hgext/evolve.py
author Pierre-Yves David <pierre-yves.david@fb.com>
Thu, 11 Feb 2016 00:39:56 +0000
branchstable
changeset 1612 bd59cc2ee203
parent 1611 5eedfe43e762
child 1615 b04d73c0a13e
permissions -rw-r--r--
merge with strange other head on stable

# Copyright 2011 Peter Arrenbrecht <peter.arrenbrecht@gmail.com>
#                Logilab SA        <contact@logilab.fr>
#                Pierre-Yves David <pierre-yves.david@ens-lyon.org>
#                Patrick Mezard <patrick@mezard.eu>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

'''extends Mercurial feature related to Changeset Evolution

This extension provides several commands to mutate history and deal with
issues it may raise.

It also:

    - enables the "Changeset Obsolescence" feature of mercurial,
    - alters core commands and extensions that rewrite history to use
      this feature,
    - improves some aspect of the early implementation in Mercurial core
'''

__version__ = '5.3.0'
testedwith = '3.4.3 3.5.2 3.6.2 3.7'
buglink = 'http://bz.selenic.com/'


evolutionhelptext = """
Obsolescence markers make it possible to mark changesets that have been
deleted or superset in a new version of the changeset.

Unlike the previous way of handling such changes, by stripping the old
changesets from the repository, obsolescence markers can be propagated
between repositories. This allows for a safe and simple way of exchanging
mutable history and altering it after the fact. Changeset phases are
respected, such that only draft and secret changesets can be altered (see
:hg:`hg phases` for details).

Obsolescence is tracked using "obsolete markers", a piece of metadata
tracking which changesets have been made obsolete, potential successors for
a given changeset, the moment the changeset was marked as obsolete, and the
user who performed the rewriting operation. The markers are stored
separately from standard changeset data can be exchanged without any of the
precursor changesets, preventing unnecessary exchange of obsolescence data.

The complete set of obsolescence markers describes a history of changeset
modifications that is orthogonal to the repository history of file
modifications. This changeset history allows for detection and automatic
resolution of edge cases arising from multiple users rewriting the same part
of history concurrently.

Current feature status
======================

This feature is still in development.  If you see this help, you have enable an
extension that turned this feature on.

Obsolescence markers will be exchanged between repositories that explicitly
assert support for the obsolescence feature (this can currently only be done
via an extension).""".strip()


import sys, os
import random
from StringIO import StringIO
import struct
import re
import collections
import socket
import errno
import struct
sha1re = re.compile(r'\b[0-9a-f]{6,40}\b')

import mercurial
from mercurial import util
from mercurial import repair

try:
    from mercurial import obsolete
    if not obsolete._enabled:
        obsolete._enabled = True
    from mercurial import wireproto
    gboptslist = getattr(wireproto, 'gboptslist', None)
    gboptsmap = getattr(wireproto, 'gboptsmap', None)
except (ImportError, AttributeError):
    gboptslist = gboptsmap = None

# Flags for enabling optional parts of evolve
commandopt = 'allnewcommands'

from mercurial import bookmarks as bookmarksmod
from mercurial import cmdutil
from mercurial import commands
from mercurial import context
from mercurial import copies
from mercurial import error
from mercurial import exchange
from mercurial import extensions
from mercurial import help
from mercurial import httppeer
from mercurial import hg
from mercurial import lock as lockmod
from mercurial import merge
from mercurial import node
from mercurial import phases
from mercurial import patch
from mercurial import revset
from mercurial import scmutil
from mercurial import templatekw
from mercurial.i18n import _
from mercurial.commands import walkopts, commitopts, commitopts2, mergetoolopts
from mercurial.node import nullid
from mercurial import wireproto
from mercurial import localrepo
from mercurial.hgweb import hgweb_mod

cmdtable = {}
command = cmdutil.command(cmdtable)

_pack = struct.pack
_unpack = struct.unpack

if gboptsmap is not None:
    memfilectx = context.memfilectx
elif gboptslist is not None:
    oldmemfilectx = context.memfilectx
    def memfilectx(repo, *args, **kwargs):
        return oldmemfilectx(*args, **kwargs)
else:
    raise ImportError('evolve needs version %s or above' %
                      min(testedwith.split()))

aliases, entry = cmdutil.findcmd('commit', commands.table)
hasinteractivemode = any(['interactive' in e for e in entry[1]])
if hasinteractivemode:
    interactiveopt = [['i', 'interactive', None, _('use interactive mode')]]
else:
    interactiveopt = []
# This extension contains the following code
#
# - Extension Helper code
# - Obsolescence cache
# - ...
# - Older format compat


#####################################################################
### Extension helper                                              ###
#####################################################################

class exthelper(object):
    """Helper for modular extension setup

    A single helper should be instantiated for each extension. Helper
    methods are then used as decorators for various purpose.

    All decorators return the original function and may be chained.
    """

    def __init__(self):
        self._uicallables = []
        self._extcallables = []
        self._repocallables = []
        self._revsetsymbols = []
        self._templatekws = []
        self._commandwrappers = []
        self._extcommandwrappers = []
        self._functionwrappers = []
        self._duckpunchers = []

    def final_uisetup(self, ui):
        """Method to be used as the extension uisetup

        The following operations belong here:

        - Changes to ui.__class__ . The ui object that will be used to run the
          command has not yet been created. Changes made here will affect ui
          objects created after this, and in particular the ui that will be
          passed to runcommand
        - Command wraps (extensions.wrapcommand)
        - Changes that need to be visible to other extensions: because
          initialization occurs in phases (all extensions run uisetup, then all
          run extsetup), a change made here will be visible to other extensions
          during extsetup
        - Monkeypatch or wrap function (extensions.wrapfunction) of dispatch
          module members
        - Setup of pre-* and post-* hooks
        - pushkey setup
        """
        for cont, funcname, func in self._duckpunchers:
            setattr(cont, funcname, func)
        for command, wrapper, opts in self._commandwrappers:
            entry = extensions.wrapcommand(commands.table, command, wrapper)
            if opts:
                for short, long, val, msg in opts:
                    entry[1].append((short, long, val, msg))
        for cont, funcname, wrapper in self._functionwrappers:
            extensions.wrapfunction(cont, funcname, wrapper)
        for c in self._uicallables:
            c(ui)

    def final_extsetup(self, ui):
        """Method to be used as a the extension extsetup

        The following operations belong here:

        - Changes depending on the status of other extensions. (if
          extensions.find('mq'))
        - Add a global option to all commands
        - Register revset functions
        """
        knownexts = {}
        for name, symbol in self._revsetsymbols:
            revset.symbols[name] = symbol
        for name, kw in self._templatekws:
            templatekw.keywords[name] = kw
        for ext, command, wrapper, opts in self._extcommandwrappers:
            if ext not in knownexts:
                try:
                    e = extensions.find(ext)
                except KeyError:
                    # Extension isn't enabled, so don't bother trying to wrap
                    # it.
                    continue
                knownexts[ext] = e.cmdtable
            entry = extensions.wrapcommand(knownexts[ext], command, wrapper)
            if opts:
                for short, long, val, msg in opts:
                    entry[1].append((short, long, val, msg))

        for c in self._extcallables:
            c(ui)

    def final_reposetup(self, ui, repo):
        """Method to be used as a the extension reposetup

        The following operations belong here:

        - All hooks but pre-* and post-*
        - Modify configuration variables
        - Changes to repo.__class__, repo.dirstate.__class__
        """
        for c in self._repocallables:
            c(ui, repo)

    def uisetup(self, call):
        """Decorated function will be executed during uisetup

        example::

            @eh.uisetup
            def setupbabar(ui):
                print 'this is uisetup!'
        """
        self._uicallables.append(call)
        return call

    def extsetup(self, call):
        """Decorated function will be executed during extsetup

        example::

            @eh.extsetup
            def setupcelestine(ui):
                print 'this is extsetup!'
        """
        self._extcallables.append(call)
        return call

    def reposetup(self, call):
        """Decorated function will be executed during reposetup

        example::

            @eh.reposetup
            def setupzephir(ui, repo):
                print 'this is reposetup!'
        """
        self._repocallables.append(call)
        return call

    def revset(self, symbolname):
        """Decorated function is a revset symbol

        The name of the symbol must be given as the decorator argument.
        The symbol is added during `extsetup`.

        example::

            @eh.revset('hidden')
            def revsetbabar(repo, subset, x):
                args = revset.getargs(x, 0, 0, 'babar accept no argument')
                return [r for r in subset if 'babar' in repo[r].description()]
        """
        def dec(symbol):
            self._revsetsymbols.append((symbolname, symbol))
            return symbol
        return dec


    def templatekw(self, keywordname):
        """Decorated function is a template keyword

        The name of the keyword must be given as the decorator argument.
        The symbol is added during `extsetup`.

        example::

            @eh.templatekw('babar')
            def kwbabar(ctx):
                return 'babar'
        """
        def dec(keyword):
            self._templatekws.append((keywordname, keyword))
            return keyword
        return dec

    def wrapcommand(self, command, extension=None, opts=[]):
        """Decorated function is a command wrapper

        The name of the command must be given as the decorator argument.
        The wrapping is installed during `uisetup`.

        If the second option `extension` argument is provided, the wrapping
        will be applied in the extension commandtable. This argument must be a
        string that will be searched using `extension.find` if not found and
        Abort error is raised. If the wrapping applies to an extension, it is
        installed during `extsetup`

        example::

            @eh.wrapcommand('summary')
            def wrapsummary(orig, ui, repo, *args, **kwargs):
                ui.note('Barry!')
                return orig(ui, repo, *args, **kwargs)

        The `opts` argument allows specifying additional arguments for the
        command.

