merge: don't overwrite untracked file at directory rename target

When a directory was renamed and a new untracked file was added in the
new directory and the remote directory added a file by the same name
in the old directory, the local untracked file gets overwritten, as
demonstrated by the broken test case in test-rename-dir-merge.

Fix by checking for unknown files for 'dg' actions too. Since
_checkunknownfile() currently expects the same filename in both
contexts, we need to add a new parameter for the remote filename to
it.

hgext/largefiles/overrides.py

@@ -1,1307 +1,1307 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, util, cmdutil, scmutil, match as match_, \
         archival, pathutil, revset
 from mercurial.i18n import _
 from mercurial.node import hex
 
 import lfutil
 import lfcommands
 import basestore
 
 # -- Utility functions: commonly/repeatedly needed functionality ---------------
 
 def composelargefilematcher(match, manifest):
     '''create a matcher that matches only the largefiles in the original
     matcher'''
     m = copy.copy(match)
     lfile = lambda f: lfutil.standin(f) in manifest
     m._files = filter(lfile, m._files)
     m._fmap = set(m._files)
     m._always = False
     origmatchfn = m.matchfn
     m.matchfn = lambda f: lfile(f) and origmatchfn(f)
     return m
 
 def composenormalfilematcher(match, manifest):
     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)
     return m
 
 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)
         return composenormalfilematcher(match, manifest)
     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 the original matchfn.'''
     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
     installmatchandpatsfn was called. No-op if scmutil.matchandpats
     is its original function.
 
     Note that n calls to installmatchandpatsfn will require n calls
     to restore the original matchfn.'''
     scmutil.matchandpats = getattr(scmutil.matchandpats, 'oldmatchandpats',
             scmutil.matchandpats)
 
 def addlargefiles(ui, repo, matcher, **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 = copy.copy(matcher)
     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 = []
 
     # 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'):
             standins = []
             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, isaddremove, *pats, **opts):
     after = opts.get('after')
     if not pats and not after:
         raise util.Abort(_('no files specified'))
     m = composelargefilematcher(scmutil.match(repo[None], pats, opts),
                                 repo[None].manifest())
     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.modified, s.added,
                                                     s.deleted, s.clean)]
 
     def warn(files, msg):
         for f in files:
             ui.warn(msg % m.rel(f))
         return int(len(files) > 0)
 
     result = 0
 
     if after:
         remove = deleted
         result = warn(modified + added + clean,
                       _('not removing %s: file still exists\n'))
     else:
         remove = 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):
         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 isaddremove:
                     ui.status(_('removing %s\n') % f)
 
             if not opts.get('dry_run'):
                 if not after:
                     util.unlinkpath(repo.wjoin(f), ignoremissing=True)
                 lfdirstate.remove(f)
 
         if opts.get('dry_run'):
             return result
 
         lfdirstate.write()
         remove = [lfutil.standin(f) for f in remove]
         # If this is being called by addremove, let the original addremove
         # function handle this.
         if not isaddremove:
             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)
     matcher = scmutil.match(repo[None], pats, opts)
     bad = addlargefiles(ui, repo, matcher, **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, False, *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)
 
         if m.always():
             # We want to match everything anyway, so there's no benefit trying
             # to add standins.
             return 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
 
     # For hg log --patch, the match object is used in two different senses:
     # (1) to determine what revisions should be printed out, and
     # (2) to determine what files to print out diffs for.
     # The magic matchandpats override should be used for case (1) but not for
     # case (2).
     def overridemakelogfilematcher(repo, pats, opts):
         pctx = repo[None]
         match, pats = oldmatchandpats(pctx, pats, opts)
         return lambda rev: match
 
     oldmatchandpats = installmatchandpatsfn(overridematchandpats)
     oldmakelogfilematcher = cmdutil._makenofollowlogfilematcher
     setattr(cmdutil, '_makenofollowlogfilematcher', overridemakelogfilematcher)
 
     try:
         return orig(ui, repo, *pats, **opts)
     finally:
         restorematchandpatsfn()
         setattr(cmdutil, '_makenofollowlogfilematcher', oldmakelogfilematcher)
 
 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:
         if opts['check']:
             lfdirstate = lfutil.openlfdirstate(ui, repo)
             unsure, s = lfdirstate.status(
                 match_.always(repo.root, repo.getcwd()),
                 [], False, False, False)
 
             mod = len(s.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'))
         return orig(ui, repo, *pats, **opts)
     finally:
         wlock.release()
 
 # Before starting the manifest merge, merge.updates will call
 # _checkunknownfile 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 calculateupdates function below.
-def overridecheckunknownfile(origfn, repo, wctx, mctx, f):
+def overridecheckunknownfile(origfn, repo, wctx, mctx, f, f2=None):
     if lfutil.standin(repo.dirstate.normalize(f)) in wctx:
         return False
-    return origfn(repo, wctx, mctx, f)
+    return origfn(repo, wctx, mctx, f, f2)
 
 # 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 calculateupdates 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, diverge, renamedelete = origfn(
         repo, p1, p2, pas, branchmerge, force, partial, acceptremote,
         followcopies)
 
     if overwrite:
         return actions, diverge, renamedelete
 
     # Convert to dictionary with filename as key and action as value.
     lfiles = set()
     for f in actions:
         splitstandin = f and lfutil.splitstandin(f)
         if splitstandin in p1:
             lfiles.add(splitstandin)
         elif lfutil.standin(f) in p1:
             lfiles.add(f)
 
     for lfile in lfiles:
         standin = lfutil.standin(lfile)
         (lm, largs, lmsg) = actions.get(lfile, (None, None, None))
         (sm, sargs, smsg) = actions.get(standin, (None, None, None))
         if sm in ('g', 'dc') and lm != 'r':
             # Case 1: normal file in the working copy, largefile in
             # the second parent
             usermsg = _('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(usermsg, 0) == 0: # pick remote largefile
                 actions[lfile] = ('r', None, 'replaced by standin')
                 actions[standin] = ('g', sargs, 'replaces standin')
             else: # keep local normal file
                 actions[lfile] = ('k', None, 'replaces standin')
                 if branchmerge:
                     actions[standin] = ('k', None, 'replaced by non-standin')
                 else:
                     actions[standin] = ('r', None, 'replaced by non-standin')
         elif lm in ('g', 'dc') and sm != 'r':
             # Case 2: largefile in the working copy, normal file in
             # the second parent
             usermsg = _('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(usermsg, 0) == 0: # keep local largefile
                 if branchmerge:
                     # largefile can be restored from standin safely
                     actions[lfile] = ('k', None, 'replaced by standin')
                     actions[standin] = ('k', None, 'replaces standin')
                 else:
                     # "lfile" should be marked as "removed" without
                     # removal of itself
                     actions[lfile] = ('lfmr', None,
                                       'forget non-standin largefile')
 
                     # linear-merge should treat this largefile as 're-added'
                     actions[standin] = ('a', None, 'keep standin')
             else: # pick remote normal file
                 actions[lfile] = ('g', largs, 'replaces standin')
                 actions[standin] = ('r', None, 'replaced by non-standin')
 
     return actions, diverge, renamedelete
 
 def mergerecordupdates(orig, repo, actions, branchmerge):
     if 'lfmr' in actions:
         # this should be executed before 'orig', to execute 'remove'
         # before all other actions
         for lfile, args, msg in actions['lfmr']:
             repo.dirstate.remove(lfile)
 
     return orig(repo, actions, branchmerge)
 
 
 # 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)
                 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)
         s = lfutil.lfdirstatestatus(lfdirstate, repo)
         lfdirstate.write()
         for lfile in s.modified:
             lfutil.updatestandin(repo, lfutil.standin(lfile))
         for lfile in s.deleted:
             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)
             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)
         # 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()
 
