fix race in localrepo.addchangegroup.

localrepo.addchangegroup writes to changelog, then manifest, then normal
files.  this breaks access ordering.  if reader reads changelog while
manifest is being written, can find pointers into places in manifest
that are not yet written.  same can happen for manifest and normal files.

fix is to make almost no change to localrepo.addchangegroup.  it must
to write changelog and manifest data early because it has to read them
while writing other files.  instead, write changelog and manifest data
to temp file that reader cannot see, then append temp data to manifest
after all normal files written, finally append temp data to changelog.

temp file code is in new appendfile module.  can be used in other places
with small changes.

much smaller race still left.  we write all new data in one write call,
but reader can maybe see partial update because python or os or filesystem
cannot always make write really atomic. file locking no help: slow, not
portable, not reliable over nfs.  only real safe other plan is write to
temp file every time and rename, but performance bad when manifest or
changelog is big.

mercurial/localrepo.py

@@ -1,1904 +1,1919 b''
 # localrepo.py - read/write repository class for mercurial
 #
 # Copyright 2005 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms
 # of the GNU General Public License, incorporated herein by reference.
 
 import os, util
 import filelog, manifest, changelog, dirstate, repo
 from node import *
 from i18n import gettext as _
 from demandload import *
 demandload(globals(), "re lock transaction tempfile stat mdiff errno ui")
-demandload(globals(), "changegroup")
+demandload(globals(), "appendfile changegroup")
 
 class localrepository(object):
     def __del__(self):
         self.transhandle = None
     def __init__(self, parentui, path=None, create=0):
         if not path:
             p = os.getcwd()
             while not os.path.isdir(os.path.join(p, ".hg")):
                 oldp = p
                 p = os.path.dirname(p)
                 if p == oldp:
                     raise repo.RepoError(_("no repo found"))
             path = p
         self.path = os.path.join(path, ".hg")
 
         if not create and not os.path.isdir(self.path):
             raise repo.RepoError(_("repository %s not found") % path)
 
         self.root = os.path.abspath(path)
         self.ui = ui.ui(parentui=parentui)
         self.opener = util.opener(self.path)
         self.wopener = util.opener(self.root)
         self.manifest = manifest.manifest(self.opener)
         self.changelog = changelog.changelog(self.opener)
         self.tagscache = None
         self.nodetagscache = None
         self.encodepats = None
         self.decodepats = None
         self.transhandle = None
 
         if create:
             os.mkdir(self.path)
             os.mkdir(self.join("data"))
 
         self.dirstate = dirstate.dirstate(self.opener, self.ui, self.root)
         try:
             self.ui.readconfig(self.join("hgrc"), self.root)
         except IOError:
             pass
 
     def hook(self, name, throw=False, **args):
         def runhook(name, cmd):
             self.ui.note(_("running hook %s: %s\n") % (name, cmd))
             env = dict([('HG_' + k.upper(), v) for k, v in args.iteritems()] +
                        [(k.upper(), v) for k, v in args.iteritems()])
             r = util.system(cmd, environ=env, cwd=self.root)
             if r:
                 desc, r = util.explain_exit(r)
                 if throw:
                     raise util.Abort(_('%s hook %s') % (name, desc))
                 self.ui.warn(_('error: %s hook %s\n') % (name, desc))
                 return False
             return True
 
         r = True
         hooks = [(hname, cmd) for hname, cmd in self.ui.configitems("hooks")
                  if hname.split(".", 1)[0] == name and cmd]
         hooks.sort()
         for hname, cmd in hooks:
             r = runhook(hname, cmd) and r
         return r
 
     def tags(self):
         '''return a mapping of tag to node'''
         if not self.tagscache:
             self.tagscache = {}
 
             def parsetag(line, context):
                 if not line:
                     return
                 s = l.split(" ", 1)
                 if len(s) != 2:
                     self.ui.warn(_("%s: ignoring invalid tag\n") % context)
                     return
                 node, key = s
                 try:
                     bin_n = bin(node)
                 except TypeError:
                     self.ui.warn(_("%s: ignoring invalid tag\n") % context)
                     return
                 if bin_n not in self.changelog.nodemap:
                     self.ui.warn(_("%s: ignoring invalid tag\n") % context)
                     return
                 self.tagscache[key.strip()] = bin_n
 
             # read each head of the tags file, ending with the tip
             # and add each tag found to the map, with "newer" ones
             # taking precedence
             fl = self.file(".hgtags")
             h = fl.heads()
             h.reverse()
             for r in h:
                 count = 0
                 for l in fl.read(r).splitlines():
                     count += 1
                     parsetag(l, ".hgtags:%d" % count)
 
             try:
                 f = self.opener("localtags")
                 count = 0
                 for l in f:
                     count += 1
                     parsetag(l, "localtags:%d" % count)
             except IOError:
                 pass
 
             self.tagscache['tip'] = self.changelog.tip()
 
         return self.tagscache
 
     def tagslist(self):
         '''return a list of tags ordered by revision'''
         l = []
         for t, n in self.tags().items():
             try:
                 r = self.changelog.rev(n)
             except:
                 r = -2 # sort to the beginning of the list if unknown
             l.append((r, t, n))
         l.sort()
         return [(t, n) for r, t, n in l]
 
     def nodetags(self, node):
         '''return the tags associated with a node'''
         if not self.nodetagscache:
             self.nodetagscache = {}
             for t, n in self.tags().items():
                 self.nodetagscache.setdefault(n, []).append(t)
         return self.nodetagscache.get(node, [])
 
     def lookup(self, key):
         try:
             return self.tags()[key]
         except KeyError:
             try:
                 return self.changelog.lookup(key)
             except:
                 raise repo.RepoError(_("unknown revision '%s'") % key)
 
     def dev(self):
         return os.stat(self.path).st_dev
 
     def local(self):
         return True
 
     def join(self, f):
         return os.path.join(self.path, f)
 
     def wjoin(self, f):
         return os.path.join(self.root, f)
 
     def file(self, f):
         if f[0] == '/':
             f = f[1:]
         return filelog.filelog(self.opener, f)
 
     def getcwd(self):
         return self.dirstate.getcwd()
 
     def wfile(self, f, mode='r'):
         return self.wopener(f, mode)
 
     def wread(self, filename):
         if self.encodepats == None:
             l = []
             for pat, cmd in self.ui.configitems("encode"):
                 mf = util.matcher(self.root, "", [pat], [], [])[1]
                 l.append((mf, cmd))
             self.encodepats = l
 
         data = self.wopener(filename, 'r').read()
 
         for mf, cmd in self.encodepats:
             if mf(filename):
                 self.ui.debug(_("filtering %s through %s\n") % (filename, cmd))
                 data = util.filter(data, cmd)
                 break
 
         return data
 
     def wwrite(self, filename, data, fd=None):
         if self.decodepats == None:
             l = []
             for pat, cmd in self.ui.configitems("decode"):
                 mf = util.matcher(self.root, "", [pat], [], [])[1]
                 l.append((mf, cmd))
             self.decodepats = l
 
         for mf, cmd in self.decodepats:
             if mf(filename):
                 self.ui.debug(_("filtering %s through %s\n") % (filename, cmd))
                 data = util.filter(data, cmd)
                 break
 
         if fd:
             return fd.write(data)
         return self.wopener(filename, 'w').write(data)
 
     def transaction(self):
         tr = self.transhandle
         if tr != None and tr.running():
             return tr.nest()
 
