annotate mercurial/localrepo.py @ 8562:e3495c399006

named branches: server branchmap wire protocol support (issue736)

The repository command, 'branchmap', returns a dictionary, branchname
-> [branchheads], and will be implemented for localrepo, httprepo and
sshrepo.

The following wire format is used for returning data:

branchname1 branch1head2 branch1head2 ...
branchname2 ...
...

Branch names are URL encoded to escape white space, and branch heads
are sent as hex encoded node ids. All branches and all their heads are
sent.

The background and motivation for this command is the desire for a
richer named branch semantics when pushing changesets. The details are
explained in the original proposal which is included below.


1. BACKGROUND

The algorithm currently implemented in Mercurial only considers the
graph theoretical heads when determining whether new heads are
created, rather than using the branch heads as a count (the algorithm
considers a branch head effectively closed when it is merged into
another branch or a new named branch is started from that point
onward).

Our particular problem with the algorithm is that we'd like to see the
following case working without forcing a push:

Upsteam has:

(0:dev) ---- (1:dev)
\
 `--- (2:stable)

Someone merges stable into dev:

(0:dev) ---- (1:dev) ------(3:dev)
\                         /
 `--- (2:stable) --------´

This can be pushed without --force (as it should).
Now someone else does some coding on stable (a bug fix, say):

(0:dev) ---- (1:dev) ------(3:dev)
\                          /
 `--- (2:stable) ---------´---------(4:stable)

This time we need --force to push.

We allow this to be pushed without using --force by getting all the
remote branch heads (by extending the wire protocol with a new
function).

We would, furthermore, also prefer if it is impossible to push a new
branch without --force (or a later --newbranch option so --force isn't
shoe-horned into too many disparate functions, if need be), except of
course in the case where the remote repository is empty.

This is what our patches accomplish.


2. ALTERNATIVES

We have, of course, considered some alternatives to reconstructing
enough information to decide whether we are creating new remote branch
heads, before we added the new wire protocol command.

2.1. LOOKUP ON REMOTE

The main alternative is to use the information from remote.heads() and
remote.lookup() to try to reconstruct enough graph information to
decide whether we are creating new heads. This is not adequate as
illustrated below.

Remember that each lookup is typically a request-response pair over
SSH or HTTP(S).

If we have a simple repository at the remote end like this:

(0:dev) ---- (1:dev) ---- (3:stable)
\
 `--- (2:dev)

then remote.heads() will yield [2, 3]. Assume we have nodes [0, 1, 2]
locally and want to create a new node, 4:dev, as a descendant from
(1:dev), which should be OK as 1:dev is a branch head.

If we do remote.lookup('dev') we will get [2]. Thus, we can get
information about whether a branch exists on the remote server or not,
but this does not solve our problem of figuring out whether we are
creating new heads or not.

Pushing 4:dev ought to be OK, since after the push, we still only have
two heads on branch a.

Using remote.lookup() and remote.heads() is thus not adequate to
consistently decide whether we are creating new remote heads (e.g. in
this situation the latter would never return 1:dev).

2.2. USING INCOMING TO RECONSTRUCT THE GRAPH

An alternative would be to use information equivalent to hg incoming
to get the full remote graph in addition to the local graph.

To do this, we would have to get a changegroup(subset) bundle
representing the remote end (which may be a substantial amount of
data), getting the branch heads from an instantiated bundlerepository,
deleting the bundle, and finally, we can compute the prepush logic.

While this is backwards compatible, it will cause a possibly
substantial slowdown of the push command as it first needs to pull in
all changes.


3. FURTHER ARGUMENTS IN FAVOUR OF THE BRANCHMAP WIRE-PROTOCOL EXTENSION

Currently, the commands incoming and pull, work based on the tip of a
given branch if used with "-r branchname", making it hard to get all
revisions of a certain branch only (if it has multiple heads). This
can be solved by requesting the remote's branchheads and letting the
revisions to be used with the command be these heads. This can be done
by extending the commands with a new option, e.g.:

hg pull -b branchname

which will be turned into the equivalent of:

hg pull -r branchhead1 -r branchhead2 -r branchhead3

We have a simple follow-up patch that can do this ready as well
(although not submitted yet as it is pending the acceptance of the
branch patch).


4. WRAP-UP

We generally find that the branchmap wire protocol extension can
provide better named branch support to Mercurial. Currently, some
things, like the initial push scenario in this mail, are fairly
counter-intuitive, and the more often you have to force push, the more
it is likely you will get a lot of spurious and unnecessary merge
nodes. Also, restricting incoming and pull to all changes on a branch
rather than changes on the tip-most head would be a sensible extension
to making named branches a first class citizen in Mercurial.
Currently, named branches sometimes feel like a late-coming unwanted
step-child.

We have run it in a production environment for a while, with fewer
multiple heads occurring in our repositories and fewer confused users
as a result.

Also, it fixes the long-standing issue 736.

Co-contributor: Sune Foldager <[e-mail]>

# localrepo.py - read/write repository class for mercurial
#
# Copyright 2005-2007 Matt Mackall <[e-mail]>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2, incorporated herein by reference.

from node import bin, hex, nullid, nullrev, short
from i18n import _
import repo, changegroup
import changelog, dirstate, filelog, manifest, context
import lock, transaction, store, encoding
import util, extensions, hook, error
import match as match_
import merge as merge_
from lock import release
import weakref, stat, errno, os, time, inspect
propertycache = util.propertycache

class localrepository(repo.repository):
    capabilities = set(('lookup', 'changegroupsubset', 'branchmap'))
    supported = set('revlogv1 store fncache'.split())

    def __init__(self, baseui, path=None, create=0):
        repo.repository.__init__(self)
        self.root = os.path.realpath(path)
        self.path = os.path.join(self.root, ".hg")
        self.origroot = path
        self.opener = util.opener(self.path)
        self.wopener = util.opener(self.root)

        if not os.path.isdir(self.path):
            if create:
                if not os.path.exists(path):
                    os.mkdir(path)
                os.mkdir(self.path)
                requirements = ["revlogv1"]
                if baseui.configbool('format', 'usestore', True):
                    os.mkdir(os.path.join(self.path, "store"))
                    requirements.append("store")
                    if baseui.configbool('format', 'usefncache', True):
                        requirements.append("fncache")
                    # create an invalid changelog
                    self.opener("00changelog.i", "a").write(
                        '\0\0\0\2' # represents revlogv2
                        ' dummy changelog to prevent using the old repo layout'
                    )
                reqfile = self.opener("requires", "w")
                for r in requirements:
                    reqfile.write("%s\n" % r)
                reqfile.close()
            else:
                raise error.RepoError(_("repository %s not found") % path)
        elif create:
            raise error.RepoError(_("repository %s already exists") % path)
        else:
            # find requirements
            requirements = set()
            try:
                requirements = set(self.opener("requires").read().splitlines())
            except IOError, inst:
                if inst.errno != errno.ENOENT:
                    raise
            for r in requirements - self.supported:
                raise error.RepoError(_("requirement '%s' not supported") % r)

        self.store = store.store(requirements, self.path, util.opener)
        self.spath = self.store.path
        self.sopener = self.store.opener
        self.sjoin = self.store.join
        self.opener.createmode = self.store.createmode

        self.baseui = baseui
        self.ui = baseui.copy()
        try:
            self.ui.readconfig(self.join("hgrc"), self.root)
            extensions.loadall(self.ui)
        except IOError:
            pass

        self.tagscache = None
        self._tagstypecache = None
        self.branchcache = None
        self._ubranchcache = None  # UTF-8 version of branchcache
        self._branchcachetip = None
        self.nodetagscache = None
        self.filterpats = {}
        self._datafilters = {}
        self._transref = self._lockref = self._wlockref = None

