largefiles: use "normallookup" on "lfdirstate" while reverting

Before this patch, largefiles gotten from revisions other than the
parent of the working directory at "hg revert" become "clean"
unexpectedly in steps below:

    1. "repo.status()" is invoked (for status check before reverting)
      1-1 "dirstate" entry for standinfile SF is "normal"-ed
     (1-2 "lfdirstate" entry of largefile LF (for SF) is "normal"-ed)

    2. "cmdutil.revert()" is invoked
      2-1 standinfile SF is updated in the working directory
      2-2 "dirstate" entry for SF is NOT updated

    3. "lfcommands.updatelfiles()" is invoked (by "overrides.overriderevert()")
      3-1 largefile LF (for SF) is updated in the working directory
      3-2 "dirstate" returns "n" and valid timestamp for SF (by 1-1 and 2-2)
      3-3 "lfdirstate" entry for LF is "normal"-ed
      3-4 "lfdirstate" is written into ".hg/largefiles/dirstate", and
          timestamp of LF is stored into "lfdirstate" file (by 3-3)
          (ASSUMPTION: timestamp of LF differs from one of "lfdirstate" file)

Then, "hs status" treats LF as "clean", even though LF is updated by
"other" revision (by 3-1), because "lfilesrepo.status()" always treats
"normal"-ed files (by 3-3 and 3-4) as "clean".

When largefiles are reverted, they should be "normallookup"-ed
forcibly.

This patch uses "normallookup" on "lfdirstate" while reverting, by
passing "True" to newly added argument "normallookup".

Forcible "normallookup"-ing is not so expensive, because list of
target largefiles is explicitly specified in this case.

This patch uses "[debug] dirstate.delaywrite" feature in the test, to
ensure that timestamp of the largefile gotten from "other" revision is
stored into ".hg/largefiles/dirstate" (for ASSUMPTION at 3-4)

hgext/largefiles/lfcommands.py

@@ -1,577 +1,578 b''
 # Copyright 2009-2010 Gregory P. Ward
 # Copyright 2009-2010 Intelerad Medical Systems Incorporated
 # Copyright 2010-2011 Fog Creek Software
 # Copyright 2010-2011 Unity Technologies
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 '''High-level command function for lfconvert, plus the cmdtable.'''
 
 import os, errno
 import shutil
 
 from mercurial import util, match as match_, hg, node, context, error, \
     cmdutil, scmutil, commands
 from mercurial.i18n import _
 from mercurial.lock import release
 
 import lfutil
 import basestore
 
 # -- Commands ----------------------------------------------------------
 
 cmdtable = {}
 command = cmdutil.command(cmdtable)
 
 @command('lfconvert',
     [('s', 'size', '',
       _('minimum size (MB) for files to be converted as largefiles'), 'SIZE'),
     ('', 'to-normal', False,
      _('convert from a largefiles repo to a normal repo')),
     ],
     _('hg lfconvert SOURCE DEST [FILE ...]'),
     norepo=True,
     inferrepo=True)
 def lfconvert(ui, src, dest, *pats, **opts):
     '''convert a normal repository to a largefiles repository
 
     Convert repository SOURCE to a new repository DEST, identical to
     SOURCE except that certain files will be converted as largefiles:
     specifically, any file that matches any PATTERN *or* whose size is
     above the minimum size threshold is converted as a largefile. The
     size used to determine whether or not to track a file as a
     largefile is the size of the first version of the file. The
     minimum size can be specified either with --size or in
     configuration as ``largefiles.size``.
 
     After running this command you will need to make sure that
     largefiles is enabled anywhere you intend to push the new
     repository.
 
     Use --to-normal to convert largefiles back to normal files; after
     this, the DEST repository can be used without largefiles at all.'''
 
     if opts['to_normal']:
         tolfile = False
     else:
         tolfile = True
         size = lfutil.getminsize(ui, True, opts.get('size'), default=None)
 
     if not hg.islocal(src):
         raise util.Abort(_('%s is not a local Mercurial repo') % src)
     if not hg.islocal(dest):
         raise util.Abort(_('%s is not a local Mercurial repo') % dest)
 
     rsrc = hg.repository(ui, src)
     ui.status(_('initializing destination %s\n') % dest)
     rdst = hg.repository(ui, dest, create=True)
 
     success = False
     dstwlock = dstlock = None
     try:
         # Lock destination to prevent modification while it is converted to.
         # Don't need to lock src because we are just reading from its history
         # which can't change.
         dstwlock = rdst.wlock()
         dstlock = rdst.lock()
 
         # Get a list of all changesets in the source.  The easy way to do this
         # is to simply walk the changelog, using changelog.nodesbetween().
         # Take a look at mercurial/revlog.py:639 for more details.
         # Use a generator instead of a list to decrease memory usage
         ctxs = (rsrc[ctx] for ctx in rsrc.changelog.nodesbetween(None,
             rsrc.heads())[0])
         revmap = {node.nullid: node.nullid}
         if tolfile:
             lfiles = set()
             normalfiles = set()
             if not pats:
                 pats = ui.configlist(lfutil.longname, 'patterns', default=[])
             if pats:
                 matcher = match_.match(rsrc.root, '', list(pats))
             else:
                 matcher = None
 
             lfiletohash = {}
             for ctx in ctxs:
                 ui.progress(_('converting revisions'), ctx.rev(),
                     unit=_('revision'), total=rsrc['tip'].rev())
                 _lfconvert_addchangeset(rsrc, rdst, ctx, revmap,
                     lfiles, normalfiles, matcher, size, lfiletohash)
             ui.progress(_('converting revisions'), None)
 
             if os.path.exists(rdst.wjoin(lfutil.shortname)):
                 shutil.rmtree(rdst.wjoin(lfutil.shortname))
 
             for f in lfiletohash.keys():
                 if os.path.isfile(rdst.wjoin(f)):
                     os.unlink(rdst.wjoin(f))
                 try:
                     os.removedirs(os.path.dirname(rdst.wjoin(f)))
                 except OSError:
                     pass
 
             # If there were any files converted to largefiles, add largefiles
             # to the destination repository's requirements.
             if lfiles:
                 rdst.requirements.add('largefiles')
                 rdst._writerequirements()
         else:
             for ctx in ctxs:
                 ui.progress(_('converting revisions'), ctx.rev(),
                     unit=_('revision'), total=rsrc['tip'].rev())
                 _addchangeset(ui, rsrc, rdst, ctx, revmap)
 
             ui.progress(_('converting revisions'), None)
         success = True
     finally:
         rdst.dirstate.clear()
         release(dstlock, dstwlock)
         if not success:
             # we failed, remove the new directory
             shutil.rmtree(rdst.root)
 
 def _addchangeset(ui, rsrc, rdst, ctx, revmap):
     # Convert src parents to dst parents
     parents = _convertparents(ctx, revmap)
 
     # Generate list of changed files
     files = _getchangedfiles(ctx, parents)
 
     def getfilectx(repo, memctx, f):
         if lfutil.standin(f) in files:
             # if the file isn't in the manifest then it was removed
             # or renamed, raise IOError to indicate this
             try:
                 fctx = ctx.filectx(lfutil.standin(f))
             except error.LookupError:
                 raise IOError
             renamed = fctx.renamed()
             if renamed:
                 renamed = lfutil.splitstandin(renamed[0])
 
             hash = fctx.data().strip()
             path = lfutil.findfile(rsrc, hash)
 
             # If one file is missing, likely all files from this rev are
             if path is None:
                 cachelfiles(ui, rsrc, ctx.node())
                 path = lfutil.findfile(rsrc, hash)
 
                 if path is None:
                     raise util.Abort(
                         _("missing largefile \'%s\' from revision %s")
                          % (f, node.hex(ctx.node())))
 
             data = ''
             fd = None
             try:
                 fd = open(path, 'rb')
                 data = fd.read()
             finally:
                 if fd:
                     fd.close()
             return context.memfilectx(repo, f, data, 'l' in fctx.flags(),
                                       'x' in fctx.flags(), renamed)
         else:
             return _getnormalcontext(repo, ctx, f, revmap)
 
     dstfiles = []
     for file in files:
         if lfutil.isstandin(file):
             dstfiles.append(lfutil.splitstandin(file))
         else:
             dstfiles.append(file)
     # Commit
     _commitcontext(rdst, parents, ctx, dstfiles, getfilectx, revmap)
 
 def _lfconvert_addchangeset(rsrc, rdst, ctx, revmap, lfiles, normalfiles,
         matcher, size, lfiletohash):
     # Convert src parents to dst parents
     parents = _convertparents(ctx, revmap)
 
