] >= 0:
                 cur = self.positions[i]
                 last_cut = cur
                 while True:
                     self.positions[i] = offset
                     i += 1
                     if i == len(self.positions) or self.positions[i] < 0:
                         break
                     offset += self.positions[i] - cur
                     cur = self.positions[i]
                 end_cut = self.data.find('\n', cur)
                 if end_cut != -1:
                     end_cut += 1
                 offset += end_cut - cur
                 l.append(self.data[last_cut:end_cut])
             else:
                 while i < len(self.positions) and self.positions[i] < 0:
                     cur = self.positions[i]
                     t = self.extradata[-cur - 1]
                     l.append(self._pack(t))
                     self.positions[i] = offset
                     if len(t[1]) > 20:
                         self.extrainfo[i] = ord(t[1][21])
                     offset += len(l[-1])
                     i += 1
         self.data = ''.join(l)
         self.extradata = []
 
     def _pack(self, d):
         return d[0] + '\x00' + hex(d[1][:20]) + d[2] + '\n'
 
     def text(self):
         self._compact()
         return self.data
 
     def diff(self, m2, clean=False):
         '''Finds changes between the current manifest and m2.'''
         # XXX think whether efficiency matters here
         diff = {}
 
         for fn, e1, flags in self.iterentries():
             if fn not in m2:
                 diff[fn] = (e1, flags), (None, '')
             else:
                 e2 = m2[fn]
                 if (e1, flags) != e2:
                     diff[fn] = (e1, flags), e2
                 elif clean:
                     diff[fn] = None
 
         for fn, e2, flags in m2.iterentries():
             if fn not in self:
                 diff[fn] = (None, ''), (e2, flags)
 
         return diff
 
     def iterentries(self):
         return lazymanifestiterentries(self)
 
     def iterkeys(self):
         return lazymanifestiter(self)
 
     def __iter__(self):
         return lazymanifestiter(self)
 
     def __len__(self):
         return len(self.positions)
 
     def filtercopy(self, filterfn):
         # XXX should be optimized
         c = _lazymanifest('')
         for f, n, fl in self.iterentries():
             if filterfn(f):
                 c[f] = n, fl
         return c
 
 try:
     _lazymanifest = parsers.lazymanifest
 except AttributeError:
     pass
 
 @interfaceutil.implementer(repository.imanifestdict)
 class manifestdict(object):
     def __init__(self, data=''):
         self._lm = _lazymanifest(data)
 
     def __getitem__(self, key):
         return self._lm[key][0]
 
     def find(self, key):
         return self._lm[key]
 
     def __len__(self):
         return len(self._lm)
 
     def __nonzero__(self):
         # nonzero is covered by the __len__ function, but implementing it here
         # makes it easier for extensions to override.
         return len(self._lm) != 0
 
     __bool__ = __nonzero__
 
     def __setitem__(self, key, node):
         self._lm[key] = node, self.flags(key, '')
 
     def __contains__(self, key):
         if key is None:
             return False
         return key in self._lm
 
     def __delitem__(self, key):
         del self._lm[key]
 
     def __iter__(self):
         return self._lm.__iter__()
 
     def iterkeys(self):
         return self._lm.iterkeys()
 
     def keys(self):
         return list(self.iterkeys())
 
     def filesnotin(self, m2, match=None):
         '''Set of files in this manifest that are not in the other'''
         if match:
             m1 = self.matches(match)
             m2 = m2.matches(match)
             return m1.filesnotin(m2)
         diff = self.diff(m2)
         files = set(filepath
                     for filepath, hashflags in diff.iteritems()
                     if hashflags[1][0] is None)
         return files
 
     @propertycache
     def _dirs(self):
         return util.dirs(self)
 
     def dirs(self):
         return self._dirs
 
     def hasdir(self, dir):
         return dir in self._dirs
 
     def _filesfastpath(self, match):
         '''Checks whether we can correctly and quickly iterate over matcher
         files instead of over manifest files.'''
         files = match.files()
         return (len(files) < 100 and (match.isexact() or
             (match.prefix() and all(fn in self for fn in files))))
 
     def walk(self, match):
         '''Generates matching file names.
 
         Equivalent to manifest.matches(match).iterkeys(), but without creating
         an entirely new manifest.
 
         It also reports nonexistent files by marking them bad with match.bad().
         '''
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         # avoid the entire walk if we're only looking for specific files
         if self._filesfastpath(match):
             for fn in sorted(fset):
                 yield fn
             return
 
         for fn in self:
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             if match(fn):
                 yield fn
 
         # for dirstate.walk, files=['.'] means "walk the whole tree".
         # follow that here, too
         fset.discard('.')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def matches(self, match):
         '''generate a new manifest filtered by the match argument'''
         if match.always():
             return self.copy()
 
         if self._filesfastpath(match):
             m = manifestdict()
             lm = self._lm
             for fn in match.files():
                 if fn in lm:
                     m._lm[fn] = lm[fn]
             return m
 
         m = manifestdict()
         m._lm = self._lm.filtercopy(match)
         return m
 
     def diff(self, m2, match=None, clean=False):
         '''Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         '''
         if match:
             m1 = self.matches(match)
             m2 = m2.matches(match)
             return m1.diff(m2, clean=clean)
         return self._lm.diff(m2._lm, clean)
 
     def setflag(self, key, flag):
         self._lm[key] = self[key], flag
 
     def get(self, key, default=None):
         try:
             return self._lm[key][0]
         except KeyError:
             return default
 
     def flags(self, key, default=''):
         try:
             return self._lm[key][1]
         except KeyError:
             return default
 
     def copy(self):
         c = manifestdict()
         c._lm = self._lm.copy()
         return c
 
     def items(self):
         return (x[:2] for x in self._lm.iterentries())
 
     def iteritems(self):
         return (x[:2] for x in self._lm.iterentries())
 
     def iterentries(self):
         return self._lm.iterentries()
 
     def text(self):
         # most likely uses native version
         return self._lm.text()
 
     def fastdelta(self, base, changes):
         """Given a base manifest text as a bytearray and a list of changes
         relative to that text, compute a delta that can be used by revlog.
         """
         delta = []
         dstart = None
         dend = None
         dline = [""]
         start = 0
         # zero copy representation of base as a buffer
         addbuf = util.buffer(base)
 
         changes = list(changes)
         if len(changes) < 1000:
             # start with a readonly loop that finds the offset of
             # each line and creates the deltas
             for f, todelete in changes:
                 # bs will either be the index of the item or the insert point
                 start, end = _msearch(addbuf, f, start)
                 if not todelete:
                     h, fl = self._lm[f]
                     l = "%s\0%s%s\n" % (f, hex(h), fl)
                 else:
                     if start == end:
                         # item we want to delete was not found, error out
                         raise AssertionError(
                                 _("failed to remove %s from manifest") % f)
                     l = ""
                 if dstart is not None and dstart <= start and dend >= start:
                     if dend < end:
                         dend = end
                     if l:
                         dline.append(l)
                 else:
                     if dstart is not None:
                         delta.append([dstart, dend, "".join(dline)])
                     dstart = start
                     dend = end
                     dline = [l]
 
             if dstart is not None:
                 delta.append([dstart, dend, "".join(dline)])
             # apply the delta to the base, and get a delta for addrevision
             deltatext, arraytext = _addlistdelta(base, delta)
         else:
             # For large changes, it's much cheaper to just build the text and
             # diff it.
             arraytext = bytearray(self.text())
             deltatext = mdiff.textdiff(
                 util.buffer(base), util.buffer(arraytext))
 
         return arraytext, deltatext
 
 def _msearch(m, s, lo=0, hi=None):
     '''return a tuple (start, end) that says where to find s within m.
 
     If the string is found m[start:end] are the line containing
     that string.  If start == end the string was not found and
     they indicate the proper sorted insertion point.
 
     m should be a buffer, a memoryview or a byte string.
     s is a byte string'''
     def advance(i, c):
         while i < lenm and m[i:i + 1] != c:
             i += 1
         return i
     if not s:
         return (lo, lo)
     lenm = len(m)
     if not hi:
         hi = lenm
     while lo < hi:
         mid = (lo + hi) // 2
         start = mid
         while start > 0 and m[start - 1:start] != '\n':
             start -= 1
         end = advance(start, '\0')
         if bytes(m[start:end]) < s:
             # we know that after the null there are 40 bytes of sha1
             # this translates to the bisect lo = mid + 1
             lo = advance(end + 40, '\n') + 1
         else:
             # this translates to the bisect hi = mid
             hi = start
     end = advance(lo, '\0')
     found = m[lo:end]
     if s == found:
         # we know that after the null there are 40 bytes of sha1
         end = advance(end + 40, '\n')
         return (lo, end + 1)
     else:
         return (lo, lo)
 
 def _checkforbidden(l):
     """Check filenames for illegal characters."""
     for f in l:
         if '\n' in f or '\r' in f:
             raise error.RevlogError(
                 _("'\\n' and '\\r' disallowed in filenames: %r")
                 % pycompat.bytestr(f))
 
 
 # apply the changes collected during the bisect loop to our addlist
 # return a delta suitable for addrevision
 def _addlistdelta(addlist, x):
     # for large addlist arrays, building a new array is cheaper
     # than repeatedly modifying the existing one
     currentposition = 0
     newaddlist = bytearray()
 
     for start, end, content in x:
         newaddlist += addlist[currentposition:start]
         if content:
             newaddlist += bytearray(content)
 
         currentposition = end
 
     newaddlist += addlist[currentposition:]
 
     deltatext = "".join(struct.pack(">lll", start, end, len(content))
                    + content for start, end, content in x)
     return deltatext, newaddlist
 
 def _splittopdir(f):
     if '/' in f:
         dir, subpath = f.split('/', 1)
         return dir + '/', subpath
     else:
         return '', f
 
 _noop = lambda s: None
 
 class treemanifest(object):
     def __init__(self, dir='', text=''):
         self._dir = dir
         self._node = nullid
         self._loadfunc = _noop
         self._copyfunc = _noop
         self._dirty = False
         self._dirs = {}
         self._lazydirs = {}
         # Using _lazymanifest here is a little slower than plain old dicts
         self._files = {}
         self._flags = {}
         if text:
             def readsubtree(subdir, subm):
                 raise AssertionError('treemanifest constructor only accepts '
                                      'flat manifests')
             self.parse(text, readsubtree)
             self._dirty = True # Mark flat manifest dirty after parsing
 
     def _subpath(self, path):
         return self._dir + path
 
     def _loadalllazy(self):
         for k, (path, node, readsubtree) in self._lazydirs.iteritems():
             self._dirs[k] = readsubtree(path, node)
         self._lazydirs = {}
 
     def _loadlazy(self, d):
         path, node, readsubtree = self._lazydirs[d]
         self._dirs[d] = readsubtree(path, node)
         del self._lazydirs[d]
 
     def _loadchildrensetlazy(self, visit):
         if not visit:
             return None
         if visit == 'all' or visit == 'this':
             self._loadalllazy()
             return None
 
         todel = []
         for k in visit:
             kslash = k + '/'
             ld = self._lazydirs.get(kslash)
             if ld:
                 path, node, readsubtree = ld
                 self._dirs[kslash] = readsubtree(path, node)
                 todel.append(kslash)
         for kslash in todel:
             del self._lazydirs[kslash]
         return visit
 
     def __len__(self):
         self._load()
         size = len(self._files)
         self._loadalllazy()
         for m in self._dirs.values():
             size += m.__len__()
         return size
 
     def __nonzero__(self):
         # Faster than "__len() != 0" since it avoids loading sub-manifests
         return not self._isempty()
 