 # after pulling changesets, we need to take some extra care to get
 # largefiles updated remotely
 def overridepull(orig, ui, repo, source=None, **opts):
     revsprepull = len(repo)
     if not source:
         source = 'default'
     repo.lfpullsource = source
     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):
     resuming = opts.get('continue')
     repo._lfcommithooks.append(lfutil.automatedcommithook(resuming))
     repo._lfstatuswriters.append(lambda *msg, **opts: None)
     try:
         return orig(ui, repo, **opts)
     finally:
         repo._lfstatuswriters.pop()
         repo._lfcommithooks.pop()
 
 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(archiver, prefix, submatch)
 
     archiver.done()
 
 def hgsubrepoarchive(orig, repo, archiver, prefix, match=None):
     repo._get(repo._state + ('hg',))
     rev = repo._state[1]
     ctx = repo._repo[rev]
 
     lfcommands.cachelfiles(repo.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(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, backout and fetch.
 def overridebailifchanged(orig, repo):
     orig(repo)
     repo.lfstatus = True
     s = repo.status()
     repo.lfstatus = False
     if s.modified or s.added or s.removed or s.deleted:
         raise util.Abort(_('uncommitted changes'))
 
 def overrideforget(orig, ui, repo, *pats, **opts):
     installnormalfilesmatchfn(repo[None].manifest())
     result = orig(ui, repo, *pats, **opts)
     restorematchfn()
     m = composelargefilematcher(scmutil.match(repo[None], pats, opts),
                                 repo[None].manifest())
 
     try:
         repo.lfstatus = True
         s = repo.status(match=m, clean=True)
     finally:
         repo.lfstatus = False
     forget = sorted(s.modified + s.added + s.deleted + s.clean)
     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, matcher, prefix, opts={}, dry_run=None,
                      similarity=None):
     if not lfutil.islfilesrepo(repo):
         return orig(repo, matcher, prefix, opts, dry_run, similarity)
     # Get the list of missing largefiles so we can remove them
     lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
     unsure, s = lfdirstate.status(match_.always(repo.root, repo.getcwd()), [],
                                   False, False, False)
 
     # 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 s.deleted:
         m = [repo.wjoin(f) for f in s.deleted]
         removelargefiles(repo.ui, repo, True, *m, **opts)
     # Call into the normal add code, and any files that *should* be added as
     # largefiles will be
     addlargefiles(repo.ui, repo, matcher, **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 passing a matcher that will ignore them.
     matcher = composenormalfilematcher(matcher, repo[None].manifest())
     return orig(repo, matcher, prefix, opts, dry_run, similarity)
 
 # 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 Monkey patching a repoview will not work. The assigned attribute will
     # be set on the unfiltered repo, but we will only lookup attributes in the
     # unfiltered repo if the lookup in the repoview object itself fails. As the
     # monkey patched method exists on the repoview class the lookup will not
     # fail. As a result, the original version will shadow the monkey patched
     # one, defeating the monkey patch.
     #
     # As a work around we use an unfiltered repo here. We should do something
     # cleaner instead.
     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)
         unknown = [f for f in r.unknown if lfdirstate[f] == '?']
         ignored = [f for f in r.ignored if lfdirstate[f] == '?']
         return scmutil.status(r.modified, r.added, r.removed, r.deleted,
                               unknown, ignored, r.clean)
     repo.status = overridestatus
     orig(ui, repo, *dirs, **opts)
     repo.status = oldstatus
 def overriderollback(orig, ui, repo, **opts):
     wlock = repo.wlock()
     try:
         before = repo.dirstate.parents()
         orphans = set(f for f in repo.dirstate
                       if lfutil.isstandin(f) and repo.dirstate[f] != 'r')
         result = orig(ui, repo, **opts)
         after = repo.dirstate.parents()
         if before == after:
             return result # no need to restore standins
 
         pctx = repo['.']
         for f in repo.dirstate:
             if lfutil.isstandin(f):
                 orphans.discard(f)
                 if repo.dirstate[f] == 'r':
                     repo.wvfs.unlinkpath(f, ignoremissing=True)
                 elif f in pctx:
                     fctx = pctx[f]
                     repo.wwrite(f, fctx.data(), fctx.flags())
                 else:
                     # content of standin is not so important in 'a',
                     # 'm' or 'n' (coming from the 2nd parent) cases
                     lfutil.writestandin(repo, f, '', False)
         for standin in orphans:
             repo.wvfs.unlinkpath(standin, ignoremissing=True)
 
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         orphans = set(lfdirstate)
         lfiles = lfutil.listlfiles(repo)
         for file in lfiles:
             lfutil.synclfdirstate(repo, lfdirstate, file, True)
             orphans.discard(file)
         for lfile in orphans:
             lfdirstate.drop(lfile)
         lfdirstate.write()
     finally:
         wlock.release()
     return result
 
 def overridetransplant(orig, ui, repo, *revs, **opts):
     resuming = opts.get('continue')
     repo._lfcommithooks.append(lfutil.automatedcommithook(resuming))
     repo._lfstatuswriters.append(lambda *msg, **opts: None)
     try:
         result = orig(ui, repo, *revs, **opts)
     finally:
         repo._lfstatuswriters.pop()
         repo._lfcommithooks.pop()
     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 mergeupdate(orig, repo, node, branchmerge, force, partial,
                 *args, **kwargs):
     wlock = repo.wlock()
     try:
         # branch |       |         |
         #  merge | force | partial | action
         # -------+-------+---------+--------------
         #    x   |   x   |    x    | linear-merge
         #    o   |   x   |    x    | branch-merge
         #    x   |   o   |    x    | overwrite (as clean update)
         #    o   |   o   |    x    | force-branch-merge (*1)
         #    x   |   x   |    o    |   (*)
         #    o   |   x   |    o    |   (*)
         #    x   |   o   |    o    | overwrite (as revert)
         #    o   |   o   |    o    |   (*)
         #
         # (*) don't care
         # (*1) deprecated, but used internally (e.g: "rebase --collapse")
 
         linearmerge = not branchmerge and not force and not partial
 
         if linearmerge or (branchmerge and force and not partial):
             # update standins for linear-merge or force-branch-merge,
             # because largefiles in the working directory may be modified
             lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
             unsure, s = lfdirstate.status(match_.always(repo.root,
                                                         repo.getcwd()),
                                           [], False, False, False)
             for lfile in unsure + s.modified + s.added:
                 lfutil.updatestandin(repo, lfutil.standin(lfile))
 
         if linearmerge:
             # Only call updatelfiles on the standins that have changed
             # to save time
             oldstandins = lfutil.getstandinsstate(repo)
 
         result = orig(repo, node, branchmerge, force, partial, *args, **kwargs)
 
         filelist = None
         if linearmerge:
             newstandins = lfutil.getstandinsstate(repo)
             filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
 
         lfcommands.updatelfiles(repo.ui, repo, filelist=filelist,
                                 normallookup=partial)
 
         return result
     finally:
         wlock.release()
 
 def scmutilmarktouched(orig, repo, files, *args, **kwargs):
     result = orig(repo, files, *args, **kwargs)
 
     filelist = [lfutil.splitstandin(f) for f in files if lfutil.isstandin(f)]
     if filelist:
         lfcommands.updatelfiles(repo.ui, repo, filelist=filelist,
                                 printmessage=False, normallookup=True)
 
     return result

mercurial/merge.py

@@ -1,1172 +1,1177 b''
 # merge.py - directory-level update/merge handling for Mercurial
 #
 # Copyright 2006, 2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 import struct
 
 from node import nullid, nullrev, hex, bin
 from i18n import _
 from mercurial import obsolete
 import error as errormod, util, filemerge, copies, subrepo, worker
 import errno, os, shutil
 
 _pack = struct.pack
 _unpack = struct.unpack
 
 def _droponode(data):
     # used for compatibility for v1
     bits = data.split('\0')
     bits = bits[:-2] + bits[-1:]
     return '\0'.join(bits)
 
 class mergestate(object):
     '''track 3-way merge state of individual files
 
     it is stored on disk when needed. Two file are used, one with an old
     format, one with a new format. Both contains similar data, but the new
     format can store new kind of field.
 