     # Generate list of changed files
     files = _getchangedfiles(ctx, parents)
 
     dstfiles = []
     for f in files:
         if f not in lfiles and f not in normalfiles:
             islfile = _islfile(f, ctx, matcher, size)
             # If this file was renamed or copied then copy
             # the largefile-ness of its predecessor
             if f in ctx.manifest():
                 fctx = ctx.filectx(f)
                 renamed = fctx.renamed()
                 renamedlfile = renamed and renamed[0] in lfiles
                 islfile |= renamedlfile
                 if 'l' in fctx.flags():
                     if renamedlfile:
                         raise util.Abort(
                             _('renamed/copied largefile %s becomes symlink')
                             % f)
                     islfile = False
             if islfile:
                 lfiles.add(f)
             else:
                 normalfiles.add(f)
 
         if f in lfiles:
             dstfiles.append(lfutil.standin(f))
             # largefile in manifest if it has not been removed/renamed
             if f in ctx.manifest():
                 fctx = ctx.filectx(f)
                 if 'l' in fctx.flags():
                     renamed = fctx.renamed()
                     if renamed and renamed[0] in lfiles:
                         raise util.Abort(_('largefile %s becomes symlink') % f)
 
                 # largefile was modified, update standins
                 m = util.sha1('')
                 m.update(ctx[f].data())
                 hash = m.hexdigest()
                 if f not in lfiletohash or lfiletohash[f] != hash:
                     rdst.wwrite(f, ctx[f].data(), ctx[f].flags())
                     executable = 'x' in ctx[f].flags()
                     lfutil.writestandin(rdst, lfutil.standin(f), hash,
                         executable)
                     lfiletohash[f] = hash
         else:
             # normal file
             dstfiles.append(f)
 
     def getfilectx(repo, memctx, f):
         if lfutil.isstandin(f):
             # if the file isn't in the manifest then it was removed
             # or renamed, raise IOError to indicate this
             srcfname = lfutil.splitstandin(f)
             try:
                 fctx = ctx.filectx(srcfname)
             except error.LookupError:
                 raise IOError
             renamed = fctx.renamed()
             if renamed:
                 # standin is always a largefile because largefile-ness
                 # doesn't change after rename or copy
                 renamed = lfutil.standin(renamed[0])
 
             return context.memfilectx(repo, f, lfiletohash[srcfname] + '\n',
                                       'l' in fctx.flags(), 'x' in fctx.flags(),
                                       renamed)
         else:
             return _getnormalcontext(repo, ctx, f, revmap)
 
     # Commit
     _commitcontext(rdst, parents, ctx, dstfiles, getfilectx, revmap)
 
 def _commitcontext(rdst, parents, ctx, dstfiles, getfilectx, revmap):
     mctx = context.memctx(rdst, parents, ctx.description(), dstfiles,
                           getfilectx, ctx.user(), ctx.date(), ctx.extra())
     ret = rdst.commitctx(mctx)
     rdst.setparents(ret)
     revmap[ctx.node()] = rdst.changelog.tip()
 
 # Generate list of changed files
 def _getchangedfiles(ctx, parents):
     files = set(ctx.files())
     if node.nullid not in parents:
         mc = ctx.manifest()
         mp1 = ctx.parents()[0].manifest()
         mp2 = ctx.parents()[1].manifest()
         files |= (set(mp1) | set(mp2)) - set(mc)
         for f in mc:
             if mc[f] != mp1.get(f, None) or mc[f] != mp2.get(f, None):
                 files.add(f)
     return files
 
 # Convert src parents to dst parents
 def _convertparents(ctx, revmap):
     parents = []
     for p in ctx.parents():
         parents.append(revmap[p.node()])
     while len(parents) < 2:
         parents.append(node.nullid)
     return parents
 
 # Get memfilectx for a normal file
 def _getnormalcontext(repo, ctx, f, revmap):
     try:
         fctx = ctx.filectx(f)
     except error.LookupError:
         raise IOError
     renamed = fctx.renamed()
     if renamed:
         renamed = renamed[0]
 
     data = fctx.data()
     if f == '.hgtags':
         data = _converttags (repo.ui, revmap, data)
     return context.memfilectx(repo, f, data, 'l' in fctx.flags(),
                               'x' in fctx.flags(), renamed)
 
 # Remap tag data using a revision map
 def _converttags(ui, revmap, data):
     newdata = []
     for line in data.splitlines():
         try:
             id, name = line.split(' ', 1)
         except ValueError:
             ui.warn(_('skipping incorrectly formatted tag %s\n')
                 % line)
             continue
         try:
             newid = node.bin(id)
         except TypeError:
             ui.warn(_('skipping incorrectly formatted id %s\n')
                 % id)
             continue
         try:
             newdata.append('%s %s\n' % (node.hex(revmap[newid]),
                 name))
         except KeyError:
             ui.warn(_('no mapping for id %s\n') % id)
             continue
     return ''.join(newdata)
 
 def _islfile(file, ctx, matcher, size):
     '''Return true if file should be considered a largefile, i.e.
     matcher matches it or it is larger than size.'''
     # never store special .hg* files as largefiles
     if file == '.hgtags' or file == '.hgignore' or file == '.hgsigs':
         return False
     if matcher and matcher(file):
         return True
     try:
         return ctx.filectx(file).size() >= size * 1024 * 1024
     except error.LookupError:
         return False
 
 def uploadlfiles(ui, rsrc, rdst, files):
     '''upload largefiles to the central store'''
 
     if not files:
         return
 
     store = basestore._openstore(rsrc, rdst, put=True)
 
     at = 0
     ui.debug("sending statlfile command for %d largefiles\n" % len(files))
     retval = store.exists(files)
     files = filter(lambda h: not retval[h], files)
     ui.debug("%d largefiles need to be uploaded\n" % len(files))
 
     for hash in files:
         ui.progress(_('uploading largefiles'), at, unit='largefile',
                     total=len(files))
         source = lfutil.findfile(rsrc, hash)
         if not source:
             raise util.Abort(_('largefile %s missing from store'
                                ' (needs to be uploaded)') % hash)
         # XXX check for errors here
         store.put(source, hash)
         at += 1
     ui.progress(_('uploading largefiles'), None)
 
 def verifylfiles(ui, repo, all=False, contents=False):
     '''Verify that every largefile revision in the current changeset
     exists in the central store.  With --contents, also verify that
     the contents of each local largefile file revision are correct (SHA-1 hash
     matches the revision ID).  With --all, check every changeset in
     this repository.'''
     if all:
         # Pass a list to the function rather than an iterator because we know a
         # list will work.
         revs = range(len(repo))
     else:
         revs = ['.']
 
     store = basestore._openstore(repo)
     return store.verify(revs, contents=contents)
 
 def cachelfiles(ui, repo, node, filelist=None):
     '''cachelfiles ensures that all largefiles needed by the specified revision
     are present in the repository's largefile cache.
 
     returns a tuple (cached, missing).  cached is the list of files downloaded
     by this operation; missing is the list of files that were needed but could
     not be found.'''
     lfiles = lfutil.listlfiles(repo, node)
     if filelist:
         lfiles = set(lfiles) & set(filelist)
     toget = []
 
     for lfile in lfiles:
         try:
             expectedhash = repo[node][lfutil.standin(lfile)].data().strip()
         except IOError, err:
             if err.errno == errno.ENOENT:
                 continue # node must be None and standin wasn't found in wctx
             raise
         if not lfutil.findfile(repo, expectedhash):
             toget.append((lfile, expectedhash))
 
     if toget:
         store = basestore._openstore(repo)
         ret = store.get(toget)
         return ret
 
     return ([], [])
 
 def downloadlfiles(ui, repo, rev=None):
     matchfn = scmutil.match(repo[None],
                             [repo.wjoin(lfutil.shortname)], {})
     def prepare(ctx, fns):
         pass
     totalsuccess = 0
     totalmissing = 0
     if rev != []: # walkchangerevs on empty list would return all revs
         for ctx in cmdutil.walkchangerevs(repo, matchfn, {'rev' : rev},
                                           prepare):
             success, missing = cachelfiles(ui, repo, ctx.node())
             totalsuccess += len(success)
             totalmissing += len(missing)
     ui.status(_("%d additional largefiles cached\n") % totalsuccess)
     if totalmissing > 0:
         ui.status(_("%d largefiles failed to download\n") % totalmissing)
     return totalsuccess, totalmissing
 