         # save dirstate for undo
         try:
             ds = self.opener("dirstate").read()
         except IOError:
             ds = ""
         self.opener("journal.dirstate", "w").write(ds)
 
         tr = transaction.transaction(self.ui.warn, self.opener,
                                        self.join("journal"),
                                        aftertrans(self.path))
         self.transhandle = tr
         return tr
 
     def recover(self):
         l = self.lock()
         if os.path.exists(self.join("journal")):
             self.ui.status(_("rolling back interrupted transaction\n"))
             transaction.rollback(self.opener, self.join("journal"))
             self.reload()
             return True
         else:
             self.ui.warn(_("no interrupted transaction available\n"))
             return False
 
     def undo(self, wlock=None):
         if not wlock:
             wlock = self.wlock()
         l = self.lock()
         if os.path.exists(self.join("undo")):
             self.ui.status(_("rolling back last transaction\n"))
             transaction.rollback(self.opener, self.join("undo"))
             util.rename(self.join("undo.dirstate"), self.join("dirstate"))
             self.reload()
             self.wreload()
         else:
             self.ui.warn(_("no undo information available\n"))
 
     def wreload(self):
         self.dirstate.read()
 
     def reload(self):
         self.changelog.load()
         self.manifest.load()
         self.tagscache = None
         self.nodetagscache = None
 
     def do_lock(self, lockname, wait, releasefn=None, acquirefn=None):
         try:
             l = lock.lock(self.join(lockname), 0, releasefn)
         except lock.LockHeld, inst:
             if not wait:
                 raise inst
             self.ui.warn(_("waiting for lock held by %s\n") % inst.args[0])
             try:
                 # default to 600 seconds timeout
                 l = lock.lock(self.join(lockname),
                               int(self.ui.config("ui", "timeout") or 600),
                               releasefn)
             except lock.LockHeld, inst:
                 raise util.Abort(_("timeout while waiting for "
                                    "lock held by %s") % inst.args[0])
         if acquirefn:
             acquirefn()
         return l
 
     def lock(self, wait=1):
         return self.do_lock("lock", wait, acquirefn=self.reload)
 
     def wlock(self, wait=1):
         return self.do_lock("wlock", wait,
                             self.dirstate.write,
                             self.wreload)
 
     def checkfilemerge(self, filename, text, filelog, manifest1, manifest2):
         "determine whether a new filenode is needed"
         fp1 = manifest1.get(filename, nullid)
         fp2 = manifest2.get(filename, nullid)
 
         if fp2 != nullid:
             # is one parent an ancestor of the other?
             fpa = filelog.ancestor(fp1, fp2)
             if fpa == fp1:
                 fp1, fp2 = fp2, nullid
             elif fpa == fp2:
                 fp2 = nullid
 
             # is the file unmodified from the parent? report existing entry
             if fp2 == nullid and text == filelog.read(fp1):
                 return (fp1, None, None)
 
         return (None, fp1, fp2)
 
     def rawcommit(self, files, text, user, date, p1=None, p2=None, wlock=None):
         orig_parent = self.dirstate.parents()[0] or nullid
         p1 = p1 or self.dirstate.parents()[0] or nullid
         p2 = p2 or self.dirstate.parents()[1] or nullid
         c1 = self.changelog.read(p1)
         c2 = self.changelog.read(p2)
         m1 = self.manifest.read(c1[0])
         mf1 = self.manifest.readflags(c1[0])
         m2 = self.manifest.read(c2[0])
         changed = []
 
         if orig_parent == p1:
             update_dirstate = 1
         else:
             update_dirstate = 0
 
         if not wlock:
             wlock = self.wlock()
         l = self.lock()
         tr = self.transaction()
         mm = m1.copy()
         mfm = mf1.copy()
         linkrev = self.changelog.count()
         for f in files:
             try:
                 t = self.wread(f)
                 tm = util.is_exec(self.wjoin(f), mfm.get(f, False))
                 r = self.file(f)
                 mfm[f] = tm
 
                 (entry, fp1, fp2) = self.checkfilemerge(f, t, r, m1, m2)
                 if entry:
                     mm[f] = entry
                     continue
 
                 mm[f] = r.add(t, {}, tr, linkrev, fp1, fp2)
                 changed.append(f)
                 if update_dirstate:
                     self.dirstate.update([f], "n")
             except IOError:
                 try:
                     del mm[f]
                     del mfm[f]
                     if update_dirstate:
                         self.dirstate.forget([f])
                 except:
                     # deleted from p2?
                     pass
 
         mnode = self.manifest.add(mm, mfm, tr, linkrev, c1[0], c2[0])
         user = user or self.ui.username()
         n = self.changelog.add(mnode, changed, text, tr, p1, p2, user, date)
         tr.close()
         if update_dirstate:
             self.dirstate.setparents(n, nullid)
 
     def commit(self, files=None, text="", user=None, date=None,
                match=util.always, force=False, lock=None, wlock=None):
         commit = []
         remove = []
         changed = []
 
         if files:
             for f in files:
                 s = self.dirstate.state(f)
                 if s in 'nmai':
                     commit.append(f)
                 elif s == 'r':
                     remove.append(f)
                 else:
                     self.ui.warn(_("%s not tracked!\n") % f)
         else:
             modified, added, removed, deleted, unknown = self.changes(match=match)
             commit = modified + added
             remove = removed
 
         p1, p2 = self.dirstate.parents()
         c1 = self.changelog.read(p1)
         c2 = self.changelog.read(p2)
         m1 = self.manifest.read(c1[0])
         mf1 = self.manifest.readflags(c1[0])
         m2 = self.manifest.read(c2[0])
 
         if not commit and not remove and not force and p2 == nullid:
             self.ui.status(_("nothing changed\n"))
             return None
 
         xp1 = hex(p1)
         if p2 == nullid: xp2 = ''
         else: xp2 = hex(p2)
 
         self.hook("precommit", throw=True, parent1=xp1, parent2=xp2)
 
         if not wlock:
             wlock = self.wlock()
         if not lock:
             lock = self.lock()
         tr = self.transaction()
 
         # check in files
         new = {}
         linkrev = self.changelog.count()
         commit.sort()
         for f in commit:
             self.ui.note(f + "\n")
             try:
                 mf1[f] = util.is_exec(self.wjoin(f), mf1.get(f, False))
                 t = self.wread(f)
             except IOError:
                 self.ui.warn(_("trouble committing %s!\n") % f)
                 raise
 
             r = self.file(f)
 
             meta = {}
             cp = self.dirstate.copied(f)
             if cp:
                 meta["copy"] = cp
                 meta["copyrev"] = hex(m1.get(cp, m2.get(cp, nullid)))
                 self.ui.debug(_(" %s: copy %s:%s\n") % (f, cp, meta["copyrev"]))
                 fp1, fp2 = nullid, nullid
             else:
                 entry, fp1, fp2 = self.checkfilemerge(f, t, r, m1, m2)
                 if entry:
                     new[f] = entry
                     continue
 
             new[f] = r.add(t, meta, tr, linkrev, fp1, fp2)
             # remember what we've added so that we can later calculate
             # the files to pull from a set of changesets
             changed.append(f)
 
         # update manifest
         m1 = m1.copy()
         m1.update(new)
         for f in remove:
             if f in m1:
                 del m1[f]
         mn = self.manifest.add(m1, mf1, tr, linkrev, c1[0], c2[0],
                                (new, remove))
 
         # add changeset
         new = new.keys()
         new.sort()
 
         user = user or self.ui.username()
         if not text:
             edittext = [""]
             if p2 != nullid:
                 edittext.append("HG: branch merge")
             edittext.extend(["HG: changed %s" % f for f in changed])
             edittext.extend(["HG: removed %s" % f for f in remove])
             if not changed and not remove:
                 edittext.append("HG: no files changed")
             edittext.append("")
             # run editor in the repository root
             olddir = os.getcwd()
             os.chdir(self.root)
             edittext = self.ui.edit("\n".join(edittext), user)
             os.chdir(olddir)
             if not edittext.rstrip():
                 return None
             text = edittext
 
         n = self.changelog.add(mn, changed + remove, text, tr, p1, p2, user, date)
         self.hook('pretxncommit', throw=True, node=hex(n), parent1=xp1,
                   parent2=xp2)
         tr.close()
 
         self.dirstate.setparents(n)
         self.dirstate.update(new, "n")
         self.dirstate.forget(remove)
 
         self.hook("commit", node=hex(n), parent1=xp1, parent2=xp2)
         return n
 
     def walk(self, node=None, files=[], match=util.always):
         if node:
             fdict = dict.fromkeys(files)
             for fn in self.manifest.read(self.changelog.read(node)[0]):
                 fdict.pop(fn, None)
                 if match(fn):
                     yield 'm', fn
             for fn in fdict:
                 self.ui.warn(_('%s: No such file in rev %s\n') % (
                     util.pathto(self.getcwd(), fn), short(node)))
         else:
             for src, fn in self.dirstate.walk(files, match):
                 yield src, fn
 
     def changes(self, node1=None, node2=None, files=[], match=util.always,
                 wlock=None):
         """return changes between two nodes or node and working directory
 