     __bool__ = __nonzero__
 
     def _isempty(self):
         self._load() # for consistency; already loaded by all callers
         # See if we can skip loading everything.
         if self._files or (self._dirs and
                            any(not m._isempty() for m in self._dirs.values())):
             return False
         self._loadalllazy()
         return (not self._dirs or
                 all(m._isempty() for m in self._dirs.values()))
 
     def __repr__(self):
         return ('<treemanifest dir=%s, node=%s, loaded=%s, dirty=%s at 0x%x>' %
                 (self._dir, hex(self._node),
                  bool(self._loadfunc is _noop),
                  self._dirty, id(self)))
 
     def dir(self):
         '''The directory that this tree manifest represents, including a
         trailing '/'. Empty string for the repo root directory.'''
         return self._dir
 
     def node(self):
         '''This node of this instance. nullid for unsaved instances. Should
         be updated when the instance is read or written from a revlog.
         '''
         assert not self._dirty
         return self._node
 
     def setnode(self, node):
         self._node = node
         self._dirty = False
 
     def iterentries(self):
         self._load()
         self._loadalllazy()
         for p, n in sorted(itertools.chain(self._dirs.items(),
                                            self._files.items())):
             if p in self._files:
                 yield self._subpath(p), n, self._flags.get(p, '')
             else:
                 for x in n.iterentries():
                     yield x
 
     def items(self):
         self._load()
         self._loadalllazy()
         for p, n in sorted(itertools.chain(self._dirs.items(),
                                            self._files.items())):
             if p in self._files:
                 yield self._subpath(p), n
             else:
                 for f, sn in n.iteritems():
                     yield f, sn
 
     iteritems = items
 
     def iterkeys(self):
         self._load()
         self._loadalllazy()
         for p in sorted(itertools.chain(self._dirs, self._files)):
             if p in self._files:
                 yield self._subpath(p)
             else:
                 for f in self._dirs[p]:
                     yield f
 
     def keys(self):
         return list(self.iterkeys())
 
     def __iter__(self):
         return self.iterkeys()
 
     def __contains__(self, f):
         if f is None:
             return False
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir in self._lazydirs:
                 self._loadlazy(dir)
 
             if dir not in self._dirs:
                 return False
 
             return self._dirs[dir].__contains__(subpath)
         else:
             return f in self._files
 
     def get(self, f, default=None):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir in self._lazydirs:
                 self._loadlazy(dir)
 
             if dir not in self._dirs:
                 return default
             return self._dirs[dir].get(subpath, default)
         else:
             return self._files.get(f, default)
 
     def __getitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir in self._lazydirs:
                 self._loadlazy(dir)
 
             return self._dirs[dir].__getitem__(subpath)
         else:
             return self._files[f]
 
     def flags(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir in self._lazydirs:
                 self._loadlazy(dir)
 
             if dir not in self._dirs:
                 return ''
             return self._dirs[dir].flags(subpath)
         else:
             if f in self._lazydirs or f in self._dirs:
                 return ''
             return self._flags.get(f, '')
 
     def find(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir in self._lazydirs:
                 self._loadlazy(dir)
 
             return self._dirs[dir].find(subpath)
         else:
             return self._files[f], self._flags.get(f, '')
 
     def __delitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir in self._lazydirs:
                 self._loadlazy(dir)
 
             self._dirs[dir].__delitem__(subpath)
             # If the directory is now empty, remove it
             if self._dirs[dir]._isempty():
                 del self._dirs[dir]
         else:
             del self._files[f]
             if f in self._flags:
                 del self._flags[f]
         self._dirty = True
 
     def __setitem__(self, f, n):
         assert n is not None
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir in self._lazydirs:
                 self._loadlazy(dir)
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].__setitem__(subpath, n)
         else:
             self._files[f] = n[:21] # to match manifestdict's behavior
         self._dirty = True
 
     def _load(self):
         if self._loadfunc is not _noop:
             lf, self._loadfunc = self._loadfunc, _noop
             lf(self)
         elif self._copyfunc is not _noop:
             cf, self._copyfunc = self._copyfunc, _noop
             cf(self)
 
     def setflag(self, f, flags):
         """Set the flags (symlink, executable) for path f."""
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             if dir in self._lazydirs:
                 self._loadlazy(dir)
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].setflag(subpath, flags)
         else:
             self._flags[f] = flags
         self._dirty = True
 
     def copy(self):
         copy = treemanifest(self._dir)
         copy._node = self._node
         copy._dirty = self._dirty
         if self._copyfunc is _noop:
             def _copyfunc(s):
                 self._load()
                 # OPT: it'd be nice to not load everything here. Unfortunately
                 # this makes a mess of the "dirty" state tracking if we don't.
                 self._loadalllazy()
                 sdirs = s._dirs
                 for d, v in self._dirs.iteritems():
                     sdirs[d] = v.copy()
                 s._files = dict.copy(self._files)
                 s._flags = dict.copy(self._flags)
             if self._loadfunc is _noop:
                 _copyfunc(copy)
             else:
                 copy._copyfunc = _copyfunc
         else:
             copy._copyfunc = self._copyfunc
         return copy
 
     def filesnotin(self, m2, match=None):
         '''Set of files in this manifest that are not in the other'''
         if match and not match.always():
             m1 = self.matches(match)
             m2 = m2.matches(match)
             return m1.filesnotin(m2)
 
         files = set()
         def _filesnotin(t1, t2):
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             t1._loadalllazy()
             t2._loadalllazy()
             for d, m1 in t1._dirs.iteritems():
                 if d in t2._dirs:
                     m2 = t2._dirs[d]
                     _filesnotin(m1, m2)
                 else:
                     files.update(m1.iterkeys())
 
             for fn in t1._files:
                 if fn not in t2._files:
                     files.add(t1._subpath(fn))
 
         _filesnotin(self, m2)
         return files
 
     @propertycache
     def _alldirs(self):
         return util.dirs(self)
 
     def dirs(self):
         return self._alldirs
 
     def hasdir(self, dir):
         self._load()
         topdir, subdir = _splittopdir(dir)
         if topdir:
             if topdir in self._lazydirs:
                 self._loadlazy(topdir)
             if topdir in self._dirs:
                 return self._dirs[topdir].hasdir(subdir)
             return False
         dirslash = dir + '/'
         return dirslash in self._dirs or dirslash in self._lazydirs
 
     def walk(self, match):
         '''Generates matching file names.
 
         Equivalent to manifest.matches(match).iterkeys(), but without creating
         an entirely new manifest.
 
         It also reports nonexistent files by marking them bad with match.bad().
         '''
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         for fn in self._walk(match):
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             yield fn
 
         # for dirstate.walk, files=['.'] means "walk the whole tree".
         # follow that here, too
         fset.discard('.')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def _walk(self, match):
         '''Recursively generates matching file names for walk().'''
         visit = match.visitchildrenset(self._dir[:-1] or '.')
         if not visit:
             return
 
         # yield this dir's files and walk its submanifests
         self._load()
         visit = self._loadchildrensetlazy(visit)
         for p in sorted(list(self._dirs) + list(self._files)):
             if p in self._files:
                 fullp = self._subpath(p)
                 if match(fullp):
                     yield fullp
             else:
                 if not visit or p[:-1] in visit:
                     for f in self._dirs[p]._walk(match):
                         yield f
 
     def matches(self, match):
         '''generate a new manifest filtered by the match argument'''
         if match.always():
             return self.copy()
 
         return self._matches(match)
 
     def _matches(self, match):
         '''recursively generate a new manifest filtered by the match argument.
         '''
 
         visit = match.visitchildrenset(self._dir[:-1] or '.')
         if visit == 'all':
             return self.copy()
         ret = treemanifest(self._dir)
         if not visit:
             return ret
 
         self._load()
         for fn in self._files:
             # While visitchildrenset *usually* lists only subdirs, this is
             # actually up to the matcher and may have some files in the set().
             # If visit == 'this', we should obviously look at the files in this
             # directory; if visit is a set, and fn is in it, we should inspect
             # fn (but no need to inspect things not in the set).
             if visit != 'this' and fn not in visit:
                 continue
             fullp = self._subpath(fn)
             # visitchildrenset isn't perfect, we still need to call the regular
             # matcher code to further filter results.
             if not match(fullp):
                 continue
             ret._files[fn] = self._files[fn]
             if fn in self._flags:
                 ret._flags[fn] = self._flags[fn]
 
         visit = self._loadchildrensetlazy(visit)
         for dir, subm in self._dirs.iteritems():
             if visit and dir[:-1] not in visit:
                 continue
             m = subm._matches(match)
             if not m._isempty():
                 ret._dirs[dir] = m
 
         if not ret._isempty():
             ret._dirty = True
         return ret
 
     def diff(self, m2, match=None, clean=False):
         '''Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         '''
         if match and not match.always():
             m1 = self.matches(match)
             m2 = m2.matches(match)
             return m1.diff(m2, clean=clean)
         result = {}
         emptytree = treemanifest()
         def _diff(t1, t2):
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             # OPT: do we need to load everything?
             t1._loadalllazy()
             t2._loadalllazy()
             for d, m1 in t1._dirs.iteritems():
                 m2 = t2._dirs.get(d, emptytree)
                 _diff(m1, m2)
 
             for d, m2 in t2._dirs.iteritems():
                 if d not in t1._dirs:
                     _diff(emptytree, m2)
 
             for fn, n1 in t1._files.iteritems():
                 fl1 = t1._flags.get(fn, '')
                 n2 = t2._files.get(fn, None)
                 fl2 = t2._flags.get(fn, '')
                 if n1 != n2 or fl1 != fl2:
                     result[t1._subpath(fn)] = ((n1, fl1), (n2, fl2))
                 elif clean:
                     result[t1._subpath(fn)] = None
 
             for fn, n2 in t2._files.iteritems():
                 if fn not in t1._files:
                     fl2 = t2._flags.get(fn, '')
                     result[t2._subpath(fn)] = ((None, ''), (n2, fl2))
 
         _diff(self, m2)
         return result
 
     def unmodifiedsince(self, m2):
         return not self._dirty and not m2._dirty and self._node == m2._node
 
     def parse(self, text, readsubtree):
         selflazy = self._lazydirs
         subpath = self._subpath
         for f, n, fl in _parse(text):
             if fl == 't':
                 f = f + '/'
                 selflazy[f] = (subpath(f), n, readsubtree)
             elif '/' in f:
                 # This is a flat manifest, so use __setitem__ and setflag rather
                 # than assigning directly to _files and _flags, so we can
                 # assign a path in a subdirectory, and to mark dirty (compared
                 # to nullid).
                 self[f] = n
                 if fl:
                     self.setflag(f, fl)
             else:
                 # Assigning to _files and _flags avoids marking as dirty,
                 # and should be a little faster.
                 self._files[f] = n
                 if fl:
                     self._flags[f] = fl
 
     def text(self):
         """Get the full data of this manifest as a bytestring."""
         self._load()
         return _text(self.iterentries())
 
     def dirtext(self):
         """Get the full data of this directory as a bytestring. Make sure that
         any submanifests have been written first, so their nodeids are correct.
         """
         self._load()
         flags = self.flags
         lazydirs = [(d[:-1], node, 't') for
                     d, (path, node, readsubtree) in self._lazydirs.iteritems()]
         dirs = [(d[:-1], self._dirs[d]._node, 't') for d in self._dirs]
         files = [(f, self._files[f], flags(f)) for f in self._files]
         return _text(sorted(dirs + files + lazydirs))
 
     def read(self, gettext, readsubtree):
         def _load_for_read(s):
             s.parse(gettext(), readsubtree)
             s._dirty = False
         self._loadfunc = _load_for_read
 
-    def writesubtrees(self, m1, m2, writesubtree):
+    def writesubtrees(self, m1, m2, writesubtree, match):
         self._load() # for consistency; should never have any effect here
         m1._load()
         m2._load()
         emptytree = treemanifest()
         def getnode(m, d):
             ld = m._lazydirs.get(d)
             if ld:
                 return ld[1]
             return m._dirs.get(d, emptytree)._node
 
+        # we should have always loaded everything by the time we get here for
+        # `self`, but possibly not in `m1` or `m2`.
+        assert not self._lazydirs
+        # let's skip investigating things that `match` says we do not need.
+        visit = match.visitchildrenset(self._dir[:-1] or '.')
+        if visit == 'this' or visit == 'all':
+            visit = None
         for d, subm in self._dirs.iteritems():
+            if visit and d[:-1] not in visit:
+                continue
             subp1 = getnode(m1, d)
             subp2 = getnode(m2, d)
             if subp1 == nullid:
                 subp1, subp2 = subp2, subp1
-            writesubtree(subm, subp1, subp2)
+            writesubtree(subm, subp1, subp2, match)
 
     def walksubtrees(self, matcher=None):
         """Returns an iterator of the subtrees of this manifest, including this
         manifest itself.
 