-def updatelfiles(ui, repo, filelist=None, printmessage=True):
+def updatelfiles(ui, repo, filelist=None, printmessage=True,
+                 normallookup=False):
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         lfiles = set(lfutil.listlfiles(repo)) | set(lfdirstate)
 
         if filelist is not None:
             lfiles = [f for f in lfiles if f in filelist]
 
         update = {}
         updated, removed = 0, 0
         for lfile in lfiles:
             abslfile = repo.wjoin(lfile)
             absstandin = repo.wjoin(lfutil.standin(lfile))
             if os.path.exists(absstandin):
                 if (os.path.exists(absstandin + '.orig') and
                     os.path.exists(abslfile)):
                     shutil.copyfile(abslfile, abslfile + '.orig')
                     util.unlinkpath(absstandin + '.orig')
                 expecthash = lfutil.readstandin(repo, lfile)
                 if (expecthash != '' and
                     (not os.path.exists(abslfile) or
                      expecthash != lfutil.hashfile(abslfile))):
                     if lfile not in repo[None]: # not switched to normal file
                         util.unlinkpath(abslfile, ignoremissing=True)
                     # use normallookup() to allocate entry in largefiles
                     # dirstate, because lack of it misleads
                     # lfilesrepo.status() into recognition that such cache
                     # missing files are REMOVED.
                     lfdirstate.normallookup(lfile)
                     update[lfile] = expecthash
             else:
                 # Remove lfiles for which the standin is deleted, unless the
                 # lfile is added to the repository again. This happens when a
                 # largefile is converted back to a normal file: the standin
                 # disappears, but a new (normal) file appears as the lfile.
                 if (os.path.exists(abslfile) and
                     repo.dirstate.normalize(lfile) not in repo[None]):
                     util.unlinkpath(abslfile)
                     removed += 1
 
         # largefile processing might be slow and be interrupted - be prepared
         lfdirstate.write()
 
         if lfiles:
             if printmessage:
                 ui.status(_('getting changed largefiles\n'))
             cachelfiles(ui, repo, None, lfiles)
 
         for lfile in lfiles:
             update1 = 0
 
             expecthash = update.get(lfile)
             if expecthash:
                 if not lfutil.copyfromcache(repo, expecthash, lfile):
                     # failed ... but already removed and set to normallookup
                     continue
                 # Synchronize largefile dirstate to the last modified
                 # time of the file
                 lfdirstate.normal(lfile)
                 update1 = 1
 
             # copy the state of largefile standin from the repository's
             # dirstate to its state in the lfdirstate.
             abslfile = repo.wjoin(lfile)
             absstandin = repo.wjoin(lfutil.standin(lfile))
             if os.path.exists(absstandin):
                 mode = os.stat(absstandin).st_mode
                 if mode != os.stat(abslfile).st_mode:
                     os.chmod(abslfile, mode)
                     update1 = 1
 
             updated += update1
 
             standin = lfutil.standin(lfile)
             if standin in repo.dirstate:
                 stat = repo.dirstate._map[standin]
                 state, mtime = stat[0], stat[3]
             else:
                 state, mtime = '?', -1
             if state == 'n':
-                if mtime < 0:
+                if normallookup or mtime < 0:
                     # state 'n' doesn't ensure 'clean' in this case
                     lfdirstate.normallookup(lfile)
                 else:
                     lfdirstate.normal(lfile)
             elif state == 'm':
                 lfdirstate.normallookup(lfile)
             elif state == 'r':
                 lfdirstate.remove(lfile)
             elif state == 'a':
                 lfdirstate.add(lfile)
             elif state == '?':
                 lfdirstate.drop(lfile)
 
         lfdirstate.write()
         if printmessage and lfiles:
             ui.status(_('%d largefiles updated, %d removed\n') % (updated,
                 removed))
     finally:
         wlock.release()
 
 @command('lfpull',
     [('r', 'rev', [], _('pull largefiles for these revisions'))
     ] + commands.remoteopts,
     _('-r REV... [-e CMD] [--remotecmd CMD] [SOURCE]'))
 def lfpull(ui, repo, source="default", **opts):
     """pull largefiles for the specified revisions from the specified source
 
     Pull largefiles that are referenced from local changesets but missing
     locally, pulling from a remote repository to the local cache.
 
     If SOURCE is omitted, the 'default' path will be used.
     See :hg:`help urls` for more information.
 
     .. container:: verbose
 
       Some examples:
 
       - pull largefiles for all branch heads::
 
           hg lfpull -r "head() and not closed()"
 
       - pull largefiles on the default branch::
 
           hg lfpull -r "branch(default)"
     """
     repo.lfpullsource = source
 
     revs = opts.get('rev', [])
     if not revs:
         raise util.Abort(_('no revisions specified'))
     revs = scmutil.revrange(repo, revs)
 
     numcached = 0
     for rev in revs:
         ui.note(_('pulling largefiles for revision %s\n') % rev)
         (cached, missing) = cachelfiles(ui, repo, rev)
         numcached += len(cached)
     ui.status(_("%d largefiles cached\n") % numcached)

hgext/largefiles/overrides.py

@@ -1,1222 +1,1228 b''
 # Copyright 2009-2010 Gregory P. Ward
 # Copyright 2009-2010 Intelerad Medical Systems Incorporated
 # Copyright 2010-2011 Fog Creek Software
 # Copyright 2010-2011 Unity Technologies
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 '''Overridden Mercurial commands and functions for the largefiles extension'''
 
 import os
 import copy
 
 from mercurial import hg, commands, util, cmdutil, scmutil, match as match_, \
         archival, merge, pathutil, revset
 from mercurial.i18n import _
 from mercurial.node import hex
 from hgext import rebase
 
 import lfutil
 import lfcommands
 import basestore
 
 # -- Utility functions: commonly/repeatedly needed functionality ---------------
 
 def installnormalfilesmatchfn(manifest):
     '''installmatchfn with a matchfn that ignores all largefiles'''
     def overridematch(ctx, pats=[], opts={}, globbed=False,
             default='relpath'):
         match = oldmatch(ctx, pats, opts, globbed, default)
         m = copy.copy(match)
         notlfile = lambda f: not (lfutil.isstandin(f) or lfutil.standin(f) in
                 manifest)
         m._files = filter(notlfile, m._files)
         m._fmap = set(m._files)
         m._always = False
         origmatchfn = m.matchfn
         m.matchfn = lambda f: notlfile(f) and origmatchfn(f) or None
         return m
     oldmatch = installmatchfn(overridematch)
 
 def installmatchfn(f):
     '''monkey patch the scmutil module with a custom match function.
     Warning: it is monkey patching the _module_ on runtime! Not thread safe!'''
     oldmatch = scmutil.match
     setattr(f, 'oldmatch', oldmatch)
     scmutil.match = f
     return oldmatch
 
 def restorematchfn():
     '''restores scmutil.match to what it was before installmatchfn
     was called.  no-op if scmutil.match is its original function.
 
     Note that n calls to installmatchfn will require n calls to
     restore matchfn to reverse'''
     scmutil.match = getattr(scmutil.match, 'oldmatch')
 
 def installmatchandpatsfn(f):
     oldmatchandpats = scmutil.matchandpats
     setattr(f, 'oldmatchandpats', oldmatchandpats)
     scmutil.matchandpats = f
     return oldmatchandpats
 
 def restorematchandpatsfn():
     '''restores scmutil.matchandpats to what it was before
     installnormalfilesmatchandpatsfn was called.  no-op if scmutil.matchandpats
     is its original function.
 
     Note that n calls to installnormalfilesmatchandpatsfn will require n calls
     to restore matchfn to reverse'''
     scmutil.matchandpats = getattr(scmutil.matchandpats, 'oldmatchandpats',
             scmutil.matchandpats)
 
 def addlargefiles(ui, repo, *pats, **opts):
     large = opts.pop('large', None)
     lfsize = lfutil.getminsize(
         ui, lfutil.islfilesrepo(repo), opts.pop('lfsize', None))
 
     lfmatcher = None
     if lfutil.islfilesrepo(repo):
         lfpats = ui.configlist(lfutil.longname, 'patterns', default=[])
         if lfpats:
             lfmatcher = match_.match(repo.root, '', list(lfpats))
 
     lfnames = []
     m = scmutil.match(repo[None], pats, opts)
     m.bad = lambda x, y: None
     wctx = repo[None]
     for f in repo.walk(m):
         exact = m.exact(f)
         lfile = lfutil.standin(f) in wctx
         nfile = f in wctx
         exists = lfile or nfile
 
         # Don't warn the user when they attempt to add a normal tracked file.
         # The normal add code will do that for us.
         if exact and exists:
             if lfile:
                 ui.warn(_('%s already a largefile\n') % f)
             continue
 
         if (exact or not exists) and not lfutil.isstandin(f):
             wfile = repo.wjoin(f)
 