         If node1 is None, use the first dirstate parent instead.
         If node2 is None, compare node1 with working directory.
         """
 
         def fcmp(fn, mf):
             t1 = self.wread(fn)
             t2 = self.file(fn).read(mf.get(fn, nullid))
             return cmp(t1, t2)
 
         def mfmatches(node):
             change = self.changelog.read(node)
             mf = dict(self.manifest.read(change[0]))
             for fn in mf.keys():
                 if not match(fn):
                     del mf[fn]
             return mf
 
         if node1:
             # read the manifest from node1 before the manifest from node2,
             # so that we'll hit the manifest cache if we're going through
             # all the revisions in parent->child order.
             mf1 = mfmatches(node1)
 
         # are we comparing the working directory?
         if not node2:
             if not wlock:
                 try:
                     wlock = self.wlock(wait=0)
                 except lock.LockException:
                     wlock = None
             lookup, modified, added, removed, deleted, unknown = (
                 self.dirstate.changes(files, match))
 
             # are we comparing working dir against its parent?
             if not node1:
                 if lookup:
                     # do a full compare of any files that might have changed
                     mf2 = mfmatches(self.dirstate.parents()[0])
                     for f in lookup:
                         if fcmp(f, mf2):
                             modified.append(f)
                         elif wlock is not None:
                             self.dirstate.update([f], "n")
             else:
                 # we are comparing working dir against non-parent
                 # generate a pseudo-manifest for the working dir
                 mf2 = mfmatches(self.dirstate.parents()[0])
                 for f in lookup + modified + added:
                     mf2[f] = ""
                 for f in removed:
                     if f in mf2:
                         del mf2[f]
         else:
             # we are comparing two revisions
             deleted, unknown = [], []
             mf2 = mfmatches(node2)
 
         if node1:
             # flush lists from dirstate before comparing manifests
             modified, added = [], []
 
             for fn in mf2:
                 if mf1.has_key(fn):
                     if mf1[fn] != mf2[fn] and (mf2[fn] != "" or fcmp(fn, mf1)):
                         modified.append(fn)
                     del mf1[fn]
                 else:
                     added.append(fn)
 
             removed = mf1.keys()
 
         # sort and return results:
         for l in modified, added, removed, deleted, unknown:
             l.sort()
         return (modified, added, removed, deleted, unknown)
 
     def add(self, list, wlock=None):
         if not wlock:
             wlock = self.wlock()
         for f in list:
             p = self.wjoin(f)
             if not os.path.exists(p):
                 self.ui.warn(_("%s does not exist!\n") % f)
             elif not os.path.isfile(p):
                 self.ui.warn(_("%s not added: only files supported currently\n")
                              % f)
             elif self.dirstate.state(f) in 'an':
                 self.ui.warn(_("%s already tracked!\n") % f)
             else:
                 self.dirstate.update([f], "a")
 
     def forget(self, list, wlock=None):
         if not wlock:
             wlock = self.wlock()
         for f in list:
             if self.dirstate.state(f) not in 'ai':
                 self.ui.warn(_("%s not added!\n") % f)
             else:
                 self.dirstate.forget([f])
 
     def remove(self, list, unlink=False, wlock=None):
         if unlink:
             for f in list:
                 try:
                     util.unlink(self.wjoin(f))
                 except OSError, inst:
                     if inst.errno != errno.ENOENT:
                         raise
         if not wlock:
             wlock = self.wlock()
         for f in list:
             p = self.wjoin(f)
             if os.path.exists(p):
                 self.ui.warn(_("%s still exists!\n") % f)
             elif self.dirstate.state(f) == 'a':
                 self.dirstate.forget([f])
             elif f not in self.dirstate:
                 self.ui.warn(_("%s not tracked!\n") % f)
             else:
                 self.dirstate.update([f], "r")
 
     def undelete(self, list, wlock=None):
         p = self.dirstate.parents()[0]
         mn = self.changelog.read(p)[0]
         mf = self.manifest.readflags(mn)
         m = self.manifest.read(mn)
         if not wlock:
             wlock = self.wlock()
         for f in list:
             if self.dirstate.state(f) not in  "r":
                 self.ui.warn("%s not removed!\n" % f)
             else:
                 t = self.file(f).read(m[f])
                 self.wwrite(f, t)
                 util.set_exec(self.wjoin(f), mf[f])
                 self.dirstate.update([f], "n")
 
     def copy(self, source, dest, wlock=None):
         p = self.wjoin(dest)
         if not os.path.exists(p):
             self.ui.warn(_("%s does not exist!\n") % dest)
         elif not os.path.isfile(p):
             self.ui.warn(_("copy failed: %s is not a file\n") % dest)
         else:
             if not wlock:
                 wlock = self.wlock()
             if self.dirstate.state(dest) == '?':
                 self.dirstate.update([dest], "a")
             self.dirstate.copy(source, dest)
 
     def heads(self, start=None):
         heads = self.changelog.heads(start)
         # sort the output in rev descending order
         heads = [(-self.changelog.rev(h), h) for h in heads]
         heads.sort()
         return [n for (r, n) in heads]
 
     # branchlookup returns a dict giving a list of branches for
     # each head.  A branch is defined as the tag of a node or
     # the branch of the node's parents.  If a node has multiple
     # branch tags, tags are eliminated if they are visible from other
     # branch tags.
     #
     # So, for this graph:  a->b->c->d->e
     #                       \         /
     #                        aa -----/
     # a has tag 2.6.12
     # d has tag 2.6.13
     # e would have branch tags for 2.6.12 and 2.6.13.  Because the node
     # for 2.6.12 can be reached from the node 2.6.13, that is eliminated
     # from the list.
     #
     # It is possible that more than one head will have the same branch tag.
     # callers need to check the result for multiple heads under the same
     # branch tag if that is a problem for them (ie checkout of a specific
     # branch).
     #
     # passing in a specific branch will limit the depth of the search
     # through the parents.  It won't limit the branches returned in the
     # result though.
     def branchlookup(self, heads=None, branch=None):
         if not heads:
             heads = self.heads()
         headt = [ h for h in heads ]
         chlog = self.changelog
         branches = {}
         merges = []
         seenmerge = {}
 