        """
        def dec(wrapper):
            if extension is None:
                self._commandwrappers.append((command, wrapper, opts))
            else:
                self._extcommandwrappers.append((extension, command, wrapper,
                    opts))
            return wrapper
        return dec

    def wrapfunction(self, container, funcname):
        """Decorated function is a function wrapper

        This function takes two arguments, the container and the name of the
        function to wrap. The wrapping is performed during `uisetup`.
        (there is no extension support)

        example::

            @eh.function(discovery, 'checkheads')
            def wrapfunction(orig, *args, **kwargs):
                ui.note('His head smashed in and his heart cut out')
                return orig(*args, **kwargs)
        """
        def dec(wrapper):
            self._functionwrappers.append((container, funcname, wrapper))
            return wrapper
        return dec

    def addattr(self, container, funcname):
        """Decorated function is to be added to the container

        This function takes two arguments, the container and the name of the
        function to wrap. The wrapping is performed during `uisetup`.

        example::

            @eh.function(context.changectx, 'babar')
            def babar(ctx):
                return 'babar' in ctx.description
        """
        def dec(func):
            self._duckpunchers.append((container, funcname, func))
            return func
        return dec

eh = exthelper()
uisetup = eh.final_uisetup
extsetup = eh.final_extsetup
reposetup = eh.final_reposetup

#####################################################################
### Option configuration                                          ###
#####################################################################

@eh.reposetup # must be the first of its kin.
def _configureoptions(ui, repo):
    # If no capabilities are specified, enable everything.
    # This is so existing evolve users don't need to change their config.
    evolveopts = ui.configlist('experimental', 'evolution')
    if not evolveopts:
        evolveopts = ['all']
        ui.setconfig('experimental', 'evolution', evolveopts)

@eh.uisetup
def _configurecmdoptions(ui):
    # Unregister evolve commands if the command capability is not specified.
    #
    # This must be in the same function as the option configuration above to
    # guarantee it happens after the above configuration, but before the
    # extsetup functions.
    evolvecommands = ui.configlist('experimental', 'evolutioncommands')
    evolveopts = ui.configlist('experimental', 'evolution')
    if evolveopts and (commandopt not in evolveopts and
                       'all' not in evolveopts):
        # We build whitelist containing the commands we want to enable
        whitelist = set()
        for cmd in evolvecommands:
            matchingevolvecommands = [e for e in cmdtable.keys() if cmd in e]
            if not matchingevolvecommands:
                raise error.Abort(_('unknown command: %s') % cmd)
            elif len(matchingevolvecommands) > 1:
                msg = _('ambiguous command specification: "%s" matches %r')
                raise error.Abort(msg % (cmd, matchingevolvecommands))
            else:
                whitelist.add(matchingevolvecommands[0])
        for disabledcmd in set(cmdtable) - whitelist:
            del cmdtable[disabledcmd]

#####################################################################
### experimental behavior                                         ###
#####################################################################

commitopts3 = [
    ('D', 'current-date', None,
     _('record the current date as commit date')),
    ('U', 'current-user', None,
     _('record the current user as committer')),
]

def _resolveoptions(ui, opts):
    """modify commit options dict to handle related options

    For now, all it does is figure out the commit date: respect -D unless
    -d was supplied.
    """
    # N.B. this is extremely similar to setupheaderopts() in mq.py
    if not opts.get('date') and opts.get('current_date'):
        opts['date'] = '%d %d' % util.makedate()
    if not opts.get('user') and opts.get('current_user'):
        opts['user'] = ui.username()

getrevs = obsolete.getrevs

#####################################################################
### Additional Utilities                                          ###
#####################################################################

# This section contains a lot of small utility function and method

# - Function to create markers
# - useful alias pstatus and pdiff (should probably go in evolve)
# - "troubles" method on changectx
# - function to travel throught the obsolescence graph
# - function to find useful changeset to stabilize


### Useful alias

@eh.uisetup
def _installalias(ui):
    if ui.config('alias', 'pstatus', None) is None:
        ui.setconfig('alias', 'pstatus', 'status --rev .^')
    if ui.config('alias', 'pdiff', None) is None:
        ui.setconfig('alias', 'pdiff', 'diff --rev .^')
    if ui.config('alias', 'olog', None) is None:
        ui.setconfig('alias', 'olog', "log -r 'precursors(.)' --hidden")
    if ui.config('alias', 'odiff', None) is None:
        ui.setconfig('alias', 'odiff',
            "diff --hidden --rev 'limit(precursors(.),1)' --rev .")
    if ui.config('alias', 'grab', None) is None:
        if os.name == 'nt':
            ui.setconfig('alias', 'grab',
                "! " + util.hgexecutable() + " rebase --dest . --rev $@ && "
                 + util.hgexecutable() + " up tip")
        else:
            ui.setconfig('alias', 'grab',
                "! $HG rebase --dest . --rev $@ && $HG up tip")


### Troubled revset symbol

@eh.revset('troubled')
def revsettroubled(repo, subset, x):
    """``troubled()``
    Changesets with troubles.
    """
    revset.getargs(x, 0, 0, 'troubled takes no arguments')
    troubled = set()
    troubled.update(getrevs(repo, 'unstable'))
    troubled.update(getrevs(repo, 'bumped'))
    troubled.update(getrevs(repo, 'divergent'))
    troubled = revset.baseset(troubled)
    troubled.sort() # set is non-ordered, enforce order
    return subset & troubled

### Obsolescence graph

# XXX SOME MAJOR CLEAN UP TO DO HERE XXX

def _precursors(repo, s):
    """Precursor of a changeset"""
    cs = set()
    nm = repo.changelog.nodemap
    markerbysubj = repo.obsstore.precursors
    node = repo.changelog.node
    for r in s:
        for p in markerbysubj.get(node(r), ()):
            pr = nm.get(p[0])
            if pr is not None:
                cs.add(pr)
    cs -= repo.changelog.filteredrevs # nodemap has no filtering
    return cs

def _allprecursors(repo, s):  # XXX we need a better naming
    """transitive precursors of a subset"""
    node = repo.changelog.node
    toproceed = [node(r) for r in s]
    seen = set()
    allsubjects = repo.obsstore.precursors
    while toproceed:
        nc = toproceed.pop()
        for mark in allsubjects.get(nc, ()):
            np = mark[0]
            if np not in seen:
                seen.add(np)
                toproceed.append(np)
    nm = repo.changelog.nodemap
    cs = set()
    for p in seen:
        pr = nm.get(p)
        if pr is not None:
            cs.add(pr)
    cs -= repo.changelog.filteredrevs # nodemap has no filtering
    return cs

def _successors(repo, s):
    """Successors of a changeset"""
    cs = set()
    node = repo.changelog.node
    nm = repo.changelog.nodemap
    markerbyobj = repo.obsstore.successors
    for r in s:
        for p in markerbyobj.get(node(r), ()):
            for sub in p[1]:
                sr = nm.get(sub)
                if sr is not None:
                    cs.add(sr)
    cs -= repo.changelog.filteredrevs # nodemap has no filtering
    return cs

def _allsuccessors(repo, s, haltonflags=0):  # XXX we need a better naming
    """transitive successors of a subset

    haltonflags allows to provide flags which prevent the evaluation of a
    marker.  """
    node = repo.changelog.node
    toproceed = [node(r) for r in s]
    seen = set()
    allobjects = repo.obsstore.successors
    while toproceed:
        nc = toproceed.pop()
        for mark in allobjects.get(nc, ()):
            if mark[2] & haltonflags:
                continue
            for sub in mark[1]:
                if sub == nullid:
                    continue # should not be here!
                if sub not in seen:
                    seen.add(sub)
                    toproceed.append(sub)
    nm = repo.changelog.nodemap
    cs = set()
    for s in seen:
        sr = nm.get(s)
        if sr is not None:
            cs.add(sr)
    cs -= repo.changelog.filteredrevs # nodemap has no filtering
    return cs




#####################################################################
### Extending revset and template                                 ###
#####################################################################

# this section add several useful revset symbol not yet in core.
# they are subject to changes


### XXX I'm not sure this revset is useful
@eh.revset('suspended')
def revsetsuspended(repo, subset, x):
    """``suspended()``
    Obsolete changesets with non-obsolete descendants.
    """
    revset.getargs(x, 0, 0, 'suspended takes no arguments')
    suspended = revset.baseset(getrevs(repo, 'suspended'))
    suspended.sort()
    return subset & suspended


@eh.revset('precursors')
def revsetprecursors(repo, subset, x):
    """``precursors(set)``
    Immediate precursors of changesets in set.
    """
    s = revset.getset(repo, revset.fullreposet(repo), x)
    s = revset.baseset(_precursors(repo, s))
    s.sort()
    return subset & s


@eh.revset('allprecursors')
def revsetallprecursors(repo, subset, x):
    """``allprecursors(set)``
    Transitive precursors of changesets in set.
    """
    s = revset.getset(repo, revset.fullreposet(repo), x)
    s = revset.baseset(_allprecursors(repo, s))
    s.sort()
    return subset & s


@eh.revset('successors')
def revsetsuccessors(repo, subset, x):
    """``successors(set)``
    Immediate successors of changesets in set.
    """
    s = revset.getset(repo, revset.fullreposet(repo), x)
    s = revset.baseset(_successors(repo, s))
    s.sort()
    return subset & s