    @propertycache
    def changelog(self):
        c = changelog.changelog(self.sopener)
        if 'HG_PENDING' in os.environ:
            p = os.environ['HG_PENDING']
            if p.startswith(self.root):
                c.readpending('00changelog.i.a')
        self.sopener.defversion = c.version
        return c

    @propertycache
    def manifest(self):
        return manifest.manifest(self.sopener)

    @propertycache
    def dirstate(self):
        return dirstate.dirstate(self.opener, self.ui, self.root)

    def __getitem__(self, changeid):
        if changeid is None:
            return context.workingctx(self)
        return context.changectx(self, changeid)

    def __nonzero__(self):
        return True

    def __len__(self):
        return len(self.changelog)

    def __iter__(self):
        for i in xrange(len(self)):
            yield i

    def url(self):
        return 'file:' + self.root

    def hook(self, name, throw=False, **args):
        return hook.hook(self.ui, self, name, throw, **args)

    tag_disallowed = ':\r\n'

    def _tag(self, names, node, message, local, user, date, extra={}):
        if isinstance(names, str):
            allchars = names
            names = (names,)
        else:
            allchars = ''.join(names)
        for c in self.tag_disallowed:
            if c in allchars:
                raise util.Abort(_('%r cannot be used in a tag name') % c)

        for name in names:
            self.hook('pretag', throw=True, node=hex(node), tag=name,
                      local=local)

        def writetags(fp, names, munge, prevtags):
            fp.seek(0, 2)
            if prevtags and prevtags[-1] != '\n':
                fp.write('\n')
            for name in names:
                m = munge and munge(name) or name
                if self._tagstypecache and name in self._tagstypecache:
                    old = self.tagscache.get(name, nullid)
                    fp.write('%s %s\n' % (hex(old), m))
                fp.write('%s %s\n' % (hex(node), m))
            fp.close()

        prevtags = ''
        if local:
            try:
                fp = self.opener('localtags', 'r+')
            except IOError:
                fp = self.opener('localtags', 'a')
            else:
                prevtags = fp.read()

            # local tags are stored in the current charset
            writetags(fp, names, None, prevtags)
            for name in names:
                self.hook('tag', node=hex(node), tag=name, local=local)
            return

        try:
            fp = self.wfile('.hgtags', 'rb+')
        except IOError:
            fp = self.wfile('.hgtags', 'ab')
        else:
            prevtags = fp.read()

        # committed tags are stored in UTF-8
        writetags(fp, names, encoding.fromlocal, prevtags)

        if '.hgtags' not in self.dirstate:
            self.add(['.hgtags'])

        tagnode = self.commit(['.hgtags'], message, user, date, extra=extra)

        for name in names:
            self.hook('tag', node=hex(node), tag=name, local=local)

        return tagnode

    def tag(self, names, node, message, local, user, date):
        '''tag a revision with one or more symbolic names.

        names is a list of strings or, when adding a single tag, names may be a
        string.

        if local is True, the tags are stored in a per-repository file.
        otherwise, they are stored in the .hgtags file, and a new
        changeset is committed with the change.

        keyword arguments:

        local: whether to store tags in non-version-controlled file
        (default False)

        message: commit message to use if committing

        user: name of user to use if committing

        date: date tuple to use if committing'''

        for x in self.status()[:5]:
            if '.hgtags' in x:
                raise util.Abort(_('working copy of .hgtags is changed '
                                   '(please commit .hgtags manually)'))

        self.tags() # instantiate the cache
        self._tag(names, node, message, local, user, date)

    def tags(self):
        '''return a mapping of tag to node'''
        if self.tagscache:
            return self.tagscache

        globaltags = {}
        tagtypes = {}

        def readtags(lines, fn, tagtype):
            filetags = {}
            count = 0

            def warn(msg):
                self.ui.warn(_("%s, line %s: %s\n") % (fn, count, msg))

            for l in lines:
                count += 1
                if not l:
                    continue
                s = l.split(" ", 1)
                if len(s) != 2:
                    warn(_("cannot parse entry"))
                    continue
                node, key = s
                key = encoding.tolocal(key.strip()) # stored in UTF-8
                try:
                    bin_n = bin(node)
                except TypeError:
                    warn(_("node '%s' is not well formed") % node)
                    continue
                if bin_n not in self.changelog.nodemap:
                    warn(_("tag '%s' refers to unknown node") % key)
                    continue

                h = []
                if key in filetags:
                    n, h = filetags[key]
                    h.append(n)
                filetags[key] = (bin_n, h)

            for k, nh in filetags.iteritems():
                if k not in globaltags:
                    globaltags[k] = nh
                    tagtypes[k] = tagtype
                    continue

                # we prefer the global tag if:
                #  it supercedes us OR
                #  mutual supercedes and it has a higher rank
                # otherwise we win because we're tip-most
                an, ah = nh
                bn, bh = globaltags[k]
                if (bn != an and an in bh and
                    (bn not in ah or len(bh) > len(ah))):
                    an = bn
                ah.extend([n for n in bh if n not in ah])
                globaltags[k] = an, ah
                tagtypes[k] = tagtype

        # read the tags file from each head, ending with the tip
        f = None
        for rev, node, fnode in self._hgtagsnodes():
            f = (f and f.filectx(fnode) or
                 self.filectx('.hgtags', fileid=fnode))
            readtags(f.data().splitlines(), f, "global")

        try:
            data = encoding.fromlocal(self.opener("localtags").read())
            # localtags are stored in the local character set
            # while the internal tag table is stored in UTF-8
            readtags(data.splitlines(), "localtags", "local")
        except IOError:
            pass

        self.tagscache = {}
        self._tagstypecache = {}
        for k, nh in globaltags.iteritems():
            n = nh[0]
            if n != nullid:
                self.tagscache[k] = n
            self._tagstypecache[k] = tagtypes[k]
        self.tagscache['tip'] = self.changelog.tip()
        return self.tagscache

    def tagtype(self, tagname):
        '''
        return the type of the given tag. result can be:

        'local'  : a local tag
        'global' : a global tag
        None     : tag does not exist
        '''

        self.tags()

        return self._tagstypecache.get(tagname)

    def _hgtagsnodes(self):
        last = {}
        ret = []
        for node in reversed(self.heads()):
            c = self[node]
            rev = c.rev()
            try:
                fnode = c.filenode('.hgtags')
            except error.LookupError:
                continue
            ret.append((rev, node, fnode))
            if fnode in last:
                ret[last[fnode]] = None
            last[fnode] = len(ret) - 1
        return [item for item in ret if item]

    def tagslist(self):
        '''return a list of tags ordered by revision'''
        l = []
        for t, n in self.tags().iteritems():
            try:
                r = self.changelog.rev(n)
            except:
                r = -2 # sort to the beginning of the list if unknown
            l.append((r, t, n))
        return [(t, n) for r, t, n in sorted(l)]

    def nodetags(self, node):
        '''return the tags associated with a node'''
        if not self.nodetagscache:
            self.nodetagscache = {}
            for t, n in self.tags().iteritems():
                self.nodetagscache.setdefault(n, []).append(t)
        return self.nodetagscache.get(node, [])

    def _branchtags(self, partial, lrev):
        # TODO: rename this function?
        tiprev = len(self) - 1
        if lrev != tiprev:
            self._updatebranchcache(partial, lrev+1, tiprev+1)
            self._writebranchcache(partial, self.changelog.tip(), tiprev)

        return partial

    def branchmap(self):
        tip = self.changelog.tip()
        if self.branchcache is not None and self._branchcachetip == tip:
            return self.branchcache

        oldtip = self._branchcachetip
        self._branchcachetip = tip
        if self.branchcache is None:
            self.branchcache = {} # avoid recursion in changectx
        else:
            self.branchcache.clear() # keep using the same dict
        if oldtip is None or oldtip not in self.changelog.nodemap:
            partial, last, lrev = self._readbranchcache()
        else:
            lrev = self.changelog.rev(oldtip)
            partial = self._ubranchcache

        self._branchtags(partial, lrev)
        # this private cache holds all heads (not just tips)
        self._ubranchcache = partial

        # the branch cache is stored on disk as UTF-8, but in the local
        # charset internally
        for k, v in partial.iteritems():
            self.branchcache[encoding.tolocal(k)] = v
        return self.branchcache


    def branchtags(self):
        '''return a dict where branch names map to the tipmost head of
        the branch, open heads come before closed'''
        bt = {}
        for bn, heads in self.branchmap().iteritems():
            head = None
            for i in range(len(heads)-1, -1, -1):
                h = heads[i]
                if 'close' not in self.changelog.read(h)[5]:
                    head = h
                    break
            # no open heads were found
            if head is None:
                head = heads[-1]
            bt[bn] = head
        return bt


    def _readbranchcache(self):
        partial = {}
        try:
            f = self.opener("branchheads.cache")
            lines = f.read().split('\n')
            f.close()
        except (IOError, OSError):
            return {}, nullid, nullrev

        try:
            last, lrev = lines.pop(0).split(" ", 1)
            last, lrev = bin(last), int(lrev)
            if lrev >= len(self) or self[lrev].node() != last:
                # invalidate the cache
                raise ValueError('invalidating branch cache (tip differs)')
            for l in lines:
                if not l: continue
                node, label = l.split(" ", 1)
                partial.setdefault(label.strip(), []).append(bin(node))
        except KeyboardInterrupt:
            raise
        except Exception, inst:
            if self.ui.debugflag:
                self.ui.warn(str(inst), '\n')
            partial, last, lrev = {}, nullid, nullrev
        return partial, last, lrev

    def _writebranchcache(self, branches, tip, tiprev):
        try:
            f = self.opener("branchheads.cache", "w", atomictemp=True)
            f.write("%s %s\n" % (hex(tip), tiprev))
            for label, nodes in branches.iteritems():
                for node in nodes:
                    f.write("%s %s\n" % (hex(node), label))
            f.rename()
        except (IOError, OSError):
            pass

    def _updatebranchcache(self, partial, start, end):
        for r in xrange(start, end):
            c = self[r]
            b = c.branch()
            bheads = partial.setdefault(b, [])
            bheads.append(c.node())
            for p in c.parents():
                pn = p.node()
                if pn in bheads:
                    bheads.remove(pn)

    def lookup(self, key):
        if isinstance(key, int):
            return self.changelog.node(key)
        elif key == '.':
            return self.dirstate.parents()[0]
        elif key == 'null':
            return nullid
        elif key == 'tip':
            return self.changelog.tip()
        n = self.changelog._match(key)
        if n:
            return n
        if key in self.tags():
            return self.tags()[key]
        if key in self.branchtags():
            return self.branchtags()[key]
        n = self.changelog._partialmatch(key)
        if n:
            return n
        try:
            if len(key) == 20:
                key = hex(key)
        except:
            pass
        raise error.RepoError(_("unknown revision '%s'") % key)

    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 rjoin(self, f):
        return os.path.join(self.root, util.pconvert(f))

    def file(self, f):
        if f[0] == '/':
            f = f[1:]
        return filelog.filelog(self.sopener, f)

    def changectx(self, changeid):
        return self[changeid]

    def parents(self, changeid=None):
        '''get list of changectxs for parents of changeid'''
        return self[changeid].parents()

    def filectx(self, path, changeid=None, fileid=None):
        """changeid can be a changeset revision, node, or tag.
           fileid can be a file revision or node."""
        return context.filectx(self, path, changeid, fileid)

    def getcwd(self):
        return self.dirstate.getcwd()

    def pathto(self, f, cwd=None):
        return self.dirstate.pathto(f, cwd)

    def wfile(self, f, mode='r'):
        return self.wopener(f, mode)

    def _link(self, f):
        return os.path.islink(self.wjoin(f))

    def _filter(self, filter, filename, data):
        if filter not in self.filterpats:
            l = []
            for pat, cmd in self.ui.configitems(filter):
                if cmd == '!':
                    continue
                mf = util.matcher(self.root, "", [pat], [], [])[1]
                fn = None
                params = cmd
                for name, filterfn in self._datafilters.iteritems():
                    if cmd.startswith(name):
                        fn = filterfn
                        params = cmd[len(name):].lstrip()
                        break
                if not fn:
                    fn = lambda s, c, **kwargs: util.filter(s, c)
                # Wrap old filters not supporting keyword arguments
                if not inspect.getargspec(fn)[2]:
                    oldfn = fn
                    fn = lambda s, c, **kwargs: oldfn(s, c)
                l.append((mf, fn, params))
            self.filterpats[filter] = l

        for mf, fn, cmd in self.filterpats[filter]:
            if mf(filename):
                self.ui.debug(_("filtering %s through %s\n") % (filename, cmd))
                data = fn(data, cmd, ui=self.ui, repo=self, filename=filename)
                break

        return data

    def adddatafilter(self, name, filter):
        self._datafilters[name] = filter

    def wread(self, filename):
        if self._link(filename):
            data = os.readlink(self.wjoin(filename))
        else:
            data = self.wopener(filename, 'r').read()
        return self._filter("encode", filename, data)

    def wwrite(self, filename, data, flags):
        data = self._filter("decode", filename, data)
        try:
            os.unlink(self.wjoin(filename))
        except OSError:
            pass
        if 'l' in flags:
            self.wopener.symlink(data, filename)
        else:
            self.wopener(filename, 'w').write(data)
            if 'x' in flags:
                util.set_flags(self.wjoin(filename), False, True)

    def wwritedata(self, filename, data):
        return self._filter("decode", filename, data)

    def transaction(self):
        tr = self._transref and self._transref() or None
        if tr and tr.running():
            return tr.nest()

        # abort here if the journal already exists
        if os.path.exists(self.sjoin("journal")):
            raise error.RepoError(_("journal already exists - run hg recover"))