             # In case the file was removed previously, but not committed
             # (issue3507)
             if not os.path.exists(wfile):
                 continue
 
             abovemin = (lfsize and
                         os.lstat(wfile).st_size >= lfsize * 1024 * 1024)
             if large or abovemin or (lfmatcher and lfmatcher(f)):
                 lfnames.append(f)
                 if ui.verbose or not exact:
                     ui.status(_('adding %s as a largefile\n') % m.rel(f))
 
     bad = []
     standins = []
 
     # Need to lock, otherwise there could be a race condition between
     # when standins are created and added to the repo.
     wlock = repo.wlock()
     try:
         if not opts.get('dry_run'):
             lfdirstate = lfutil.openlfdirstate(ui, repo)
             for f in lfnames:
                 standinname = lfutil.standin(f)
                 lfutil.writestandin(repo, standinname, hash='',
                     executable=lfutil.getexecutable(repo.wjoin(f)))
                 standins.append(standinname)
                 if lfdirstate[f] == 'r':
                     lfdirstate.normallookup(f)
                 else:
                     lfdirstate.add(f)
             lfdirstate.write()
             bad += [lfutil.splitstandin(f)
                     for f in repo[None].add(standins)
                     if f in m.files()]
     finally:
         wlock.release()
     return bad
 
 def removelargefiles(ui, repo, *pats, **opts):
     after = opts.get('after')
     if not pats and not after:
         raise util.Abort(_('no files specified'))
     m = scmutil.match(repo[None], pats, opts)
     try:
         repo.lfstatus = True
         s = repo.status(match=m, clean=True)
     finally:
         repo.lfstatus = False
     manifest = repo[None].manifest()
     modified, added, deleted, clean = [[f for f in list
                                         if lfutil.standin(f) in manifest]
                                        for list in [s[0], s[1], s[3], s[6]]]
 
     def warn(files, msg):
         for f in files:
             ui.warn(msg % m.rel(f))
         return int(len(files) > 0)
 
     result = 0
 
     if after:
         remove, forget = deleted, []
         result = warn(modified + added + clean,
                       _('not removing %s: file still exists\n'))
     else:
         remove, forget = deleted + clean, []
         result = warn(modified, _('not removing %s: file is modified (use -f'
                                   ' to force removal)\n'))
         result = warn(added, _('not removing %s: file has been marked for add'
                                ' (use forget to undo)\n')) or result
 
     for f in sorted(remove + forget):
         if ui.verbose or not m.exact(f):
             ui.status(_('removing %s\n') % m.rel(f))
 
     # Need to lock because standin files are deleted then removed from the
     # repository and we could race in-between.
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for f in remove:
             if not after:
                 # If this is being called by addremove, notify the user that we
                 # are removing the file.
                 if getattr(repo, "_isaddremove", False):
                     ui.status(_('removing %s\n') % f)
                 util.unlinkpath(repo.wjoin(f), ignoremissing=True)
             lfdirstate.remove(f)
         lfdirstate.write()
         forget = [lfutil.standin(f) for f in forget]
         remove = [lfutil.standin(f) for f in remove]
         repo[None].forget(forget)
         # If this is being called by addremove, let the original addremove
         # function handle this.
         if not getattr(repo, "_isaddremove", False):
             for f in remove:
                 util.unlinkpath(repo.wjoin(f), ignoremissing=True)
         repo[None].forget(remove)
     finally:
         wlock.release()
 
     return result
 
 # For overriding mercurial.hgweb.webcommands so that largefiles will
 # appear at their right place in the manifests.
 def decodepath(orig, path):
     return lfutil.splitstandin(path) or path
 
 # -- Wrappers: modify existing commands --------------------------------
 
 # Add works by going through the files that the user wanted to add and
 # checking if they should be added as largefiles. Then it makes a new
 # matcher which matches only the normal files and runs the original
 # version of add.
 def overrideadd(orig, ui, repo, *pats, **opts):
     normal = opts.pop('normal')
     if normal:
         if opts.get('large'):
             raise util.Abort(_('--normal cannot be used with --large'))
         return orig(ui, repo, *pats, **opts)
     bad = addlargefiles(ui, repo, *pats, **opts)
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(ui, repo, *pats, **opts)
     restorematchfn()
 
     return (result == 1 or bad) and 1 or 0
 
 def overrideremove(orig, ui, repo, *pats, **opts):
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(ui, repo, *pats, **opts)
     restorematchfn()
     return removelargefiles(ui, repo, *pats, **opts) or result
 
 def overridestatusfn(orig, repo, rev2, **opts):
     try:
         repo._repo.lfstatus = True
         return orig(repo, rev2, **opts)
     finally:
         repo._repo.lfstatus = False
 
 def overridestatus(orig, ui, repo, *pats, **opts):
     try:
         repo.lfstatus = True
         return orig(ui, repo, *pats, **opts)
     finally:
         repo.lfstatus = False
 
 def overridedirty(orig, repo, ignoreupdate=False):
     try:
         repo._repo.lfstatus = True
         return orig(repo, ignoreupdate)
     finally:
         repo._repo.lfstatus = False
 
 def overridelog(orig, ui, repo, *pats, **opts):
     def overridematchandpats(ctx, pats=[], opts={}, globbed=False,
             default='relpath'):
         """Matcher that merges root directory with .hglf, suitable for log.
         It is still possible to match .hglf directly.
         For any listed files run log on the standin too.
         matchfn tries both the given filename and with .hglf stripped.
         """
         matchandpats = oldmatchandpats(ctx, pats, opts, globbed, default)
         m, p = copy.copy(matchandpats)
 
         pats = set(p)
         # TODO: handling of patterns in both cases below
         if m._cwd:
             if os.path.isabs(m._cwd):
                 # TODO: handle largefile magic when invoked from other cwd
                 return matchandpats
             back = (m._cwd.count('/') + 1) * '../'
             pats.update(back + lfutil.standin(m._cwd + '/' + f) for f in p)
         else:
             pats.update(lfutil.standin(f) for f in p)
 
         for i in range(0, len(m._files)):
             standin = lfutil.standin(m._files[i])
             if standin in repo[ctx.node()]:
                 m._files[i] = standin
             elif m._files[i] not in repo[ctx.node()]:
                 m._files.append(standin)
             pats.add(standin)
 
         m._fmap = set(m._files)
         m._always = False
         origmatchfn = m.matchfn
         def lfmatchfn(f):
             lf = lfutil.splitstandin(f)
             if lf is not None and origmatchfn(lf):
                 return True
             r = origmatchfn(f)
             return r
         m.matchfn = lfmatchfn
 
         return m, pats
 
     oldmatchandpats = installmatchandpatsfn(overridematchandpats)
     try:
         repo.lfstatus = True
         return orig(ui, repo, *pats, **opts)
     finally:
         repo.lfstatus = False
         restorematchandpatsfn()
 
 def overrideverify(orig, ui, repo, *pats, **opts):
     large = opts.pop('large', False)
     all = opts.pop('lfa', False)
     contents = opts.pop('lfc', False)
 
     result = orig(ui, repo, *pats, **opts)
     if large or all or contents:
         result = result or lfcommands.verifylfiles(ui, repo, all, contents)
     return result
 
 def overridedebugstate(orig, ui, repo, *pats, **opts):
     large = opts.pop('large', False)
     if large:
         class fakerepo(object):
             dirstate = lfutil.openlfdirstate(ui, repo)
         orig(ui, fakerepo, *pats, **opts)
     else:
         orig(ui, repo, *pats, **opts)
 
 # Override needs to refresh standins so that update's normal merge
 # will go through properly. Then the other update hook (overriding repo.update)
 # will get the new files. Filemerge is also overridden so that the merge
 # will merge standins correctly.
 def overrideupdate(orig, ui, repo, *pats, **opts):
     # Need to lock between the standins getting updated and their
     # largefiles getting updated
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         s = lfdirstate.status(match_.always(repo.root, repo.getcwd()),
             [], False, False, False)
         (unsure, modified, added, removed, missing, unknown, ignored, clean) = s
 
         if opts['check']:
             mod = len(modified) > 0
             for lfile in unsure:
                 standin = lfutil.standin(lfile)
                 if repo['.'][standin].data().strip() != \
                         lfutil.hashfile(repo.wjoin(lfile)):
                     mod = True
                 else:
                     lfdirstate.normal(lfile)
             lfdirstate.write()
             if mod:
                 raise util.Abort(_('uncommitted changes'))
         # XXX handle removed differently
         if not opts['clean']:
             for lfile in unsure + modified + added:
                 lfutil.updatestandin(repo, lfutil.standin(lfile))
         return orig(ui, repo, *pats, **opts)
     finally:
         wlock.release()
 