@eh.revset('allsuccessors')
def revsetallsuccessors(repo, subset, x):
    """``allsuccessors(set)``
    Transitive successors of changesets in set.
    """
    s = revset.getset(repo, revset.fullreposet(repo), x)
    s = revset.baseset(_allsuccessors(repo, s))
    s.sort()
    return subset & s

### template keywords
# XXX it does not handle troubles well :-/

@eh.templatekw('obsolete')
def obsoletekw(repo, ctx, templ, **args):
    """:obsolete: String. The obsolescence level of the node, could be
    ``stable``, ``unstable``, ``suspended`` or ``extinct``.
    """
    if ctx.obsolete():
        if ctx.extinct():
            return 'extinct'
        else:
            return 'suspended'
    elif ctx.unstable():
        return 'unstable'
    return 'stable'

@eh.templatekw('troubles')
def showtroubles(repo, ctx, **args):
    """:troubles: List of strings. Evolution troubles affecting the changeset
    (zero or more of "unstable", "divergent" or "bumped")."""
    return templatekw.showlist('trouble', ctx.troubles(), plural='troubles',
                               **args)

#####################################################################
### Various trouble warning                                       ###
#####################################################################

# This section take care of issue warning to the user when troubles appear


def _warnobsoletewc(ui, repo):
    if repo['.'].obsolete():
        ui.warn(_('working directory parent is obsolete!\n'))
        if (not ui.quiet) and obsolete.isenabled(repo, commandopt):
            ui.warn(_('(use "hg evolve" to update to its successor)\n'))

@eh.wrapcommand("update")
@eh.wrapcommand("pull")
def wrapmayobsoletewc(origfn, ui, repo, *args, **opts):
    """Warn that the working directory parent is an obsolete changeset"""
    def warnobsolete():
        _warnobsoletewc(ui, repo)
    wlock = None
    try:
        wlock = repo.wlock()
        repo._afterlock(warnobsolete)
        res = origfn(ui, repo, *args, **opts)
    finally:
        lockmod.release(wlock)
    return res

@eh.wrapcommand("parents")
def wrapparents(origfn, ui, repo, *args, **opts):
    res = origfn(ui, repo, *args, **opts)
    _warnobsoletewc(ui, repo)
    return res

# XXX this could wrap transaction code
# XXX (but this is a bit a layer violation)
@eh.wrapcommand("commit")
@eh.wrapcommand("import")
@eh.wrapcommand("push")
@eh.wrapcommand("pull")
@eh.wrapcommand("graft")
@eh.wrapcommand("phase")
@eh.wrapcommand("unbundle")
def warnobserrors(orig, ui, repo, *args, **kwargs):
    """display warning is the command resulted in more instable changeset"""
    # part of the troubled stuff may be filtered (stash ?)
    # This needs a better implementation but will probably wait for core.
    filtered = repo.changelog.filteredrevs
    priorunstables = len(set(getrevs(repo, 'unstable')) - filtered)
    priorbumpeds = len(set(getrevs(repo, 'bumped')) - filtered)
    priordivergents = len(set(getrevs(repo, 'divergent')) - filtered)
    ret = orig(ui, repo, *args, **kwargs)
    # workaround phase stupidity
    #phases._filterunknown(ui, repo.changelog, repo._phasecache.phaseroots)
    filtered = repo.changelog.filteredrevs
    newunstables = \
        len(set(getrevs(repo, 'unstable')) - filtered) - priorunstables
    newbumpeds = \
        len(set(getrevs(repo, 'bumped')) - filtered) - priorbumpeds
    newdivergents = \
        len(set(getrevs(repo, 'divergent')) - filtered) - priordivergents
    if newunstables > 0:
        ui.warn(_('%i new unstable changesets\n') % newunstables)
    if newbumpeds > 0:
        ui.warn(_('%i new bumped changesets\n') % newbumpeds)
    if newdivergents > 0:
        ui.warn(_('%i new divergent changesets\n') % newdivergents)
    return ret

@eh.wrapfunction(mercurial.exchange, 'push')
def push(orig, repo, *args, **opts):
    """Add a hint for "hg evolve" when troubles make push fails
    """
    try:
        return orig(repo, *args, **opts)
    except error.Abort as ex:
        hint = _("use 'hg evolve' to get a stable history "
                 "or --force to ignore warnings")
        if (len(ex.args) >= 1
            and ex.args[0].startswith('push includes ')
            and ex.hint is None):
            ex.hint = hint
        raise

def summaryhook(ui, repo):
    def write(fmt, count):
        s = fmt % count
        if count:
            ui.write(s)
        else:
            ui.note(s)

    # util.versiontuple was introduced in 3.6.2
    if not util.safehasattr(util, 'versiontuple'):
        nbunstable = len(getrevs(repo, 'unstable'))
        nbbumped = len(getrevs(repo, 'bumped'))
        nbdivergent = len(getrevs(repo, 'divergent'))
        write('unstable: %i changesets\n', nbunstable)
        write('bumped: %i changesets\n', nbbumped)
        write('divergent: %i changesets\n', nbdivergent)
    else:
        # In 3.6.2, summary in core gained this feature, no need to display it
        pass

@eh.extsetup
def obssummarysetup(ui):
    cmdutil.summaryhooks.add('evolve', summaryhook)


#####################################################################
### Core Other extension compat                                   ###
#####################################################################


@eh.extsetup
def _rebasewrapping(ui):
    # warning about more obsolete
    try:
        rebase = extensions.find('rebase')
        if rebase:
            extensions.wrapcommand(rebase.cmdtable, 'rebase', warnobserrors)
    except KeyError:
        pass # rebase not found
    try:
        histedit = extensions.find('histedit')
        if histedit:
            extensions.wrapcommand(histedit.cmdtable, 'histedit', warnobserrors)
    except KeyError:
        pass # histedit not found

#####################################################################
### Old Evolve extension content                                  ###
#####################################################################

# XXX need clean up and proper sorting in other section

### util function
#############################

### changeset rewriting logic
#############################

def rewrite(repo, old, updates, head, newbases, commitopts):
    """Return (nodeid, created) where nodeid is the identifier of the
    changeset generated by the rewrite process, and created is True if
    nodeid was actually created. If created is False, nodeid
    references a changeset existing before the rewrite call.
    """
    wlock = lock = tr = None
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        tr = repo.transaction('rewrite')
        if len(old.parents()) > 1: #XXX remove this unecessary limitation.
            raise error.Abort(_('cannot amend merge changesets'))
        base = old.p1()
        updatebookmarks = _bookmarksupdater(repo, old.node(), tr)

        # commit a new version of the old changeset, including the update
        # collect all files which might be affected
        files = set(old.files())
        for u in updates:
            files.update(u.files())

        # Recompute copies (avoid recording a -> b -> a)
        copied = copies.pathcopies(base, head)


        # prune files which were reverted by the updates
        def samefile(f):
            if f in head.manifest():
                a = head.filectx(f)
                if f in base.manifest():
                    b = base.filectx(f)
                    return (a.data() == b.data()
                            and a.flags() == b.flags())
                else:
                    return False
            else:
                return f not in base.manifest()
        files = [f for f in files if not samefile(f)]
        # commit version of these files as defined by head
        headmf = head.manifest()
        def filectxfn(repo, ctx, path):
            if path in headmf:
                fctx = head[path]
                flags = fctx.flags()
                mctx = memfilectx(repo, fctx.path(), fctx.data(),
                                  islink='l' in flags,
                                  isexec='x' in flags,
                                  copied=copied.get(path))
                return mctx
            return None

        message = cmdutil.logmessage(repo.ui, commitopts)
        if not message:
            message = old.description()

        user = commitopts.get('user') or old.user()
        date = commitopts.get('date') or None # old.date()
        extra = dict(commitopts.get('extra', old.extra()))
        extra['branch'] = head.branch()

        new = context.memctx(repo,
                             parents=newbases,
                             text=message,
                             files=files,
                             filectxfn=filectxfn,
                             user=user,
                             date=date,
                             extra=extra)

        if commitopts.get('edit'):
            new._text = cmdutil.commitforceeditor(repo, new, [])
        revcount = len(repo)
        newid = repo.commitctx(new)
        new = repo[newid]
        created = len(repo) != revcount
        updatebookmarks(newid)

        tr.close()
        return newid, created
    finally:
        lockmod.release(tr, lock, wlock)

class MergeFailure(error.Abort):
    pass

def relocate(repo, orig, dest, pctx=None, keepbranch=False):
    """rewrite <rev> on dest"""
    if orig.rev() == dest.rev():
        raise error.Abort(_('tried to relocate a node on top of itself'),
                         hint=_("This shouldn't happen. If you still "
                                "need to move changesets, please do so "
                                "manually with nothing to rebase - working "
                                "directory parent is also destination"))

    if pctx is None:
        if len(orig.parents()) == 2:
            raise error.Abort(_("tried to relocate a merge commit without "
                                "specifying which parent should be moved"),
                              hint=_("Specify the parent by passing in pctx"))
        pctx = orig.p1()

    destbookmarks = repo.nodebookmarks(dest.node())
    nodesrc = orig.node()
    destphase = repo[nodesrc].phase()
    commitmsg = orig.description()

    cache = {}
    sha1s = re.findall(sha1re, commitmsg)
    unfi = repo.unfiltered()
    for sha1 in sha1s:
        ctx = None
        try:
            ctx = unfi[sha1]
        except error.RepoLookupError:
            continue

        if not ctx.obsolete():
            continue

        successors = obsolete.successorssets(repo, ctx.node(), cache)