         # traverse the tree once for each head, recording in the branches
         # dict which tags are visible from this head.  The branches
         # dict also records which tags are visible from each tag
         # while we traverse.
         while headt or merges:
             if merges:
                 n, found = merges.pop()
                 visit = [n]
             else:
                 h = headt.pop()
                 visit = [h]
                 found = [h]
                 seen = {}
             while visit:
                 n = visit.pop()
                 if n in seen:
                     continue
                 pp = chlog.parents(n)
                 tags = self.nodetags(n)
                 if tags:
                     for x in tags:
                         if x == 'tip':
                             continue
                         for f in found:
                             branches.setdefault(f, {})[n] = 1
                         branches.setdefault(n, {})[n] = 1
                         break
                     if n not in found:
                         found.append(n)
                     if branch in tags:
                         continue
                 seen[n] = 1
                 if pp[1] != nullid and n not in seenmerge:
                     merges.append((pp[1], [x for x in found]))
                     seenmerge[n] = 1
                 if pp[0] != nullid:
                     visit.append(pp[0])
         # traverse the branches dict, eliminating branch tags from each
         # head that are visible from another branch tag for that head.
         out = {}
         viscache = {}
         for h in heads:
             def visible(node):
                 if node in viscache:
                     return viscache[node]
                 ret = {}
                 visit = [node]
                 while visit:
                     x = visit.pop()
                     if x in viscache:
                         ret.update(viscache[x])
                     elif x not in ret:
                         ret[x] = 1
                         if x in branches:
                             visit[len(visit):] = branches[x].keys()
                 viscache[node] = ret
                 return ret
             if h not in branches:
                 continue
             # O(n^2), but somewhat limited.  This only searches the
             # tags visible from a specific head, not all the tags in the
             # whole repo.
             for b in branches[h]:
                 vis = False
                 for bb in branches[h].keys():
                     if b != bb:
                         if b in visible(bb):
                             vis = True
                             break
                 if not vis:
                     l = out.setdefault(h, [])
                     l[len(l):] = self.nodetags(b)
         return out
 
     def branches(self, nodes):
         if not nodes:
             nodes = [self.changelog.tip()]
         b = []
         for n in nodes:
             t = n
             while n:
                 p = self.changelog.parents(n)
                 if p[1] != nullid or p[0] == nullid:
                     b.append((t, n, p[0], p[1]))
                     break
                 n = p[0]
         return b
 
     def between(self, pairs):
         r = []
 
         for top, bottom in pairs:
             n, l, i = top, [], 0
             f = 1
 
             while n != bottom:
                 p = self.changelog.parents(n)[0]
                 if i == f:
                     l.append(n)
                     f = f * 2
                 n = p
                 i += 1
 
             r.append(l)
 
         return r
 
     def findincoming(self, remote, base=None, heads=None, force=False):
         m = self.changelog.nodemap
         search = []
         fetch = {}
         seen = {}
         seenbranch = {}
         if base == None:
             base = {}
 
         # assume we're closer to the tip than the root
         # and start by examining the heads
         self.ui.status(_("searching for changes\n"))
 
         if not heads:
             heads = remote.heads()
 
         unknown = []
         for h in heads:
             if h not in m:
                 unknown.append(h)
             else:
                 base[h] = 1
 
         if not unknown:
             return []
 
         rep = {}
         reqcnt = 0
 
         # search through remote branches
         # a 'branch' here is a linear segment of history, with four parts:
         # head, root, first parent, second parent
         # (a branch always has two parents (or none) by definition)
         unknown = remote.branches(unknown)
         while unknown:
             r = []
             while unknown:
                 n = unknown.pop(0)
                 if n[0] in seen:
                     continue
 
                 self.ui.debug(_("examining %s:%s\n")
                               % (short(n[0]), short(n[1])))
                 if n[0] == nullid:
                     break
                 if n in seenbranch:
                     self.ui.debug(_("branch already found\n"))
                     continue
                 if n[1] and n[1] in m: # do we know the base?
                     self.ui.debug(_("found incomplete branch %s:%s\n")
                                   % (short(n[0]), short(n[1])))
                     search.append(n) # schedule branch range for scanning
                     seenbranch[n] = 1
                 else:
                     if n[1] not in seen and n[1] not in fetch:
                         if n[2] in m and n[3] in m:
                             self.ui.debug(_("found new changeset %s\n") %
                                           short(n[1]))
                             fetch[n[1]] = 1 # earliest unknown
                             base[n[2]] = 1 # latest known
                             continue
 
                     for a in n[2:4]:
                         if a not in rep:
                             r.append(a)
                             rep[a] = 1
 
                 seen[n[0]] = 1
 
             if r:
                 reqcnt += 1
                 self.ui.debug(_("request %d: %s\n") %
                             (reqcnt, " ".join(map(short, r))))
                 for p in range(0, len(r), 10):
                     for b in remote.branches(r[p:p+10]):
                         self.ui.debug(_("received %s:%s\n") %
                                       (short(b[0]), short(b[1])))
                         if b[0] in m:
                             self.ui.debug(_("found base node %s\n")
                                           % short(b[0]))
                             base[b[0]] = 1
                         elif b[0] not in seen:
                             unknown.append(b)
 
         # do binary search on the branches we found
         while search:
             n = search.pop(0)
             reqcnt += 1
             l = remote.between([(n[0], n[1])])[0]
             l.append(n[1])
             p = n[0]
             f = 1
             for i in l:
                 self.ui.debug(_("narrowing %d:%d %s\n") % (f, len(l), short(i)))
                 if i in m:
                     if f <= 2:
                         self.ui.debug(_("found new branch changeset %s\n") %
                                           short(p))
                         fetch[p] = 1
                         base[i] = 1
                     else:
                         self.ui.debug(_("narrowed branch search to %s:%s\n")
                                       % (short(p), short(i)))
                         search.append((p, i))
                     break
                 p, f = i, f * 2
 
         # sanity check our fetch list
         for f in fetch.keys():
             if f in m:
                 raise repo.RepoError(_("already have changeset ") + short(f[:4]))
 
         if base.keys() == [nullid]:
             if force:
                 self.ui.warn(_("warning: repository is unrelated\n"))
             else:
                 raise util.Abort(_("repository is unrelated"))
 
         self.ui.note(_("found new changesets starting at ") +
                      " ".join([short(f) for f in fetch]) + "\n")
 
         self.ui.debug(_("%d total queries\n") % reqcnt)
 
         return fetch.keys()
 
     def findoutgoing(self, remote, base=None, heads=None, force=False):
         if base == None:
             base = {}
             self.findincoming(remote, base, heads, force=force)
 
         self.ui.debug(_("common changesets up to ")
                       + " ".join(map(short, base.keys())) + "\n")
 
         remain = dict.fromkeys(self.changelog.nodemap)
 
         # prune everything remote has from the tree
         del remain[nullid]
         remove = base.keys()
         while remove:
             n = remove.pop(0)
             if n in remain:
                 del remain[n]
                 for p in self.changelog.parents(n):
                     remove.append(p)
 
         # find every node whose parents have been pruned
         subset = []
         for n in remain:
             p1, p2 = self.changelog.parents(n)
             if p1 not in remain and p2 not in remain:
                 subset.append(n)
 
         # this is the set of all roots we have to push
         return subset
 
     def pull(self, remote, heads=None, force=False):
         l = self.lock()
 
         # if we have an empty repo, fetch everything
         if self.changelog.tip() == nullid:
             self.ui.status(_("requesting all changes\n"))
             fetch = [nullid]
         else:
             fetch = self.findincoming(remote, force=force)
 
         if not fetch:
             self.ui.status(_("no changes found\n"))
             return 1
 
         if heads is None:
             cg = remote.changegroup(fetch, 'pull')
         else:
             cg = remote.changegroupsubset(fetch, heads, 'pull')
         return self.addchangegroup(cg)
 
     def push(self, remote, force=False, revs=None):
         lock = remote.lock()
 
         base = {}
         heads = remote.heads()
         inc = self.findincoming(remote, base, heads, force=force)
         if not force and inc:
             self.ui.warn(_("abort: unsynced remote changes!\n"))
             self.ui.status(_("(did you forget to sync? use push -f to force)\n"))
             return 1
 
         update = self.findoutgoing(remote, base)
         if revs is not None:
             msng_cl, bases, heads = self.changelog.nodesbetween(update, revs)
         else:
             bases, heads = update, self.changelog.heads()
 
         if not bases:
             self.ui.status(_("no changes found\n"))
             return 1
         elif not force:
             if len(bases) < len(heads):
                 self.ui.warn(_("abort: push creates new remote branches!\n"))
                 self.ui.status(_("(did you forget to merge?"
                                  " use push -f to force)\n"))
                 return 1
 
         if revs is None:
             cg = self.changegroup(update, 'push')
         else:
             cg = self.changegroupsubset(update, revs, 'push')
         return remote.addchangegroup(cg)
 
     def changegroupsubset(self, bases, heads, source):
         """This function generates a changegroup consisting of all the nodes
         that are descendents of any of the bases, and ancestors of any of
         the heads.
 