         If `matcher` is provided, it only returns subtrees that match.
         """
         if matcher and not matcher.visitdir(self._dir[:-1] or '.'):
             return
         if not matcher or matcher(self._dir[:-1]):
             yield self
 
         self._load()
         # OPT: use visitchildrenset to avoid loading everything.
         self._loadalllazy()
         for d, subm in self._dirs.iteritems():
             for subtree in subm.walksubtrees(matcher=matcher):
                 yield subtree
 
 class manifestfulltextcache(util.lrucachedict):
     """File-backed LRU cache for the manifest cache
 
     File consists of entries, up to EOF:
 
     - 20 bytes node, 4 bytes length, <length> manifest data
 
     These are written in reverse cache order (oldest to newest).
 
     """
     def __init__(self, max):
         super(manifestfulltextcache, self).__init__(max)
         self._dirty = False
         self._read = False
         self._opener = None
 
     def read(self):
         if self._read or self._opener is None:
             return
 
         try:
             with self._opener('manifestfulltextcache') as fp:
                 set = super(manifestfulltextcache, self).__setitem__
                 # ignore trailing data, this is a cache, corruption is skipped
                 while True:
                     node = fp.read(20)
                     if len(node) < 20:
                         break
                     try:
                         size = struct.unpack('>L', fp.read(4))[0]
                     except struct.error:
                         break
                     value = bytearray(fp.read(size))
                     if len(value) != size:
                         break
                     set(node, value)
         except IOError:
             # the file is allowed to be missing
             pass
 
         self._read = True
         self._dirty = False
 
     def write(self):
         if not self._dirty or self._opener is None:
             return
         # rotate backwards to the first used node
         with self._opener(
                 'manifestfulltextcache', 'w', atomictemp=True, checkambig=True
             ) as fp:
             node = self._head.prev
             while True:
                 if node.key in self._cache:
                     fp.write(node.key)
                     fp.write(struct.pack('>L', len(node.value)))
                     fp.write(node.value)
                 if node is self._head:
                     break
                 node = node.prev
 
     def __len__(self):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).__len__()
 
     def __contains__(self, k):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).__contains__(k)
 
     def __iter__(self):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).__iter__()
 
     def __getitem__(self, k):
         if not self._read:
             self.read()
         # the cache lru order can change on read
         setdirty = self._cache.get(k) is not self._head
         value = super(manifestfulltextcache, self).__getitem__(k)
         if setdirty:
             self._dirty = True
         return value
 
     def __setitem__(self, k, v):
         if not self._read:
             self.read()
         super(manifestfulltextcache, self).__setitem__(k, v)
         self._dirty = True
 
     def __delitem__(self, k):
         if not self._read:
             self.read()
         super(manifestfulltextcache, self).__delitem__(k)
         self._dirty = True
 
     def get(self, k, default=None):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).get(k, default=default)
 
     def clear(self, clear_persisted_data=False):
         super(manifestfulltextcache, self).clear()
         if clear_persisted_data:
             self._dirty = True
             self.write()
         self._read = False
 
 @interfaceutil.implementer(repository.imanifeststorage)
 class manifestrevlog(object):
     '''A revlog that stores manifest texts. This is responsible for caching the
     full-text manifest contents.
     '''
     def __init__(self, opener, tree='', dirlogcache=None, indexfile=None,
                  treemanifest=False):
         """Constructs a new manifest revlog
 
         `indexfile` - used by extensions to have two manifests at once, like
         when transitioning between flatmanifeset and treemanifests.
 
         `treemanifest` - used to indicate this is a tree manifest revlog. Opener
         options can also be used to make this a tree manifest revlog. The opener
         option takes precedence, so if it is set to True, we ignore whatever
         value is passed in to the constructor.
         """
         # During normal operations, we expect to deal with not more than four
         # revs at a time (such as during commit --amend). When rebasing large
         # stacks of commits, the number can go up, hence the config knob below.
         cachesize = 4
         optiontreemanifest = False
         opts = getattr(opener, 'options', None)
         if opts is not None:
             cachesize = opts.get('manifestcachesize', cachesize)
             optiontreemanifest = opts.get('treemanifest', False)
 
         self._treeondisk = optiontreemanifest or treemanifest
 
         self._fulltextcache = manifestfulltextcache(cachesize)
 
         if tree:
             assert self._treeondisk, 'opts is %r' % opts
 
         if indexfile is None:
             indexfile = '00manifest.i'
             if tree:
                 indexfile = "meta/" + tree + indexfile
 
         self.tree = tree
 
         # The dirlogcache is kept on the root manifest log
         if tree:
             self._dirlogcache = dirlogcache
         else:
             self._dirlogcache = {'': self}
 
         self._revlog = revlog.revlog(opener, indexfile,
                                      # only root indexfile is cached
                                      checkambig=not bool(tree),
                                      mmaplargeindex=True)
 
         self.index = self._revlog.index
         self.version = self._revlog.version
         self._generaldelta = self._revlog._generaldelta
 
     def _setupmanifestcachehooks(self, repo):
         """Persist the manifestfulltextcache on lock release"""
         if not util.safehasattr(repo, '_lockref'):
             return
 
         self._fulltextcache._opener = repo.cachevfs
         reporef = weakref.ref(repo)
         manifestrevlogref = weakref.ref(self)
 
         def persistmanifestcache():
             repo = reporef()
             self = manifestrevlogref()
             if repo is None or self is None:
                 return
             if repo.manifestlog.getstorage(b'') is not self:
                 # there's a different manifest in play now, abort
                 return
             self._fulltextcache.write()
 
         if repo._currentlock(repo._lockref) is not None:
             repo._afterlock(persistmanifestcache)
 
     @property
     def fulltextcache(self):
         return self._fulltextcache
 
     def clearcaches(self, clear_persisted_data=False):
         self._revlog.clearcaches()
         self._fulltextcache.clear(clear_persisted_data=clear_persisted_data)
         self._dirlogcache = {self.tree: self}
 
     def dirlog(self, d):
         if d:
             assert self._treeondisk
         if d not in self._dirlogcache:
             mfrevlog = manifestrevlog(self.opener, d,
                                       self._dirlogcache,
                                       treemanifest=self._treeondisk)
             self._dirlogcache[d] = mfrevlog
         return self._dirlogcache[d]
 
-    def add(self, m, transaction, link, p1, p2, added, removed, readtree=None):
+    def add(self, m, transaction, link, p1, p2, added, removed, readtree=None,
+            match=None):
         if p1 in self.fulltextcache and util.safehasattr(m, 'fastdelta'):
             # If our first parent is in the manifest cache, we can
             # compute a delta here using properties we know about the
             # manifest up-front, which may save time later for the
             # revlog layer.
 
             _checkforbidden(added)
             # combine the changed lists into one sorted iterator
             work = heapq.merge([(x, False) for x in added],
                                [(x, True) for x in removed])
 
             arraytext, deltatext = m.fastdelta(self.fulltextcache[p1], work)
             cachedelta = self._revlog.rev(p1), deltatext
             text = util.buffer(arraytext)
             n = self._revlog.addrevision(text, transaction, link, p1, p2,
                                          cachedelta)
         else:
             # The first parent manifest isn't already loaded, so we'll
             # just encode a fulltext of the manifest and pass that
             # through to the revlog layer, and let it handle the delta
             # process.
             if self._treeondisk:
                 assert readtree, "readtree must be set for treemanifest writes"
+                assert match, "match must be specified for treemanifest writes"
                 m1 = readtree(self.tree, p1)
                 m2 = readtree(self.tree, p2)
-                n = self._addtree(m, transaction, link, m1, m2, readtree)
+                n = self._addtree(m, transaction, link, m1, m2, readtree,
+                                  match=match)
                 arraytext = None
             else:
                 text = m.text()
                 n = self._revlog.addrevision(text, transaction, link, p1, p2)
                 arraytext = bytearray(text)
 
         if arraytext is not None:
             self.fulltextcache[n] = arraytext
 
         return n
 
-    def _addtree(self, m, transaction, link, m1, m2, readtree):
+    def _addtree(self, m, transaction, link, m1, m2, readtree, match):
         # If the manifest is unchanged compared to one parent,
         # don't write a new revision
         if self.tree != '' and (m.unmodifiedsince(m1) or m.unmodifiedsince(
             m2)):
             return m.node()
-        def writesubtree(subm, subp1, subp2):
+        def writesubtree(subm, subp1, subp2, match):
             sublog = self.dirlog(subm.dir())
             sublog.add(subm, transaction, link, subp1, subp2, None, None,
-                       readtree=readtree)
-        m.writesubtrees(m1, m2, writesubtree)
+                       readtree=readtree, match=match)
+        m.writesubtrees(m1, m2, writesubtree, match)
         text = m.dirtext()
         n = None
         if self.tree != '':
             # Double-check whether contents are unchanged to one parent
             if text == m1.dirtext():
                 n = m1.node()
             elif text == m2.dirtext():
                 n = m2.node()
 
         if not n:
             n = self._revlog.addrevision(text, transaction, link, m1.node(),
                                          m2.node())
 
         # Save nodeid so parent manifest can calculate its nodeid
         m.setnode(n)
         return n
 
     def __len__(self):
         return len(self._revlog)
 
     def __iter__(self):
         return self._revlog.__iter__()
 
     def rev(self, node):
         return self._revlog.rev(node)
 
     def node(self, rev):
         return self._revlog.node(rev)
 
     def lookup(self, value):
         return self._revlog.lookup(value)
 
     def parentrevs(self, rev):
         return self._revlog.parentrevs(rev)
 
     def parents(self, node):
         return self._revlog.parents(node)
 
     def linkrev(self, rev):
         return self._revlog.linkrev(rev)
 
     def checksize(self):
         return self._revlog.checksize()
 
     def revision(self, node, _df=None, raw=False):
         return self._revlog.revision(node, _df=_df, raw=raw)
 
     def revdiff(self, rev1, rev2):
         return self._revlog.revdiff(rev1, rev2)
 
     def cmp(self, node, text):
         return self._revlog.cmp(node, text)
 
     def deltaparent(self, rev):
         return self._revlog.deltaparent(rev)
 
     def emitrevisiondeltas(self, requests):
         return self._revlog.emitrevisiondeltas(requests)
 
     def addgroup(self, deltas, linkmapper, transaction, addrevisioncb=None):
         return self._revlog.addgroup(deltas, linkmapper, transaction,
                                      addrevisioncb=addrevisioncb)
 
     def getstrippoint(self, minlink):
         return self._revlog.getstrippoint(minlink)
 
     def strip(self, minlink, transaction):
         return self._revlog.strip(minlink, transaction)
 
     def files(self):
         return self._revlog.files()
 
     def clone(self, tr, destrevlog, **kwargs):
         if not isinstance(destrevlog, manifestrevlog):
             raise error.ProgrammingError('expected manifestrevlog to clone()')
 
         return self._revlog.clone(tr, destrevlog._revlog, **kwargs)
 
     @property
     def indexfile(self):
         return self._revlog.indexfile
 
     @indexfile.setter
     def indexfile(self, value):
         self._revlog.indexfile = value
 
     @property
     def opener(self):
         return self._revlog.opener
 
     @opener.setter
     def opener(self, value):
         self._revlog.opener = value
 
 @interfaceutil.implementer(repository.imanifestlog)
 class manifestlog(object):
     """A collection class representing the collection of manifest snapshots
     referenced by commits in the repository.
 