     Current new format is a list of arbitrary record of the form:
 
         [type][length][content]
 
     Type is a single character, length is a 4 bytes integer, content is an
     arbitrary suites of bytes of length `length`.
 
     Type should be a letter. Capital letter are mandatory record, Mercurial
     should abort if they are unknown. lower case record can be safely ignored.
 
     Currently known record:
 
     L: the node of the "local" part of the merge (hexified version)
     O: the node of the "other" part of the merge (hexified version)
     F: a file to be merged entry
     '''
     statepathv1 = 'merge/state'
     statepathv2 = 'merge/state2'
 
     def __init__(self, repo):
         self._repo = repo
         self._dirty = False
         self._read()
 
     def reset(self, node=None, other=None):
         self._state = {}
         self._local = None
         self._other = None
         if node:
             self._local = node
             self._other = other
         shutil.rmtree(self._repo.join('merge'), True)
         self._dirty = False
 
     def _read(self):
         """Analyse each record content to restore a serialized state from disk
 
         This function process "record" entry produced by the de-serialization
         of on disk file.
         """
         self._state = {}
         self._local = None
         self._other = None
         records = self._readrecords()
         for rtype, record in records:
             if rtype == 'L':
                 self._local = bin(record)
             elif rtype == 'O':
                 self._other = bin(record)
             elif rtype == 'F':
                 bits = record.split('\0')
                 self._state[bits[0]] = bits[1:]
             elif not rtype.islower():
                 raise util.Abort(_('unsupported merge state record: %s')
                                    % rtype)
         self._dirty = False
 
     def _readrecords(self):
         """Read merge state from disk and return a list of record (TYPE, data)
 
         We read data from both v1 and v2 files and decide which one to use.
 
         V1 has been used by version prior to 2.9.1 and contains less data than
         v2. We read both versions and check if no data in v2 contradicts
         v1. If there is not contradiction we can safely assume that both v1
         and v2 were written at the same time and use the extract data in v2. If
         there is contradiction we ignore v2 content as we assume an old version
         of Mercurial has overwritten the mergestate file and left an old v2
         file around.
 
         returns list of record [(TYPE, data), ...]"""
         v1records = self._readrecordsv1()
         v2records = self._readrecordsv2()
         oldv2 = set() # old format version of v2 record
         for rec in v2records:
             if rec[0] == 'L':
                 oldv2.add(rec)
             elif rec[0] == 'F':
                 # drop the onode data (not contained in v1)
                 oldv2.add(('F', _droponode(rec[1])))
         for rec in v1records:
             if rec not in oldv2:
                 # v1 file is newer than v2 file, use it
                 # we have to infer the "other" changeset of the merge
                 # we cannot do better than that with v1 of the format
                 mctx = self._repo[None].parents()[-1]
                 v1records.append(('O', mctx.hex()))
                 # add place holder "other" file node information
                 # nobody is using it yet so we do no need to fetch the data
                 # if mctx was wrong `mctx[bits[-2]]` may fails.
                 for idx, r in enumerate(v1records):
                     if r[0] == 'F':
                         bits = r[1].split('\0')
                         bits.insert(-2, '')
                         v1records[idx] = (r[0], '\0'.join(bits))
                 return v1records
         else:
             return v2records
 
     def _readrecordsv1(self):
         """read on disk merge state for version 1 file
 
         returns list of record [(TYPE, data), ...]
 
         Note: the "F" data from this file are one entry short
               (no "other file node" entry)
         """
         records = []
         try:
             f = self._repo.opener(self.statepathv1)
             for i, l in enumerate(f):
                 if i == 0:
                     records.append(('L', l[:-1]))
                 else:
                     records.append(('F', l[:-1]))
             f.close()
         except IOError, err:
             if err.errno != errno.ENOENT:
                 raise
         return records
 
     def _readrecordsv2(self):
         """read on disk merge state for version 2 file
 
         returns list of record [(TYPE, data), ...]
         """
         records = []
         try:
             f = self._repo.opener(self.statepathv2)
             data = f.read()
             off = 0
             end = len(data)
             while off < end:
                 rtype = data[off]
                 off += 1
                 length = _unpack('>I', data[off:(off + 4)])[0]
                 off += 4
                 record = data[off:(off + length)]
                 off += length
                 records.append((rtype, record))
             f.close()
         except IOError, err:
             if err.errno != errno.ENOENT:
                 raise
         return records
 
     def active(self):
         """Whether mergestate is active.
 
         Returns True if there appears to be mergestate. This is a rough proxy
         for "is a merge in progress."
         """
         # Check local variables before looking at filesystem for performance
         # reasons.
         return bool(self._local) or bool(self._state) or \
                self._repo.opener.exists(self.statepathv1) or \
                self._repo.opener.exists(self.statepathv2)
 
     def commit(self):
         """Write current state on disk (if necessary)"""
         if self._dirty:
             records = []
             records.append(('L', hex(self._local)))
             records.append(('O', hex(self._other)))
             for d, v in self._state.iteritems():
                 records.append(('F', '\0'.join([d] + v)))
             self._writerecords(records)
             self._dirty = False
 
     def _writerecords(self, records):
         """Write current state on disk (both v1 and v2)"""
         self._writerecordsv1(records)
         self._writerecordsv2(records)
 
     def _writerecordsv1(self, records):
         """Write current state on disk in a version 1 file"""
         f = self._repo.opener(self.statepathv1, 'w')
         irecords = iter(records)
         lrecords = irecords.next()
         assert lrecords[0] == 'L'
         f.write(hex(self._local) + '\n')
         for rtype, data in irecords:
             if rtype == 'F':
                 f.write('%s\n' % _droponode(data))
         f.close()
 
     def _writerecordsv2(self, records):
         """Write current state on disk in a version 2 file"""
         f = self._repo.opener(self.statepathv2, 'w')
         for key, data in records:
             assert len(key) == 1
             format = '>sI%is' % len(data)
             f.write(_pack(format, key, len(data), data))
         f.close()
 
     def add(self, fcl, fco, fca, fd):
         """add a new (potentially?) conflicting file the merge state
         fcl: file context for local,
         fco: file context for remote,
         fca: file context for ancestors,
         fd:  file path of the resulting merge.
 
         note: also write the local version to the `.hg/merge` directory.
         """
         hash = util.sha1(fcl.path()).hexdigest()
         self._repo.opener.write('merge/' + hash, fcl.data())
         self._state[fd] = ['u', hash, fcl.path(),
                            fca.path(), hex(fca.filenode()),
                            fco.path(), hex(fco.filenode()),
                            fcl.flags()]
         self._dirty = True
 
     def __contains__(self, dfile):
         return dfile in self._state
 
     def __getitem__(self, dfile):
         return self._state[dfile][0]
 
     def __iter__(self):
         return iter(sorted(self._state))
 
     def files(self):
         return self._state.keys()
 
     def mark(self, dfile, state):
         self._state[dfile][0] = state
         self._dirty = True
 
     def unresolved(self):
         """Obtain the paths of unresolved files."""
 
         for f, entry in self._state.items():
             if entry[0] == 'u':
                 yield f
 
     def resolve(self, dfile, wctx, labels=None):
         """rerun merge process for file path `dfile`"""
         if self[dfile] == 'r':
             return 0
         stateentry = self._state[dfile]
         state, hash, lfile, afile, anode, ofile, onode, flags = stateentry
         octx = self._repo[self._other]
         fcd = wctx[dfile]
         fco = octx[ofile]
         fca = self._repo.filectx(afile, fileid=anode)
         # "premerge" x flags
         flo = fco.flags()
         fla = fca.flags()
         if 'x' in flags + flo + fla and 'l' not in flags + flo + fla:
             if fca.node() == nullid:
                 self._repo.ui.warn(_('warning: cannot merge flags for %s\n') %
                                    afile)
             elif flags == fla:
                 flags = flo
         # restore local
         f = self._repo.opener('merge/' + hash)
         self._repo.wwrite(dfile, f.read(), flags)
         f.close()
         r = filemerge.filemerge(self._repo, self._local, lfile, fcd, fco, fca,
                                 labels=labels)
         if r is None:
             # no real conflict
             del self._state[dfile]
             self._dirty = True
         elif not r:
             self.mark(dfile, 'r')
         return r
 