        # We can't make any assumptions about how to update the hash if the
        # cset in question was split or diverged.
        if len(successors) == 1 and len(successors[0]) == 1:
            newsha1 = node.hex(successors[0][0])
            commitmsg = commitmsg.replace(sha1, newsha1[:len(sha1)])
        else:
            repo.ui.note(_('The stale commit message reference to %s could '
                           'not be updated\n') % sha1)

    tr = repo.currenttransaction()
    assert tr is not None
    try:
        try:
            r = _evolvemerge(repo, orig, dest, pctx, keepbranch)
            if r[-1]:  #some conflict
                raise error.Abort(
                        'unresolved merge conflicts (see hg help resolve)')
            nodenew = _relocatecommit(repo, orig, commitmsg)
        except error.Abort as exc:
            repo.dirstate.beginparentchange()
            repo.setparents(repo['.'].node(), nullid)
            writedirstate(repo.dirstate, tr)
            # fix up dirstate for copies and renames
            copies.duplicatecopies(repo, dest.rev(), orig.p1().rev())
            repo.dirstate.endparentchange()
            class LocalMergeFailure(MergeFailure, exc.__class__):
                pass
            exc.__class__ = LocalMergeFailure
            tr.close() # to keep changes in this transaction (e.g. dirstate)
            raise
        oldbookmarks = repo.nodebookmarks(nodesrc)
        _finalizerelocate(repo, orig, dest, nodenew, tr)
    finally:
        pass # TODO: remove this redundant try/finally block
    return nodenew

def _bookmarksupdater(repo, oldid, tr):
    """Return a callable update(newid) updating the current bookmark
    and bookmarks bound to oldid to newid.
    """
    def updatebookmarks(newid):
        dirty = False
        oldbookmarks = repo.nodebookmarks(oldid)
        if oldbookmarks:
            for b in oldbookmarks:
                repo._bookmarks[b] = newid
            dirty = True
        if dirty:
            repo._bookmarks.recordchange(tr)
    return updatebookmarks

### bookmarks api compatibility layer ###
def bmdeactivate(repo):
    try:
        return bookmarksmod.deactivate(repo)
    except AttributeError:
        return bookmarksmod.unsetcurrent(repo)
def bmactivate(repo, book):
    try:
        return bookmarksmod.activate(repo, book)
    except AttributeError:
        return bookmarksmod.setcurrent(repo, book)

def bmactive(repo):
    try:
        return repo._activebookmark
    except AttributeError:
        return repo._bookmarkcurrent

### dirstate compatibility layer < hg 3.6

def writedirstate(dirstate, tr):
    if dirstate.write.func_defaults is not None: # mercurial 3.6 and above
        return dirstate.write(tr)
    return dirstate.write()



### new command
#############################
metadataopts = [
    ('d', 'date', '',
     _('record the specified date in metadata'), _('DATE')),
    ('u', 'user', '',
     _('record the specified user in metadata'), _('USER')),
]

@eh.uisetup
def _installimportobsolete(ui):
    entry = cmdutil.findcmd('import', commands.table)[1]
    entry[1].append(('', 'obsolete', False,
                    _('mark the old node as obsoleted by '
                      'the created commit')))

@eh.wrapfunction(mercurial.cmdutil, 'tryimportone')
def tryimportone(orig, ui, repo, hunk, parents, opts, *args, **kwargs):
    extracted = patch.extract(ui, hunk)
    if util.safehasattr(extracted, 'get'):
        # mercurial 3.6 return a dictionary there
        expected = extracted.get('nodeid')
    else:
        expected = extracted[5]
    if expected is not None:
        expected = node.bin(expected)
    oldextract = patch.extract
    try:
        patch.extract = lambda ui, hunk: extracted
        ret = orig(ui, repo, hunk, parents, opts, *args, **kwargs)
    finally:
        patch.extract = oldextract
    created = ret[1]
    if (opts['obsolete'] and None not in (created, expected)
        and created != expected):
            tr = repo.transaction('import-obs')
            try:
                metadata = {'user': ui.username()}
                repo.obsstore.create(tr, expected, (created,),
                                     metadata=metadata)
                tr.close()
            finally:
                tr.release()
    return ret


def _deprecatealias(oldalias, newalias):
    '''Deprecates an alias for a command in favour of another

    Creates a new entry in the command table for the old alias. It creates a
    wrapper that has its synopsis set to show that is has been deprecated.
    The documentation will be replace with a pointer to the new alias.
    If a user invokes the command a deprecation warning will be printed and
    the command of the *new* alias will be invoked.

    This function is loosely based on the extensions.wrapcommand function.
    '''
    try:
        aliases, entry = cmdutil.findcmd(newalias, cmdtable)
    except error.UnknownCommand:
        # Commands may be disabled
        return
    for alias, e in cmdtable.iteritems():
        if e is entry:
            break

    synopsis = '(DEPRECATED)'
    if len(entry) > 2:
        fn, opts, _syn = entry
    else:
        fn, opts, = entry
    deprecationwarning = _('%s have been deprecated in favor of %s\n') % (
        oldalias, newalias)
    def newfn(*args, **kwargs):
        ui = args[0]
        ui.warn(deprecationwarning)
        util.checksignature(fn)(*args, **kwargs)
    newfn.__doc__  = deprecationwarning
    cmdwrapper = command(oldalias, opts, synopsis)
    cmdwrapper(newfn)

@eh.extsetup
def deprecatealiases(ui):
    _deprecatealias('gup', 'next')
    _deprecatealias('gdown', 'previous')

@command('debugrecordpruneparents', [], '')
def cmddebugrecordpruneparents(ui, repo):
    """add parent data to prune markers when possible

    This command searches the repo for prune markers without parent information.
    If the pruned node is locally known, it creates a new marker with parent
    data.
    """
    pgop = 'reading markers'

    # lock from the beginning to prevent race
    wlock = lock = tr = None
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        tr = repo.transaction('recordpruneparents')
        unfi = repo.unfiltered()
        nm = unfi.changelog.nodemap
        store = repo.obsstore
        pgtotal = len(store._all)
        for idx, mark in enumerate(list(store._all)):
            if not mark[1]:
                rev = nm.get(mark[0])
                if rev is not None:
                    ctx = unfi[rev]
                    parents = tuple(p.node() for p in ctx.parents())
                    before = len(store._all)
                    store.create(tr, mark[0], mark[1], mark[2], marks[3],
                                 parents=parents)
                    if len(store._all) - before:
                        ui.write(_('created new markers for %i\n') % rev)
            ui.progress(pgop, idx, total=pgtotal)
        tr.close()
        ui.progress(pgop, None)
    finally:
        lockmod.release(tr, lock, wlock)

@command('debugobsstorestat', [], '')
def cmddebugobsstorestat(ui, repo):
    """print statistics about obsolescence markers in the repo"""
    def _updateclustermap(nodes, mark, clustersmap):
        c = (set(nodes), set([mark]))
        toproceed = set(nodes)
        while toproceed:
            n = toproceed.pop()
            other = clustersmap.get(n)
            if (other is not None
                and other is not c):
                other[0].update(c[0])
                other[1].update(c[1])
                for on in c[0]:
                    if on in toproceed:
                        continue
                    clustersmap[on] = other
                c = other
            clustersmap[n] = c

    store = repo.obsstore
    unfi = repo.unfiltered()
    nm = unfi.changelog.nodemap
    ui.write(_('markers total:              %9i\n') % len(store._all))
    sucscount = [0, 0 , 0, 0]
    known = 0
    parentsdata = 0
    metakeys = {}
    # node -> cluster mapping
    #   a cluster is a (set(nodes), set(markers)) tuple
    clustersmap = {}
    # same data using parent information
    pclustersmap = {}
    for mark in store:
        if mark[0] in nm:
            known += 1
        nbsucs = len(mark[1])
        sucscount[min(nbsucs, 3)] += 1
        meta = mark[3]
        for key, value in meta:
            metakeys.setdefault(key, 0)
            metakeys[key] += 1
        meta = dict(meta)
        parents = [meta.get('p1'), meta.get('p2')]
        parents = [node.bin(p) for p in parents if p is not None]
        if parents:
            parentsdata += 1
        # cluster handling
        nodes = set(mark[1])
        nodes.add(mark[0])
        _updateclustermap(nodes, mark, clustersmap)
        # same with parent data
        nodes.update(parents)
        _updateclustermap(nodes, mark, pclustersmap)