     In this situation, 'manifest' refers to the abstract concept of a snapshot
     of the list of files in the given commit. Consumers of the output of this
     class do not care about the implementation details of the actual manifests
     they receive (i.e. tree or flat or lazily loaded, etc)."""
     def __init__(self, opener, repo):
         usetreemanifest = False
         cachesize = 4
 
         opts = getattr(opener, 'options', None)
         if opts is not None:
             usetreemanifest = opts.get('treemanifest', usetreemanifest)
             cachesize = opts.get('manifestcachesize', cachesize)
 
         self._treemanifests = usetreemanifest
 
         self._rootstore = repo._constructmanifest()
         self._rootstore._setupmanifestcachehooks(repo)
         self._narrowmatch = repo.narrowmatch()
 
         # A cache of the manifestctx or treemanifestctx for each directory
         self._dirmancache = {}
         self._dirmancache[''] = util.lrucachedict(cachesize)
 
         self._cachesize = cachesize
 
     def __getitem__(self, node):
         """Retrieves the manifest instance for the given node. Throws a
         LookupError if not found.
         """
         return self.get('', node)
 
     def get(self, tree, node, verify=True):
         """Retrieves the manifest instance for the given node. Throws a
         LookupError if not found.
 
         `verify` - if True an exception will be thrown if the node is not in
                    the revlog
         """
         if node in self._dirmancache.get(tree, ()):
             return self._dirmancache[tree][node]
 
         if not self._narrowmatch.always():
             if not self._narrowmatch.visitdir(tree[:-1] or '.'):
                 return excludeddirmanifestctx(tree, node)
         if tree:
             if self._rootstore._treeondisk:
                 if verify:
                     # Side-effect is LookupError is raised if node doesn't
                     # exist.
                     self.getstorage(tree).rev(node)
 
                 m = treemanifestctx(self, tree, node)
             else:
                 raise error.Abort(
                         _("cannot ask for manifest directory '%s' in a flat "
                           "manifest") % tree)
         else:
             if verify:
                 # Side-effect is LookupError is raised if node doesn't exist.
                 self._rootstore.rev(node)
 
             if self._treemanifests:
                 m = treemanifestctx(self, '', node)
             else:
                 m = manifestctx(self, node)
 
         if node != nullid:
             mancache = self._dirmancache.get(tree)
             if not mancache:
                 mancache = util.lrucachedict(self._cachesize)
                 self._dirmancache[tree] = mancache
             mancache[node] = m
         return m
 
     def getstorage(self, tree):
         return self._rootstore.dirlog(tree)
 
     def clearcaches(self, clear_persisted_data=False):
         self._dirmancache.clear()
         self._rootstore.clearcaches(clear_persisted_data=clear_persisted_data)
 
     def rev(self, node):
         return self._rootstore.rev(node)
 
 @interfaceutil.implementer(repository.imanifestrevisionwritable)
 class memmanifestctx(object):
     def __init__(self, manifestlog):
         self._manifestlog = manifestlog
         self._manifestdict = manifestdict()
 
     def _storage(self):
         return self._manifestlog.getstorage(b'')
 
     def new(self):
         return memmanifestctx(self._manifestlog)
 
     def copy(self):
         memmf = memmanifestctx(self._manifestlog)
         memmf._manifestdict = self.read().copy()
         return memmf
 
     def read(self):
         return self._manifestdict
 
-    def write(self, transaction, link, p1, p2, added, removed):
+    def write(self, transaction, link, p1, p2, added, removed, match=None):
         return self._storage().add(self._manifestdict, transaction, link,
-                                   p1, p2, added, removed)
+                                   p1, p2, added, removed, match=match)
 
 @interfaceutil.implementer(repository.imanifestrevisionstored)
 class manifestctx(object):
     """A class representing a single revision of a manifest, including its
     contents, its parent revs, and its linkrev.
     """
     def __init__(self, manifestlog, node):
         self._manifestlog = manifestlog
         self._data = None
 
         self._node = node
 
         # TODO: We eventually want p1, p2, and linkrev exposed on this class,
         # but let's add it later when something needs it and we can load it
         # lazily.
         #self.p1, self.p2 = store.parents(node)
         #rev = store.rev(node)
         #self.linkrev = store.linkrev(rev)
 
     def _storage(self):
         return self._manifestlog.getstorage(b'')
 
     def node(self):
         return self._node
 
     def new(self):
         return memmanifestctx(self._manifestlog)
 
     def copy(self):
         memmf = memmanifestctx(self._manifestlog)
         memmf._manifestdict = self.read().copy()
         return memmf
 
     @propertycache
     def parents(self):
         return self._storage().parents(self._node)
 
     def read(self):
         if self._data is None:
             if self._node == nullid:
                 self._data = manifestdict()
             else:
                 store = self._storage()
                 if self._node in store.fulltextcache:
                     text = pycompat.bytestr(store.fulltextcache[self._node])
                 else:
                     text = store.revision(self._node)
                     arraytext = bytearray(text)
                     store.fulltextcache[self._node] = arraytext
                 self._data = manifestdict(text)
         return self._data
 
     def readfast(self, shallow=False):
         '''Calls either readdelta or read, based on which would be less work.
         readdelta is called if the delta is against the p1, and therefore can be
         read quickly.
 
         If `shallow` is True, nothing changes since this is a flat manifest.
         '''
         store = self._storage()
         r = store.rev(self._node)
         deltaparent = store.deltaparent(r)
         if deltaparent != nullrev and deltaparent in store.parentrevs(r):
             return self.readdelta()
         return self.read()
 
     def readdelta(self, shallow=False):
         '''Returns a manifest containing just the entries that are present
         in this manifest, but not in its p1 manifest. This is efficient to read
         if the revlog delta is already p1.
 
         Changing the value of `shallow` has no effect on flat manifests.
         '''
         store = self._storage()
         r = store.rev(self._node)
         d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r))
         return manifestdict(d)
 
     def find(self, key):
         return self.read().find(key)
 
 @interfaceutil.implementer(repository.imanifestrevisionwritable)
 class memtreemanifestctx(object):
     def __init__(self, manifestlog, dir=''):
         self._manifestlog = manifestlog
         self._dir = dir
         self._treemanifest = treemanifest()
 
     def _storage(self):
         return self._manifestlog.getstorage(b'')
 
     def new(self, dir=''):
         return memtreemanifestctx(self._manifestlog, dir=dir)
 
     def copy(self):
         memmf = memtreemanifestctx(self._manifestlog, dir=self._dir)
         memmf._treemanifest = self._treemanifest.copy()
         return memmf
 
     def read(self):
         return self._treemanifest
 
-    def write(self, transaction, link, p1, p2, added, removed):
+    def write(self, transaction, link, p1, p2, added, removed, match=None):
         def readtree(dir, node):
             return self._manifestlog.get(dir, node).read()
         return self._storage().add(self._treemanifest, transaction, link,
-                                   p1, p2, added, removed, readtree=readtree)
+                                   p1, p2, added, removed, readtree=readtree,
+                                   match=match)
 
 @interfaceutil.implementer(repository.imanifestrevisionstored)
 class treemanifestctx(object):
     def __init__(self, manifestlog, dir, node):
         self._manifestlog = manifestlog
         self._dir = dir
         self._data = None
 
         self._node = node
 
         # TODO: Load p1/p2/linkrev lazily. They need to be lazily loaded so that
         # we can instantiate treemanifestctx objects for directories we don't
         # have on disk.
         #self.p1, self.p2 = store.parents(node)
         #rev = store.rev(node)
         #self.linkrev = store.linkrev(rev)
 
     def _storage(self):
         narrowmatch = self._manifestlog._narrowmatch
         if not narrowmatch.always():
             if not narrowmatch.visitdir(self._dir[:-1] or '.'):
                 return excludedmanifestrevlog(self._dir)
         return self._manifestlog.getstorage(self._dir)
 
     def read(self):
         if self._data is None:
             store = self._storage()
             if self._node == nullid:
                 self._data = treemanifest()
             # TODO accessing non-public API
             elif store._treeondisk:
                 m = treemanifest(dir=self._dir)
                 def gettext():
                     return store.revision(self._node)
                 def readsubtree(dir, subm):
                     # Set verify to False since we need to be able to create
                     # subtrees for trees that don't exist on disk.
                     return self._manifestlog.get(dir, subm, verify=False).read()
                 m.read(gettext, readsubtree)
                 m.setnode(self._node)
                 self._data = m
             else:
                 if self._node in store.fulltextcache:
                     text = pycompat.bytestr(store.fulltextcache[self._node])
                 else:
                     text = store.revision(self._node)
                     arraytext = bytearray(text)
                     store.fulltextcache[self._node] = arraytext
                 self._data = treemanifest(dir=self._dir, text=text)
 
         return self._data
 
     def node(self):
         return self._node
 
     def new(self, dir=''):
         return memtreemanifestctx(self._manifestlog, dir=dir)
 
     def copy(self):
         memmf = memtreemanifestctx(self._manifestlog, dir=self._dir)
         memmf._treemanifest = self.read().copy()
         return memmf
 
     @propertycache
     def parents(self):
         return self._storage().parents(self._node)
 
     def readdelta(self, shallow=False):
         '''Returns a manifest containing just the entries that are present
         in this manifest, but not in its p1 manifest. This is efficient to read
         if the revlog delta is already p1.
 
         If `shallow` is True, this will read the delta for this directory,
         without recursively reading subdirectory manifests. Instead, any
         subdirectory entry will be reported as it appears in the manifest, i.e.
         the subdirectory will be reported among files and distinguished only by
         its 't' flag.
         '''
         store = self._storage()
         if shallow:
             r = store.rev(self._node)
             d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r))
             return manifestdict(d)
         else:
             # Need to perform a slow delta
             r0 = store.deltaparent(store.rev(self._node))
             m0 = self._manifestlog.get(self._dir, store.node(r0)).read()
             m1 = self.read()
             md = treemanifest(dir=self._dir)
             for f, ((n0, fl0), (n1, fl1)) in m0.diff(m1).iteritems():
                 if n1:
                     md[f] = n1
                     if fl1:
                         md.setflag(f, fl1)
             return md
 
     def readfast(self, shallow=False):
         '''Calls either readdelta or read, based on which would be less work.
         readdelta is called if the delta is against the p1, and therefore can be
         read quickly.
 