-def _checkunknownfile(repo, wctx, mctx, f):
+def _checkunknownfile(repo, wctx, mctx, f, f2=None):
+    if f2 is None:
+        f2 = f
     return (os.path.isfile(repo.wjoin(f))
         and repo.wopener.audit.check(f)
         and repo.dirstate.normalize(f) not in repo.dirstate
-        and mctx[f].cmp(wctx[f]))
+        and mctx[f2].cmp(wctx[f]))
 
 def _forgetremoved(wctx, mctx, branchmerge):
     """
     Forget removed files
 
     If we're jumping between revisions (as opposed to merging), and if
     neither the working directory nor the target rev has the file,
     then we need to remove it from the dirstate, to prevent the
     dirstate from listing the file when it is no longer in the
     manifest.
 
     If we're merging, and the other revision has removed a file
     that is not present in the working directory, we need to mark it
     as removed.
     """
 
     actions = {}
     m = 'f'
     if branchmerge:
         m = 'r'
     for f in wctx.deleted():
         if f not in mctx:
             actions[f] = m, None, "forget deleted"
 
     if not branchmerge:
         for f in wctx.removed():
             if f not in mctx:
                 actions[f] = 'f', None, "forget removed"
 
     return actions
 
 def _checkcollision(repo, wmf, actions):
     # build provisional merged manifest up
     pmmf = set(wmf)
 
     if actions:
         # k, dr, e and rd are no-op
         for m in 'a', 'f', 'g', 'cd', 'dc':
             for f, args, msg in actions[m]:
                 pmmf.add(f)
         for f, args, msg in actions['r']:
             pmmf.discard(f)
         for f, args, msg in actions['dm']:
             f2, flags = args
             pmmf.discard(f2)
             pmmf.add(f)
         for f, args, msg in actions['dg']:
             pmmf.add(f)
         for f, args, msg in actions['m']:
             f1, f2, fa, move, anc = args
             if move:
                 pmmf.discard(f1)
             pmmf.add(f)
 
     # check case-folding collision in provisional merged manifest
     foldmap = {}
     for f in sorted(pmmf):
         fold = util.normcase(f)
         if fold in foldmap:
             raise util.Abort(_("case-folding collision between %s and %s")
                              % (f, foldmap[fold]))
         foldmap[fold] = f
 
 def manifestmerge(repo, wctx, p2, pa, branchmerge, force, partial,
                   acceptremote, followcopies):
     """
     Merge p1 and p2 with ancestor pa and generate merge action list
 
     branchmerge and force are as passed in to update
     partial = function to filter file lists
     acceptremote = accept the incoming changes without prompting
     """
 
     copy, movewithdir, diverge, renamedelete = {}, {}, {}, {}
 
     # manifests fetched in order are going to be faster, so prime the caches
     [x.manifest() for x in
      sorted(wctx.parents() + [p2, pa], key=lambda x: x.rev())]
 
     if followcopies:
         ret = copies.mergecopies(repo, wctx, p2, pa)
         copy, movewithdir, diverge, renamedelete = ret
 
     repo.ui.note(_("resolving manifests\n"))
     repo.ui.debug(" branchmerge: %s, force: %s, partial: %s\n"
                   % (bool(branchmerge), bool(force), bool(partial)))
     repo.ui.debug(" ancestor: %s, local: %s, remote: %s\n" % (pa, wctx, p2))
 
     m1, m2, ma = wctx.manifest(), p2.manifest(), pa.manifest()
     copied = set(copy.values())
     copied.update(movewithdir.values())
 
     if '.hgsubstate' in m1:
         # check whether sub state is modified
         for s in sorted(wctx.substate):
             if wctx.sub(s).dirty():
                 m1['.hgsubstate'] += '+'
                 break
 
     # Compare manifests
     diff = m1.diff(m2)
 
     actions = {}
     for f, ((n1, fl1), (n2, fl2)) in diff.iteritems():
         if partial and not partial(f):
             continue
         if n1 and n2: # file exists on both local and remote side
             if f not in ma:
                 fa = copy.get(f, None)
                 if fa is not None:
                     actions[f] = ('m', (f, f, fa, False, pa.node()),
                                   "both renamed from " + fa)
                 else:
                     actions[f] = ('m', (f, f, None, False, pa.node()),
                                   "both created")
             else:
                 a = ma[f]
                 fla = ma.flags(f)
                 nol = 'l' not in fl1 + fl2 + fla
                 if n2 == a and fl2 == fla:
                     actions[f] = ('k' , (), "remote unchanged")
                 elif n1 == a and fl1 == fla: # local unchanged - use remote
                     if n1 == n2: # optimization: keep local content
                         actions[f] = ('e', (fl2,), "update permissions")
                     else:
                         actions[f] = ('g', (fl2,), "remote is newer")
                 elif nol and n2 == a: # remote only changed 'x'
                     actions[f] = ('e', (fl2,), "update permissions")
                 elif nol and n1 == a: # local only changed 'x'
                     actions[f] = ('g', (fl1,), "remote is newer")
                 else: # both changed something
                     actions[f] = ('m', (f, f, f, False, pa.node()),
                                    "versions differ")
         elif n1: # file exists only on local side
             if f in copied:
                 pass # we'll deal with it on m2 side
             elif f in movewithdir: # directory rename, move local
                 f2 = movewithdir[f]
                 if f2 in m2:
                     actions[f2] = ('m', (f, f2, None, True, pa.node()),
                                    "remote directory rename, both created")
                 else:
                     actions[f2] = ('dm', (f, fl1),
                                    "remote directory rename - move from " + f)
             elif f in copy:
                 f2 = copy[f]
                 actions[f] = ('m', (f, f2, f2, False, pa.node()),
                               "local copied/moved from " + f2)
             elif f in ma: # clean, a different, no remote
                 if n1 != ma[f]:
                     if acceptremote:
                         actions[f] = ('r', None, "remote delete")
                     else:
                         actions[f] = ('cd', None,  "prompt changed/deleted")
                 elif n1[20:] == 'a':
                     # This extra 'a' is added by working copy manifest to mark
                     # the file as locally added. We should forget it instead of
                     # deleting it.
                     actions[f] = ('f', None, "remote deleted")
                 else:
                     actions[f] = ('r', None, "other deleted")
         elif n2: # file exists only on remote side
             if f in copied:
                 pass # we'll deal with it on m1 side
             elif f in movewithdir:
                 f2 = movewithdir[f]
                 if f2 in m1:
                     actions[f2] = ('m', (f2, f, None, False, pa.node()),
                                    "local directory rename, both created")
                 else:
                     actions[f2] = ('dg', (f, fl2),
                                    "local directory rename - get from " + f)
             elif f in copy:
                 f2 = copy[f]
                 if f2 in m2:
                     actions[f] = ('m', (f2, f, f2, False, pa.node()),
                                   "remote copied from " + f2)
                 else:
                     actions[f] = ('m', (f2, f, f2, True, pa.node()),
                                   "remote moved from " + f2)
             elif f not in ma:
                 # local unknown, remote created: the logic is described by the
                 # following table:
                 #
                 # force  branchmerge  different  |  action
                 #   n         *           *      |   create
                 #   y         n           *      |   create
                 #   y         y           n      |   create
                 #   y         y           y      |   merge
                 #
                 # Checking whether the files are different is expensive, so we
                 # don't do that when we can avoid it.
                 if not force:
                     actions[f] = ('c', (fl2,), "remote created")
                 elif not branchmerge:
                     actions[f] = ('c', (fl2,), "remote created")
                 else:
                     different = _checkunknownfile(repo, wctx, p2, f)
                     if different:
                         actions[f] = ('m', (f, f, None, False, pa.node()),
                                       "remote differs from untracked local")
                     else:
                         actions[f] = ('g', (fl2,), "remote created")
             elif n2 != ma[f]:
                 if acceptremote:
                     actions[f] = ('c', (fl2,), "remote recreating")
                 else:
                     actions[f] = ('dc', (fl2,), "prompt deleted/changed")
 