        # save dirstate for rollback
        try:
            ds = self.opener("dirstate").read()
        except IOError:
            ds = ""
        self.opener("journal.dirstate", "w").write(ds)
        self.opener("journal.branch", "w").write(self.dirstate.branch())

        renames = [(self.sjoin("journal"), self.sjoin("undo")),
                   (self.join("journal.dirstate"), self.join("undo.dirstate")),
                   (self.join("journal.branch"), self.join("undo.branch"))]
        tr = transaction.transaction(self.ui.warn, self.sopener,
                                     self.sjoin("journal"),
                                     aftertrans(renames),
                                     self.store.createmode)
        self._transref = weakref.ref(tr)
        return tr

    def recover(self):
        lock = self.lock()
        try:
            if os.path.exists(self.sjoin("journal")):
                self.ui.status(_("rolling back interrupted transaction\n"))
                transaction.rollback(self.sopener, self.sjoin("journal"), self.ui.warn)
                self.invalidate()
                return True
            else:
                self.ui.warn(_("no interrupted transaction available\n"))
                return False
        finally:
            lock.release()

    def rollback(self):
        wlock = lock = None
        try:
            wlock = self.wlock()
            lock = self.lock()
            if os.path.exists(self.sjoin("undo")):
                self.ui.status(_("rolling back last transaction\n"))
                transaction.rollback(self.sopener, self.sjoin("undo"), self.ui.warn)
                util.rename(self.join("undo.dirstate"), self.join("dirstate"))
                try:
                    branch = self.opener("undo.branch").read()
                    self.dirstate.setbranch(branch)
                except IOError:
                    self.ui.warn(_("Named branch could not be reset, "
                                   "current branch still is: %s\n")
                                 % encoding.tolocal(self.dirstate.branch()))
                self.invalidate()
                self.dirstate.invalidate()
            else:
                self.ui.warn(_("no rollback information available\n"))
        finally:
            release(lock, wlock)

    def invalidate(self):
        for a in "changelog manifest".split():
            if a in self.__dict__:
                delattr(self, a)
        self.tagscache = None
        self._tagstypecache = None
        self.nodetagscache = None
        self.branchcache = None
        self._ubranchcache = None
        self._branchcachetip = None

    def _lock(self, lockname, wait, releasefn, acquirefn, desc):
        try:
            l = lock.lock(lockname, 0, releasefn, desc=desc)
        except error.LockHeld, inst:
            if not wait:
                raise
            self.ui.warn(_("waiting for lock on %s held by %r\n") %
                         (desc, inst.locker))
            # default to 600 seconds timeout
            l = lock.lock(lockname, int(self.ui.config("ui", "timeout", "600")),
                          releasefn, desc=desc)
        if acquirefn:
            acquirefn()
        return l

    def lock(self, wait=True):
        l = self._lockref and self._lockref()
        if l is not None and l.held:
            l.lock()
            return l

        l = self._lock(self.sjoin("lock"), wait, None, self.invalidate,
                       _('repository %s') % self.origroot)
        self._lockref = weakref.ref(l)
        return l

    def wlock(self, wait=True):
        l = self._wlockref and self._wlockref()
        if l is not None and l.held:
            l.lock()
            return l

        l = self._lock(self.join("wlock"), wait, self.dirstate.write,
                       self.dirstate.invalidate, _('working directory of %s') %
                       self.origroot)
        self._wlockref = weakref.ref(l)
        return l

    def _filecommit(self, fctx, manifest1, manifest2, linkrev, tr, changelist):
        """
        commit an individual file as part of a larger transaction
        """

        fname = fctx.path()
        text = fctx.data()
        flog = self.file(fname)
        fparent1 = manifest1.get(fname, nullid)
        fparent2 = fparent2o = manifest2.get(fname, nullid)

        meta = {}
        copy = fctx.renamed()
        if copy and copy[0] != fname:
            # Mark the new revision of this file as a copy of another
            # file.  This copy data will effectively act as a parent
            # of this new revision.  If this is a merge, the first
            # parent will be the nullid (meaning "look up the copy data")
            # and the second one will be the other parent.  For example:
            #
            # 0 --- 1 --- 3   rev1 changes file foo
            #   \       /     rev2 renames foo to bar and changes it
            #    \- 2 -/      rev3 should have bar with all changes and
            #                      should record that bar descends from
            #                      bar in rev2 and foo in rev1
            #
            # this allows this merge to succeed:
            #
            # 0 --- 1 --- 3   rev4 reverts the content change from rev2
            #   \       /     merging rev3 and rev4 should use bar@rev2
            #    \- 2 --- 4        as the merge base
            #

            cfname = copy[0]
            crev = manifest1.get(cfname)
            newfparent = fparent2

            if manifest2: # branch merge
                if fparent2 == nullid or crev is None: # copied on remote side
                    if cfname in manifest2:
                        crev = manifest2[cfname]
                        newfparent = fparent1

            # find source in nearest ancestor if we've lost track
            if not crev:
                self.ui.debug(_(" %s: searching for copy revision for %s\n") %
                              (fname, cfname))
                for ancestor in self['.'].ancestors():
                    if cfname in ancestor:
                        crev = ancestor[cfname].filenode()
                        break

            self.ui.debug(_(" %s: copy %s:%s\n") % (fname, cfname, hex(crev)))
            meta["copy"] = cfname
            meta["copyrev"] = hex(crev)
            fparent1, fparent2 = nullid, newfparent
        elif fparent2 != nullid:
            # is one parent an ancestor of the other?
            fparentancestor = flog.ancestor(fparent1, fparent2)
            if fparentancestor == fparent1:
                fparent1, fparent2 = fparent2, nullid
            elif fparentancestor == fparent2:
                fparent2 = nullid

        # is the file changed?
        if fparent2 != nullid or flog.cmp(fparent1, text) or meta:
            changelist.append(fname)
            return flog.add(text, meta, tr, linkrev, fparent1, fparent2)

        # are just the flags changed during merge?
        if fparent1 !=  fparent2o and manifest1.flags(fname) != fctx.flags():
            changelist.append(fname)

        return fparent1

    def commit(self, files=None, text="", user=None, date=None, match=None,
               force=False, editor=False, extra={}):
        """Add a new revision to current repository.

        Revision information is gathered from the working directory, files and
        match can be used to filter the committed files.
        If editor is supplied, it is called to get a commit message.
        """
        wlock = self.wlock()
        try:
            p1, p2 = self.dirstate.parents()

            if (not force and p2 != nullid and match and
                (match.files() or match.anypats())):
                raise util.Abort(_('cannot partially commit a merge '
                                   '(do not specify files or patterns)'))

            if files:
                modified, removed = [], []
                for f in sorted(set(files)):
                    s = self.dirstate[f]
                    if s in 'nma':
                        modified.append(f)
                    elif s == 'r':
                        removed.append(f)
                    else:
                        self.ui.warn(_("%s not tracked!\n") % f)
                changes = [modified, [], removed, [], []]
            else:
                changes = self.status(match=match)

            if (not force and not extra.get("close") and p2 == nullid
                and not (changes[0] or changes[1] or changes[2])
                and self[None].branch() == self['.'].branch()):
                self.ui.status(_("nothing changed\n"))
                return None

            ms = merge_.mergestate(self)
            for f in changes[0]:
                if f in ms and ms[f] == 'u':
                    raise util.Abort(_("unresolved merge conflicts "
                                                    "(see hg resolve)"))

            wctx = context.workingctx(self, (p1, p2), text, user, date,
                                      extra, changes)
            if editor:
                wctx._text = editor(self, wctx,
                                    changes[1], changes[0], changes[2])
            ret = self.commitctx(wctx, True)

            # update dirstate and mergestate
            for f in changes[0] + changes[1]:
                self.dirstate.normal(f)
            for f in changes[2]:
                self.dirstate.forget(f)
            self.dirstate.setparents(ret)
            ms.reset()

            return ret

        finally:
            wlock.release()

    def commitctx(self, ctx, error=False):
        """Add a new revision to current repository.