    # freezing the result
    for c in clustersmap.values():
        fc = (frozenset(c[0]), frozenset(c[1]))
        for n in fc[0]:
            clustersmap[n] = fc
    # same with parent data
    for c in pclustersmap.values():
        fc = (frozenset(c[0]), frozenset(c[1]))
        for n in fc[0]:
            pclustersmap[n] = fc
    ui.write(('    for known precursors:   %9i\n' % known))
    ui.write(('    with parents data:      %9i\n' % parentsdata))
    # successors data
    ui.write(('markers with no successors: %9i\n' % sucscount[0]))
    ui.write(('              1 successors: %9i\n' % sucscount[1]))
    ui.write(('              2 successors: %9i\n' % sucscount[2]))
    ui.write(('    more than 2 successors: %9i\n' % sucscount[3]))
    # meta data info
    ui.write(('    available  keys:\n'))
    for key in sorted(metakeys):
        ui.write(('    %15s:        %9i\n' % (key, metakeys[key])))

    allclusters = list(set(clustersmap.values()))
    allclusters.sort(key=lambda x: len(x[1]))
    ui.write(('disconnected clusters:      %9i\n' % len(allclusters)))

    ui.write('        any known node:     %9i\n'
             % len([c for c in allclusters
                    if [n for n in c[0] if nm.get(n) is not None]]))
    if allclusters:
        nbcluster = len(allclusters)
        ui.write(('        smallest length:    %9i\n' % len(allclusters[0][1])))
        ui.write(('        longer length:      %9i\n'
                 % len(allclusters[-1][1])))
        median = len(allclusters[nbcluster//2][1])
        ui.write(('        median length:      %9i\n' % median))
        mean = sum(len(x[1]) for x in allclusters) // nbcluster
        ui.write(('        mean length:        %9i\n' % mean))
    allpclusters = list(set(pclustersmap.values()))
    allpclusters.sort(key=lambda x: len(x[1]))
    ui.write(('    using parents data:     %9i\n' % len(allpclusters)))
    ui.write('        any known node:     %9i\n'
             % len([c for c in allclusters
                    if [n for n in c[0] if nm.get(n) is not None]]))
    if allpclusters:
        nbcluster = len(allpclusters)
        ui.write(('        smallest length:    %9i\n'
                 % len(allpclusters[0][1])))
        ui.write(('        longer length:      %9i\n'
                 % len(allpclusters[-1][1])))
        median = len(allpclusters[nbcluster//2][1])
        ui.write(('        median length:      %9i\n' % median))
        mean = sum(len(x[1]) for x in allpclusters) // nbcluster
        ui.write(('        mean length:        %9i\n' % mean))

def _solveone(ui, repo, ctx, dryrun, confirm, progresscb, category):
    """Resolve the troubles affecting one revision"""
    wlock = lock = tr = None
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        tr = repo.transaction("evolve")
        if 'unstable' == category:
            result = _solveunstable(ui, repo, ctx, dryrun, confirm, progresscb)
        elif 'bumped' == category:
            result = _solvebumped(ui, repo, ctx, dryrun, confirm, progresscb)
        elif 'divergent' == category:
            result = _solvedivergent(ui, repo, ctx, dryrun, confirm,
                                   progresscb)
        else:
            assert False, "unknown trouble category: %s" % (category)
        tr.close()
        return result
    finally:
        lockmod.release(tr, lock, wlock)

def _handlenotrouble(ui, repo, allopt, revopt, anyopt, targetcat):
    """Used by the evolve function to display an error message when
    no troubles can be resolved"""
    troublecategories = ['bumped', 'divergent', 'unstable']
    unselectedcategories = [c for c in troublecategories if c != targetcat]
    msg = None
    hint = None

    troubled = {
            "unstable": repo.revs("unstable()"),
            "divergent": repo.revs("divergent()"),
            "bumped": repo.revs("bumped()"),
            "all": repo.revs("troubled()"),
    }


    hintmap = {
            'bumped': _("do you want to use --bumped"),
            'bumped+divergent': _("do you want to use --bumped or --divergent"),
            'bumped+unstable': _("do you want to use --bumped or --unstable"),
            'divergent': _("do you want to use --divergent"),
            'divergent+unstable': _("do you want to use --divergent"
                                    " or --unstable"),
            'unstable': _("do you want to use --unstable"),
            'any+bumped': _("do you want to use --any (or --rev) and --bumped"),
            'any+bumped+divergent': _("do you want to use --any (or --rev) and"
                                      " --bumped or --divergent"),
            'any+bumped+unstable': _("do you want to use --any (or --rev) and"
                                     "--bumped or --unstable"),
            'any+divergent': _("do you want to use --any (or --rev) and"
                               " --divergent"),
            'any+divergent+unstable': _("do you want to use --any (or --rev)"
                                        " and --divergent or --unstable"),
            'any+unstable': _("do you want to use --any (or --rev)"
                              "and --unstable"),
    }

    if revopt:
        revs = scmutil.revrange(repo, revopt)
        if not revs:
            msg = _("set of specified revisions is empty")
        else:
            msg = _("no %s changesets in specified revisions") % targetcat
            othertroubles = []
            for cat in unselectedcategories:
                if revs & troubled[cat]:
                    othertroubles.append(cat)
            if othertroubles:
                hint = hintmap['+'.join(othertroubles)]

    elif anyopt:
        msg = _("no %s changesets to evolve") % targetcat
        othertroubles = []
        for cat in unselectedcategories:
            if troubled[cat]:
                othertroubles.append(cat)
        if othertroubles:
            hint = hintmap['+'.join(othertroubles)]

    else:
        # evolve without any option = relative to the current wdir
        if targetcat == 'unstable':
            msg = _("nothing to evolve on current working copy parent")
        else:
            msg = _("current working copy parent is not %s") % targetcat

        p1 = repo['.'].rev()
        othertroubles = []
        for cat in unselectedcategories:
            if p1 in troubled[cat]:
                othertroubles.append(cat)
        if othertroubles:
            hint = hintmap['+'.join(othertroubles)]
        else:
            l = len(troubled[targetcat])
            if l:
                hint = _("%d other %s in the repository, do you want --any "
                        "or --rev") % (l, targetcat)
            else:
                othertroubles = []
                for cat in unselectedcategories:
                    if troubled[cat]:
                        othertroubles.append(cat)
                if othertroubles:
                    hint = hintmap['any+'+('+'.join(othertroubles))]
                else:
                    msg = _("no troubled changesets")

    assert msg is not None
    ui.write_err(msg+"\n")
    if hint:
        ui.write_err("("+hint+")\n")
        return 2
    else:
        return 1

def _cleanup(ui, repo, startnode, showprogress):
    if showprogress:
        ui.progress(_('evolve'), None)
    if repo['.'] != startnode:
        ui.status(_('working directory is now at %s\n') % repo['.'])

class MultipleSuccessorsError(RuntimeError):
    """Exception raised by _singlesuccessor when multiple successor sets exists

    The object contains the list of successorssets in its 'successorssets'
    attribute to call to easily recover.
    """

    def __init__(self, successorssets):
        self.successorssets = successorssets

def _singlesuccessor(repo, p):
    """returns p (as rev) if not obsolete or its unique latest successors

    fail if there are no such successor"""

    if not p.obsolete():
        return p.rev()
    obs = repo[p]
    ui = repo.ui
    newer = obsolete.successorssets(repo, obs.node())
    # search of a parent which is not killed
    while not newer:
        ui.debug("stabilize target %s is plain dead,"
                 " trying to stabilize on its parent\n" %
                 obs)
        obs = obs.parents()[0]
        newer = obsolete.successorssets(repo, obs.node())
    if len(newer) > 1 or len(newer[0]) > 1:
        raise MultipleSuccessorsError(newer)

    return repo[newer[0][0]].rev()

def builddependencies(repo, revs):
    """returns dependency graphs giving an order to solve instability of revs
    (see _orderrevs for more information on usage)"""

    # For each troubled revision we keep track of what instability if any should
    # be resolved in order to resolve it. Example:
    # dependencies = {3: [6], 6:[]}
    # Means that: 6 has no dependency, 3 depends on 6 to be solved
    dependencies = {}
    # rdependencies is the inverted dict of dependencies
    rdependencies = collections.defaultdict(set)

    for r in revs:
        dependencies[r] = set()
        for p in repo[r].parents():
            try:
                succ = _singlesuccessor(repo, p)
            except MultipleSuccessorsError as exc:
                dependencies[r] = exc.successorssets
                continue
            if succ in revs:
                dependencies[r].add(succ)
                rdependencies[succ].add(r)
    return dependencies, rdependencies

def _dedupedivergents(repo, revs):
    """Dedupe the divergents revs in revs to get one from each group with the
    lowest revision numbers
    """
    repo = repo.unfiltered()
    res = set()
    # To not reevaluate divergents of the same group once one is encountered
    discarded = set()
    for rev in revs:
        if rev in discarded:
            continue
        divergent = repo[rev]
        base, others = divergentdata(divergent)
        othersrevs = [o.rev() for o in others]
        res.add(min([divergent.rev()] + othersrevs))
        discarded.update(othersrevs)
    return res

def _selectrevs(repo, allopt, revopt, anyopt, targetcat):
    """select troubles in repo matching according to given options"""
    revs = set()
    if allopt or revopt:
        revs = repo.revs(targetcat+'()')
        if revopt:
            revs = scmutil.revrange(repo, revopt) & revs
        elif not anyopt and targetcat == 'unstable':
            revs = set(_aspiringdescendant(repo,
                                           repo.revs('(.::) - obsolete()::')))
        if targetcat == 'divergent':
            # Pick one divergent per group of divergents
            revs = _dedupedivergents(repo, revs)
    elif anyopt:
        revs = repo.revs('first(%s())' % (targetcat))
    elif targetcat == 'unstable':
        revs = set(_aspiringchildren(repo, repo.revs('(.::) - obsolete()::')))
        if 1 < len(revs):
            msg = "multiple evolve candidates"
            hint = (_("select one of %s with --rev")
                    % ', '.join([str(repo[r]) for r in sorted(revs)]))
            raise error.Abort(msg, hint=hint)
    elif targetcat in repo['.'].troubles():
        revs = set([repo['.'].rev()])
    return revs


def _orderrevs(repo, revs):
    """Compute an ordering to solve instability for the given revs

    revs is a list of unstable revisions.

    Returns the same revisions ordered to solve their instability from the
    bottom to the top of the stack that the stabilization process will produce
    eventually.