 # Before starting the manifest merge, merge.updates will call
 # _checkunknown to check if there are any files in the merged-in
 # changeset that collide with unknown files in the working copy.
 #
 # The largefiles are seen as unknown, so this prevents us from merging
 # in a file 'foo' if we already have a largefile with the same name.
 #
 # The overridden function filters the unknown files by removing any
 # largefiles. This makes the merge proceed and we can then handle this
 # case further in the overridden manifestmerge function below.
 def overridecheckunknownfile(origfn, repo, wctx, mctx, f):
     if lfutil.standin(repo.dirstate.normalize(f)) in wctx:
         return False
     return origfn(repo, wctx, mctx, f)
 
 # The manifest merge handles conflicts on the manifest level. We want
 # to handle changes in largefile-ness of files at this level too.
 #
 # The strategy is to run the original manifestmerge and then process
 # the action list it outputs. There are two cases we need to deal with:
 #
 # 1. Normal file in p1, largefile in p2. Here the largefile is
 #    detected via its standin file, which will enter the working copy
 #    with a "get" action. It is not "merge" since the standin is all
 #    Mercurial is concerned with at this level -- the link to the
 #    existing normal file is not relevant here.
 #
 # 2. Largefile in p1, normal file in p2. Here we get a "merge" action
 #    since the largefile will be present in the working copy and
 #    different from the normal file in p2. Mercurial therefore
 #    triggers a merge action.
 #
 # In both cases, we prompt the user and emit new actions to either
 # remove the standin (if the normal file was kept) or to remove the
 # normal file and get the standin (if the largefile was kept). The
 # default prompt answer is to use the largefile version since it was
 # presumably changed on purpose.
 #
 # Finally, the merge.applyupdates function will then take care of
 # writing the files into the working copy and lfcommands.updatelfiles
 # will update the largefiles.
 def overridecalculateupdates(origfn, repo, p1, p2, pas, branchmerge, force,
                              partial, acceptremote, followcopies):
     overwrite = force and not branchmerge
     actions = origfn(repo, p1, p2, pas, branchmerge, force, partial,
                      acceptremote, followcopies)
 
     if overwrite:
         return actions
 
     removes = set(a[0] for a in actions['r'])
 
     newglist = []
     for action in actions['g']:
         f, args, msg = action
         splitstandin = f and lfutil.splitstandin(f)
         if (splitstandin is not None and
             splitstandin in p1 and splitstandin not in removes):
             # Case 1: normal file in the working copy, largefile in
             # the second parent
             lfile = splitstandin
             standin = f
             msg = _('remote turned local normal file %s into a largefile\n'
                     'use (l)argefile or keep (n)ormal file?'
                     '$$ &Largefile $$ &Normal file') % lfile
             if repo.ui.promptchoice(msg, 0) == 0:
                 actions['r'].append((lfile, None, msg))
                 newglist.append((standin, (p2.flags(standin),), msg))
             else:
                 actions['r'].append((standin, None, msg))
         elif lfutil.standin(f) in p1 and lfutil.standin(f) not in removes:
             # Case 2: largefile in the working copy, normal file in
             # the second parent
             standin = lfutil.standin(f)
             lfile = f
             msg = _('remote turned local largefile %s into a normal file\n'
                     'keep (l)argefile or use (n)ormal file?'
                     '$$ &Largefile $$ &Normal file') % lfile
             if repo.ui.promptchoice(msg, 0) == 0:
                 actions['r'].append((lfile, None, msg))
             else:
                 actions['r'].append((standin, None, msg))
                 newglist.append((lfile, (p2.flags(lfile),), msg))
         else:
             newglist.append(action)
 
     newglist.sort()
     actions['g'] = newglist
 
     return actions
 
 # Override filemerge to prompt the user about how they wish to merge
 # largefiles. This will handle identical edits without prompting the user.
 def overridefilemerge(origfn, repo, mynode, orig, fcd, fco, fca, labels=None):
     if not lfutil.isstandin(orig):
         return origfn(repo, mynode, orig, fcd, fco, fca, labels=labels)
 
     ahash = fca.data().strip().lower()
     dhash = fcd.data().strip().lower()
     ohash = fco.data().strip().lower()
     if (ohash != ahash and
         ohash != dhash and
         (dhash == ahash or
          repo.ui.promptchoice(
              _('largefile %s has a merge conflict\nancestor was %s\n'
                'keep (l)ocal %s or\ntake (o)ther %s?'
                '$$ &Local $$ &Other') %
                (lfutil.splitstandin(orig), ahash, dhash, ohash),
              0) == 1)):
         repo.wwrite(fcd.path(), fco.data(), fco.flags())
     return 0
 
 # Copy first changes the matchers to match standins instead of
 # largefiles.  Then it overrides util.copyfile in that function it
 # checks if the destination largefile already exists. It also keeps a
 # list of copied files so that the largefiles can be copied and the
 # dirstate updated.
 def overridecopy(orig, ui, repo, pats, opts, rename=False):
     # doesn't remove largefile on rename
     if len(pats) < 2:
         # this isn't legal, let the original function deal with it
         return orig(ui, repo, pats, opts, rename)
 
     def makestandin(relpath):
         path = pathutil.canonpath(repo.root, repo.getcwd(), relpath)
         return os.path.join(repo.wjoin(lfutil.standin(path)))
 
     fullpats = scmutil.expandpats(pats)
     dest = fullpats[-1]
 
     if os.path.isdir(dest):
         if not os.path.isdir(makestandin(dest)):
             os.makedirs(makestandin(dest))
     # This could copy both lfiles and normal files in one command,
     # but we don't want to do that. First replace their matcher to
     # only match normal files and run it, then replace it to just
     # match largefiles and run it again.
     nonormalfiles = False
     nolfiles = False
     installnormalfilesmatchfn(repo[None].manifest())
     try:
         try:
             result = orig(ui, repo, pats, opts, rename)
         except util.Abort, e:
             if str(e) != _('no files to copy'):
                 raise e
             else:
                 nonormalfiles = True
             result = 0
     finally:
         restorematchfn()
 
     # The first rename can cause our current working directory to be removed.
     # In that case there is nothing left to copy/rename so just quit.
     try:
         repo.getcwd()
     except OSError:
         return result
 
     try:
         try:
             # When we call orig below it creates the standins but we don't add
             # them to the dir state until later so lock during that time.
             wlock = repo.wlock()
 
             manifest = repo[None].manifest()
             def overridematch(ctx, pats=[], opts={}, globbed=False,
                     default='relpath'):
                 newpats = []
                 # The patterns were previously mangled to add the standin
                 # directory; we need to remove that now
                 for pat in pats:
                     if match_.patkind(pat) is None and lfutil.shortname in pat:
                         newpats.append(pat.replace(lfutil.shortname, ''))
                     else:
                         newpats.append(pat)
                 match = oldmatch(ctx, newpats, opts, globbed, default)
                 m = copy.copy(match)
                 lfile = lambda f: lfutil.standin(f) in manifest
                 m._files = [lfutil.standin(f) for f in m._files if lfile(f)]
                 m._fmap = set(m._files)
                 m._always = False
                 origmatchfn = m.matchfn
                 m.matchfn = lambda f: (lfutil.isstandin(f) and
                                     (f in manifest) and
                                     origmatchfn(lfutil.splitstandin(f)) or
                                     None)
                 return m
             oldmatch = installmatchfn(overridematch)
             listpats = []
             for pat in pats:
                 if match_.patkind(pat) is not None:
                     listpats.append(pat)
                 else:
                     listpats.append(makestandin(pat))
 
             try:
                 origcopyfile = util.copyfile
                 copiedfiles = []
                 def overridecopyfile(src, dest):
                     if (lfutil.shortname in src and
                         dest.startswith(repo.wjoin(lfutil.shortname))):
                         destlfile = dest.replace(lfutil.shortname, '')
                         if not opts['force'] and os.path.exists(destlfile):
                             raise IOError('',
                                 _('destination largefile already exists'))
                     copiedfiles.append((src, dest))
                     origcopyfile(src, dest)
 
                 util.copyfile = overridecopyfile
                 result += orig(ui, repo, listpats, opts, rename)
             finally:
                 util.copyfile = origcopyfile
 
             lfdirstate = lfutil.openlfdirstate(ui, repo)
             for (src, dest) in copiedfiles:
                 if (lfutil.shortname in src and
                     dest.startswith(repo.wjoin(lfutil.shortname))):
                     srclfile = src.replace(repo.wjoin(lfutil.standin('')), '')
                     destlfile = dest.replace(repo.wjoin(lfutil.standin('')), '')
                     destlfiledir = os.path.dirname(repo.wjoin(destlfile)) or '.'
                     if not os.path.isdir(destlfiledir):
                         os.makedirs(destlfiledir)
                     if rename:
                         os.rename(repo.wjoin(srclfile), repo.wjoin(destlfile))
 