        Revision information is passed via the context argument.
        """

        tr = lock = None
        removed = ctx.removed()
        p1, p2 = ctx.p1(), ctx.p2()
        m1 = p1.manifest().copy()
        m2 = p2.manifest()
        user = ctx.user()

        xp1, xp2 = p1.hex(), p2 and p2.hex() or ''
        self.hook("precommit", throw=True, parent1=xp1, parent2=xp2)

        lock = self.lock()
        try:
            tr = self.transaction()
            trp = weakref.proxy(tr)

            # check in files
            new = {}
            changed = []
            linkrev = len(self)
            for f in sorted(ctx.modified() + ctx.added()):
                self.ui.note(f + "\n")
                try:
                    fctx = ctx[f]
                    new[f] = self._filecommit(fctx, m1, m2, linkrev, trp,
                                              changed)
                    m1.set(f, fctx.flags())
                except (OSError, IOError):
                    if error:
                        self.ui.warn(_("trouble committing %s!\n") % f)
                        raise
                    else:
                        removed.append(f)

            # update manifest
            m1.update(new)
            removed = [f for f in sorted(removed) if f in m1 or f in m2]
            drop = [f for f in removed if f in m1]
            for f in drop:
                del m1[f]
            mn = self.manifest.add(m1, trp, linkrev, p1.manifestnode(),
                                   p2.manifestnode(), (new, drop))

            # update changelog
            self.changelog.delayupdate()
            n = self.changelog.add(mn, changed + removed, ctx.description(),
                                   trp, p1.node(), p2.node(),
                                   user, ctx.date(), ctx.extra().copy())
            p = lambda: self.changelog.writepending() and self.root or ""
            self.hook('pretxncommit', throw=True, node=hex(n), parent1=xp1,
                      parent2=xp2, pending=p)
            self.changelog.finalize(trp)
            tr.close()

            if self.branchcache:
                self.branchtags()

            self.hook("commit", node=hex(n), parent1=xp1, parent2=xp2)
            return n
        finally:
            del tr
            lock.release()

    def walk(self, match, node=None):
        '''
        walk recursively through the directory tree or a given
        changeset, finding all files matched by the match
        function
        '''
        return self[node].walk(match)

    def status(self, node1='.', node2=None, match=None,
               ignored=False, clean=False, unknown=False):
        """return status of files 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 mfmatches(ctx):
            mf = ctx.manifest().copy()
            for fn in mf.keys():
                if not match(fn):
                    del mf[fn]
            return mf

        if isinstance(node1, context.changectx):
            ctx1 = node1
        else:
            ctx1 = self[node1]
        if isinstance(node2, context.changectx):
            ctx2 = node2
        else:
            ctx2 = self[node2]

        working = ctx2.rev() is None
        parentworking = working and ctx1 == self['.']
        match = match or match_.always(self.root, self.getcwd())
        listignored, listclean, listunknown = ignored, clean, unknown

        # load earliest manifest first for caching reasons
        if not working and ctx2.rev() < ctx1.rev():
            ctx2.manifest()

        if not parentworking:
            def bad(f, msg):
                if f not in ctx1:
                    self.ui.warn('%s: %s\n' % (self.dirstate.pathto(f), msg))
                return False
            match.bad = bad

        if working: # we need to scan the working dir
            s = self.dirstate.status(match, listignored, listclean, listunknown)
            cmp, modified, added, removed, deleted, unknown, ignored, clean = s

            # check for any possibly clean files
            if parentworking and cmp:
                fixup = []
                # do a full compare of any files that might have changed
                for f in sorted(cmp):
                    if (f not in ctx1 or ctx2.flags(f) != ctx1.flags(f)
                        or ctx1[f].cmp(ctx2[f].data())):
                        modified.append(f)
                    else:
                        fixup.append(f)

                if listclean:
                    clean += fixup

                # update dirstate for files that are actually clean
                if fixup:
                    wlock = None
                    try:
                        try:
                            # updating the dirstate is optional
                            # so we don't wait on the lock
                            wlock = self.wlock(False)
                            for f in fixup:
                                self.dirstate.normal(f)
                        except error.LockError:
                            pass
                    finally:
                        release(wlock)

        if not parentworking:
            mf1 = mfmatches(ctx1)
            if working:
                # we are comparing working dir against non-parent
                # generate a pseudo-manifest for the working dir
                mf2 = mfmatches(self['.'])
                for f in cmp + modified + added:
                    mf2[f] = None
                    mf2.set(f, ctx2.flags(f))
                for f in removed:
                    if f in mf2:
                        del mf2[f]
            else:
                # we are comparing two revisions
                deleted, unknown, ignored = [], [], []
                mf2 = mfmatches(ctx2)

            modified, added, clean = [], [], []
            for fn in mf2:
                if fn in mf1:
                    if (mf1.flags(fn) != mf2.flags(fn) or
                        (mf1[fn] != mf2[fn] and
                         (mf2[fn] or ctx1[fn].cmp(ctx2[fn].data())))):
                        modified.append(fn)
                    elif listclean:
                        clean.append(fn)
                    del mf1[fn]
                else:
                    added.append(fn)
            removed = mf1.keys()

        r = modified, added, removed, deleted, unknown, ignored, clean
        [l.sort() for l in r]
        return r

    def add(self, list):
        wlock = self.wlock()
        try:
            rejected = []
            for f in list:
                p = self.wjoin(f)
                try:
                    st = os.lstat(p)
                except:
                    self.ui.warn(_("%s does not exist!\n") % f)
                    rejected.append(f)
                    continue
                if st.st_size > 10000000:
                    self.ui.warn(_("%s: files over 10MB may cause memory and"
                                   " performance problems\n"
                                   "(use 'hg revert %s' to unadd the file)\n")
                                   % (f, f))
                if not (stat.S_ISREG(st.st_mode) or stat.S_ISLNK(st.st_mode)):
                    self.ui.warn(_("%s not added: only files and symlinks "
                                   "supported currently\n") % f)
                    rejected.append(p)
                elif self.dirstate[f] in 'amn':
                    self.ui.warn(_("%s already tracked!\n") % f)
                elif self.dirstate[f] == 'r':
                    self.dirstate.normallookup(f)
                else:
                    self.dirstate.add(f)
            return rejected
        finally:
            wlock.release()

    def forget(self, list):
        wlock = self.wlock()
        try:
            for f in list:
                if self.dirstate[f] != 'a':
                    self.ui.warn(_("%s not added!\n") % f)
                else:
                    self.dirstate.forget(f)
        finally:
            wlock.release()