     aborts = []
     if not force:
         for f, (m, args, msg) in actions.iteritems():
             if m in ('c', 'dc'):
                 if _checkunknownfile(repo, wctx, p2, f):
                     aborts.append(f)
+            elif m == 'dg':
+                if _checkunknownfile(repo, wctx, p2, f, args[0]):
+                    aborts.append(f)
 
     for f in sorted(aborts):
         repo.ui.warn(_("%s: untracked file differs\n") % f)
     if aborts:
         raise util.Abort(_("untracked files in working directory differ "
                            "from files in requested revision"))
 
     for f, (m, args, msg) in actions.iteritems():
         if m == 'c':
             actions[f] = ('g', args, msg)
 
     return actions, diverge, renamedelete
 
 def _resolvetrivial(repo, wctx, mctx, ancestor, actions):
     """Resolves false conflicts where the nodeid changed but the content
        remained the same."""
 
     for f, (m, args, msg) in actions.items():
         if m == 'cd' and f in ancestor and not wctx[f].cmp(ancestor[f]):
             # local did change but ended up with same content
             actions[f] = 'r', None, "prompt same"
         elif m == 'dc' and f in ancestor and not mctx[f].cmp(ancestor[f]):
             # remote did change but ended up with same content
             del actions[f] # don't get = keep local deleted
 
 def calculateupdates(repo, wctx, mctx, ancestors, branchmerge, force, partial,
                      acceptremote, followcopies):
     "Calculate the actions needed to merge mctx into wctx using ancestors"
 
     if len(ancestors) == 1: # default
         actions, diverge, renamedelete = manifestmerge(
             repo, wctx, mctx, ancestors[0], branchmerge, force, partial,
             acceptremote, followcopies)
 
     else: # only when merge.preferancestor=* - the default
         repo.ui.note(
             _("note: merging %s and %s using bids from ancestors %s\n") %
             (wctx, mctx, _(' and ').join(str(anc) for anc in ancestors)))
 
         # Call for bids
         fbids = {} # mapping filename to bids (action method to list af actions)
         diverge, renamedelete = None, None
         for ancestor in ancestors:
             repo.ui.note(_('\ncalculating bids for ancestor %s\n') % ancestor)
             actions, diverge1, renamedelete1 = manifestmerge(
                 repo, wctx, mctx, ancestor, branchmerge, force, partial,
                 acceptremote, followcopies)
             if diverge is None: # and renamedelete is None.
                 # Arbitrarily pick warnings from first iteration
                 diverge = diverge1
                 renamedelete = renamedelete1
             for f, a in sorted(actions.iteritems()):
                 m, args, msg = a
                 repo.ui.debug(' %s: %s -> %s\n' % (f, msg, m))
                 if f in fbids:
                     d = fbids[f]
                     if m in d:
                         d[m].append(a)
                     else:
                         d[m] = [a]
                 else:
                     fbids[f] = {m: [a]}
 
         # Pick the best bid for each file
         repo.ui.note(_('\nauction for merging merge bids\n'))
         actions = {}
         for f, bids in sorted(fbids.items()):
             # bids is a mapping from action method to list af actions
             # Consensus?
             if len(bids) == 1: # all bids are the same kind of method
                 m, l = bids.items()[0]
                 if util.all(a == l[0] for a in l[1:]): # len(bids) is > 1
                     repo.ui.note(" %s: consensus for %s\n" % (f, m))
                     actions[f] = l[0]
                     continue
             # If keep is an option, just do it.
             if 'k' in bids:
                 repo.ui.note(" %s: picking 'keep' action\n" % f)
                 actions[f] = bids['k'][0]
                 continue
             # If there are gets and they all agree [how could they not?], do it.
             if 'g' in bids:
                 ga0 = bids['g'][0]
                 if util.all(a == ga0 for a in bids['g'][1:]):
                     repo.ui.note(" %s: picking 'get' action\n" % f)
                     actions[f] = ga0
                     continue
             # TODO: Consider other simple actions such as mode changes
             # Handle inefficient democrazy.
             repo.ui.note(_(' %s: multiple bids for merge action:\n') % f)
             for m, l in sorted(bids.items()):
                 for _f, args, msg in l:
                     repo.ui.note('  %s -> %s\n' % (msg, m))
             # Pick random action. TODO: Instead, prompt user when resolving
             m, l = bids.items()[0]
             repo.ui.warn(_(' %s: ambiguous merge - picked %s action\n') %
                          (f, m))
             actions[f] = l[0]
             continue
         repo.ui.note(_('end of auction\n\n'))
 
     _resolvetrivial(repo, wctx, mctx, ancestors[0], actions)
 
     if wctx.rev() is None:
         fractions = _forgetremoved(wctx, mctx, branchmerge)
         actions.update(fractions)
 
     return actions, diverge, renamedelete
 
 def batchremove(repo, actions):
     """apply removes to the working directory
 
     yields tuples for progress updates
     """
     verbose = repo.ui.verbose
     unlink = util.unlinkpath
     wjoin = repo.wjoin
     audit = repo.wopener.audit
     i = 0
     for f, args, msg in actions:
         repo.ui.debug(" %s: %s -> r\n" % (f, msg))
         if verbose:
             repo.ui.note(_("removing %s\n") % f)
         audit(f)
         try:
             unlink(wjoin(f), ignoremissing=True)
         except OSError, inst:
             repo.ui.warn(_("update failed to remove %s: %s!\n") %
                          (f, inst.strerror))
         if i == 100:
             yield i, f
             i = 0
         i += 1
     if i > 0:
         yield i, f
 
 def batchget(repo, mctx, actions):
     """apply gets to the working directory
 
     mctx is the context to get from
 
     yields tuples for progress updates
     """
     verbose = repo.ui.verbose
     fctx = mctx.filectx
     wwrite = repo.wwrite
     i = 0
     for f, args, msg in actions:
         repo.ui.debug(" %s: %s -> g\n" % (f, msg))
         if verbose:
             repo.ui.note(_("getting %s\n") % f)
         wwrite(f, fctx(f).data(), args[0])
         if i == 100:
             yield i, f
             i = 0
         i += 1
     if i > 0:
         yield i, f
 
 def applyupdates(repo, actions, wctx, mctx, overwrite, labels=None):
     """apply the merge action list to the working directory
 
     wctx is the working copy context
     mctx is the context to be merged into the working copy
 
     Return a tuple of counts (updated, merged, removed, unresolved) that
     describes how many files were affected by the update.
     """
 
     updated, merged, removed, unresolved = 0, 0, 0, 0
     ms = mergestate(repo)
     ms.reset(wctx.p1().node(), mctx.node())
     moves = []
     for m, l in actions.items():
         l.sort()
 
     # prescan for merges
     for f, args, msg in actions['m']:
         f1, f2, fa, move, anc = args
         if f == '.hgsubstate': # merged internally
             continue
         repo.ui.debug(" preserving %s for resolve of %s\n" % (f1, f))
         fcl = wctx[f1]
         fco = mctx[f2]
         actx = repo[anc]
         if fa in actx:
             fca = actx[fa]
         else:
             fca = repo.filectx(f1, fileid=nullrev)
         ms.add(fcl, fco, fca, f)
         if f1 != f and move:
             moves.append(f1)
 
     audit = repo.wopener.audit
     _updating = _('updating')
     _files = _('files')
     progress = repo.ui.progress
 