         If `shallow` is True, it only returns the entries from this manifest,
         and not any submanifests.
         '''
         store = self._storage()
         r = store.rev(self._node)
         deltaparent = store.deltaparent(r)
         if (deltaparent != nullrev and
             deltaparent in store.parentrevs(r)):
             return self.readdelta(shallow=shallow)
 
         if shallow:
             return manifestdict(store.revision(self._node))
         else:
             return self.read()
 
     def find(self, key):
         return self.read().find(key)
 
 class excludeddir(treemanifest):
     """Stand-in for a directory that is excluded from the repository.
 
     With narrowing active on a repository that uses treemanifests,
     some of the directory revlogs will be excluded from the resulting
     clone. This is a huge storage win for clients, but means we need
     some sort of pseudo-manifest to surface to internals so we can
     detect a merge conflict outside the narrowspec. That's what this
     class is: it stands in for a directory whose node is known, but
     whose contents are unknown.
     """
     def __init__(self, dir, node):
         super(excludeddir, self).__init__(dir)
         self._node = node
         # Add an empty file, which will be included by iterators and such,
         # appearing as the directory itself (i.e. something like "dir/")
         self._files[''] = node
         self._flags[''] = 't'
 
     # Manifests outside the narrowspec should never be modified, so avoid
     # copying. This makes a noticeable difference when there are very many
     # directories outside the narrowspec. Also, it makes sense for the copy to
     # be of the same type as the original, which would not happen with the
     # super type's copy().
     def copy(self):
         return self
 
 class excludeddirmanifestctx(treemanifestctx):
     """context wrapper for excludeddir - see that docstring for rationale"""
     def __init__(self, dir, node):
         self._dir = dir
         self._node = node
 
     def read(self):
         return excludeddir(self._dir, self._node)
 
     def write(self, *args):
         raise error.ProgrammingError(
             'attempt to write manifest from excluded dir %s' % self._dir)
 
 class excludedmanifestrevlog(manifestrevlog):
     """Stand-in for excluded treemanifest revlogs.
 
     When narrowing is active on a treemanifest repository, we'll have
     references to directories we can't see due to the revlog being
     skipped. This class exists to conform to the manifestrevlog
     interface for those directories and proactively prevent writes to
     outside the narrowspec.
     """
 
     def __init__(self, dir):
         self._dir = dir
 
     def __len__(self):
         raise error.ProgrammingError(
             'attempt to get length of excluded dir %s' % self._dir)
 
     def rev(self, node):
         raise error.ProgrammingError(
             'attempt to get rev from excluded dir %s' % self._dir)
 
     def linkrev(self, node):
         raise error.ProgrammingError(
             'attempt to get linkrev from excluded dir %s' % self._dir)
 
     def node(self, rev):
         raise error.ProgrammingError(
             'attempt to get node from excluded dir %s' % self._dir)
 
     def add(self, *args, **kwargs):
         # We should never write entries in dirlogs outside the narrow clone.
         # However, the method still gets called from writesubtree() in
         # _addtree(), so we need to handle it. We should possibly make that
         # avoid calling add() with a clean manifest (_dirty is always False
         # in excludeddir instances).
         pass

mercurial/repository.py

@@ -1,1565 +1,1580 b''
 # repository.py - Interfaces and base classes for repositories and peers.
 #
 # Copyright 2017 Gregory Szorc <[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
 
 from .i18n import _
 from . import (
     error,
 )
 from .utils import (
     interfaceutil,
 )
 
 # When narrowing is finalized and no longer subject to format changes,
 # we should move this to just "narrow" or similar.
 NARROW_REQUIREMENT = 'narrowhg-experimental'
 
 class ipeerconnection(interfaceutil.Interface):
     """Represents a "connection" to a repository.
 
     This is the base interface for representing a connection to a repository.
     It holds basic properties and methods applicable to all peer types.
 
     This is not a complete interface definition and should not be used
     outside of this module.
     """
     ui = interfaceutil.Attribute("""ui.ui instance""")
 
     def url():
         """Returns a URL string representing this peer.
 
         Currently, implementations expose the raw URL used to construct the
         instance. It may contain credentials as part of the URL. The
         expectations of the value aren't well-defined and this could lead to
         data leakage.
 
         TODO audit/clean consumers and more clearly define the contents of this
         value.
         """
 
     def local():
         """Returns a local repository instance.
 
         If the peer represents a local repository, returns an object that
         can be used to interface with it. Otherwise returns ``None``.
         """
 
     def peer():
         """Returns an object conforming to this interface.
 
         Most implementations will ``return self``.
         """
 
     def canpush():
         """Returns a boolean indicating if this peer can be pushed to."""
 
     def close():
         """Close the connection to this peer.
 
         This is called when the peer will no longer be used. Resources
         associated with the peer should be cleaned up.
         """
 
 class ipeercapabilities(interfaceutil.Interface):
     """Peer sub-interface related to capabilities."""
 
     def capable(name):
         """Determine support for a named capability.
 
         Returns ``False`` if capability not supported.
 
         Returns ``True`` if boolean capability is supported. Returns a string
         if capability support is non-boolean.
 
         Capability strings may or may not map to wire protocol capabilities.
         """
 
     def requirecap(name, purpose):
         """Require a capability to be present.
 
         Raises a ``CapabilityError`` if the capability isn't present.
         """
 
 class ipeercommands(interfaceutil.Interface):
     """Client-side interface for communicating over the wire protocol.
 
     This interface is used as a gateway to the Mercurial wire protocol.
     methods commonly call wire protocol commands of the same name.
     """
 
     def branchmap():
         """Obtain heads in named branches.
 
         Returns a dict mapping branch name to an iterable of nodes that are
         heads on that branch.
         """
 
     def capabilities():
         """Obtain capabilities of the peer.
 
         Returns a set of string capabilities.
         """
 
     def clonebundles():
         """Obtains the clone bundles manifest for the repo.
 
         Returns the manifest as unparsed bytes.
         """
 
     def debugwireargs(one, two, three=None, four=None, five=None):
         """Used to facilitate debugging of arguments passed over the wire."""
 
     def getbundle(source, **kwargs):
         """Obtain remote repository data as a bundle.
 
         This command is how the bulk of repository data is transferred from
         the peer to the local repository
 
         Returns a generator of bundle data.
         """
 
     def heads():
         """Determine all known head revisions in the peer.
 
         Returns an iterable of binary nodes.
         """
 
     def known(nodes):
         """Determine whether multiple nodes are known.
 
         Accepts an iterable of nodes whose presence to check for.
 
         Returns an iterable of booleans indicating of the corresponding node
         at that index is known to the peer.
         """
 
     def listkeys(namespace):
         """Obtain all keys in a pushkey namespace.
 
         Returns an iterable of key names.
         """
 
     def lookup(key):
         """Resolve a value to a known revision.
 
         Returns a binary node of the resolved revision on success.
         """
 
     def pushkey(namespace, key, old, new):
         """Set a value using the ``pushkey`` protocol.
 
         Arguments correspond to the pushkey namespace and key to operate on and
         the old and new values for that key.
 
         Returns a string with the peer result. The value inside varies by the
         namespace.
         """
 
     def stream_out():
         """Obtain streaming clone data.
 
         Successful result should be a generator of data chunks.
         """
 
     def unbundle(bundle, heads, url):
         """Transfer repository data to the peer.
 
         This is how the bulk of data during a push is transferred.
 
         Returns the integer number of heads added to the peer.
         """
 
 class ipeerlegacycommands(interfaceutil.Interface):
     """Interface for implementing support for legacy wire protocol commands.
 
     Wire protocol commands transition to legacy status when they are no longer
     used by modern clients. To facilitate identifying which commands are
     legacy, the interfaces are split.
     """
 
     def between(pairs):
         """Obtain nodes between pairs of nodes.
 
         ``pairs`` is an iterable of node pairs.
 
         Returns an iterable of iterables of nodes corresponding to each
         requested pair.
         """
 
     def branches(nodes):
         """Obtain ancestor changesets of specific nodes back to a branch point.
 
         For each requested node, the peer finds the first ancestor node that is
         a DAG root or is a merge.
 
         Returns an iterable of iterables with the resolved values for each node.
         """
 
     def changegroup(nodes, source):
         """Obtain a changegroup with data for descendants of specified nodes."""
 
     def changegroupsubset(bases, heads, source):
         pass
 
 class ipeercommandexecutor(interfaceutil.Interface):
     """Represents a mechanism to execute remote commands.
 
     This is the primary interface for requesting that wire protocol commands
     be executed. Instances of this interface are active in a context manager
     and have a well-defined lifetime. When the context manager exits, all
     outstanding requests are waited on.
     """
 
     def callcommand(name, args):
         """Request that a named command be executed.
 
         Receives the command name and a dictionary of command arguments.
 
         Returns a ``concurrent.futures.Future`` that will resolve to the
         result of that command request. That exact value is left up to
         the implementation and possibly varies by command.
 
         Not all commands can coexist with other commands in an executor
         instance: it depends on the underlying wire protocol transport being
         used and the command itself.
 
         Implementations MAY call ``sendcommands()`` automatically if the
         requested command can not coexist with other commands in this executor.
 
         Implementations MAY call ``sendcommands()`` automatically when the
         future's ``result()`` is called. So, consumers using multiple
         commands with an executor MUST ensure that ``result()`` is not called
         until all command requests have been issued.
         """
 
     def sendcommands():
         """Trigger submission of queued command requests.
 
         Not all transports submit commands as soon as they are requested to
         run. When called, this method forces queued command requests to be
         issued. It will no-op if all commands have already been sent.
 
         When called, no more new commands may be issued with this executor.
         """
 
     def close():
         """Signal that this command request is finished.
 
         When called, no more new commands may be issued. All outstanding
         commands that have previously been issued are waited on before
         returning. This not only includes waiting for the futures to resolve,
         but also waiting for all response data to arrive. In other words,
         calling this waits for all on-wire state for issued command requests
         to finish.
 
         When used as a context manager, this method is called when exiting the
         context manager.
 
         This method may call ``sendcommands()`` if there are buffered commands.
         """
 
 class ipeerrequests(interfaceutil.Interface):
     """Interface for executing commands on a peer."""
 
     def commandexecutor():
         """A context manager that resolves to an ipeercommandexecutor.
 
         The object this resolves to can be used to issue command requests
         to the peer.
 
         Callers should call its ``callcommand`` method to issue command
         requests.
 
         A new executor should be obtained for each distinct set of commands
         (possibly just a single command) that the consumer wants to execute
         as part of a single operation or round trip. This is because some
         peers are half-duplex and/or don't support persistent connections.
         e.g. in the case of HTTP peers, commands sent to an executor represent
         a single HTTP request. While some peers may support multiple command
         sends over the wire per executor, consumers need to code to the least
         capable peer. So it should be assumed that command executors buffer
         called commands until they are told to send them and that each
         command executor could result in a new connection or wire-level request
         being issued.
         """
 
 class ipeerbase(ipeerconnection, ipeercapabilities, ipeerrequests):
     """Unified interface for peer repositories.
 
     All peer instances must conform to this interface.
     """
 
 @interfaceutil.implementer(ipeerbase)
 class peer(object):
     """Base class for peer repositories."""
 
     def capable(self, name):
         caps = self.capabilities()
         if name in caps:
             return True
 
         name = '%s=' % name
         for cap in caps:
             if cap.startswith(name):
                 return cap[len(name):]
 
         return False
 
     def requirecap(self, name, purpose):
         if self.capable(name):
             return
 
         raise error.CapabilityError(
             _('cannot %s; remote repository does not support the %r '
               'capability') % (purpose, name))
 
 class irevisiondelta(interfaceutil.Interface):
     """Represents a delta between one revision and another.
 