    def remove(self, list, unlink=False):
        wlock = None
        try:
            if unlink:
                for f in list:
                    try:
                        util.unlink(self.wjoin(f))
                    except OSError, inst:
                        if inst.errno != errno.ENOENT:
                            raise
            wlock = self.wlock()
            for f in list:
                if unlink and os.path.exists(self.wjoin(f)):
                    self.ui.warn(_("%s still exists!\n") % f)
                elif self.dirstate[f] == 'a':
                    self.dirstate.forget(f)
                elif f not in self.dirstate:
                    self.ui.warn(_("%s not tracked!\n") % f)
                else:
                    self.dirstate.remove(f)
        finally:
            release(wlock)

    def undelete(self, list):
        manifests = [self.manifest.read(self.changelog.read(p)[0])
                     for p in self.dirstate.parents() if p != nullid]
        wlock = self.wlock()
        try:
            for f in list:
                if self.dirstate[f] != 'r':
                    self.ui.warn(_("%s not removed!\n") % f)
                else:
                    m = f in manifests[0] and manifests[0] or manifests[1]
                    t = self.file(f).read(m[f])
                    self.wwrite(f, t, m.flags(f))
                    self.dirstate.normal(f)
        finally:
            wlock.release()

    def copy(self, source, dest):
        p = self.wjoin(dest)
        if not (os.path.exists(p) or os.path.islink(p)):
            self.ui.warn(_("%s does not exist!\n") % dest)
        elif not (os.path.isfile(p) or os.path.islink(p)):
            self.ui.warn(_("copy failed: %s is not a file or a "
                           "symbolic link\n") % dest)
        else:
            wlock = self.wlock()
            try:
                if self.dirstate[dest] in '?r':
                    self.dirstate.add(dest)
                self.dirstate.copy(source, dest)
            finally:
                wlock.release()

    def heads(self, start=None, closed=True):
        heads = self.changelog.heads(start)
        def display(head):
            if closed:
                return True
            extras = self.changelog.read(head)[5]
            return ('close' not in extras)
        # sort the output in rev descending order
        heads = [(-self.changelog.rev(h), h) for h in heads if display(h)]
        return [n for (r, n) in sorted(heads)]

    def branchheads(self, branch=None, start=None, closed=True):
        if branch is None:
            branch = self[None].branch()
        branches = self.branchmap()
        if branch not in branches:
            return []
        bheads = branches[branch]
        # the cache returns heads ordered lowest to highest
        bheads.reverse()
        if start is not None:
            # filter out the heads that cannot be reached from startrev
            bheads = self.changelog.nodesbetween([start], bheads)[2]
        if not closed:
            bheads = [h for h in bheads if
                      ('close' not in self.changelog.read(h)[5])]
        return bheads

    def branches(self, nodes):
        if not nodes:
            nodes = [self.changelog.tip()]
        b = []
        for n in nodes:
            t = n
            while 1:
                p = self.changelog.parents(n)
                if p[1] != nullid or p[0] == nullid:
                    b.append((t, n, p[0], p[1]))
                    break
                n = p[0]
        return b

    def between(self, pairs):
        r = []

        for top, bottom in pairs:
            n, l, i = top, [], 0
            f = 1

            while n != bottom and n != nullid:
                p = self.changelog.parents(n)[0]
                if i == f:
                    l.append(n)
                    f = f * 2
                n = p
                i += 1

            r.append(l)

        return r

    def findincoming(self, remote, base=None, heads=None, force=False):
        """Return list of roots of the subsets of missing nodes from remote

        If base dict is specified, assume that these nodes and their parents
        exist on the remote side and that no child of a node of base exists
        in both remote and self.
        Furthermore base will be updated to include the nodes that exists
        in self and remote but no children exists in self and remote.
        If a list of heads is specified, return only nodes which are heads
        or ancestors of these heads.

        All the ancestors of base are in self and in remote.
        All the descendants of the list returned are missing in self.
        (and so we know that the rest of the nodes are missing in remote, see
        outgoing)
        """
        return self.findcommonincoming(remote, base, heads, force)[1]

    def findcommonincoming(self, remote, base=None, heads=None, force=False):
        """Return a tuple (common, missing roots, heads) used to identify
        missing nodes from remote.

        If base dict is specified, assume that these nodes and their parents
        exist on the remote side and that no child of a node of base exists
        in both remote and self.
        Furthermore base will be updated to include the nodes that exists
        in self and remote but no children exists in self and remote.
        If a list of heads is specified, return only nodes which are heads
        or ancestors of these heads.

        All the ancestors of base are in self and in remote.
        """
        m = self.changelog.nodemap
        search = []
        fetch = set()
        seen = set()
        seenbranch = set()
        if base is None:
            base = {}

        if not heads:
            heads = remote.heads()

        if self.changelog.tip() == nullid:
            base[nullid] = 1
            if heads != [nullid]:
                return [nullid], [nullid], list(heads)
            return [nullid], [], []

        # assume we're closer to the tip than the root
        # and start by examining the heads
        self.ui.status(_("searching for changes\n"))

        unknown = []
        for h in heads:
            if h not in m:
                unknown.append(h)
            else:
                base[h] = 1

        heads = unknown
        if not unknown:
            return base.keys(), [], []

        req = set(unknown)
        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: # found the end of the branch
                    pass
                elif n in seenbranch:
                    self.ui.debug(_("branch already found\n"))
                    continue
                elif 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[0:2]) # schedule branch range for scanning
                    seenbranch.add(n)
                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.add(n[1]) # earliest unknown
                        for p in n[2:4]:
                            if p in m:
                                base[p] = 1 # latest known

                    for p in n[2:4]:
                        if p not in req and p not in m:
                            r.append(p)
                            req.add(p)
                seen.add(n[0])

            if r:
                reqcnt += 1
                self.ui.debug(_("request %d: %s\n") %
                            (reqcnt, " ".join(map(short, r))))
                for p in xrange(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])))
                        unknown.append(b)

        # do binary search on the branches we found
        while search:
            newsearch = []
            reqcnt += 1
            for n, l in zip(search, remote.between(search)):
                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.add(p)
                            base[i] = 1
                        else:
                            self.ui.debug(_("narrowed branch search to %s:%s\n")
                                          % (short(p), short(i)))
                            newsearch.append((p, i))
                        break
                    p, f = i, f * 2
                search = newsearch

        # sanity check our fetch list
        for f in fetch:
            if f in m:
                raise error.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.debug(_("found new changesets starting at ") +
                     " ".join([short(f) for f in fetch]) + "\n")

        self.ui.debug(_("%d total queries\n") % reqcnt)

        return base.keys(), list(fetch), heads

    def findoutgoing(self, remote, base=None, heads=None, force=False):
        """Return list of nodes that are roots of subsets not in remote

        If base dict is specified, assume that these nodes and their parents
        exist on the remote side.
        If a list of heads is specified, return only nodes which are heads
        or ancestors of these heads, and return a second element which
        contains all remote heads which get new children.
        """
        if base is None:
            base = {}
            self.findincoming(remote, base, heads, force=force)

        self.ui.debug(_("common changesets up to ")
                      + " ".join(map(short, base.keys())) + "\n")

        remain = set(self.changelog.nodemap)

        # prune everything remote has from the tree
        remain.remove(nullid)
        remove = base.keys()
        while remove:
            n = remove.pop(0)
            if n in remain:
                remain.remove(n)
                for p in self.changelog.parents(n):
                    remove.append(p)