     # remove renamed files after safely stored
     for f in moves:
         if os.path.lexists(repo.wjoin(f)):
             repo.ui.debug("removing %s\n" % f)
             audit(f)
             util.unlinkpath(repo.wjoin(f))
 
     numupdates = sum(len(l) for m, l in actions.items() if m != 'k')
 
     if [a for a in actions['r'] if a[0] == '.hgsubstate']:
         subrepo.submerge(repo, wctx, mctx, wctx, overwrite)
 
     # remove in parallel (must come first)
     z = 0
     prog = worker.worker(repo.ui, 0.001, batchremove, (repo,), actions['r'])
     for i, item in prog:
         z += i
         progress(_updating, z, item=item, total=numupdates, unit=_files)
     removed = len(actions['r'])
 
     # get in parallel
     prog = worker.worker(repo.ui, 0.001, batchget, (repo, mctx), actions['g'])
     for i, item in prog:
         z += i
         progress(_updating, z, item=item, total=numupdates, unit=_files)
     updated = len(actions['g'])
 
     if [a for a in actions['g'] if a[0] == '.hgsubstate']:
         subrepo.submerge(repo, wctx, mctx, wctx, overwrite)
 
     # forget (manifest only, just log it) (must come first)
     for f, args, msg in actions['f']:
         repo.ui.debug(" %s: %s -> f\n" % (f, msg))
         z += 1
         progress(_updating, z, item=f, total=numupdates, unit=_files)
 
     # re-add (manifest only, just log it)
     for f, args, msg in actions['a']:
         repo.ui.debug(" %s: %s -> a\n" % (f, msg))
         z += 1
         progress(_updating, z, item=f, total=numupdates, unit=_files)
 
     # keep (noop, just log it)
     for f, args, msg in actions['k']:
         repo.ui.debug(" %s: %s -> k\n" % (f, msg))
         # no progress
 
     # merge
     for f, args, msg in actions['m']:
         repo.ui.debug(" %s: %s -> m\n" % (f, msg))
         z += 1
         progress(_updating, z, item=f, total=numupdates, unit=_files)
         if f == '.hgsubstate': # subrepo states need updating
             subrepo.submerge(repo, wctx, mctx, wctx.ancestor(mctx),
                              overwrite)
             continue
         audit(f)
         r = ms.resolve(f, wctx, labels=labels)
         if r is not None and r > 0:
             unresolved += 1
         else:
             if r is None:
                 updated += 1
             else:
                 merged += 1
 
     # directory rename, move local
     for f, args, msg in actions['dm']:
         repo.ui.debug(" %s: %s -> dm\n" % (f, msg))
         z += 1
         progress(_updating, z, item=f, total=numupdates, unit=_files)
         f0, flags = args
         repo.ui.note(_("moving %s to %s\n") % (f0, f))
         audit(f)
         repo.wwrite(f, wctx.filectx(f0).data(), flags)
         util.unlinkpath(repo.wjoin(f0))
         updated += 1
 
     # local directory rename, get
     for f, args, msg in actions['dg']:
         repo.ui.debug(" %s: %s -> dg\n" % (f, msg))
         z += 1
         progress(_updating, z, item=f, total=numupdates, unit=_files)
         f0, flags = args
         repo.ui.note(_("getting %s to %s\n") % (f0, f))
         repo.wwrite(f, mctx.filectx(f0).data(), flags)
         updated += 1
 
     # exec
     for f, args, msg in actions['e']:
         repo.ui.debug(" %s: %s -> e\n" % (f, msg))
         z += 1
         progress(_updating, z, item=f, total=numupdates, unit=_files)
         flags, = args
         audit(f)
         util.setflags(repo.wjoin(f), 'l' in flags, 'x' in flags)
         updated += 1
 
     ms.commit()
     progress(_updating, None, total=numupdates, unit=_files)
 
     return updated, merged, removed, unresolved
 
 def recordupdates(repo, actions, branchmerge):
     "record merge actions to the dirstate"
     # remove (must come first)
     for f, args, msg in actions['r']:
         if branchmerge:
             repo.dirstate.remove(f)
         else:
             repo.dirstate.drop(f)
 
     # forget (must come first)
     for f, args, msg in actions['f']:
         repo.dirstate.drop(f)
 
     # re-add
     for f, args, msg in actions['a']:
         if not branchmerge:
             repo.dirstate.add(f)
 
     # exec change
     for f, args, msg in actions['e']:
         repo.dirstate.normallookup(f)
 
     # keep
     for f, args, msg in actions['k']:
         pass
 
     # get
     for f, args, msg in actions['g']:
         if branchmerge:
             repo.dirstate.otherparent(f)
         else:
             repo.dirstate.normal(f)
 
     # merge
     for f, args, msg in actions['m']:
         f1, f2, fa, move, anc = args
         if branchmerge:
             # We've done a branch merge, mark this file as merged
             # so that we properly record the merger later
             repo.dirstate.merge(f)
             if f1 != f2: # copy/rename
                 if move:
                     repo.dirstate.remove(f1)
                 if f1 != f:
                     repo.dirstate.copy(f1, f)
                 else:
                     repo.dirstate.copy(f2, f)
         else:
             # We've update-merged a locally modified file, so
             # we set the dirstate to emulate a normal checkout
             # of that file some time in the past. Thus our
             # merge will appear as a normal local file
             # modification.
             if f2 == f: # file not locally copied/moved
                 repo.dirstate.normallookup(f)
             if move:
                 repo.dirstate.drop(f1)
 
     # directory rename, move local
     for f, args, msg in actions['dm']:
         f0, flag = args
         if branchmerge:
             repo.dirstate.add(f)
             repo.dirstate.remove(f0)
             repo.dirstate.copy(f0, f)
         else:
             repo.dirstate.normal(f)
             repo.dirstate.drop(f0)
 
     # directory rename, get
     for f, args, msg in actions['dg']:
         f0, flag = args
         if branchmerge:
             repo.dirstate.add(f)
             repo.dirstate.copy(f0, f)
         else:
             repo.dirstate.normal(f)
 
 def update(repo, node, branchmerge, force, partial, ancestor=None,
            mergeancestor=False, labels=None):
     """
     Perform a merge between the working directory and the given node
 
     node = the node to update to, or None if unspecified
     branchmerge = whether to merge between branches
     force = whether to force branch merging or file overwriting
     partial = a function to filter file lists (dirstate not updated)
     mergeancestor = whether it is merging with an ancestor. If true,
       we should accept the incoming changes for any prompts that occur.
       If false, merging with an ancestor (fast-forward) is only allowed
       between different named branches. This flag is used by rebase extension
       as a temporary fix and should be avoided in general.
 
     The table below shows all the behaviors of the update command
     given the -c and -C or no options, whether the working directory
     is dirty, whether a revision is specified, and the relationship of
     the parent rev to the target rev (linear, on the same named
     branch, or on another named branch).
 
     This logic is tested by test-update-branches.t.
 