     Instances convey enough information to allow a revision to be exchanged
     with another repository.
 
     Instances represent the fulltext revision data or a delta against
     another revision. Therefore the ``revision`` and ``delta`` attributes
     are mutually exclusive.
 
     Typically used for changegroup generation.
     """
 
     node = interfaceutil.Attribute(
         """20 byte node of this revision.""")
 
     p1node = interfaceutil.Attribute(
         """20 byte node of 1st parent of this revision.""")
 
     p2node = interfaceutil.Attribute(
         """20 byte node of 2nd parent of this revision.""")
 
     linknode = interfaceutil.Attribute(
         """20 byte node of the changelog revision this node is linked to.""")
 
     flags = interfaceutil.Attribute(
         """2 bytes of integer flags that apply to this revision.""")
 
     basenode = interfaceutil.Attribute(
         """20 byte node of the revision this data is a delta against.
 
         ``nullid`` indicates that the revision is a full revision and not
         a delta.
         """)
 
     baserevisionsize = interfaceutil.Attribute(
         """Size of base revision this delta is against.
 
         May be ``None`` if ``basenode`` is ``nullid``.
         """)
 
     revision = interfaceutil.Attribute(
         """Raw fulltext of revision data for this node.""")
 
     delta = interfaceutil.Attribute(
         """Delta between ``basenode`` and ``node``.
 
         Stored in the bdiff delta format.
         """)
 
 class irevisiondeltarequest(interfaceutil.Interface):
     """Represents a request to generate an ``irevisiondelta``."""
 
     node = interfaceutil.Attribute(
         """20 byte node of revision being requested.""")
 
     p1node = interfaceutil.Attribute(
         """20 byte node of 1st parent of revision.""")
 
     p2node = interfaceutil.Attribute(
         """20 byte node of 2nd parent of revision.""")
 
     linknode = interfaceutil.Attribute(
         """20 byte node to store in ``linknode`` attribute.""")
 
     basenode = interfaceutil.Attribute(
         """Base revision that delta should be generated against.
 
         If ``nullid``, the derived ``irevisiondelta`` should have its
         ``revision`` field populated and no delta should be generated.
 
         If ``None``, the delta may be generated against any revision that
         is an ancestor of this revision. Or a full revision may be used.
 
         If any other value, the delta should be produced against that
         revision.
         """)
 
     ellipsis = interfaceutil.Attribute(
         """Boolean on whether the ellipsis flag should be set.""")
 
 class ifilerevisionssequence(interfaceutil.Interface):
     """Contains index data for all revisions of a file.
 
     Types implementing this behave like lists of tuples. The index
     in the list corresponds to the revision number. The values contain
     index metadata.
 
     The *null* revision (revision number -1) is always the last item
     in the index.
     """
 
     def __len__():
         """The total number of revisions."""
 
     def __getitem__(rev):
         """Returns the object having a specific revision number.
 
         Returns an 8-tuple with the following fields:
 
         offset+flags
            Contains the offset and flags for the revision. 64-bit unsigned
            integer where first 6 bytes are the offset and the next 2 bytes
            are flags. The offset can be 0 if it is not used by the store.
         compressed size
             Size of the revision data in the store. It can be 0 if it isn't
             needed by the store.
         uncompressed size
             Fulltext size. It can be 0 if it isn't needed by the store.
         base revision
             Revision number of revision the delta for storage is encoded
             against. -1 indicates not encoded against a base revision.
         link revision
             Revision number of changelog revision this entry is related to.
         p1 revision
             Revision number of 1st parent. -1 if no 1st parent.
         p2 revision
             Revision number of 2nd parent. -1 if no 1st parent.
         node
             Binary node value for this revision number.
 
         Negative values should index off the end of the sequence. ``-1``
         should return the null revision. ``-2`` should return the most
         recent revision.
         """
 
     def __contains__(rev):
         """Whether a revision number exists."""
 
     def insert(self, i, entry):
         """Add an item to the index at specific revision."""
 
 class ifileindex(interfaceutil.Interface):
     """Storage interface for index data of a single file.
 
     File storage data is divided into index metadata and data storage.
     This interface defines the index portion of the interface.
 
     The index logically consists of:
 
     * A mapping between revision numbers and nodes.
     * DAG data (storing and querying the relationship between nodes).
     * Metadata to facilitate storage.
     """
     index = interfaceutil.Attribute(
         """An ``ifilerevisionssequence`` instance.""")
 
     def __len__():
         """Obtain the number of revisions stored for this file."""
 
     def __iter__():
         """Iterate over revision numbers for this file."""
 
     def revs(start=0, stop=None):
         """Iterate over revision numbers for this file, with control."""
 
     def parents(node):
         """Returns a 2-tuple of parent nodes for a revision.
 
         Values will be ``nullid`` if the parent is empty.
         """
 
     def parentrevs(rev):
         """Like parents() but operates on revision numbers."""
 
     def rev(node):
         """Obtain the revision number given a node.
 
         Raises ``error.LookupError`` if the node is not known.
         """
 
     def node(rev):
         """Obtain the node value given a revision number.
 
         Raises ``IndexError`` if the node is not known.
         """
 
     def lookup(node):
         """Attempt to resolve a value to a node.
 
         Value can be a binary node, hex node, revision number, or a string
         that can be converted to an integer.
 
         Raises ``error.LookupError`` if a node could not be resolved.
         """
 
     def linkrev(rev):
         """Obtain the changeset revision number a revision is linked to."""
 
     def flags(rev):
         """Obtain flags used to affect storage of a revision."""
 
     def iscensored(rev):
         """Return whether a revision's content has been censored."""
 
     def commonancestorsheads(node1, node2):
         """Obtain an iterable of nodes containing heads of common ancestors.
 
         See ``ancestor.commonancestorsheads()``.
         """
 
     def descendants(revs):
         """Obtain descendant revision numbers for a set of revision numbers.
 
         If ``nullrev`` is in the set, this is equivalent to ``revs()``.
         """
 
     def headrevs():
         """Obtain a list of revision numbers that are DAG heads.
 
         The list is sorted oldest to newest.
 
         TODO determine if sorting is required.
         """
 
     def heads(start=None, stop=None):
         """Obtain a list of nodes that are DAG heads, with control.
 
         The set of revisions examined can be limited by specifying
         ``start`` and ``stop``. ``start`` is a node. ``stop`` is an
         iterable of nodes. DAG traversal starts at earlier revision
         ``start`` and iterates forward until any node in ``stop`` is
         encountered.
         """
 
     def children(node):
         """Obtain nodes that are children of a node.
 
         Returns a list of nodes.
         """
 
     def deltaparent(rev):
         """"Return the revision that is a suitable parent to delta against."""
 
 class ifiledata(interfaceutil.Interface):
     """Storage interface for data storage of a specific file.
 
     This complements ``ifileindex`` and provides an interface for accessing
     data for a tracked file.
     """
     def rawsize(rev):
         """The size of the fulltext data for a revision as stored."""
 
     def size(rev):
         """Obtain the fulltext size of file data.
 
         Any metadata is excluded from size measurements. Use ``rawsize()`` if
         metadata size is important.
         """
 
     def checkhash(fulltext, node, p1=None, p2=None, rev=None):
         """Validate the stored hash of a given fulltext and node.
 
         Raises ``error.RevlogError`` is hash validation fails.
         """
 
     def revision(node, raw=False):
         """"Obtain fulltext data for a node.
 
         By default, any storage transformations are applied before the data
         is returned. If ``raw`` is True, non-raw storage transformations
         are not applied.
 
         The fulltext data may contain a header containing metadata. Most
         consumers should use ``read()`` to obtain the actual file data.
         """
 
     def read(node):
         """Resolve file fulltext data.
 
         This is similar to ``revision()`` except any metadata in the data
         headers is stripped.
         """
 
     def renamed(node):
         """Obtain copy metadata for a node.
 
         Returns ``False`` if no copy metadata is stored or a 2-tuple of
         (path, node) from which this revision was copied.
         """
 
     def cmp(node, fulltext):
         """Compare fulltext to another revision.
 
         Returns True if the fulltext is different from what is stored.
 
         This takes copy metadata into account.
 
         TODO better document the copy metadata and censoring logic.
         """
 
     def revdiff(rev1, rev2):
         """Obtain a delta between two revision numbers.
 
         Operates on raw data in the store (``revision(node, raw=True)``).
 
         The returned data is the result of ``bdiff.bdiff`` on the raw
         revision data.
         """
 
     def emitrevisiondeltas(requests):
         """Produce ``irevisiondelta`` from ``irevisiondeltarequest``s.
 
         Given an iterable of objects conforming to the ``irevisiondeltarequest``
         interface, emits objects conforming to the ``irevisiondelta``
         interface.
 
         This method is a generator.
 
         ``irevisiondelta`` should be emitted in the same order of
         ``irevisiondeltarequest`` that was passed in.
 
         The emitted objects MUST conform by the results of
         ``irevisiondeltarequest``. Namely, they must respect any requests
         for building a delta from a specific ``basenode`` if defined.
 
         When sending deltas, implementations must take into account whether
         the client has the base delta before encoding a delta against that
         revision. A revision encountered previously in ``requests`` is
         always a suitable base revision. An example of a bad delta is a delta
         against a non-ancestor revision. Another example of a bad delta is a
         delta against a censored revision.
         """
 
 class ifilemutation(interfaceutil.Interface):
     """Storage interface for mutation events of a tracked file."""
 
     def add(filedata, meta, transaction, linkrev, p1, p2):
         """Add a new revision to the store.
 
         Takes file data, dictionary of metadata, a transaction, linkrev,
         and parent nodes.
 
         Returns the node that was added.
 
         May no-op if a revision matching the supplied data is already stored.
         """
 
     def addrevision(revisiondata, transaction, linkrev, p1, p2, node=None,
                     flags=0, cachedelta=None):
         """Add a new revision to the store.
 
         This is similar to ``add()`` except it operates at a lower level.
 
         The data passed in already contains a metadata header, if any.
 
         ``node`` and ``flags`` can be used to define the expected node and
         the flags to use with storage.
 
         ``add()`` is usually called when adding files from e.g. the working
         directory. ``addrevision()`` is often called by ``add()`` and for
         scenarios where revision data has already been computed, such as when
         applying raw data from a peer repo.
         """
 
     def addgroup(deltas, linkmapper, transaction, addrevisioncb=None):
         """Process a series of deltas for storage.
 
         ``deltas`` is an iterable of 7-tuples of
         (node, p1, p2, linknode, deltabase, delta, flags) defining revisions
         to add.
 
         The ``delta`` field contains ``mpatch`` data to apply to a base
         revision, identified by ``deltabase``. The base node can be
         ``nullid``, in which case the header from the delta can be ignored
         and the delta used as the fulltext.
 
         ``addrevisioncb`` should be called for each node as it is committed.
 
         Returns a list of nodes that were processed. A node will be in the list
         even if it existed in the store previously.
         """
 
     def getstrippoint(minlink):
         """Find the minimum revision that must be stripped to strip a linkrev.
 
         Returns a 2-tuple containing the minimum revision number and a set
         of all revisions numbers that would be broken by this strip.
 