         It is fairly complex as determining which filenodes and which
         manifest nodes need to be included for the changeset to be complete
         is non-trivial.
 
         Another wrinkle is doing the reverse, figuring out which changeset in
         the changegroup a particular filenode or manifestnode belongs to."""
 
         self.hook('preoutgoing', throw=True, source=source)
 
         # Set up some initial variables
         # Make it easy to refer to self.changelog
         cl = self.changelog
         # msng is short for missing - compute the list of changesets in this
         # changegroup.
         msng_cl_lst, bases, heads = cl.nodesbetween(bases, heads)
         # Some bases may turn out to be superfluous, and some heads may be
         # too.  nodesbetween will return the minimal set of bases and heads
         # necessary to re-create the changegroup.
 
         # Known heads are the list of heads that it is assumed the recipient
         # of this changegroup will know about.
         knownheads = {}
         # We assume that all parents of bases are known heads.
         for n in bases:
             for p in cl.parents(n):
                 if p != nullid:
                     knownheads[p] = 1
         knownheads = knownheads.keys()
         if knownheads:
             # Now that we know what heads are known, we can compute which
             # changesets are known.  The recipient must know about all
             # changesets required to reach the known heads from the null
             # changeset.
             has_cl_set, junk, junk = cl.nodesbetween(None, knownheads)
             junk = None
             # Transform the list into an ersatz set.
             has_cl_set = dict.fromkeys(has_cl_set)
         else:
             # If there were no known heads, the recipient cannot be assumed to
             # know about any changesets.
             has_cl_set = {}
 
         # Make it easy to refer to self.manifest
         mnfst = self.manifest
         # We don't know which manifests are missing yet
         msng_mnfst_set = {}
         # Nor do we know which filenodes are missing.
         msng_filenode_set = {}
 
         junk = mnfst.index[mnfst.count() - 1] # Get around a bug in lazyindex
         junk = None
 
         # A changeset always belongs to itself, so the changenode lookup
         # function for a changenode is identity.
         def identity(x):
             return x
 
         # A function generating function.  Sets up an environment for the
         # inner function.
         def cmp_by_rev_func(revlog):
             # Compare two nodes by their revision number in the environment's
             # revision history.  Since the revision number both represents the
             # most efficient order to read the nodes in, and represents a
             # topological sorting of the nodes, this function is often useful.
             def cmp_by_rev(a, b):
                 return cmp(revlog.rev(a), revlog.rev(b))
             return cmp_by_rev
 
         # If we determine that a particular file or manifest node must be a
         # node that the recipient of the changegroup will already have, we can
         # also assume the recipient will have all the parents.  This function
         # prunes them from the set of missing nodes.
         def prune_parents(revlog, hasset, msngset):
             haslst = hasset.keys()
             haslst.sort(cmp_by_rev_func(revlog))
             for node in haslst:
                 parentlst = [p for p in revlog.parents(node) if p != nullid]
                 while parentlst:
                     n = parentlst.pop()
                     if n not in hasset:
                         hasset[n] = 1
                         p = [p for p in revlog.parents(n) if p != nullid]
                         parentlst.extend(p)
             for n in hasset:
                 msngset.pop(n, None)
 
         # This is a function generating function used to set up an environment
         # for the inner function to execute in.
         def manifest_and_file_collector(changedfileset):
             # This is an information gathering function that gathers
             # information from each changeset node that goes out as part of
             # the changegroup.  The information gathered is a list of which
             # manifest nodes are potentially required (the recipient may
             # already have them) and total list of all files which were
             # changed in any changeset in the changegroup.
             #
             # We also remember the first changenode we saw any manifest
             # referenced by so we can later determine which changenode 'owns'
             # the manifest.
             def collect_manifests_and_files(clnode):
                 c = cl.read(clnode)
                 for f in c[3]:
                     # This is to make sure we only have one instance of each
                     # filename string for each filename.
                     changedfileset.setdefault(f, f)
                 msng_mnfst_set.setdefault(c[0], clnode)
             return collect_manifests_and_files
 
         # Figure out which manifest nodes (of the ones we think might be part
         # of the changegroup) the recipient must know about and remove them
         # from the changegroup.
         def prune_manifests():
             has_mnfst_set = {}
             for n in msng_mnfst_set:
                 # If a 'missing' manifest thinks it belongs to a changenode
                 # the recipient is assumed to have, obviously the recipient
                 # must have that manifest.
                 linknode = cl.node(mnfst.linkrev(n))
                 if linknode in has_cl_set:
                     has_mnfst_set[n] = 1
             prune_parents(mnfst, has_mnfst_set, msng_mnfst_set)
 
         # Use the information collected in collect_manifests_and_files to say
         # which changenode any manifestnode belongs to.
         def lookup_manifest_link(mnfstnode):
             return msng_mnfst_set[mnfstnode]
 
         # A function generating function that sets up the initial environment
         # the inner function.
         def filenode_collector(changedfiles):
             next_rev = [0]
             # This gathers information from each manifestnode included in the
             # changegroup about which filenodes the manifest node references
             # so we can include those in the changegroup too.
             #
             # It also remembers which changenode each filenode belongs to.  It
             # does this by assuming the a filenode belongs to the changenode
             # the first manifest that references it belongs to.
             def collect_msng_filenodes(mnfstnode):
                 r = mnfst.rev(mnfstnode)
                 if r == next_rev[0]:
                     # If the last rev we looked at was the one just previous,
                     # we only need to see a diff.
                     delta = mdiff.patchtext(mnfst.delta(mnfstnode))
                     # For each line in the delta
                     for dline in delta.splitlines():
                         # get the filename and filenode for that line
                         f, fnode = dline.split('\0')
                         fnode = bin(fnode[:40])
                         f = changedfiles.get(f, None)
                         # And if the file is in the list of files we care
                         # about.
                         if f is not None:
                             # Get the changenode this manifest belongs to
                             clnode = msng_mnfst_set[mnfstnode]
                             # Create the set of filenodes for the file if
                             # there isn't one already.
                             ndset = msng_filenode_set.setdefault(f, {})
                             # And set the filenode's changelog node to the
                             # manifest's if it hasn't been set already.
                             ndset.setdefault(fnode, clnode)
                 else:
                     # Otherwise we need a full manifest.
                     m = mnfst.read(mnfstnode)
                     # For every file in we care about.
                     for f in changedfiles:
                         fnode = m.get(f, None)
                         # If it's in the manifest
                         if fnode is not None:
                             # See comments above.
                             clnode = msng_mnfst_set[mnfstnode]
                             ndset = msng_filenode_set.setdefault(f, {})
                             ndset.setdefault(fnode, clnode)
                 # Remember the revision we hope to see next.
                 next_rev[0] = r + 1
             return collect_msng_filenodes
 
         # We have a list of filenodes we think we need for a file, lets remove
         # all those we now the recipient must have.
         def prune_filenodes(f, filerevlog):
             msngset = msng_filenode_set[f]
             hasset = {}
             # If a 'missing' filenode thinks it belongs to a changenode we
             # assume the recipient must have, then the recipient must have
             # that filenode.
             for n in msngset:
                 clnode = cl.node(filerevlog.linkrev(n))
                 if clnode in has_cl_set:
                     hasset[n] = 1
             prune_parents(filerevlog, hasset, msngset)
 
         # A function generator function that sets up the a context for the
         # inner function.
         def lookup_filenode_link_func(fname):
             msngset = msng_filenode_set[fname]
             # Lookup the changenode the filenode belongs to.
             def lookup_filenode_link(fnode):
                 return msngset[fnode]
             return lookup_filenode_link
 