     -c  -C  dirty  rev  |  linear   same  cross
      n   n    n     n   |    ok     (1)     x
      n   n    n     y   |    ok     ok     ok
      n   n    y     n   |   merge   (2)    (2)
      n   n    y     y   |   merge   (3)    (3)
      n   y    *     *   |    ---  discard  ---
      y   n    y     *   |    ---    (4)    ---
      y   n    n     *   |    ---    ok     ---
      y   y    *     *   |    ---    (5)    ---
 
     x = can't happen
     * = don't-care
     1 = abort: not a linear update (merge or update --check to force update)
     2 = abort: uncommitted changes (commit and merge, or update --clean to
                  discard changes)
     3 = abort: uncommitted changes (commit or update --clean to discard changes)
     4 = abort: uncommitted changes (checked in commands.py)
     5 = incompatible options (checked in commands.py)
 
     Return the same tuple as applyupdates().
     """
 
     onode = node
     wlock = repo.wlock()
     try:
         wc = repo[None]
         pl = wc.parents()
         p1 = pl[0]
         pas = [None]
         if ancestor is not None:
             pas = [repo[ancestor]]
 
         if node is None:
             # Here is where we should consider bookmarks, divergent bookmarks,
             # foreground changesets (successors), and tip of current branch;
             # but currently we are only checking the branch tips.
             try:
                 node = repo.branchtip(wc.branch())
             except errormod.RepoLookupError:
                 if wc.branch() == 'default': # no default branch!
                     node = repo.lookup('tip') # update to tip
                 else:
                     raise util.Abort(_("branch %s not found") % wc.branch())
 
             if p1.obsolete() and not p1.children():
                 # allow updating to successors
                 successors = obsolete.successorssets(repo, p1.node())
 
                 # behavior of certain cases is as follows,
                 #
                 # divergent changesets: update to highest rev, similar to what
                 #     is currently done when there are more than one head
                 #     (i.e. 'tip')
                 #
                 # replaced changesets: same as divergent except we know there
                 # is no conflict
                 #
                 # pruned changeset: no update is done; though, we could
                 #     consider updating to the first non-obsolete parent,
                 #     similar to what is current done for 'hg prune'
 
                 if successors:
                     # flatten the list here handles both divergent (len > 1)
                     # and the usual case (len = 1)
                     successors = [n for sub in successors for n in sub]
 
                     # get the max revision for the given successors set,
                     # i.e. the 'tip' of a set
                     node = repo.revs('max(%ln)', successors).first()
                     pas = [p1]
 
         overwrite = force and not branchmerge
 
         p2 = repo[node]
         if pas[0] is None:
             if repo.ui.config('merge', 'preferancestor', '*') == '*':
                 cahs = repo.changelog.commonancestorsheads(p1.node(), p2.node())
                 pas = [repo[anc] for anc in (sorted(cahs) or [nullid])]
             else:
                 pas = [p1.ancestor(p2, warn=branchmerge)]
 
         fp1, fp2, xp1, xp2 = p1.node(), p2.node(), str(p1), str(p2)
 
         ### check phase
         if not overwrite and len(pl) > 1:
             raise util.Abort(_("outstanding uncommitted merge"))
         if branchmerge:
             if pas == [p2]:
                 raise util.Abort(_("merging with a working directory ancestor"
                                    " has no effect"))
             elif pas == [p1]:
                 if not mergeancestor and p1.branch() == p2.branch():
                     raise util.Abort(_("nothing to merge"),
                                      hint=_("use 'hg update' "
                                             "or check 'hg heads'"))
             if not force and (wc.files() or wc.deleted()):
                 raise util.Abort(_("uncommitted changes"),
                                  hint=_("use 'hg status' to list changes"))
             for s in sorted(wc.substate):
                 if wc.sub(s).dirty():
                     raise util.Abort(_("uncommitted changes in "
                                        "subrepository '%s'") % s)
 
         elif not overwrite:
             if p1 == p2: # no-op update
                 # call the hooks and exit early
                 repo.hook('preupdate', throw=True, parent1=xp2, parent2='')
                 repo.hook('update', parent1=xp2, parent2='', error=0)
                 return 0, 0, 0, 0
 
             if pas not in ([p1], [p2]):  # nonlinear
                 dirty = wc.dirty(missing=True)
                 if dirty or onode is None:
                     # Branching is a bit strange to ensure we do the minimal
                     # amount of call to obsolete.background.
                     foreground = obsolete.foreground(repo, [p1.node()])
                     # note: the <node> variable contains a random identifier
                     if repo[node].node() in foreground:
                         pas = [p1]  # allow updating to successors
                     elif dirty:
                         msg = _("uncommitted changes")
                         if onode is None:
                             hint = _("commit and merge, or update --clean to"
                                      " discard changes")
                         else:
                             hint = _("commit or update --clean to discard"
                                      " changes")
                         raise util.Abort(msg, hint=hint)
                     else:  # node is none
                         msg = _("not a linear update")
                         hint = _("merge or update --check to force update")
                         raise util.Abort(msg, hint=hint)
                 else:
                     # Allow jumping branches if clean and specific rev given
                     pas = [p1]
 
         followcopies = False
         if overwrite:
             pas = [wc]
         elif pas == [p2]: # backwards
             pas = [wc.p1()]
         elif not branchmerge and not wc.dirty(missing=True):
             pass
         elif pas[0] and repo.ui.configbool('merge', 'followcopies', True):
             followcopies = True
 
         ### calculate phase
         actionbyfile, diverge, renamedelete = calculateupdates(
             repo, wc, p2, pas, branchmerge, force, partial, mergeancestor,
             followcopies)
         # Convert to dictionary-of-lists format
         actions = dict((m, []) for m in 'a f g cd dc r dm dg m e k'.split())
         for f, (m, args, msg) in actionbyfile.iteritems():
             if m not in actions:
                 actions[m] = []
             actions[m].append((f, args, msg))
 
         if not util.checkcase(repo.path):
             # check collision between files only in p2 for clean update
             if (not branchmerge and
                 (force or not wc.dirty(missing=True, branch=False))):
                 _checkcollision(repo, p2.manifest(), None)
             else:
                 _checkcollision(repo, wc.manifest(), actions)
 
         # Prompt and create actions. TODO: Move this towards resolve phase.
         for f, args, msg in sorted(actions['cd']):
             if repo.ui.promptchoice(
                 _("local changed %s which remote deleted\n"
                   "use (c)hanged version or (d)elete?"
                   "$$ &Changed $$ &Delete") % f, 0):
                 actions['r'].append((f, None, "prompt delete"))
             else:
                 actions['a'].append((f, None, "prompt keep"))
         del actions['cd'][:]
 
         for f, args, msg in sorted(actions['dc']):
             flags, = args
             if repo.ui.promptchoice(
                 _("remote changed %s which local deleted\n"
                   "use (c)hanged version or leave (d)eleted?"
                   "$$ &Changed $$ &Deleted") % f, 0) == 0:
                 actions['g'].append((f, (flags,), "prompt recreating"))
         del actions['dc'][:]
 
         ### apply phase
         if not branchmerge: # just jump to the new rev
             fp1, fp2, xp1, xp2 = fp2, nullid, xp2, ''
         if not partial:
             repo.hook('preupdate', throw=True, parent1=xp1, parent2=xp2)
             # note that we're in the middle of an update
             repo.vfs.write('updatestate', p2.hex())
 
         stats = applyupdates(repo, actions, wc, p2, overwrite, labels=labels)
 
         # divergent renames
         for f, fl in sorted(diverge.iteritems()):
             repo.ui.warn(_("note: possible conflict - %s was renamed "
                            "multiple times to:\n") % f)
             for nf in fl:
                 repo.ui.warn(" %s\n" % nf)
 
         # rename and delete
         for f, fl in sorted(renamedelete.iteritems()):
             repo.ui.warn(_("note: possible conflict - %s was deleted "
                            "and renamed to:\n") % f)
             for nf in fl:
                 repo.ui.warn(" %s\n" % nf)
 
         if not partial:
             repo.dirstate.beginparentchange()
             repo.setparents(fp1, fp2)
             recordupdates(repo, actions, branchmerge)
             # update completed, clear state
             util.unlink(repo.join('updatestate'))
 
             if not branchmerge:
                 repo.dirstate.setbranch(p2.branch())
             repo.dirstate.endparentchange()
     finally:
         wlock.release()
 
     if not partial:
         repo.hook('update', parent1=xp1, parent2=xp2, error=stats[3])
     return stats
 
 def graft(repo, ctx, pctx, labels):
     """Do a graft-like merge.
 