         TODO this is highly revlog centric and should be abstracted into
         a higher-level deletion API. ``repair.strip()`` relies on this.
         """
 
     def strip(minlink, transaction):
         """Remove storage of items starting at a linkrev.
 
         This uses ``getstrippoint()`` to determine the first node to remove.
         Then it effectively truncates storage for all revisions after that.
 
         TODO this is highly revlog centric and should be abstracted into a
         higher-level deletion API.
         """
 
 class ifilestorage(ifileindex, ifiledata, ifilemutation):
     """Complete storage interface for a single tracked file."""
 
     version = interfaceutil.Attribute(
         """Version number of storage.
 
         TODO this feels revlog centric and could likely be removed.
         """)
 
     _generaldelta = interfaceutil.Attribute(
         """Whether deltas can be against any parent revision.
 
         TODO this is used by changegroup code and it could probably be
         folded into another API.
         """)
 
     def files():
         """Obtain paths that are backing storage for this file.
 
         TODO this is used heavily by verify code and there should probably
         be a better API for that.
         """
 
     def checksize():
         """Obtain the expected sizes of backing files.
 
         TODO this is used by verify and it should not be part of the interface.
         """
 
 class idirs(interfaceutil.Interface):
     """Interface representing a collection of directories from paths.
 
     This interface is essentially a derived data structure representing
     directories from a collection of paths.
     """
 
     def addpath(path):
         """Add a path to the collection.
 
         All directories in the path will be added to the collection.
         """
 
     def delpath(path):
         """Remove a path from the collection.
 
         If the removal was the last path in a particular directory, the
         directory is removed from the collection.
         """
 
     def __iter__():
         """Iterate over the directories in this collection of paths."""
 
     def __contains__(path):
         """Whether a specific directory is in this collection."""
 
 class imanifestdict(interfaceutil.Interface):
     """Interface representing a manifest data structure.
 
     A manifest is effectively a dict mapping paths to entries. Each entry
     consists of a binary node and extra flags affecting that entry.
     """
 
     def __getitem__(path):
         """Returns the binary node value for a path in the manifest.
 
         Raises ``KeyError`` if the path does not exist in the manifest.
 
         Equivalent to ``self.find(path)[0]``.
         """
 
     def find(path):
         """Returns the entry for a path in the manifest.
 
         Returns a 2-tuple of (node, flags).
 
         Raises ``KeyError`` if the path does not exist in the manifest.
         """
 
     def __len__():
         """Return the number of entries in the manifest."""
 
     def __nonzero__():
         """Returns True if the manifest has entries, False otherwise."""
 
     __bool__ = __nonzero__
 
     def __setitem__(path, node):
         """Define the node value for a path in the manifest.
 
         If the path is already in the manifest, its flags will be copied to
         the new entry.
         """
 
     def __contains__(path):
         """Whether a path exists in the manifest."""
 
     def __delitem__(path):
         """Remove a path from the manifest.
 
         Raises ``KeyError`` if the path is not in the manifest.
         """
 
     def __iter__():
         """Iterate over paths in the manifest."""
 
     def iterkeys():
         """Iterate over paths in the manifest."""
 
     def keys():
         """Obtain a list of paths in the manifest."""
 
     def filesnotin(other, match=None):
         """Obtain the set of paths in this manifest but not in another.
 
         ``match`` is an optional matcher function to be applied to both
         manifests.
 
         Returns a set of paths.
         """
 
     def dirs():
         """Returns an object implementing the ``idirs`` interface."""
 
     def hasdir(dir):
         """Returns a bool indicating if a directory is in this manifest."""
 
     def matches(match):
         """Generate a new manifest filtered through a matcher.
 
         Returns an object conforming to the ``imanifestdict`` interface.
         """
 
     def walk(match):
         """Generator of paths in manifest satisfying a matcher.
 
         This is equivalent to ``self.matches(match).iterkeys()`` except a new
         manifest object is not created.
 
         If the matcher has explicit files listed and they don't exist in
         the manifest, ``match.bad()`` is called for each missing file.
         """
 
     def diff(other, match=None, clean=False):
         """Find differences between this manifest and another.
 
         This manifest is compared to ``other``.
 
         If ``match`` is provided, the two manifests are filtered against this
         matcher and only entries satisfying the matcher are compared.
 
         If ``clean`` is True, unchanged files are included in the returned
         object.
 
         Returns a dict with paths as keys and values of 2-tuples of 2-tuples of
         the form ``((node1, flag1), (node2, flag2))`` where ``(node1, flag1)``
         represents the node and flags for this manifest and ``(node2, flag2)``
         are the same for the other manifest.
         """
 
     def setflag(path, flag):
         """Set the flag value for a given path.
 
         Raises ``KeyError`` if the path is not already in the manifest.
         """
 
     def get(path, default=None):
         """Obtain the node value for a path or a default value if missing."""
 
     def flags(path, default=''):
         """Return the flags value for a path or a default value if missing."""
 
     def copy():
         """Return a copy of this manifest."""
 
     def items():
         """Returns an iterable of (path, node) for items in this manifest."""
 
     def iteritems():
         """Identical to items()."""
 
     def iterentries():
         """Returns an iterable of (path, node, flags) for this manifest.
 
         Similar to ``iteritems()`` except items are a 3-tuple and include
         flags.
         """
 
     def text():
         """Obtain the raw data representation for this manifest.
 
         Result is used to create a manifest revision.
         """
 
     def fastdelta(base, changes):
         """Obtain a delta between this manifest and another given changes.
 
         ``base`` in the raw data representation for another manifest.
 
         ``changes`` is an iterable of ``(path, to_delete)``.
 
         Returns a 2-tuple containing ``bytearray(self.text())`` and the
         delta between ``base`` and this manifest.
         """
 
 class imanifestrevisionbase(interfaceutil.Interface):
     """Base interface representing a single revision of a manifest.
 
     Should not be used as a primary interface: should always be inherited
     as part of a larger interface.
     """
 
     def new():
         """Obtain a new manifest instance.
 
         Returns an object conforming to the ``imanifestrevisionwritable``
         interface. The instance will be associated with the same
         ``imanifestlog`` collection as this instance.
         """
 
     def copy():
         """Obtain a copy of this manifest instance.
 
         Returns an object conforming to the ``imanifestrevisionwritable``
         interface. The instance will be associated with the same
         ``imanifestlog`` collection as this instance.
         """
 
     def read():
         """Obtain the parsed manifest data structure.
 
         The returned object conforms to the ``imanifestdict`` interface.
         """
 
 class imanifestrevisionstored(imanifestrevisionbase):
     """Interface representing a manifest revision committed to storage."""
 
     def node():
         """The binary node for this manifest."""
 
     parents = interfaceutil.Attribute(
         """List of binary nodes that are parents for this manifest revision."""
     )
 
     def readdelta(shallow=False):
         """Obtain the manifest data structure representing changes from parent.
 
         This manifest is compared to its 1st parent. A new manifest representing
         those differences is constructed.
 
         The returned object conforms to the ``imanifestdict`` interface.
         """
 
     def readfast(shallow=False):
         """Calls either ``read()`` or ``readdelta()``.
 
         The faster of the two options is called.
         """
 
     def find(key):
         """Calls self.read().find(key)``.
 
         Returns a 2-tuple of ``(node, flags)`` or raises ``KeyError``.
         """
 
 class imanifestrevisionwritable(imanifestrevisionbase):
     """Interface representing a manifest revision that can be committed."""
 
-    def write(transaction, linkrev, p1node, p2node, added, removed):
+    def write(transaction, linkrev, p1node, p2node, added, removed, match=None):
         """Add this revision to storage.
 
         Takes a transaction object, the changeset revision number it will
         be associated with, its parent nodes, and lists of added and
         removed paths.
 
+        If match is provided, storage can choose not to inspect or write out
+        items that do not match. Storage is still required to be able to provide
+        the full manifest in the future for any directories written (these
+        manifests should not be "narrowed on disk").
+
         Returns the binary node of the created revision.
         """
 
 class imanifeststorage(interfaceutil.Interface):
     """Storage interface for manifest data."""
 
     tree = interfaceutil.Attribute(
         """The path to the directory this manifest tracks.
 
         The empty bytestring represents the root manifest.
         """)
 
     index = interfaceutil.Attribute(
         """An ``ifilerevisionssequence`` instance.""")
 
     indexfile = interfaceutil.Attribute(
         """Path of revlog index file.
 
         TODO this is revlog specific and should not be exposed.
         """)
 
     opener = interfaceutil.Attribute(
         """VFS opener to use to access underlying files used for storage.
 
         TODO this is revlog specific and should not be exposed.
         """)
 
     version = interfaceutil.Attribute(
         """Revlog version number.
 
         TODO this is revlog specific and should not be exposed.
         """)
 
     _generaldelta = interfaceutil.Attribute(
         """Whether generaldelta storage is being used.
 
         TODO this is revlog specific and should not be exposed.
         """)
 
     fulltextcache = interfaceutil.Attribute(
         """Dict with cache of fulltexts.
 
         TODO this doesn't feel appropriate for the storage interface.
         """)
 
     def __len__():
         """Obtain the number of revisions stored for this manifest."""
 
     def __iter__():
         """Iterate over revision numbers for this manifest."""
 
     def rev(node):
         """Obtain the revision number given a binary node.
 
         Raises ``error.LookupError`` if the node is not known.
         """
 
     def node(rev):
         """Obtain the node value given a revision number.
 
         Raises ``error.LookupError`` if the revision is not known.
         """
 
     def lookup(value):
         """Attempt to resolve a value to a node.
 
         Value can be a binary node, hex node, revision number, or a bytes
         that can be converted to an integer.
 
         Raises ``error.LookupError`` if a ndoe could not be resolved.
 
         TODO this is only used by debug* commands and can probably be deleted
         easily.
         """
 
     def parents(node):
         """Returns a 2-tuple of parent nodes for a node.
 
         Values will be ``nullid`` if the parent is empty.
         """
 
     def parentrevs(rev):
         """Like parents() but operates on revision numbers."""
 
     def linkrev(rev):
         """Obtain the changeset revision number a revision is linked to."""
 
     def revision(node, _df=None, raw=False):
         """Obtain fulltext data for a node."""
 
     def revdiff(rev1, rev2):
         """Obtain a delta between two revision numbers.
 
         The returned data is the result of ``bdiff.bdiff()`` on the raw
         revision data.
         """
 
     def cmp(node, fulltext):
         """Compare fulltext to another revision.
 
         Returns True if the fulltext is different from what is stored.
         """
 
     def emitrevisiondeltas(requests):
         """Produce ``irevisiondelta`` from ``irevisiondeltarequest``s.
 
         See the documentation for ``ifiledata`` for more.
         """
 
     def addgroup(deltas, linkmapper, transaction, addrevisioncb=None):
         """Process a series of deltas for storage.
 
         See the documentation in ``ifilemutation`` for more.
         """
 
     def getstrippoint(minlink):
         """Find minimum revision that must be stripped to strip a linkrev.
 
         See the documentation in ``ifilemutation`` for more.
         """
 
     def strip(minlink, transaction):
         """Remove storage of items starting at a linkrev.
 
         See the documentation in ``ifilemutation`` for more.
         """
 
     def checksize():
         """Obtain the expected sizes of backing files.
 
         TODO this is used by verify and it should not be part of the interface.
         """
 
     def files():
         """Obtain paths that are backing storage for this manifest.
 
         TODO this is used by verify and there should probably be a better API
         for this functionality.
         """
 
     def deltaparent(rev):
         """Obtain the revision that a revision is delta'd against.
 