         # Now that we have all theses utility functions to help out and
         # logically divide up the task, generate the group.
         def gengroup():
             # The set of changed files starts empty.
             changedfiles = {}
             # Create a changenode group generator that will call our functions
             # back to lookup the owning changenode and collect information.
             group = cl.group(msng_cl_lst, identity,
                              manifest_and_file_collector(changedfiles))
             for chnk in group:
                 yield chnk
 
             # The list of manifests has been collected by the generator
             # calling our functions back.
             prune_manifests()
             msng_mnfst_lst = msng_mnfst_set.keys()
             # Sort the manifestnodes by revision number.
             msng_mnfst_lst.sort(cmp_by_rev_func(mnfst))
             # Create a generator for the manifestnodes that calls our lookup
             # and data collection functions back.
             group = mnfst.group(msng_mnfst_lst, lookup_manifest_link,
                                 filenode_collector(changedfiles))
             for chnk in group:
                 yield chnk
 
             # These are no longer needed, dereference and toss the memory for
             # them.
             msng_mnfst_lst = None
             msng_mnfst_set.clear()
 
             changedfiles = changedfiles.keys()
             changedfiles.sort()
             # Go through all our files in order sorted by name.
             for fname in changedfiles:
                 filerevlog = self.file(fname)
                 # Toss out the filenodes that the recipient isn't really
                 # missing.
                 if msng_filenode_set.has_key(fname):
                     prune_filenodes(fname, filerevlog)
                     msng_filenode_lst = msng_filenode_set[fname].keys()
                 else:
                     msng_filenode_lst = []
                 # If any filenodes are left, generate the group for them,
                 # otherwise don't bother.
                 if len(msng_filenode_lst) > 0:
                     yield changegroup.genchunk(fname)
                     # Sort the filenodes by their revision #
                     msng_filenode_lst.sort(cmp_by_rev_func(filerevlog))
                     # Create a group generator and only pass in a changenode
                     # lookup function as we need to collect no information
                     # from filenodes.
                     group = filerevlog.group(msng_filenode_lst,
                                              lookup_filenode_link_func(fname))
                     for chnk in group:
                         yield chnk
                 if msng_filenode_set.has_key(fname):
                     # Don't need this anymore, toss it to free memory.
                     del msng_filenode_set[fname]
             # Signal that no more groups are left.
             yield changegroup.closechunk()
 
             self.hook('outgoing', node=hex(msng_cl_lst[0]), source=source)
 
         return util.chunkbuffer(gengroup())
 
     def changegroup(self, basenodes, source):
         """Generate a changegroup of all nodes that we have that a recipient
         doesn't.
 
         This is much easier than the previous function as we can assume that
         the recipient has any changenode we aren't sending them."""
 
         self.hook('preoutgoing', throw=True, source=source)
 
         cl = self.changelog
         nodes = cl.nodesbetween(basenodes, None)[0]
         revset = dict.fromkeys([cl.rev(n) for n in nodes])
 
         def identity(x):
             return x
 
         def gennodelst(revlog):
             for r in xrange(0, revlog.count()):
                 n = revlog.node(r)
                 if revlog.linkrev(n) in revset:
                     yield n
 
         def changed_file_collector(changedfileset):
             def collect_changed_files(clnode):
                 c = cl.read(clnode)
                 for fname in c[3]:
                     changedfileset[fname] = 1
             return collect_changed_files
 
         def lookuprevlink_func(revlog):
             def lookuprevlink(n):
                 return cl.node(revlog.linkrev(n))
             return lookuprevlink
 
         def gengroup():
             # construct a list of all changed files
             changedfiles = {}
 
             for chnk in cl.group(nodes, identity,
                                  changed_file_collector(changedfiles)):
                 yield chnk
             changedfiles = changedfiles.keys()
             changedfiles.sort()
 
             mnfst = self.manifest
             nodeiter = gennodelst(mnfst)
             for chnk in mnfst.group(nodeiter, lookuprevlink_func(mnfst)):
                 yield chnk
 
             for fname in changedfiles:
                 filerevlog = self.file(fname)
                 nodeiter = gennodelst(filerevlog)
                 nodeiter = list(nodeiter)
                 if nodeiter:
                     yield changegroup.genchunk(fname)
                     lookup = lookuprevlink_func(filerevlog)
                     for chnk in filerevlog.group(nodeiter, lookup):
                         yield chnk
 
             yield changegroup.closechunk()
             self.hook('outgoing', node=hex(nodes[0]), source=source)
 
         return util.chunkbuffer(gengroup())
 
     def addchangegroup(self, source):
 
         def csmap(x):
             self.ui.debug(_("add changeset %s\n") % short(x))
-            return self.changelog.count()
+            return cl.count()
 
         def revmap(x):
-            return self.changelog.rev(x)
+            return cl.rev(x)
 
         if not source:
             return
 
         self.hook('prechangegroup', throw=True)
 
         changesets = files = revisions = 0
 
         tr = self.transaction()
 
-        oldheads = len(self.changelog.heads())
+        # write changelog and manifest data to temp files so
+        # concurrent readers will not see inconsistent view
+        cl = appendfile.appendchangelog(self.opener)
+
+        oldheads = len(cl.heads())
 
         # pull off the changeset group
         self.ui.status(_("adding changesets\n"))
-        co = self.changelog.tip()
+        co = cl.tip()
         chunkiter = changegroup.chunkiter(source)
-        cn = self.changelog.addgroup(chunkiter, csmap, tr, 1) # unique
-        cnr, cor = map(self.changelog.rev, (cn, co))
+        cn = cl.addgroup(chunkiter, csmap, tr, 1) # unique
+        cnr, cor = map(cl.rev, (cn, co))
         if cn == nullid:
             cnr = cor
         changesets = cnr - cor
 
+        mf = appendfile.appendmanifest(self.opener)
+
         # pull off the manifest group
         self.ui.status(_("adding manifests\n"))
-        mm = self.manifest.tip()
+        mm = mf.tip()
         chunkiter = changegroup.chunkiter(source)
-        mo = self.manifest.addgroup(chunkiter, revmap, tr)
+        mo = mf.addgroup(chunkiter, revmap, tr)
 
         # process the files
         self.ui.status(_("adding file changes\n"))
         while 1:
             f = changegroup.getchunk(source)
             if not f:
                 break
             self.ui.debug(_("adding %s revisions\n") % f)
             fl = self.file(f)
             o = fl.count()
             chunkiter = changegroup.chunkiter(source)
             n = fl.addgroup(chunkiter, revmap, tr)
             revisions += fl.count() - o
             files += 1
 
+        # write order here is important so concurrent readers will see
+        # consistent view of repo
+        mf.writedata()
+        cl.writedata()
+
+        # make changelog and manifest see real files again
+        self.changelog = changelog.changelog(self.opener)
+        self.manifest = manifest.manifest(self.opener)
+
         newheads = len(self.changelog.heads())
         heads = ""
         if oldheads and newheads > oldheads:
             heads = _(" (+%d heads)") % (newheads - oldheads)
 
         self.ui.status(_("added %d changesets"
                          " with %d changes to %d files%s\n")
                          % (changesets, revisions, files, heads))
 
         self.hook('pretxnchangegroup', throw=True,
                   node=hex(self.changelog.node(cor+1)))
 
         tr.close()
 
         if changesets > 0:
             self.hook("changegroup", node=hex(self.changelog.node(cor+1)))
 
             for i in range(cor + 1, cnr + 1):
                 self.hook("incoming", node=hex(self.changelog.node(i)))
 
     def update(self, node, allow=False, force=False, choose=None,
                moddirstate=True, forcemerge=False, wlock=None):
         pl = self.dirstate.parents()
         if not force and pl[1] != nullid:
             self.ui.warn(_("aborting: outstanding uncommitted merges\n"))
             return 1
 
         err = False
 
         p1, p2 = pl[0], node
         pa = self.changelog.ancestor(p1, p2)
         m1n = self.changelog.read(p1)[0]
         m2n = self.changelog.read(p2)[0]
         man = self.manifest.ancestor(m1n, m2n)
         m1 = self.manifest.read(m1n)
         mf1 = self.manifest.readflags(m1n)
         m2 = self.manifest.read(m2n).copy()
         mf2 = self.manifest.readflags(m2n)
         ma = self.manifest.read(man)
         mfa = self.manifest.readflags(man)
 
         modified, added, removed, deleted, unknown = self.changes()
 