    This ensures the minimal number of stabilizations, as we can stabilize each
    revision on its final stabilized destination.
    """
    # Step 1: Build the dependency graph
    dependencies, rdependencies = builddependencies(repo, revs)
    # Step 2: Build the ordering
    # Remove the revisions with no dependency(A) and add them to the ordering.
    # Removing these revisions leads to new revisions with no dependency (the
    # one depending on A) that we can remove from the dependency graph and add
    # to the ordering. We progress in a similar fashion until the ordering is
    # built
    solvablerevs = collections.deque([r for r in sorted(dependencies.keys())
                                      if not dependencies[r]])
    ordering = []
    while solvablerevs:
        rev = solvablerevs.popleft()
        for dependent in rdependencies[rev]:
            dependencies[dependent].remove(rev)
            if not dependencies[dependent]:
                solvablerevs.append(dependent)
        del dependencies[rev]
        ordering.append(rev)

    ordering.extend(sorted(dependencies))
    return ordering

@command('^evolve|stabilize|solve',
    [('n', 'dry-run', False,
        _('do not perform actions, just print what would be done')),
     ('', 'confirm', False,
        _('ask for confirmation before performing the action')),
    ('A', 'any', False,
        _('also consider troubled changesets unrelated to current working '
          'directory')),
    ('r', 'rev', [], _('solves troubles of these revisions')),
    ('', 'bumped', False, _('solves only bumped changesets')),
    ('', 'divergent', False, _('solves only divergent changesets')),
    ('', 'unstable', False, _('solves only unstable changesets (default)')),
    ('a', 'all', False, _('evolve all troubled changesets related to the '
                          'current  working directory and its descendants')),
    ('c', 'continue', False, _('continue an interrupted evolution')),
    ] + mergetoolopts,
    _('[OPTIONS]...'))
def evolve(ui, repo, **opts):
    """solve troubled changesets in your repository

    Modifying history can lead to various types of troubled changesets:
    unstable, bumped, or divergent. The evolve command resolves your troubles
    by executing one of the following actions:

    - update working copy to a successor
    - rebase an unstable changeset
    - extract the desired changes from a bumped changeset
    - fuse divergent changesets back together

    If you pass no arguments, evolve works in automatic mode: it will execute a
    single action to reduce instability related to your working copy. There are
    two cases for this action. First, if the parent of your working copy is
    obsolete, evolve updates to the parent's successor. Second, if the working
    copy parent is not obsolete but has obsolete predecessors, then evolve
    determines if there is an unstable changeset that can be rebased onto the
    working copy parent in order to reduce instability.
    If so, evolve rebases that changeset. If not, evolve refuses to guess your
    intention, and gives a hint about what you might want to do next.

    Any time evolve creates a changeset, it updates the working copy to the new
    changeset. (Currently, every successful evolve operation involves an update
    as well; this may change in future.)

    Automatic mode only handles common use cases. For example, it avoids taking
    action in the case of ambiguity, and it ignores unstable changesets that
    are not related to your working copy.
    It also refuses to solve bumped or divergent changesets unless you explicity
    request such behavior (see below).

    Eliminating all instability around your working copy may require multiple
    invocations of :hg:`evolve`. Alternately, use ``--all`` to recursively
    select and evolve all unstable changesets that can be rebased onto the
    working copy parent.
    This is more powerful than successive invocations, since ``--all`` handles
    ambiguous cases (e.g. unstable changesets with multiple children) by
    evolving all branches.

    When your repository cannot be handled by automatic mode, you might need to
    use ``--rev`` to specify a changeset to evolve. For example, if you have
    an unstable changeset that is not related to the working copy parent,
    you could use ``--rev`` to evolve it. Or, if some changeset has multiple
    unstable children, evolve in automatic mode refuses to guess which one to
    evolve; you have to use ``--rev`` in that case.

    Alternately, ``--any`` makes evolve search for the next evolvable changeset
    regardless of whether it is related to the working copy parent.

    You can supply multiple revisions to evolve multiple troubled changesets
    in a single invocation. In revset terms, ``--any`` is equivalent to ``--rev
    first(unstable())``. ``--rev`` and ``--all`` are mutually exclusive, as are
    ``--rev`` and ``--any``.

    ``hg evolve --any --all`` is useful for cleaning up instability across all
    branches, letting evolve figure out the appropriate order and destination.

    When you have troubled changesets that are not unstable, :hg:`evolve`
    refuses to consider them unless you specify the category of trouble you
    wish to resolve, with ``--bumped`` or ``--divergent``. These options are
    currently mutually exclusive with each other and with ``--unstable``
    (the default). You can combine ``--bumped`` or ``--divergent`` with
    ``--rev``, ``--all``, or ``--any``.

    """

    # Options
    contopt = opts['continue']
    anyopt = opts['any']
    allopt = opts['all']
    startnode = repo['.']
    dryrunopt = opts['dry_run']
    confirmopt = opts['confirm']
    revopt = opts['rev']
    troublecategories = ['bumped', 'divergent', 'unstable']
    specifiedcategories = [t for t in troublecategories if opts[t]]
    targetcat = 'unstable'
    if 1 < len(specifiedcategories):
        msg = _('cannot specify more than one trouble category to solve (yet)')
        raise error.Abort(msg)
    elif len(specifiedcategories) == 1:
        targetcat = specifiedcategories[0]
    elif repo['.'].obsolete():
        displayer = cmdutil.show_changeset(ui, repo,
                                           {'template': shorttemplate})
        # no args and parent is obsolete, update to successors
        try:
            ctx = repo[_singlesuccessor(repo, repo['.'])]
        except MultipleSuccessorsError as exc:
            repo.ui.write_err('parent is obsolete with multiple successors:\n')
            for ln in exc.successorssets:
                for n in ln:
                    displayer.show(repo[n])
            return 2


        ui.status(_('update:'))
        if not ui.quiet:
            displayer.show(ctx)

        if dryrunopt:
            return 0
        res = hg.update(repo, ctx.rev())
        if ctx != startnode:
            ui.status(_('working directory is now at %s\n') % ctx)
        return res

    ui.setconfig('ui', 'forcemerge', opts.get('tool', ''), 'evolve')
    troubled = set(repo.revs('troubled()'))

    # Progress handling
    seen = 1
    count = allopt and len(troubled) or 1
    showprogress = allopt

    def progresscb():
        if revopt or allopt:
            ui.progress(_('evolve'), seen, unit='changesets', total=count)

    # Continuation handling
    if contopt:
        if anyopt:
            raise error.Abort('cannot specify both "--any" and "--continue"')
        if allopt:
            raise error.Abort('cannot specify both "--all" and "--continue"')
        state = _evolvestateread(repo)
        if state is None:
            raise error.Abort('no evolve to continue')
        orig = repo[state['current']]
        # XXX This is a terrible terrible hack, please get rid of it.
        repo.opener.write('graftstate', orig.hex() + '\n')
        try:
            graftcmd = commands.table['graft'][0]
            ret = graftcmd(ui, repo, old_obsolete=True, **{'continue': True})
            _evolvestatedelete(repo)
            return ret
        finally:
            util.unlinkpath(repo.join('graftstate'), ignoremissing=True)
    cmdutil.bailifchanged(repo)


    if revopt and allopt:
        raise error.Abort('cannot specify both "--rev" and "--all"')
    if revopt and anyopt:
        raise error.Abort('cannot specify both "--rev" and "--any"')

    revs = _selectrevs(repo, allopt, revopt, anyopt, targetcat)

    if not revs:
        return _handlenotrouble(ui, repo, allopt, revopt, anyopt, targetcat)

    # For the progress bar to show
    count = len(revs)
    # Order the revisions
    if targetcat == 'unstable':
        revs = _orderrevs(repo, revs)
    for rev in revs:
        progresscb()
        _solveone(ui, repo, repo[rev], dryrunopt, confirmopt,
                progresscb, targetcat)
        seen += 1
    progresscb()
    _cleanup(ui, repo, startnode, showprogress)

def _possibledestination(repo, rev):
    """return all changesets that may be a new parent for REV"""
    tonode = repo.changelog.node
    parents = repo.changelog.parentrevs
    torev = repo.changelog.rev
    dest = set()
    tovisit = list(parents(rev))
    while tovisit:
        r = tovisit.pop()
        succsets = obsolete.successorssets(repo, tonode(r))
        if not succsets:
            tovisit.extend(parents(r))
        else:
            # We should probably pick only one destination from split
            # (case where '1 < len(ss)'), This could be the currently tipmost
            # but logic is less clear when result of the split are now on
            # multiple branches.
            for ss in succsets:
                for n in ss:
                    dest.add(torev(n))
    return dest

def _aspiringchildren(repo, revs):
    """Return a list of changectx which can be stabilized on top of pctx or
    one of its descendants. Empty list if none can be found."""
    target = set(revs)
    result = []
    for r in repo.revs('unstable() - %ld', revs):
        dest = _possibledestination(repo, r)
        if target & dest:
            result.append(r)
    return result

def _aspiringdescendant(repo, revs):
    """Return a list of changectx which can be stabilized on top of pctx or
    one of its descendants recursively. Empty list if none can be found."""
    target = set(revs)
    result = set(target)
    paths = collections.defaultdict(set)
    for r in repo.revs('unstable() - %ld', revs):
        for d in _possibledestination(repo, r):
            paths[d].add(r)

    result = set(target)
    tovisit = list(revs)
    while tovisit:
        base = tovisit.pop()
        for unstable in paths[base]:
            if unstable not in result:
                tovisit.append(unstable)
                result.add(unstable)
    return sorted(result - target)