         TODO delta encoding is an implementation detail of storage and should
         not be exposed to the storage interface.
         """
 
     def clone(tr, dest, **kwargs):
         """Clone this instance to another."""
 
     def clearcaches(clear_persisted_data=False):
         """Clear any caches associated with this instance."""
 
     def dirlog(d):
         """Obtain a manifest storage instance for a tree."""
 
-    def add(m, transaction, link, p1, p2, added, removed, readtree=None):
+    def add(m, transaction, link, p1, p2, added, removed, readtree=None,
+            match=None):
         """Add a revision to storage.
 
         ``m`` is an object conforming to ``imanifestdict``.
 
         ``link`` is the linkrev revision number.
 
         ``p1`` and ``p2`` are the parent revision numbers.
 
         ``added`` and ``removed`` are iterables of added and removed paths,
         respectively.
+
+        ``readtree`` is a function that can be used to read the child tree(s)
+        when recursively writing the full tree structure when using
+        treemanifets.
+
+        ``match`` is a matcher that can be used to hint to storage that not all
+        paths must be inspected; this is an optimization and can be safely
+        ignored. Note that the storage must still be able to reproduce a full
+        manifest including files that did not match.
         """
 
 class imanifestlog(interfaceutil.Interface):
     """Interface representing a collection of manifest snapshots.
 
     Represents the root manifest in a repository.
 
     Also serves as a means to access nested tree manifests and to cache
     tree manifests.
     """
 
     def __getitem__(node):
         """Obtain a manifest instance for a given binary node.
 
         Equivalent to calling ``self.get('', node)``.
 
         The returned object conforms to the ``imanifestrevisionstored``
         interface.
         """
 
     def get(tree, node, verify=True):
         """Retrieve the manifest instance for a given directory and binary node.
 
         ``node`` always refers to the node of the root manifest (which will be
         the only manifest if flat manifests are being used).
 
         If ``tree`` is the empty string, the root manifest is returned.
         Otherwise the manifest for the specified directory will be returned
         (requires tree manifests).
 
         If ``verify`` is True, ``LookupError`` is raised if the node is not
         known.
 
         The returned object conforms to the ``imanifestrevisionstored``
         interface.
         """
 
     def getstorage(tree):
         """Retrieve an interface to storage for a particular tree.
 
         If ``tree`` is the empty bytestring, storage for the root manifest will
         be returned. Otherwise storage for a tree manifest is returned.
 
         TODO formalize interface for returned object.
         """
 
     def clearcaches():
         """Clear caches associated with this collection."""
 
     def rev(node):
         """Obtain the revision number for a binary node.
 
         Raises ``error.LookupError`` if the node is not known.
         """
 
 class completelocalrepository(interfaceutil.Interface):
     """Monolithic interface for local repositories.
 
     This currently captures the reality of things - not how things should be.
     """
 
     supportedformats = interfaceutil.Attribute(
         """Set of requirements that apply to stream clone.
 
         This is actually a class attribute and is shared among all instances.
         """)
 
     openerreqs = interfaceutil.Attribute(
         """Set of requirements that are passed to the opener.
 
         This is actually a class attribute and is shared among all instances.
         """)
 
     supported = interfaceutil.Attribute(
         """Set of requirements that this repo is capable of opening.""")
 
     requirements = interfaceutil.Attribute(
         """Set of requirements this repo uses.""")
 
     filtername = interfaceutil.Attribute(
         """Name of the repoview that is active on this repo.""")
 
     wvfs = interfaceutil.Attribute(
         """VFS used to access the working directory.""")
 
     vfs = interfaceutil.Attribute(
         """VFS rooted at the .hg directory.
 
         Used to access repository data not in the store.
         """)
 
     svfs = interfaceutil.Attribute(
         """VFS rooted at the store.
 
         Used to access repository data in the store. Typically .hg/store.
         But can point elsewhere if the store is shared.
         """)
 
     root = interfaceutil.Attribute(
         """Path to the root of the working directory.""")
 
     path = interfaceutil.Attribute(
         """Path to the .hg directory.""")
 
     origroot = interfaceutil.Attribute(
         """The filesystem path that was used to construct the repo.""")
 
     auditor = interfaceutil.Attribute(
         """A pathauditor for the working directory.
 
         This checks if a path refers to a nested repository.
 
         Operates on the filesystem.
         """)
 
     nofsauditor = interfaceutil.Attribute(
         """A pathauditor for the working directory.
 
         This is like ``auditor`` except it doesn't do filesystem checks.
         """)
 
     baseui = interfaceutil.Attribute(
         """Original ui instance passed into constructor.""")
 
     ui = interfaceutil.Attribute(
         """Main ui instance for this instance.""")
 
     sharedpath = interfaceutil.Attribute(
         """Path to the .hg directory of the repo this repo was shared from.""")
 
     store = interfaceutil.Attribute(
         """A store instance.""")
 
     spath = interfaceutil.Attribute(
         """Path to the store.""")
 
     sjoin = interfaceutil.Attribute(
         """Alias to self.store.join.""")
 
     cachevfs = interfaceutil.Attribute(
         """A VFS used to access the cache directory.
 
         Typically .hg/cache.
         """)
 
     filteredrevcache = interfaceutil.Attribute(
         """Holds sets of revisions to be filtered.""")
 
     names = interfaceutil.Attribute(
         """A ``namespaces`` instance.""")
 
     def close():
         """Close the handle on this repository."""
 
     def peer():
         """Obtain an object conforming to the ``peer`` interface."""
 
     def unfiltered():
         """Obtain an unfiltered/raw view of this repo."""
 
     def filtered(name, visibilityexceptions=None):
         """Obtain a named view of this repository."""
 
     obsstore = interfaceutil.Attribute(
         """A store of obsolescence data.""")
 
     changelog = interfaceutil.Attribute(
         """A handle on the changelog revlog.""")
 
     manifestlog = interfaceutil.Attribute(
         """An instance conforming to the ``imanifestlog`` interface.
 
         Provides access to manifests for the repository.
         """)
 
     dirstate = interfaceutil.Attribute(
         """Working directory state.""")
 
     narrowpats = interfaceutil.Attribute(
         """Matcher patterns for this repository's narrowspec.""")
 
     def narrowmatch():
         """Obtain a matcher for the narrowspec."""
 
     def setnarrowpats(newincludes, newexcludes):
         """Define the narrowspec for this repository."""
 
     def __getitem__(changeid):
         """Try to resolve a changectx."""
 
     def __contains__(changeid):
         """Whether a changeset exists."""
 
     def __nonzero__():
         """Always returns True."""
         return True
 
     __bool__ = __nonzero__
 
     def __len__():
         """Returns the number of changesets in the repo."""
 
     def __iter__():
         """Iterate over revisions in the changelog."""
 
     def revs(expr, *args):
         """Evaluate a revset.
 
         Emits revisions.
         """
 
     def set(expr, *args):
         """Evaluate a revset.
 
         Emits changectx instances.
         """
 
     def anyrevs(specs, user=False, localalias=None):
         """Find revisions matching one of the given revsets."""
 
     def url():
         """Returns a string representing the location of this repo."""
 
     def hook(name, throw=False, **args):
         """Call a hook."""
 
     def tags():
         """Return a mapping of tag to node."""
 
     def tagtype(tagname):
         """Return the type of a given tag."""
 
     def tagslist():
         """Return a list of tags ordered by revision."""
 
     def nodetags(node):
         """Return the tags associated with a node."""
 
     def nodebookmarks(node):
         """Return the list of bookmarks pointing to the specified node."""
 
     def branchmap():
         """Return a mapping of branch to heads in that branch."""
 
     def revbranchcache():
         pass
 
     def branchtip(branchtip, ignoremissing=False):
         """Return the tip node for a given branch."""
 
     def lookup(key):
         """Resolve the node for a revision."""
 
     def lookupbranch(key):
         """Look up the branch name of the given revision or branch name."""
 
     def known(nodes):
         """Determine whether a series of nodes is known.
 
         Returns a list of bools.
         """
 
     def local():
         """Whether the repository is local."""
         return True
 
     def publishing():
         """Whether the repository is a publishing repository."""
 
     def cancopy():
         pass
 
     def shared():
         """The type of shared repository or None."""
 
     def wjoin(f, *insidef):
         """Calls self.vfs.reljoin(self.root, f, *insidef)"""
 
     def file(f):
         """Obtain a filelog for a tracked path.
 
         The returned type conforms to the ``ifilestorage`` interface.
         """
 
     def setparents(p1, p2):
         """Set the parent nodes of the working directory."""
 
     def filectx(path, changeid=None, fileid=None):
         """Obtain a filectx for the given file revision."""
 
     def getcwd():
         """Obtain the current working directory from the dirstate."""
 
     def pathto(f, cwd=None):
         """Obtain the relative path to a file."""
 
     def adddatafilter(name, fltr):
         pass
 
     def wread(filename):
         """Read a file from wvfs, using data filters."""
 
     def wwrite(filename, data, flags, backgroundclose=False, **kwargs):
         """Write data to a file in the wvfs, using data filters."""
 
     def wwritedata(filename, data):
         """Resolve data for writing to the wvfs, using data filters."""
 
     def currenttransaction():
         """Obtain the current transaction instance or None."""
 
     def transaction(desc, report=None):
         """Open a new transaction to write to the repository."""
 
     def undofiles():
         """Returns a list of (vfs, path) for files to undo transactions."""
 
     def recover():
         """Roll back an interrupted transaction."""
 
     def rollback(dryrun=False, force=False):
         """Undo the last transaction.
 
         DANGEROUS.
         """
 
     def updatecaches(tr=None, full=False):
         """Warm repo caches."""
 
     def invalidatecaches():
         """Invalidate cached data due to the repository mutating."""
 
     def invalidatevolatilesets():
         pass
 
     def invalidatedirstate():
         """Invalidate the dirstate."""
 
     def invalidate(clearfilecache=False):
         pass
 
     def invalidateall():
         pass
 
     def lock(wait=True):
         """Lock the repository store and return a lock instance."""
 
     def wlock(wait=True):
         """Lock the non-store parts of the repository."""
 
     def currentwlock():
         """Return the wlock if it's held or None."""
 
     def checkcommitpatterns(wctx, vdirs, match, status, fail):
         pass
 
     def commit(text='', user=None, date=None, match=None, force=False,
                editor=False, extra=None):
         """Add a new revision to the repository."""
 
     def commitctx(ctx, error=False):
         """Commit a commitctx instance to the repository."""
 
     def destroying():
         """Inform the repository that nodes are about to be destroyed."""
 
     def destroyed():
         """Inform the repository that nodes have been destroyed."""
 
     def status(node1='.', node2=None, match=None, ignored=False,
                clean=False, unknown=False, listsubrepos=False):
         """Convenience method to call repo[x].status()."""
 
     def addpostdsstatus(ps):
         pass
 
     def postdsstatus():
         pass
 
     def clearpostdsstatus():
         pass
 
     def heads(start=None):
         """Obtain list of nodes that are DAG heads."""
 
     def branchheads(branch=None, start=None, closed=False):
         pass
 
     def branches(nodes):
         pass
 
     def between(pairs):
         pass
 
     def checkpush(pushop):
         pass
 
     prepushoutgoinghooks = interfaceutil.Attribute(
         """util.hooks instance.""")
 
     def pushkey(namespace, key, old, new):
         pass
 
     def listkeys(namespace):
         pass
 
     def debugwireargs(one, two, three=None, four=None, five=None):
         pass
 
     def savecommitmessage(text):
         pass