         # is this a jump, or a merge?  i.e. is there a linear path
         # from p1 to p2?
         linear_path = (pa == p1 or pa == p2)
 
         if allow and linear_path:
             raise util.Abort(_("there is nothing to merge, "
                                "just use 'hg update'"))
         if allow and not forcemerge:
             if modified or added or removed:
                 raise util.Abort(_("outstanding uncommitted changes"))
         if not forcemerge and not force:
             for f in unknown:
                 if f in m2:
                     t1 = self.wread(f)
                     t2 = self.file(f).read(m2[f])
                     if cmp(t1, t2) != 0:
                         raise util.Abort(_("'%s' already exists in the working"
                                            " dir and differs from remote") % f)
 
         # resolve the manifest to determine which files
         # we care about merging
         self.ui.note(_("resolving manifests\n"))
         self.ui.debug(_(" force %s allow %s moddirstate %s linear %s\n") %
                       (force, allow, moddirstate, linear_path))
         self.ui.debug(_(" ancestor %s local %s remote %s\n") %
                       (short(man), short(m1n), short(m2n)))
 
         merge = {}
         get = {}
         remove = []
 
         # construct a working dir manifest
         mw = m1.copy()
         mfw = mf1.copy()
         umap = dict.fromkeys(unknown)
 
         for f in added + modified + unknown:
             mw[f] = ""
             mfw[f] = util.is_exec(self.wjoin(f), mfw.get(f, False))
 
         if moddirstate and not wlock:
             wlock = self.wlock()
 
         for f in deleted + removed:
             if f in mw:
                 del mw[f]
 
             # 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 moddirstate and linear_path and f not in m2:
                 self.dirstate.forget((f,))
 
         # Compare manifests
         for f, n in mw.iteritems():
             if choose and not choose(f):
                 continue
             if f in m2:
                 s = 0
 
                 # is the wfile new since m1, and match m2?
                 if f not in m1:
                     t1 = self.wread(f)
                     t2 = self.file(f).read(m2[f])
                     if cmp(t1, t2) == 0:
                         n = m2[f]
                     del t1, t2
 
                 # are files different?
                 if n != m2[f]:
                     a = ma.get(f, nullid)
                     # are both different from the ancestor?
                     if n != a and m2[f] != a:
                         self.ui.debug(_(" %s versions differ, resolve\n") % f)
                         # merge executable bits
                         # "if we changed or they changed, change in merge"
                         a, b, c = mfa.get(f, 0), mfw[f], mf2[f]
                         mode = ((a^b) | (a^c)) ^ a
                         merge[f] = (m1.get(f, nullid), m2[f], mode)
                         s = 1
                     # are we clobbering?
                     # is remote's version newer?
                     # or are we going back in time?
                     elif force or m2[f] != a or (p2 == pa and mw[f] == m1[f]):
                         self.ui.debug(_(" remote %s is newer, get\n") % f)
                         get[f] = m2[f]
                         s = 1
                 elif f in umap:
                     # this unknown file is the same as the checkout
                     get[f] = m2[f]
 
                 if not s and mfw[f] != mf2[f]:
                     if force:
                         self.ui.debug(_(" updating permissions for %s\n") % f)
                         util.set_exec(self.wjoin(f), mf2[f])
                     else:
                         a, b, c = mfa.get(f, 0), mfw[f], mf2[f]
                         mode = ((a^b) | (a^c)) ^ a
                         if mode != b:
                             self.ui.debug(_(" updating permissions for %s\n")
                                           % f)
                             util.set_exec(self.wjoin(f), mode)
                 del m2[f]
             elif f in ma:
                 if n != ma[f]:
                     r = _("d")
                     if not force and (linear_path or allow):
                         r = self.ui.prompt(
                             (_(" local changed %s which remote deleted\n") % f) +
                              _("(k)eep or (d)elete?"), _("[kd]"), _("k"))
                     if r == _("d"):
                         remove.append(f)
                 else:
                     self.ui.debug(_("other deleted %s\n") % f)
                     remove.append(f) # other deleted it
             else:
                 # file is created on branch or in working directory
                 if force and f not in umap:
                     self.ui.debug(_("remote deleted %s, clobbering\n") % f)
                     remove.append(f)
                 elif n == m1.get(f, nullid): # same as parent
                     if p2 == pa: # going backwards?
                         self.ui.debug(_("remote deleted %s\n") % f)
                         remove.append(f)
                     else:
                         self.ui.debug(_("local modified %s, keeping\n") % f)
                 else:
                     self.ui.debug(_("working dir created %s, keeping\n") % f)
 
         for f, n in m2.iteritems():
             if choose and not choose(f):
                 continue
             if f[0] == "/":
                 continue
             if f in ma and n != ma[f]:
                 r = _("k")
                 if not force and (linear_path or allow):
                     r = self.ui.prompt(
                         (_("remote changed %s which local deleted\n") % f) +
                          _("(k)eep or (d)elete?"), _("[kd]"), _("k"))
                 if r == _("k"):
                     get[f] = n
             elif f not in ma:
                 self.ui.debug(_("remote created %s\n") % f)
                 get[f] = n
             else:
                 if force or p2 == pa: # going backwards?
                     self.ui.debug(_("local deleted %s, recreating\n") % f)
                     get[f] = n
                 else:
                     self.ui.debug(_("local deleted %s\n") % f)
 
         del mw, m1, m2, ma
 
         if force:
             for f in merge:
                 get[f] = merge[f][1]
             merge = {}
 
         if linear_path or force:
             # we don't need to do any magic, just jump to the new rev
             branch_merge = False
             p1, p2 = p2, nullid
         else:
             if not allow:
                 self.ui.status(_("this update spans a branch"
                                  " affecting the following files:\n"))
                 fl = merge.keys() + get.keys()
                 fl.sort()
                 for f in fl:
                     cf = ""
                     if f in merge:
                         cf = _(" (resolve)")
                     self.ui.status(" %s%s\n" % (f, cf))
                 self.ui.warn(_("aborting update spanning branches!\n"))
                 self.ui.status(_("(use update -m to merge across branches"
                                  " or -C to lose changes)\n"))
                 return 1
             branch_merge = True
 
         # get the files we don't need to change
         files = get.keys()
         files.sort()
         for f in files:
             if f[0] == "/":
                 continue
             self.ui.note(_("getting %s\n") % f)
             t = self.file(f).read(get[f])
             self.wwrite(f, t)
             util.set_exec(self.wjoin(f), mf2[f])
             if moddirstate:
                 if branch_merge:
                     self.dirstate.update([f], 'n', st_mtime=-1)
                 else:
                     self.dirstate.update([f], 'n')
 