        # find every node whose parents have been pruned
        subset = []
        # find every remote head that will get new children
        updated_heads = set()
        for n in remain:
            p1, p2 = self.changelog.parents(n)
            if p1 not in remain and p2 not in remain:
                subset.append(n)
            if heads:
                if p1 in heads:
                    updated_heads.add(p1)
                if p2 in heads:
                    updated_heads.add(p2)

        # this is the set of all roots we have to push
        if heads:
            return subset, list(updated_heads)
        else:
            return subset

    def pull(self, remote, heads=None, force=False):
        lock = self.lock()
        try:
            common, fetch, rheads = self.findcommonincoming(remote, heads=heads,
                                                            force=force)
            if fetch == [nullid]:
                self.ui.status(_("requesting all changes\n"))

            if not fetch:
                self.ui.status(_("no changes found\n"))
                return 0

            if heads is None and remote.capable('changegroupsubset'):
                heads = rheads

            if heads is None:
                cg = remote.changegroup(fetch, 'pull')
            else:
                if not remote.capable('changegroupsubset'):
                    raise util.Abort(_("Partial pull cannot be done because other repository doesn't support changegroupsubset."))
                cg = remote.changegroupsubset(fetch, heads, 'pull')
            return self.addchangegroup(cg, 'pull', remote.url())
        finally:
            lock.release()

    def push(self, remote, force=False, revs=None):
        # there are two ways to push to remote repo:
        #
        # addchangegroup assumes local user can lock remote
        # repo (local filesystem, old ssh servers).
        #
        # unbundle assumes local user cannot lock remote repo (new ssh
        # servers, http servers).

        if remote.capable('unbundle'):
            return self.push_unbundle(remote, force, revs)
        return self.push_addchangegroup(remote, force, revs)

    def prepush(self, remote, force, revs):
        common = {}
        remote_heads = remote.heads()
        inc = self.findincoming(remote, common, remote_heads, force=force)

        update, updated_heads = self.findoutgoing(remote, common, remote_heads)
        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 None, 1
        elif not force:
            # check if we're creating new remote heads
            # to be a remote head after push, node must be either
            # - unknown locally
            # - a local outgoing head descended from update
            # - a remote head that's known locally and not
            #   ancestral to an outgoing head

            warn = 0

            if remote_heads == [nullid]:
                warn = 0
            elif not revs and len(heads) > len(remote_heads):
                warn = 1
            else:
                newheads = list(heads)
                for r in remote_heads:
                    if r in self.changelog.nodemap:
                        desc = self.changelog.heads(r, heads)
                        l = [h for h in heads if h in desc]
                        if not l:
                            newheads.append(r)
                    else:
                        newheads.append(r)
                if len(newheads) > len(remote_heads):
                    warn = 1

            if warn:
                self.ui.warn(_("abort: push creates new remote heads!\n"))
                self.ui.status(_("(did you forget to merge?"
                                 " use push -f to force)\n"))
                return None, 0
            elif inc:
                self.ui.warn(_("note: unsynced remote changes!\n"))


        if revs is None:
            # use the fast path, no race possible on push
            cg = self._changegroup(common.keys(), 'push')
        else:
            cg = self.changegroupsubset(update, revs, 'push')
        return cg, remote_heads

    def push_addchangegroup(self, remote, force, revs):
        lock = remote.lock()
        try:
            ret = self.prepush(remote, force, revs)
            if ret[0] is not None:
                cg, remote_heads = ret
                return remote.addchangegroup(cg, 'push', self.url())
            return ret[1]
        finally:
            lock.release()

    def push_unbundle(self, remote, force, revs):
        # local repo finds heads on server, finds out what revs it
        # must push.  once revs transferred, if server finds it has
        # different heads (someone else won commit/push race), server
        # aborts.

        ret = self.prepush(remote, force, revs)
        if ret[0] is not None:
            cg, remote_heads = ret
            if force: remote_heads = ['force']
            return remote.unbundle(cg, remote_heads, 'push')
        return ret[1]

    def changegroupinfo(self, nodes, source):
        if self.ui.verbose or source == 'bundle':
            self.ui.status(_("%d changesets found\n") % len(nodes))
        if self.ui.debugflag:
            self.ui.debug(_("list of changesets:\n"))
            for node in nodes:
                self.ui.debug("%s\n" % hex(node))

    def changegroupsubset(self, bases, heads, source, extranodes=None):
        """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.

        The caller can specify some nodes that must be included in the
        changegroup using the extranodes argument.  It should be a dict
        where the keys are the filenames (or 1 for the manifest), and the
        values are lists of (node, linknode) tuples, where node is a wanted
        node and linknode is the changelog node that should be transmitted as
        the linkrev.
        """

        if extranodes is None:
            # can we go through the fast path ?
            heads.sort()
            allheads = self.heads()
            allheads.sort()
            if heads == allheads:
                common = []
                # parents of bases are known from both sides
                for n in bases:
                    for p in self.changelog.parents(n):
                        if p != nullid:
                            common.append(p)
                return self._changegroup(common, source)

        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)
        self.changegroupinfo(msng_cl_lst, source)
        # 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 = set()
        # We assume that all parents of bases are known heads.
        for n in bases:
            knownheads.update(cl.parents(n))
        knownheads.discard(nullid)
        knownheads = list(knownheads)
        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 a set.
            has_cl_set = set(has_cl_set)
        else:
            # If there were no known heads, the recipient cannot be assumed to
            # know about any changesets.
            has_cl_set = 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[len(mnfst) - 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 = list(hasset)
            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.add(n)
                        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 = 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(mnfst.rev(n)))
                if linknode in has_cl_set:
                    has_mnfst_set.add(n)
            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.
                    deltamf = mnfst.readdelta(mnfstnode)
                    # For each line in the delta
                    for f, fnode in deltamf.iteritems():
                        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 know the recipient must have.
        def prune_filenodes(f, filerevlog):
            msngset = msng_filenode_set[f]
            hasset = set()
            # 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(filerevlog.rev(n)))
                if clnode in has_cl_set:
                    hasset.add(n)
            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

        # Add the nodes that were explicitly requested.
        def add_extra_nodes(name, nodes):
            if not extranodes or name not in extranodes:
                return

            for node, linknode in extranodes[name]:
                if node not in nodes:
                    nodes[node] = linknode

        # 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()
            add_extra_nodes(1, msng_mnfst_set)
            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()

            if extranodes:
                for fname in extranodes:
                    if isinstance(fname, int):
                        continue
                    msng_filenode_set.setdefault(fname, {})
                    changedfiles[fname] = 1
            # Go through all our files in order sorted by name.
            for fname in sorted(changedfiles):
                filerevlog = self.file(fname)
                if not len(filerevlog):
                    raise util.Abort(_("empty or missing revlog for %s") % fname)
                # Toss out the filenodes that the recipient isn't really
                # missing.
                if fname in msng_filenode_set:
                    prune_filenodes(fname, filerevlog)
                    add_extra_nodes(fname, msng_filenode_set[fname])
                    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.chunkheader(len(fname))
                    yield 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 fname in msng_filenode_set:
                    # 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()

            if msng_cl_lst:
                self.hook('outgoing', node=hex(msng_cl_lst[0]), source=source)

        return util.chunkbuffer(gengroup())

    def changegroup(self, basenodes, source):
        # to avoid a race we use changegroupsubset() (issue1320)
        return self.changegroupsubset(basenodes, self.heads(), source)

    def _changegroup(self, common, 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.