def _solveunstable(ui, repo, orig, dryrun=False, confirm=False,
                   progresscb=None):
    """Stabilize an unstable changeset"""
    pctx = orig.p1()
    if len(orig.parents()) == 2:
        if not pctx.obsolete():
            pctx = orig.p2()  # second parent is obsolete ?
        elif orig.p2().obsolete():
            raise error.Abort(_("no support for evolving merge changesets "
                                "with two obsolete parents yet"),
                              hint=_("Redo the merge and use `hg prune <old> "
                                   "--succ <new>` to obsolete the old one"))

    if not pctx.obsolete():
        ui.warn(_("cannot solve instability of %s, skipping\n") % orig)
        return False
    obs = pctx
    newer = obsolete.successorssets(repo, obs.node())
    # search of a parent which is not killed
    while not newer or newer == [()]:
        ui.debug("stabilize target %s is plain dead,"
                 " trying to stabilize on its parent\n" %
                 obs)
        obs = obs.parents()[0]
        newer = obsolete.successorssets(repo, obs.node())
    if len(newer) > 1:
        msg = _("skipping %s: divergent rewriting. can't choose "
                "destination\n") % obs
        ui.write_err(msg)
        return 2
    targets = newer[0]
    assert targets
    if len(targets) > 1:
        # split target, figure out which one to pick, are they all in line?
        targetrevs = [repo[r].rev() for r in targets]
        roots = repo.revs('roots(%ld)', targetrevs)
        heads = repo.revs('heads(%ld)', targetrevs)
        if len(roots) > 1 or len(heads) > 1:
            msg = "cannot solve split accross two branches\n"
            ui.write_err(msg)
            return 2
        target = repo[heads.first()]
    else:
        target = targets[0]
    displayer = cmdutil.show_changeset(ui, repo, {'template': shorttemplate})
    target = repo[target]
    if not ui.quiet or confirm:
        repo.ui.write(_('move:'))
        displayer.show(orig)
        repo.ui.write(_('atop:'))
        displayer.show(target)
    if confirm and ui.prompt('perform evolve? [Ny]', 'n') != 'y':
            raise error.Abort(_('evolve aborted by user'))
    if progresscb: progresscb()
    todo = 'hg rebase -r %s -d %s\n' % (orig, target)
    if dryrun:
        repo.ui.write(todo)
    else:
        repo.ui.note(todo)
        if progresscb: progresscb()
        keepbranch = orig.p1().branch() != orig.branch()
        try:
            relocate(repo, orig, target, pctx, keepbranch)
        except MergeFailure:
            _evolvestatewrite(repo, {'current': orig.node()})
            repo.ui.write_err(_('evolve failed!\n'))
            repo.ui.write_err(
                _('fix conflict and run "hg evolve --continue"'
                  ' or use "hg update -C" to abort\n'))
            raise

def _solvebumped(ui, repo, bumped, dryrun=False, confirm=False,
                 progresscb=None):
    """Stabilize a bumped changeset"""
    repo = repo.unfiltered()
    bumped = repo[bumped.rev()]
    # For now we deny bumped merge
    if len(bumped.parents()) > 1:
        msg = _('skipping %s : we do not handle merge yet\n') % bumped
        ui.write_err(msg)
        return 2
    prec = repo.set('last(allprecursors(%d) and public())', bumped).next()
    # For now we deny target merge
    if len(prec.parents()) > 1:
        msg = _('skipping: %s: public version is a merge, '
                'this is not handled yet\n') % prec
        ui.write_err(msg)
        return 2

    displayer = cmdutil.show_changeset(ui, repo, {'template': shorttemplate})
    if not ui.quiet or confirm:
        repo.ui.write(_('recreate:'))
        displayer.show(bumped)
        repo.ui.write(_('atop:'))
        displayer.show(prec)
    if confirm and ui.prompt('perform evolve? [Ny]', 'n') != 'y':
        raise error.Abort(_('evolve aborted by user'))
    if dryrun:
        todo = 'hg rebase --rev %s --dest %s;\n' % (bumped, prec.p1())
        repo.ui.write(todo)
        repo.ui.write(('hg update %s;\n' % prec))
        repo.ui.write(('hg revert --all --rev %s;\n' % bumped))
        repo.ui.write(('hg commit --msg "bumped update to %s"'))
        return 0
    if progresscb: progresscb()
    newid = tmpctx = None
    tmpctx = bumped
    # Basic check for common parent. Far too complicated and fragile
    tr = repo.currenttransaction()
    assert tr is not None
    bmupdate = _bookmarksupdater(repo, bumped.node(), tr)
    try:
        if not list(repo.set('parents(%d) and parents(%d)', bumped, prec)):
            # Need to rebase the changeset at the right place
            repo.ui.status(
                _('rebasing to destination parent: %s\n') % prec.p1())
            try:
                tmpid = relocate(repo, bumped, prec.p1())
                if tmpid is not None:
                    tmpctx = repo[tmpid]
                    obsolete.createmarkers(repo, [(bumped, (tmpctx,))])
            except MergeFailure:
                repo.opener.write('graftstate', bumped.hex() + '\n')
                repo.ui.write_err(_('evolution failed!\n'))
                repo.ui.write_err(
                    _('fix conflict and run "hg evolve --continue"\n'))
                raise
        # Create the new commit context
        repo.ui.status(_('computing new diff\n'))
        files = set()
        copied = copies.pathcopies(prec, bumped)
        precmanifest = prec.manifest()
        # 3.3.2 needs a list.
        # future 3.4 don't detect the size change during iteration
        # this is fishy
        for key, val in list(bumped.manifest().iteritems()):
            precvalue = precmanifest.get(key, None)
            if precvalue is not None:
                del precmanifest[key]
            if precvalue != val:
                files.add(key)
        files.update(precmanifest)  # add missing files
        # commit it
        if files: # something to commit!
            def filectxfn(repo, ctx, path):
                if path in bumped:
                    fctx = bumped[path]
                    flags = fctx.flags()
                    mctx = memfilectx(repo, fctx.path(), fctx.data(),
                                      islink='l' in flags,
                                      isexec='x' in flags,
                                      copied=copied.get(path))
                    return mctx
                return None
            text = 'bumped update to %s:\n\n' % prec
            text += bumped.description()

            new = context.memctx(repo,
                                 parents=[prec.node(), node.nullid],
                                 text=text,
                                 files=files,
                                 filectxfn=filectxfn,
                                 user=bumped.user(),
                                 date=bumped.date(),
                                 extra=bumped.extra())

            newid = repo.commitctx(new)
        if newid is None:
            obsolete.createmarkers(repo, [(tmpctx, ())])
            newid = prec.node()
        else:
            phases.retractboundary(repo, tr, bumped.phase(), [newid])
            obsolete.createmarkers(repo, [(tmpctx, (repo[newid],))],
                                   flag=obsolete.bumpedfix)
        bmupdate(newid)
        repo.ui.status(_('committed as %s\n') % node.short(newid))
    finally:
        pass # TODO: remove this redundant try/finally block
    # reroute the working copy parent to the new changeset
    repo.dirstate.beginparentchange()
    repo.dirstate.setparents(newid, node.nullid)
    repo.dirstate.endparentchange()

def _solvedivergent(ui, repo, divergent, dryrun=False, confirm=False,
                    progresscb=None):
    repo = repo.unfiltered()
    divergent = repo[divergent.rev()]
    base, others = divergentdata(divergent)
    if len(others) > 1:
        othersstr = "[%s]" % (','.join([str(i) for i in others]))
        msg = _("skipping %d:divergent with a changeset that got splitted"
                " into multiple ones:\n"
                 "|[%s]\n"
                 "| This is not handled by automatic evolution yet\n"
                 "| You have to fallback to manual handling with commands "
                 "such as:\n"
                 "| - hg touch -D\n"
                 "| - hg prune\n"
                 "| \n"
                 "| You should contact your local evolution Guru for help.\n"
                 ) % (divergent, othersstr)
        ui.write_err(msg)
        return 2
    other = others[0]
    if len(other.parents()) > 1:
        msg = _("skipping %s: divergent changeset can't be "
                "a merge (yet)\n") % divergent
        ui.write_err(msg)
        hint = _("You have to fallback to solving this by hand...\n"
                 "| This probably means redoing the merge and using \n"
                 "| `hg prune` to kill older version.\n")
        ui.write_err(hint)
        return 2
    if other.p1() not in divergent.parents():
        msg = _("skipping %s: have a different parent than %s "
                "(not handled yet)\n") % (divergent, other)
        hint = _("| %(d)s, %(o)s are not based on the same changeset.\n"
                 "| With the current state of its implementation, \n"
                 "| evolve does not work in that case.\n"
                 "| rebase one of them next to the other and run \n"
                 "| this command again.\n"
                 "| - either: hg rebase --dest 'p1(%(d)s)' -r %(o)s\n"
                 "| - or:     hg rebase --dest 'p1(%(o)s)' -r %(d)s\n"
                 ) % {'d': divergent, 'o': other}
        ui.write_err(msg)
        ui.write_err(hint)
        return 2