                         # The file is gone, but this deletes any empty parent
                         # directories as a side-effect.
                         util.unlinkpath(repo.wjoin(srclfile), True)
                         lfdirstate.remove(srclfile)
                     else:
                         util.copyfile(repo.wjoin(srclfile),
                                       repo.wjoin(destlfile))
 
                     lfdirstate.add(destlfile)
             lfdirstate.write()
         except util.Abort, e:
             if str(e) != _('no files to copy'):
                 raise e
             else:
                 nolfiles = True
     finally:
         restorematchfn()
         wlock.release()
 
     if nolfiles and nonormalfiles:
         raise util.Abort(_('no files to copy'))
 
     return result
 
 # When the user calls revert, we have to be careful to not revert any
 # changes to other largefiles accidentally. This means we have to keep
 # track of the largefiles that are being reverted so we only pull down
 # the necessary largefiles.
 #
 # Standins are only updated (to match the hash of largefiles) before
 # commits. Update the standins then run the original revert, changing
 # the matcher to hit standins instead of largefiles. Based on the
 # resulting standins update the largefiles.
 def overriderevert(orig, ui, repo, *pats, **opts):
     # Because we put the standins in a bad state (by updating them)
     # and then return them to a correct state we need to lock to
     # prevent others from changing them in their incorrect state.
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         (modified, added, removed, missing, unknown, ignored, clean) = \
             lfutil.lfdirstatestatus(lfdirstate, repo, repo['.'].rev())
         lfdirstate.write()
         for lfile in modified:
             lfutil.updatestandin(repo, lfutil.standin(lfile))
         for lfile in missing:
             if (os.path.exists(repo.wjoin(lfutil.standin(lfile)))):
                 os.unlink(repo.wjoin(lfutil.standin(lfile)))
 
         oldstandins = lfutil.getstandinsstate(repo)
 
         def overridematch(ctx, pats=[], opts={}, globbed=False,
                 default='relpath'):
             match = oldmatch(ctx, pats, opts, globbed, default)
             m = copy.copy(match)
             def tostandin(f):
                 if lfutil.standin(f) in ctx:
                     return lfutil.standin(f)
                 elif lfutil.standin(f) in repo[None]:
                     return None
                 return f
             m._files = [tostandin(f) for f in m._files]
             m._files = [f for f in m._files if f is not None]
             m._fmap = set(m._files)
             m._always = False
             origmatchfn = m.matchfn
             def matchfn(f):
                 if lfutil.isstandin(f):
                     return (origmatchfn(lfutil.splitstandin(f)) and
                             (f in repo[None] or f in ctx))
                 return origmatchfn(f)
             m.matchfn = matchfn
             return m
         oldmatch = installmatchfn(overridematch)
         try:
             orig(ui, repo, *pats, **opts)
         finally:
             restorematchfn()
 
         newstandins = lfutil.getstandinsstate(repo)
         filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
-        lfcommands.updatelfiles(ui, repo, filelist, printmessage=False)
+        # lfdirstate should be 'normallookup'-ed for updated files,
+        # because reverting doesn't touch dirstate for 'normal' files
+        # when target revision is explicitly specified: in such case,
+        # 'n' and valid timestamp in dirstate doesn't ensure 'clean'
+        # of target (standin) file.
+        lfcommands.updatelfiles(ui, repo, filelist, printmessage=False,
+                                normallookup=True)
 
     finally:
         wlock.release()
 
 def hgupdaterepo(orig, repo, node, overwrite):
     if not overwrite:
         # Only call updatelfiles on the standins that have changed to save time
         oldstandins = lfutil.getstandinsstate(repo)
 
     result = orig(repo, node, overwrite)
 
     filelist = None
     if not overwrite:
         newstandins = lfutil.getstandinsstate(repo)
         filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
     lfcommands.updatelfiles(repo.ui, repo, filelist=filelist)
     return result
 
 def hgmerge(orig, repo, node, force=None, remind=True):
     result = orig(repo, node, force, remind)
     lfcommands.updatelfiles(repo.ui, repo)
     return result
 
 # When we rebase a repository with remotely changed largefiles, we need to
 # take some extra care so that the largefiles are correctly updated in the
 # working copy
 def overridepull(orig, ui, repo, source=None, **opts):
     revsprepull = len(repo)
     if not source:
         source = 'default'
     repo.lfpullsource = source
     if opts.get('rebase', False):
         repo._isrebasing = True
         try:
             if opts.get('update'):
                 del opts['update']
                 ui.debug('--update and --rebase are not compatible, ignoring '
                          'the update flag\n')
             del opts['rebase']
             origpostincoming = commands.postincoming
             def _dummy(*args, **kwargs):
                 pass
             commands.postincoming = _dummy
             try:
                 result = commands.pull(ui, repo, source, **opts)
             finally:
                 commands.postincoming = origpostincoming
             revspostpull = len(repo)
             if revspostpull > revsprepull:
                 result = result or rebase.rebase(ui, repo)
         finally:
             repo._isrebasing = False
     else:
         result = orig(ui, repo, source, **opts)
     revspostpull = len(repo)
     lfrevs = opts.get('lfrev', [])
     if opts.get('all_largefiles'):
         lfrevs.append('pulled()')
     if lfrevs and revspostpull > revsprepull:
         numcached = 0
         repo.firstpulled = revsprepull # for pulled() revset expression
         try:
             for rev in scmutil.revrange(repo, lfrevs):
                 ui.note(_('pulling largefiles for revision %s\n') % rev)
                 (cached, missing) = lfcommands.cachelfiles(ui, repo, rev)
                 numcached += len(cached)
         finally:
             del repo.firstpulled
         ui.status(_("%d largefiles cached\n") % numcached)
     return result
 
 def pulledrevsetsymbol(repo, subset, x):
     """``pulled()``
     Changesets that just has been pulled.
 
     Only available with largefiles from pull --lfrev expressions.
 
     .. container:: verbose
 
       Some examples:
 
       - pull largefiles for all new changesets::
 
           hg pull -lfrev "pulled()"
 
       - pull largefiles for all new branch heads::
 
           hg pull -lfrev "head(pulled()) and not closed()"
 
     """
 
     try:
         firstpulled = repo.firstpulled
     except AttributeError:
         raise util.Abort(_("pulled() only available in --lfrev"))
     return revset.baseset([r for r in subset if r >= firstpulled])
 
 def overrideclone(orig, ui, source, dest=None, **opts):
     d = dest
     if d is None:
         d = hg.defaultdest(source)
     if opts.get('all_largefiles') and not hg.islocal(d):
             raise util.Abort(_(
             '--all-largefiles is incompatible with non-local destination %s') %
             d)
 
     return orig(ui, source, dest, **opts)
 
 def hgclone(orig, ui, opts, *args, **kwargs):
     result = orig(ui, opts, *args, **kwargs)
 
     if result is not None:
         sourcerepo, destrepo = result
         repo = destrepo.local()
 
         # Caching is implicitly limited to 'rev' option, since the dest repo was
         # truncated at that point.  The user may expect a download count with
         # this option, so attempt whether or not this is a largefile repo.
         if opts.get('all_largefiles'):
             success, missing = lfcommands.downloadlfiles(ui, repo, None)
 
             if missing != 0:
                 return None
 
     return result
 
 def overriderebase(orig, ui, repo, **opts):
     repo._isrebasing = True
     try:
         return orig(ui, repo, **opts)
     finally:
         repo._isrebasing = False
 
 def overridearchive(orig, repo, dest, node, kind, decode=True, matchfn=None,
             prefix=None, mtime=None, subrepos=None):
     # No need to lock because we are only reading history and
     # largefile caches, neither of which are modified.
     lfcommands.cachelfiles(repo.ui, repo, node)
 
     if kind not in archival.archivers:
         raise util.Abort(_("unknown archive type '%s'") % kind)
 
     ctx = repo[node]
 
     if kind == 'files':
         if prefix:
             raise util.Abort(
                 _('cannot give prefix when archiving to files'))
     else:
         prefix = archival.tidyprefix(dest, kind, prefix)
 
     def write(name, mode, islink, getdata):
         if matchfn and not matchfn(name):
             return
         data = getdata()
         if decode:
             data = repo.wwritedata(name, data)
         archiver.addfile(prefix + name, mode, islink, data)
 
     archiver = archival.archivers[kind](dest, mtime or ctx.date()[0])
 