        common is the set of common nodes between remote and self"""

        self.hook('preoutgoing', throw=True, source=source)

        cl = self.changelog
        nodes = cl.findmissing(common)
        revset = set([cl.rev(n) for n in nodes])
        self.changegroupinfo(nodes, source)

        def identity(x):
            return x

        def gennodelst(log):
            for r in log:
                if log.linkrev(r) in revset:
                    yield log.node(r)

        def changed_file_collector(changedfileset):
            def collect_changed_files(clnode):
                c = cl.read(clnode)
                changedfileset.update(c[3])
            return collect_changed_files

        def lookuprevlink_func(revlog):
            def lookuprevlink(n):
                return cl.node(revlog.linkrev(revlog.rev(n)))
            return lookuprevlink

        def gengroup():
            # construct a list of all changed files
            changedfiles = set()

            for chnk in cl.group(nodes, identity,
                                 changed_file_collector(changedfiles)):
                yield chnk

            mnfst = self.manifest
            nodeiter = gennodelst(mnfst)
            for chnk in mnfst.group(nodeiter, lookuprevlink_func(mnfst)):
                yield chnk

            for fname in sorted(changedfiles):
                filerevlog = self.file(fname)
                if not len(filerevlog):
                    raise util.Abort(_("empty or missing revlog for %s") % fname)
                nodeiter = gennodelst(filerevlog)
                nodeiter = list(nodeiter)
                if nodeiter:
                    yield changegroup.chunkheader(len(fname))
                    yield fname
                    lookup = lookuprevlink_func(filerevlog)
                    for chnk in filerevlog.group(nodeiter, lookup):
                        yield chnk

            yield changegroup.closechunk()

            if nodes:
                self.hook('outgoing', node=hex(nodes[0]), source=source)

        return util.chunkbuffer(gengroup())

    def addchangegroup(self, source, srctype, url, emptyok=False):
        """add changegroup to repo.

        return values:
        - nothing changed or no source: 0
        - more heads than before: 1+added heads (2..n)
        - less heads than before: -1-removed heads (-2..-n)
        - number of heads stays the same: 1
        """
        def csmap(x):
            self.ui.debug(_("add changeset %s\n") % short(x))
            return len(cl)

        def revmap(x):
            return cl.rev(x)

        if not source:
            return 0

        self.hook('prechangegroup', throw=True, source=srctype, url=url)

        changesets = files = revisions = 0

        # write changelog data to temp files so concurrent readers will not see
        # inconsistent view
        cl = self.changelog
        cl.delayupdate()
        oldheads = len(cl.heads())

        tr = self.transaction()
        try:
            trp = weakref.proxy(tr)
            # pull off the changeset group
            self.ui.status(_("adding changesets\n"))
            clstart = len(cl)
            chunkiter = changegroup.chunkiter(source)
            if cl.addgroup(chunkiter, csmap, trp) is None and not emptyok:
                raise util.Abort(_("received changelog group is empty"))
            clend = len(cl)
            changesets = clend - clstart

            # pull off the manifest group
            self.ui.status(_("adding manifests\n"))
            chunkiter = changegroup.chunkiter(source)
            # no need to check for empty manifest group here:
            # if the result of the merge of 1 and 2 is the same in 3 and 4,
            # no new manifest will be created and the manifest group will
            # be empty during the pull
            self.manifest.addgroup(chunkiter, revmap, trp)

            # 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 = len(fl)
                chunkiter = changegroup.chunkiter(source)
                if fl.addgroup(chunkiter, revmap, trp) is None:
                    raise util.Abort(_("received file revlog group is empty"))
                revisions += len(fl) - o
                files += 1

            newheads = len(cl.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))

            if changesets > 0:
                p = lambda: cl.writepending() and self.root or ""
                self.hook('pretxnchangegroup', throw=True,
                          node=hex(cl.node(clstart)), source=srctype,
                          url=url, pending=p)

            # make changelog see real files again
            cl.finalize(trp)

            tr.close()
        finally:
            del tr

        if changesets > 0:
            # forcefully update the on-disk branch cache
            self.ui.debug(_("updating the branch cache\n"))
            self.branchtags()
            self.hook("changegroup", node=hex(cl.node(clstart)),
                      source=srctype, url=url)

            for i in xrange(clstart, clend):
                self.hook("incoming", node=hex(cl.node(i)),
                          source=srctype, url=url)

        # never return 0 here:
        if newheads < oldheads:
            return newheads - oldheads - 1
        else:
            return newheads - oldheads + 1


    def stream_in(self, remote):
        fp = remote.stream_out()
        l = fp.readline()
        try:
            resp = int(l)
        except ValueError:
            raise error.ResponseError(
                _('Unexpected response from remote server:'), l)
        if resp == 1:
            raise util.Abort(_('operation forbidden by server'))
        elif resp == 2:
            raise util.Abort(_('locking the remote repository failed'))
        elif resp != 0:
            raise util.Abort(_('the server sent an unknown error code'))
        self.ui.status(_('streaming all changes\n'))
        l = fp.readline()
        try:
            total_files, total_bytes = map(int, l.split(' ', 1))
        except (ValueError, TypeError):
            raise error.ResponseError(
                _('Unexpected response from remote server:'), l)
        self.ui.status(_('%d files to transfer, %s of data\n') %
                       (total_files, util.bytecount(total_bytes)))
        start = time.time()
        for i in xrange(total_files):
            # XXX doesn't support '\n' or '\r' in filenames
            l = fp.readline()
            try:
                name, size = l.split('\0', 1)
                size = int(size)
            except (ValueError, TypeError):
                raise error.ResponseError(
                    _('Unexpected response from remote server:'), l)
            self.ui.debug(_('adding %s (%s)\n') % (name, util.bytecount(size)))
            # for backwards compat, name was partially encoded
            ofp = self.sopener(store.decodedir(name), 'w')
            for chunk in util.filechunkiter(fp, limit=size):
                ofp.write(chunk)
            ofp.close()
        elapsed = time.time() - start
        if elapsed <= 0:
            elapsed = 0.001
        self.ui.status(_('transferred %s in %.1f seconds (%s/sec)\n') %
                       (util.bytecount(total_bytes), elapsed,
                        util.bytecount(total_bytes / elapsed)))
        self.invalidate()
        return len(self.heads()) + 1

    def clone(self, remote, heads=[], stream=False):
        '''clone remote repository.

        keyword arguments:
        heads: list of revs to clone (forces use of pull)
        stream: use streaming clone if possible'''

        # now, all clients that can request uncompressed clones can
        # read repo formats supported by all servers that can serve
        # them.

        # if revlog format changes, client will have to check version
        # and format flags on "stream" capability, and use
        # uncompressed only if compatible.

        if stream and not heads and remote.capable('stream'):
            return self.stream_in(remote)
        return self.pull(remote, heads)

# used to avoid circular references so destructors work
def aftertrans(files):
    renamefiles = [tuple(t) for t in files]
    def a():
        for src, dest in renamefiles:
            util.rename(src, dest)
    return a

def instance(ui, path, create):
    return localrepository(ui, util.drop_scheme('file', path), create)

def islocal(path):
    return True