     This is a merge where the merge ancestor is chosen such that one
     or more changesets are grafted onto the current changeset. In
     addition to the merge, this fixes up the dirstate to include only
     a single parent and tries to duplicate any renames/copies
     appropriately.
 
     ctx - changeset to rebase
     pctx - merge base, usually ctx.p1()
     labels - merge labels eg ['local', 'graft']
 
     """
 
     stats = update(repo, ctx.node(), True, True, False, pctx.node(),
                    labels=labels)
     # drop the second merge parent
     repo.dirstate.beginparentchange()
     repo.setparents(repo['.'].node(), nullid)
     repo.dirstate.write()
     # fix up dirstate for copies and renames
     copies.duplicatecopies(repo, ctx.rev(), pctx.rev())
     repo.dirstate.endparentchange()
     return stats

tests/test-rename-dir-merge.t

@@ -1,237 +1,239 b''
   $ hg init t
   $ cd t
 
   $ mkdir a
   $ echo foo > a/a
   $ echo bar > a/b
   $ hg ci -Am "0"
   adding a/a
   adding a/b
 
   $ hg co -C 0
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg mv a b
   moving a/a to b/a (glob)
   moving a/b to b/b (glob)
   $ hg ci -m "1 mv a/ b/"
 
   $ hg co -C 0
   2 files updated, 0 files merged, 2 files removed, 0 files unresolved
   $ echo baz > a/c
   $ echo quux > a/d
   $ hg add a/c
   $ hg ci -m "2 add a/c"
   created new head
 
   $ hg merge --debug 1
     searching for copies back to rev 1
     unmatched files in local:
      a/c
     unmatched files in other:
      b/a
      b/b
     all copies found (* = to merge, ! = divergent, % = renamed and deleted):
      src: 'a/a' -> dst: 'b/a' 
      src: 'a/b' -> dst: 'b/b' 
     checking for directory renames
      discovered dir src: 'a/' -> dst: 'b/'
      pending file src: 'a/c' -> dst: 'b/c'
   resolving manifests
    branchmerge: True, force: False, partial: False
    ancestor: f9b20c0d4c51, local: ce36d17b18fb+, remote: 397f8b00a740
    a/a: other deleted -> r
   removing a/a
    a/b: other deleted -> r
   removing a/b
   updating: a/b 2/5 files (40.00%)
    b/a: remote created -> g
   getting b/a
    b/b: remote created -> g
   getting b/b
   updating: b/b 4/5 files (80.00%)
    b/c: remote directory rename - move from a/c -> dm
   updating: b/c 5/5 files (100.00%)
   moving a/c to b/c (glob)
   3 files updated, 0 files merged, 2 files removed, 0 files unresolved
   (branch merge, don't forget to commit)
 
   $ echo a/* b/*
   a/d b/a b/b b/c
   $ hg st -C
   M b/a
   M b/b
   A b/c
     a/c
   R a/a
   R a/b
   R a/c
   ? a/d
   $ hg ci -m "3 merge 2+1"
   $ hg debugrename b/c
   b/c renamed from a/c:354ae8da6e890359ef49ade27b68bbc361f3ca88 (glob)
 
   $ hg co -C 1
   0 files updated, 0 files merged, 1 files removed, 0 files unresolved
   $ hg merge --debug 2
     searching for copies back to rev 1
     unmatched files in local:
      b/a
      b/b
     unmatched files in other:
      a/c
     all copies found (* = to merge, ! = divergent, % = renamed and deleted):
      src: 'a/a' -> dst: 'b/a' 
      src: 'a/b' -> dst: 'b/b' 
     checking for directory renames
      discovered dir src: 'a/' -> dst: 'b/'
      pending file src: 'a/c' -> dst: 'b/c'
   resolving manifests
    branchmerge: True, force: False, partial: False
    ancestor: f9b20c0d4c51, local: 397f8b00a740+, remote: ce36d17b18fb
    b/c: local directory rename - get from a/c -> dg
   updating: b/c 1/1 files (100.00%)
   getting a/c to b/c
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   (branch merge, don't forget to commit)
 
   $ echo a/* b/*
   a/d b/a b/b b/c
   $ hg st -C
   A b/c
     a/c
   ? a/d
   $ hg ci -m "4 merge 1+2"
   created new head
   $ hg debugrename b/c
   b/c renamed from a/c:354ae8da6e890359ef49ade27b68bbc361f3ca88 (glob)
 
 Local directory rename with conflicting file added in remote source directory
 and untracked in local target directory.
 
-BROKEN: the uncommitted file is overwritten; we should abort
-
   $ hg co -qC 1
   $ echo target > b/c
   $ hg merge 2
+  b/c: untracked file differs
+  abort: untracked files in working directory differ from files in requested revision
+  [255]
+  $ cat b/c
+  target
+but it should succeed if the content matches
+  $ hg cat -r 2 a/c > b/c
+  $ hg merge 2
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   (branch merge, don't forget to commit)
-  $ hg st -A
+  $ hg st -C
   A b/c
     a/c
   ? a/d
-  C b/a
-  C b/b
-  $ cat b/c
-  baz
 
 Local directory rename with conflicting file added in remote source directory
 and committed in local target directory.
 
   $ hg co -qC 1
   $ echo target > b/c
   $ hg add b/c
   $ hg commit -qm 'new file in target directory'
   $ hg merge 2
   merging b/c and a/c to b/c
   warning: conflicts during merge.
   merging b/c incomplete! (edit conflicts, then use 'hg resolve --mark')
   0 files updated, 0 files merged, 0 files removed, 1 files unresolved
   use 'hg resolve' to retry unresolved file merges or 'hg update -C .' to abandon
   [1]
   $ hg st -A
   M b/c
     a/c
   ? a/d
   ? b/c.orig
   C b/a
   C b/b
   $ cat b/c
   <<<<<<< local: f1c50ca4f127 - test: new file in target directory
   target
   =======
   baz
   >>>>>>> other: ce36d17b18fb  - test: 2 add a/c
   $ rm b/c.orig
 
 Remote directory rename with conflicting file added in remote target directory
 and committed in local source directory.
 
   $ hg co -qC 2
   $ hg st -A
   ? a/d
   C a/a
   C a/b
   C a/c
   $ hg merge 5
   merging a/c and b/c to b/c
   warning: conflicts during merge.
   merging b/c incomplete! (edit conflicts, then use 'hg resolve --mark')
   2 files updated, 0 files merged, 2 files removed, 1 files unresolved
   use 'hg resolve' to retry unresolved file merges or 'hg update -C .' to abandon
   [1]
   $ hg st -A
   M b/a
   M b/b
   M b/c
     a/c
   R a/a
   R a/b
   R a/c
   ? a/d
   ? b/c.orig
   $ cat b/c
   <<<<<<< local: ce36d17b18fb  - test: 2 add a/c
   baz
   =======
   target
   >>>>>>> other: f1c50ca4f127 - test: new file in target directory
 
 Second scenario with two repos:
 
   $ cd ..
   $ hg init r1
   $ cd r1
   $ mkdir a
   $ echo foo > a/f
   $ hg add a
   adding a/f (glob)
   $ hg ci -m "a/f == foo"
   $ cd ..
 
   $ hg clone r1 r2
   updating to branch default
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ cd r2
   $ hg mv a b
   moving a/f to b/f (glob)
   $ echo foo1 > b/f
   $ hg ci -m" a -> b, b/f == foo1"
   $ cd ..
 
   $ cd r1
   $ mkdir a/aa
   $ echo bar > a/aa/g
   $ hg add a/aa
   adding a/aa/g (glob)
   $ hg ci -m "a/aa/g"
   $ hg pull ../r2
   pulling from ../r2
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   added 1 changesets with 1 changes to 1 files (+1 heads)
   (run 'hg heads' to see heads, 'hg merge' to merge)
 
   $ hg merge
   2 files updated, 0 files merged, 1 files removed, 0 files unresolved
   (branch merge, don't forget to commit)
 
   $ hg st -C
   M b/f
   A b/aa/g
     a/aa/g
   R a/aa/g
   R a/f
 
   $ cd ..