     if repo.ui.configbool("ui", "archivemeta", True):
         def metadata():
             base = 'repo: %s\nnode: %s\nbranch: %s\n' % (
                 hex(repo.changelog.node(0)), hex(node), ctx.branch())
 
             tags = ''.join('tag: %s\n' % t for t in ctx.tags()
                            if repo.tagtype(t) == 'global')
             if not tags:
                 repo.ui.pushbuffer()
                 opts = {'template': '{latesttag}\n{latesttagdistance}',
                         'style': '', 'patch': None, 'git': None}
                 cmdutil.show_changeset(repo.ui, repo, opts).show(ctx)
                 ltags, dist = repo.ui.popbuffer().split('\n')
                 tags = ''.join('latesttag: %s\n' % t for t in ltags.split(':'))
                 tags += 'latesttagdistance: %s\n' % dist
 
             return base + tags
 
         write('.hg_archival.txt', 0644, False, metadata)
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         if lfutil.isstandin(f):
             path = lfutil.findfile(repo, getdata().strip())
             if path is None:
                 raise util.Abort(
                     _('largefile %s not found in repo store or system cache')
                     % lfutil.splitstandin(f))
             f = lfutil.splitstandin(f)
 
             def getdatafn():
                 fd = None
                 try:
                     fd = open(path, 'rb')
                     return fd.read()
                 finally:
                     if fd:
                         fd.close()
 
             getdata = getdatafn
         write(f, 'x' in ff and 0755 or 0644, 'l' in ff, getdata)
 
     if subrepos:
         for subpath in sorted(ctx.substate):
             sub = ctx.sub(subpath)
             submatch = match_.narrowmatcher(subpath, matchfn)
             sub.archive(repo.ui, archiver, prefix, submatch)
 
     archiver.done()
 
 def hgsubrepoarchive(orig, repo, ui, archiver, prefix, match=None):
     repo._get(repo._state + ('hg',))
     rev = repo._state[1]
     ctx = repo._repo[rev]
 
     lfcommands.cachelfiles(ui, repo._repo, ctx.node())
 
     def write(name, mode, islink, getdata):
         # At this point, the standin has been replaced with the largefile name,
         # so the normal matcher works here without the lfutil variants.
         if match and not match(f):
             return
         data = getdata()
 
         archiver.addfile(prefix + repo._path + '/' + name, mode, islink, data)
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         if lfutil.isstandin(f):
             path = lfutil.findfile(repo._repo, getdata().strip())
             if path is None:
                 raise util.Abort(
                     _('largefile %s not found in repo store or system cache')
                     % lfutil.splitstandin(f))
             f = lfutil.splitstandin(f)
 
             def getdatafn():
                 fd = None
                 try:
                     fd = open(os.path.join(prefix, path), 'rb')
                     return fd.read()
                 finally:
                     if fd:
                         fd.close()
 
             getdata = getdatafn
 
         write(f, 'x' in ff and 0755 or 0644, 'l' in ff, getdata)
 
     for subpath in sorted(ctx.substate):
         sub = ctx.sub(subpath)
         submatch = match_.narrowmatcher(subpath, match)
         sub.archive(ui, archiver, os.path.join(prefix, repo._path) + '/',
                     submatch)
 
 # If a largefile is modified, the change is not reflected in its
 # standin until a commit. cmdutil.bailifchanged() raises an exception
 # if the repo has uncommitted changes. Wrap it to also check if
 # largefiles were changed. This is used by bisect and backout.
 def overridebailifchanged(orig, repo):
     orig(repo)
     repo.lfstatus = True
     modified, added, removed, deleted = repo.status()[:4]
     repo.lfstatus = False
     if modified or added or removed or deleted:
         raise util.Abort(_('uncommitted changes'))
 
 # Fetch doesn't use cmdutil.bailifchanged so override it to add the check
 def overridefetch(orig, ui, repo, *pats, **opts):
     repo.lfstatus = True
     modified, added, removed, deleted = repo.status()[:4]
     repo.lfstatus = False
     if modified or added or removed or deleted:
         raise util.Abort(_('uncommitted changes'))
     return orig(ui, repo, *pats, **opts)
 
 def overrideforget(orig, ui, repo, *pats, **opts):
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(ui, repo, *pats, **opts)
     restorematchfn()
     m = scmutil.match(repo[None], pats, opts)
 
     try:
         repo.lfstatus = True
         s = repo.status(match=m, clean=True)
     finally:
         repo.lfstatus = False
     forget = sorted(s[0] + s[1] + s[3] + s[6])
     forget = [f for f in forget if lfutil.standin(f) in repo[None].manifest()]
 
     for f in forget:
         if lfutil.standin(f) not in repo.dirstate and not \
                 os.path.isdir(m.rel(lfutil.standin(f))):
             ui.warn(_('not removing %s: file is already untracked\n')
                     % m.rel(f))
             result = 1
 
     for f in forget:
         if ui.verbose or not m.exact(f):
             ui.status(_('removing %s\n') % m.rel(f))
 
     # Need to lock because standin files are deleted then removed from the
     # repository and we could race in-between.
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for f in forget:
             if lfdirstate[f] == 'a':
                 lfdirstate.drop(f)
             else:
                 lfdirstate.remove(f)
         lfdirstate.write()
         standins = [lfutil.standin(f) for f in forget]
         for f in standins:
             util.unlinkpath(repo.wjoin(f), ignoremissing=True)
         repo[None].forget(standins)
     finally:
         wlock.release()
 
     return result
 
 def _getoutgoings(repo, other, missing, addfunc):
     """get pairs of filename and largefile hash in outgoing revisions
     in 'missing'.
 
     largefiles already existing on 'other' repository are ignored.
 
     'addfunc' is invoked with each unique pairs of filename and
     largefile hash value.
     """
     knowns = set()
     lfhashes = set()
     def dedup(fn, lfhash):
         k = (fn, lfhash)
         if k not in knowns:
             knowns.add(k)
             lfhashes.add(lfhash)
     lfutil.getlfilestoupload(repo, missing, dedup)
     if lfhashes:
         lfexists = basestore._openstore(repo, other).exists(lfhashes)
         for fn, lfhash in knowns:
             if not lfexists[lfhash]: # lfhash doesn't exist on "other"
                 addfunc(fn, lfhash)
 
 def outgoinghook(ui, repo, other, opts, missing):
     if opts.pop('large', None):
         lfhashes = set()
         if ui.debugflag:
             toupload = {}
             def addfunc(fn, lfhash):
                 if fn not in toupload:
                     toupload[fn] = []
                 toupload[fn].append(lfhash)
                 lfhashes.add(lfhash)
             def showhashes(fn):
                 for lfhash in sorted(toupload[fn]):
                     ui.debug('    %s\n' % (lfhash))
         else:
             toupload = set()
             def addfunc(fn, lfhash):
                 toupload.add(fn)
                 lfhashes.add(lfhash)
             def showhashes(fn):
                 pass
         _getoutgoings(repo, other, missing, addfunc)
 
         if not toupload:
             ui.status(_('largefiles: no files to upload\n'))
         else:
             ui.status(_('largefiles to upload (%d entities):\n')
                       % (len(lfhashes)))
             for file in sorted(toupload):
                 ui.status(lfutil.splitstandin(file) + '\n')
                 showhashes(file)
             ui.status('\n')
 
 def summaryremotehook(ui, repo, opts, changes):
     largeopt = opts.get('large', False)
     if changes is None:
         if largeopt:
             return (False, True) # only outgoing check is needed
         else:
             return (False, False)
     elif largeopt:
         url, branch, peer, outgoing = changes[1]
         if peer is None:
             # i18n: column positioning for "hg summary"
             ui.status(_('largefiles: (no remote repo)\n'))
             return
 
         toupload = set()
         lfhashes = set()
         def addfunc(fn, lfhash):
             toupload.add(fn)
             lfhashes.add(lfhash)
         _getoutgoings(repo, peer, outgoing.missing, addfunc)
 
         if not toupload:
             # i18n: column positioning for "hg summary"
             ui.status(_('largefiles: (no files to upload)\n'))
         else:
             # i18n: column positioning for "hg summary"
             ui.status(_('largefiles: %d entities for %d files to upload\n')
                       % (len(lfhashes), len(toupload)))
 
 def overridesummary(orig, ui, repo, *pats, **opts):
     try:
         repo.lfstatus = True
         orig(ui, repo, *pats, **opts)
     finally:
         repo.lfstatus = False
 
 def scmutiladdremove(orig, repo, pats=[], opts={}, dry_run=None,
                      similarity=None):
     if not lfutil.islfilesrepo(repo):
         return orig(repo, pats, opts, dry_run, similarity)
     # Get the list of missing largefiles so we can remove them
     lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
     s = lfdirstate.status(match_.always(repo.root, repo.getcwd()), [], False,
         False, False)
     (unsure, modified, added, removed, missing, unknown, ignored, clean) = s
 