         # merge the tricky bits
         failedmerge = []
         files = merge.keys()
         files.sort()
         xp1 = hex(p1)
         xp2 = hex(p2)
         for f in files:
             self.ui.status(_("merging %s\n") % f)
             my, other, flag = merge[f]
             ret = self.merge3(f, my, other, xp1, xp2)
             if ret:
                 err = True
                 failedmerge.append(f)
             util.set_exec(self.wjoin(f), flag)
             if moddirstate:
                 if branch_merge:
                     # We've done a branch merge, mark this file as merged
                     # so that we properly record the merger later
                     self.dirstate.update([f], 'm')
                 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.
                     f_len = len(self.file(f).read(other))
                     self.dirstate.update([f], 'n', st_size=f_len, st_mtime=-1)
 
         remove.sort()
         for f in remove:
             self.ui.note(_("removing %s\n") % f)
             util.audit_path(f)
             try:
                 util.unlink(self.wjoin(f))
             except OSError, inst:
                 if inst.errno != errno.ENOENT:
                     self.ui.warn(_("update failed to remove %s: %s!\n") %
                                  (f, inst.strerror))
         if moddirstate:
             if branch_merge:
                 self.dirstate.update(remove, 'r')
             else:
                 self.dirstate.forget(remove)
 
         if moddirstate:
             self.dirstate.setparents(p1, p2)
 
         stat = ((len(get), _("updated")),
                 (len(merge) - len(failedmerge), _("merged")),
                 (len(remove), _("removed")),
                 (len(failedmerge), _("unresolved")))
         note = ", ".join([_("%d files %s") % s for s in stat])
         self.ui.note("%s\n" % note)
         if moddirstate and branch_merge:
             self.ui.note(_("(branch merge, don't forget to commit)\n"))
 
         return err
 
     def merge3(self, fn, my, other, p1, p2):
         """perform a 3-way merge in the working directory"""
 
         def temp(prefix, node):
             pre = "%s~%s." % (os.path.basename(fn), prefix)
             (fd, name) = tempfile.mkstemp("", pre)
             f = os.fdopen(fd, "wb")
             self.wwrite(fn, fl.read(node), f)
             f.close()
             return name
 
         fl = self.file(fn)
         base = fl.ancestor(my, other)
         a = self.wjoin(fn)
         b = temp("base", base)
         c = temp("other", other)
 
         self.ui.note(_("resolving %s\n") % fn)
         self.ui.debug(_("file %s: my %s other %s ancestor %s\n") %
                               (fn, short(my), short(other), short(base)))
 
         cmd = (os.environ.get("HGMERGE") or self.ui.config("ui", "merge")
                or "hgmerge")
         r = util.system('%s "%s" "%s" "%s"' % (cmd, a, b, c), cwd=self.root,
                         environ={'HG_FILE': fn,
                                  'HG_MY_NODE': p1,
                                  'HG_OTHER_NODE': p2,
                                  'HG_FILE_MY_NODE': hex(my),
                                  'HG_FILE_OTHER_NODE': hex(other),
                                  'HG_FILE_BASE_NODE': hex(base)})
         if r:
             self.ui.warn(_("merging %s failed!\n") % fn)
 
         os.unlink(b)
         os.unlink(c)
         return r
 
     def verify(self):
         filelinkrevs = {}
         filenodes = {}
         changesets = revisions = files = 0
         errors = [0]
         neededmanifests = {}
 
         def err(msg):
             self.ui.warn(msg + "\n")
             errors[0] += 1
 
         def checksize(obj, name):
             d = obj.checksize()
             if d[0]:
                 err(_("%s data length off by %d bytes") % (name, d[0]))
             if d[1]:
                 err(_("%s index contains %d extra bytes") % (name, d[1]))
 
         seen = {}
         self.ui.status(_("checking changesets\n"))
         checksize(self.changelog, "changelog")
 
         for i in range(self.changelog.count()):
             changesets += 1
             n = self.changelog.node(i)
             l = self.changelog.linkrev(n)
             if l != i:
                 err(_("incorrect link (%d) for changeset revision %d") %(l, i))
             if n in seen:
                 err(_("duplicate changeset at revision %d") % i)
             seen[n] = 1
 
             for p in self.changelog.parents(n):
                 if p not in self.changelog.nodemap:
                     err(_("changeset %s has unknown parent %s") %
                                  (short(n), short(p)))
             try:
                 changes = self.changelog.read(n)
             except KeyboardInterrupt:
                 self.ui.warn(_("interrupted"))
                 raise
             except Exception, inst:
                 err(_("unpacking changeset %s: %s") % (short(n), inst))
                 continue
 
             neededmanifests[changes[0]] = n
 
             for f in changes[3]:
                 filelinkrevs.setdefault(f, []).append(i)
 
         seen = {}
         self.ui.status(_("checking manifests\n"))
         checksize(self.manifest, "manifest")
 
         for i in range(self.manifest.count()):
             n = self.manifest.node(i)
             l = self.manifest.linkrev(n)
 
             if l < 0 or l >= self.changelog.count():
                 err(_("bad manifest link (%d) at revision %d") % (l, i))
 
             if n in neededmanifests:
                 del neededmanifests[n]
 
             if n in seen:
                 err(_("duplicate manifest at revision %d") % i)
 
             seen[n] = 1
 
             for p in self.manifest.parents(n):
                 if p not in self.manifest.nodemap:
                     err(_("manifest %s has unknown parent %s") %
                         (short(n), short(p)))
 
             try:
                 delta = mdiff.patchtext(self.manifest.delta(n))
             except KeyboardInterrupt:
                 self.ui.warn(_("interrupted"))
                 raise
             except Exception, inst:
                 err(_("unpacking manifest %s: %s") % (short(n), inst))
                 continue
 
             try:
                 ff = [ l.split('\0') for l in delta.splitlines() ]
                 for f, fn in ff:
                     filenodes.setdefault(f, {})[bin(fn[:40])] = 1
             except (ValueError, TypeError), inst:
                 err(_("broken delta in manifest %s: %s") % (short(n), inst))
 
         self.ui.status(_("crosschecking files in changesets and manifests\n"))
 
         for m, c in neededmanifests.items():
             err(_("Changeset %s refers to unknown manifest %s") %
                 (short(m), short(c)))
         del neededmanifests
 
         for f in filenodes:
             if f not in filelinkrevs:
                 err(_("file %s in manifest but not in changesets") % f)
 
         for f in filelinkrevs:
             if f not in filenodes:
                 err(_("file %s in changeset but not in manifest") % f)
 
         self.ui.status(_("checking files\n"))
         ff = filenodes.keys()
         ff.sort()
         for f in ff:
             if f == "/dev/null":
                 continue
             files += 1
             if not f:
                 err(_("file without name in manifest %s") % short(n))
                 continue
             fl = self.file(f)
             checksize(fl, f)
 
             nodes = {nullid: 1}
             seen = {}
             for i in range(fl.count()):
                 revisions += 1
                 n = fl.node(i)
 
                 if n in seen:
                     err(_("%s: duplicate revision %d") % (f, i))
                 if n not in filenodes[f]:
                     err(_("%s: %d:%s not in manifests") % (f, i, short(n)))
                 else:
                     del filenodes[f][n]
 
                 flr = fl.linkrev(n)
                 if flr not in filelinkrevs.get(f, []):
                     err(_("%s:%s points to unexpected changeset %d")
                             % (f, short(n), flr))
                 else:
                     filelinkrevs[f].remove(flr)
 
                 # verify contents
                 try:
                     t = fl.read(n)
                 except KeyboardInterrupt:
                     self.ui.warn(_("interrupted"))
                     raise
                 except Exception, inst:
                     err(_("unpacking file %s %s: %s") % (f, short(n), inst))
 
                 # verify parents
                 (p1, p2) = fl.parents(n)
                 if p1 not in nodes:
                     err(_("file %s:%s unknown parent 1 %s") %
                         (f, short(n), short(p1)))
                 if p2 not in nodes:
                     err(_("file %s:%s unknown parent 2 %s") %
                             (f, short(n), short(p1)))
                 nodes[n] = 1
 
             # cross-check
             for node in filenodes[f]:
                 err(_("node %s in manifests not in %s") % (hex(node), f))
 
         self.ui.status(_("%d files, %d changesets, %d total revisions\n") %
                        (files, changesets, revisions))
 
         if errors[0]:
             self.ui.warn(_("%d integrity errors encountered!\n") % errors[0])
             return 1
 
 # used to avoid circular references so destructors work
 def aftertrans(base):
     p = base
     def a():
         util.rename(os.path.join(p, "journal"), os.path.join(p, "undo"))
         util.rename(os.path.join(p, "journal.dirstate"),
                     os.path.join(p, "undo.dirstate"))
     return a