     # Call into the normal remove code, but the removing of the standin, we want
     # to have handled by original addremove.  Monkey patching here makes sure
     # we don't remove the standin in the largefiles code, preventing a very
     # confused state later.
     if missing:
         m = [repo.wjoin(f) for f in missing]
         repo._isaddremove = True
         removelargefiles(repo.ui, repo, *m, **opts)
         repo._isaddremove = False
     # Call into the normal add code, and any files that *should* be added as
     # largefiles will be
     addlargefiles(repo.ui, repo, *pats, **opts)
     # Now that we've handled largefiles, hand off to the original addremove
     # function to take care of the rest.  Make sure it doesn't do anything with
     # largefiles by installing a matcher that will ignore them.
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(repo, pats, opts, dry_run, similarity)
     restorematchfn()
     return result
 
 # Calling purge with --all will cause the largefiles to be deleted.
 # Override repo.status to prevent this from happening.
 def overridepurge(orig, ui, repo, *dirs, **opts):
     # XXX large file status is buggy when used on repo proxy.
     # XXX this needs to be investigate.
     repo = repo.unfiltered()
     oldstatus = repo.status
     def overridestatus(node1='.', node2=None, match=None, ignored=False,
                         clean=False, unknown=False, listsubrepos=False):
         r = oldstatus(node1, node2, match, ignored, clean, unknown,
                       listsubrepos)
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         modified, added, removed, deleted, unknown, ignored, clean = r
         unknown = [f for f in unknown if lfdirstate[f] == '?']
         ignored = [f for f in ignored if lfdirstate[f] == '?']
         return modified, added, removed, deleted, unknown, ignored, clean
     repo.status = overridestatus
     orig(ui, repo, *dirs, **opts)
     repo.status = oldstatus
 
 def overriderollback(orig, ui, repo, **opts):
     result = orig(ui, repo, **opts)
     merge.update(repo, node=None, branchmerge=False, force=True,
         partial=lfutil.isstandin)
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         lfiles = lfutil.listlfiles(repo)
         oldlfiles = lfutil.listlfiles(repo, repo[None].parents()[0].rev())
         for file in lfiles:
             if file in oldlfiles:
                 lfdirstate.normallookup(file)
             else:
                 lfdirstate.add(file)
         lfdirstate.write()
     finally:
         wlock.release()
     return result
 
 def overridetransplant(orig, ui, repo, *revs, **opts):
     try:
         oldstandins = lfutil.getstandinsstate(repo)
         repo._istransplanting = True
         result = orig(ui, repo, *revs, **opts)
         newstandins = lfutil.getstandinsstate(repo)
         filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
         lfcommands.updatelfiles(repo.ui, repo, filelist=filelist,
                                 printmessage=True)
     finally:
         repo._istransplanting = False
     return result
 
 def overridecat(orig, ui, repo, file1, *pats, **opts):
     ctx = scmutil.revsingle(repo, opts.get('rev'))
     err = 1
     notbad = set()
     m = scmutil.match(ctx, (file1,) + pats, opts)
     origmatchfn = m.matchfn
     def lfmatchfn(f):
         if origmatchfn(f):
             return True
         lf = lfutil.splitstandin(f)
         if lf is None:
             return False
         notbad.add(lf)
         return origmatchfn(lf)
     m.matchfn = lfmatchfn
     origbadfn = m.bad
     def lfbadfn(f, msg):
         if not f in notbad:
             origbadfn(f, msg)
     m.bad = lfbadfn
     for f in ctx.walk(m):
         fp = cmdutil.makefileobj(repo, opts.get('output'), ctx.node(),
                                  pathname=f)
         lf = lfutil.splitstandin(f)
         if lf is None or origmatchfn(f):
             # duplicating unreachable code from commands.cat
             data = ctx[f].data()
             if opts.get('decode'):
                 data = repo.wwritedata(f, data)
             fp.write(data)
         else:
             hash = lfutil.readstandin(repo, lf, ctx.rev())
             if not lfutil.inusercache(repo.ui, hash):
                 store = basestore._openstore(repo)
                 success, missing = store.get([(lf, hash)])
                 if len(success) != 1:
                     raise util.Abort(
                         _('largefile %s is not in cache and could not be '
                           'downloaded')  % lf)
             path = lfutil.usercachepath(repo.ui, hash)
             fpin = open(path, "rb")
             for chunk in util.filechunkiter(fpin, 128 * 1024):
                 fp.write(chunk)
             fpin.close()
         fp.close()
         err = 0
     return err
 
 def mercurialsinkbefore(orig, sink):
     sink.repo._isconverting = True
     orig(sink)
 
 def mercurialsinkafter(orig, sink):
     sink.repo._isconverting = False
     orig(sink)

tests/test-largefiles-update.t

@@ -1,82 +1,102 b''
 This file focuses mainly on updating largefiles in the working
 directory (and ".hg/largefiles/dirstate")
 
   $ cat >> $HGRCPATH <<EOF
   > [ui]
   > merge = internal:fail
   > [extensions]
   > largefiles =
   > EOF
 
   $ hg init repo
   $ cd repo
 
   $ echo large1 > large1
   $ echo large2 > large2
   $ hg add --large large1 large2
   $ echo normal1 > normal1
   $ hg add normal1
   $ hg commit -m '#0'
   $ echo 'large1 in #1' > large1
   $ echo 'normal1 in #1' > normal1
   $ hg commit -m '#1'
   $ hg update -q -C 0
   $ echo 'large2 in #2' > large2
   $ hg commit -m '#2'
   created new head
 
 Test that "hg merge" updates largefiles from "other" correctly
 
 (getting largefiles from "other" normally)
 
   $ hg status -A large1
   C large1
   $ cat large1
   large1
   $ cat .hglf/large1
   4669e532d5b2c093a78eca010077e708a071bb64
   $ hg merge --config debug.dirstate.delaywrite=2
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   (branch merge, don't forget to commit)
   getting changed largefiles
   1 largefiles updated, 0 removed
   $ hg status -A large1
   M large1
   $ cat large1
   large1 in #1
   $ cat .hglf/large1
   58e24f733a964da346e2407a2bee99d9001184f5
   $ hg diff -c 1 --nodates .hglf/large1 | grep '^[+-][0-9a-z]'
   -4669e532d5b2c093a78eca010077e708a071bb64
   +58e24f733a964da346e2407a2bee99d9001184f5
 
 (getting largefiles from "other" via conflict prompt)
 
   $ hg update -q -C 2
   $ echo 'large1 in #3' > large1
   $ echo 'normal1 in #3' > normal1
   $ hg commit -m '#3'
   $ cat .hglf/large1
   e5bb990443d6a92aaf7223813720f7566c9dd05b
   $ hg merge --config debug.dirstate.delaywrite=2 --config ui.interactive=True <<EOF
   > o
   > EOF
   largefile large1 has a merge conflict
   ancestor was 4669e532d5b2c093a78eca010077e708a071bb64
   keep (l)ocal e5bb990443d6a92aaf7223813720f7566c9dd05b or
   take (o)ther 58e24f733a964da346e2407a2bee99d9001184f5? merging normal1
   warning: conflicts during merge.
   merging normal1 incomplete! (edit conflicts, then use 'hg resolve --mark')
   0 files updated, 1 files merged, 0 files removed, 1 files unresolved
   use 'hg resolve' to retry unresolved file merges or 'hg update -C .' to abandon
   getting changed largefiles
   1 largefiles updated, 0 removed
   [1]
   $ hg status -A large1
   M large1
   $ cat large1
   large1 in #1
   $ cat .hglf/large1
   58e24f733a964da346e2407a2bee99d9001184f5
 
+Test that "hg revert -r REV" updates largefiles from "REV" correctly
+
+  $ hg update -q -C 3
+  $ hg status -A large1
+  C large1
+  $ cat large1
+  large1 in #3
+  $ cat .hglf/large1
+  e5bb990443d6a92aaf7223813720f7566c9dd05b
+  $ hg diff -c 1 --nodates .hglf/large1 | grep '^[+-][0-9a-z]'
+  -4669e532d5b2c093a78eca010077e708a071bb64
+  +58e24f733a964da346e2407a2bee99d9001184f5
+  $ hg revert --no-backup -r 1 --config debug.dirstate.delaywrite=2 large1
+  $ hg status -A large1
+  M large1
+  $ cat large1
+  large1 in #1
+  $ cat .hglf/large1
+  58e24f733a964da346e2407a2bee99d9001184f5
+
   $ cd ..