    displayer = cmdutil.show_changeset(ui, repo, {'template': shorttemplate})
    if not ui.quiet or confirm:
        ui.write(_('merge:'))
        displayer.show(divergent)
        ui.write(_('with: '))
        displayer.show(other)
        ui.write(_('base: '))
        displayer.show(base)
    if confirm and ui.prompt(_('perform evolve? [Ny]'), 'n') != 'y':
        raise error.Abort(_('evolve aborted by user'))
    if dryrun:
        ui.write(('hg update -c %s &&\n' % divergent))
        ui.write(('hg merge %s &&\n' % other))
        ui.write(('hg commit -m "auto merge resolving conflict between '
                 '%s and %s"&&\n' % (divergent, other)))
        ui.write(('hg up -C %s &&\n' % base))
        ui.write(('hg revert --all --rev tip &&\n'))
        ui.write(('hg commit -m "`hg log -r %s --template={desc}`";\n'
                 % divergent))
        return
    if divergent not in repo[None].parents():
        repo.ui.status(_('updating to "local" conflict\n'))
        hg.update(repo, divergent.rev())
    repo.ui.note(_('merging divergent changeset\n'))
    if progresscb: progresscb()
    if 'partial' in merge.update.__doc__:
        # Mercurial  < 43c00ca887d1 (3.7)
        stats = merge.update(repo,
                             other.node(),
                             branchmerge=True,
                             force=False,
                             partial=None,
                             ancestor=base.node(),
                             mergeancestor=True)
    else:
        stats = merge.update(repo,
                             other.node(),
                             branchmerge=True,
                             force=False,
                             ancestor=base.node(),
                             mergeancestor=True)

    hg._showstats(repo, stats)
    if stats[3]:
        repo.ui.status(_("use 'hg resolve' to retry unresolved file merges "
                         "or 'hg update -C .' to abandon\n"))
    if stats[3] > 0:
        raise error.Abort('merge conflict between several amendments '
            '(this is not automated yet)',
            hint="""/!\ You can try:
/!\ * manual merge + resolve => new cset X
/!\ * hg up to the parent of the amended changeset (which are named W and Z)
/!\ * hg revert --all -r X
/!\ * hg ci -m "same message as the amended changeset" => new cset Y
/!\ * hg kill -n Y W Z
""")
    if progresscb: progresscb()
    emtpycommitallowed = repo.ui.backupconfig('ui', 'allowemptycommit')
    tr = repo.currenttransaction()
    assert tr is not None
    try:
        repo.ui.setconfig('ui', 'allowemptycommit', True)
        repo.dirstate.beginparentchange()
        repo.dirstate.setparents(divergent.node(), node.nullid)
        repo.dirstate.endparentchange()
        oldlen = len(repo)
        amend(ui, repo, message='', logfile='')
        if oldlen == len(repo):
            new = divergent
            # no changes
        else:
            new = repo['.']
        obsolete.createmarkers(repo, [(other, (new,))])
        phases.retractboundary(repo, tr, other.phase(), [new.node()])
    finally:
        repo.ui.restoreconfig(emtpycommitallowed)

def divergentdata(ctx):
    """return base, other part of a conflict

    This only return the first one.

    XXX this woobly function won't survive XXX
    """
    repo = ctx._repo.unfiltered()
    for base in repo.set('reverse(allprecursors(%d))', ctx):
        newer = obsolete.successorssets(ctx._repo, base.node())
        # drop filter and solution including the original ctx
        newer = [n for n in newer if n and ctx.node() not in n]
        if newer:
            return base, tuple(ctx._repo[o] for o in newer[0])
    raise error.Abort("base of divergent changeset %s not found" % ctx,
                     hint='this case is not yet handled')



shorttemplate = '[{rev}] {desc|firstline}\n'

@command('^previous',
         [('B', 'move-bookmark', False,
             _('move active bookmark after update')),
          ('', 'merge', False, _('bring uncommitted change along')),
          ('n', 'dry-run', False,
             _('do not perform actions, just print what would be done'))],
         '[OPTION]...')
def cmdprevious(ui, repo, **opts):
    """update to parent revision

    Displays the summary line of the destination for clarity."""
    wkctx = repo[None]
    wparents = wkctx.parents()
    dryrunopt = opts['dry_run']
    if len(wparents) != 1:
        raise error.Abort('merge in progress')
    if not opts['merge']:
        try:
            cmdutil.bailifchanged(repo)
        except error.Abort as exc:
            exc.hint = _('do you want --merge?')
            raise

    parents = wparents[0].parents()
    displayer = cmdutil.show_changeset(ui, repo, {'template': shorttemplate})
    if len(parents) == 1:
        p = parents[0]
        bm = bmactive(repo)
        shouldmove = opts.get('move_bookmark') and bm is not None
        if dryrunopt:
            ui.write(('hg update %s;\n' % p.rev()))
            if shouldmove:
                ui.write(('hg bookmark %s -r %s;\n' % (bm, p.rev())))
        else:
            ret = hg.update(repo, p.rev())
            if not ret:
                tr = lock = None
                wlock = repo.wlock()
                try:
                    lock = repo.lock()
                    tr = repo.transaction('previous')
                    if shouldmove:
                        repo._bookmarks[bm] = p.node()
                        repo._bookmarks.recordchange(tr)
                    else:
                        bmdeactivate(repo)
                    tr.close()
                finally:
                    lockmod.release(tr, lock, wlock)
        displayer.show(p)
        return 0
    else:
        for p in parents:
            displayer.show(p)
        ui.warn(_('multiple parents, explicitly update to one\n'))
        return 1

@command('^next',
         [('B', 'move-bookmark', False,
             _('move active bookmark after update')),
          ('', 'merge', False, _('bring uncommitted change along')),
          ('', 'evolve', False, _('evolve the next changeset if necessary')),
          ('n', 'dry-run', False,
              _('do not perform actions, just print what would be done'))],
              '[OPTION]...')
def cmdnext(ui, repo, **opts):
    """update to next child revision

    Use the ``--evolve`` flag to evolve unstable children on demand.

    Displays the summary line of the destination for clarity.
    """
    wkctx = repo[None]
    wparents = wkctx.parents()
    dryrunopt = opts['dry_run']
    if len(wparents) != 1:
        raise error.Abort('merge in progress')
    if not opts['merge']:
        try:
            cmdutil.bailifchanged(repo)
        except error.Abort as exc:
            exc.hint = _('do you want --merge?')
            raise

    children = [ctx for ctx in wparents[0].children() if not ctx.obsolete()]
    displayer = cmdutil.show_changeset(ui, repo, {'template': shorttemplate})
    if len(children) == 1:
        c = children[0]
        bm = bmactive(repo)
        shouldmove = opts.get('move_bookmark') and bm is not None
        if dryrunopt:
            ui.write(('hg update %s;\n' % c.rev()))
            if shouldmove:
                ui.write(('hg bookmark %s -r %s;\n' % (bm, c.rev())))
        else:
            ret = hg.update(repo, c.rev())
            if not ret:
                lock = tr = None
                wlock = repo.wlock()
                try:
                    lock = repo.lock()
                    tr = repo.transaction('next')
                    if shouldmove:
                        repo._bookmarks[bm] = c.node()
                        repo._bookmarks.recordchange(tr)
                    else:
                        bmdeactivate(repo)
                    tr.close()
                finally:
                    lockmod.release(tr, lock, wlock)
        displayer.show(c)
        result = 0
    elif children:
        ui.warn(_("ambigious next changeset:\n"))
        for c in children:
            displayer.show(c)
        ui.warn(_('explicitly update to one of them\n'))
        result = 1
    else:
        aspchildren = _aspiringchildren(repo, [repo['.'].rev()])
        if not opts['evolve'] or not aspchildren:
            ui.warn(_('no children\n'))
            if aspchildren:
                msg = _('(%i unstable changesets to be evolved here, '
                        'do you want --evolve?)\n')
                ui.warn(msg % len(aspchildren))
            result = 1
        elif 1 < len(aspchildren):
            ui.warn(_("ambigious next (unstable) changeset:\n"))
            for c in aspchildren:
                displayer.show(repo[c])
            ui.warn(_('(run "hg evolve --rev REV" on one of them)\n'))
            return 1
        else:
            cmdutil.bailifchanged(repo)
            result = _solveone(ui, repo, repo[aspchildren[0]], dryrunopt,
                               False, lambda:None, category='unstable')
            if not result:
                ui.status(_('working directory now at %s\n') % repo['.'])
            return result
        return 1
    return result

def _reachablefrombookmark(repo, revs, bookmarks):
    """filter revisions and bookmarks reachable from the given bookmark
    yoinked from mq.py
    """
    repomarks = repo._bookmarks
    if not bookmarks.issubset(repomarks):
        raise error.Abort(_("bookmark '%s' not found") %
            ','.join(sorted(bookmarks - set(repomarks.keys()))))

    # If the requested bookmark is not the only one pointing to a
    # a revision we have to only delete the bookmark and not strip
    # anything. revsets cannot detect that case.
    nodetobookmarks = {}
    for mark, node in repomarks.iteritems():
        nodetobookmarks.setdefault(node, []).append(mark)
    for marks in nodetobookmarks.values():
        if bookmarks.issuperset(marks):
           if util.safehasattr(repair, 'stripbmrevset'):
               rsrevs = repair.stripbmrevset(repo, marks[0])
           else:
               rsrevs = repo.revs("ancestors(bookmark(%s)) - "
                                  "ancestors(head() and not bookmark(%s)) - "
                                  "ancestors(bookmark() and not bookmark(%s)) - "
                                  "obsolete()",
                                  marks[0], marks[0], marks[0])
           revs = set(revs)
           revs.update(set(rsrevs))
           revs = sorted(revs)
    return repomarks, revs

def _deletebookmark(repo, repomarks, bookmarks):
    wlock = lock = tr = None
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        tr = repo.transaction('prune')
        for bookmark in bookmarks:
            del repomarks[bookmark]
        repomarks.recordchange(tr)
        tr.close()
        for bookmark in sorted(bookmarks):
            repo.ui.write(_("bookmark '%s' deleted\n") % bookmark)
    finally:
        lockmod.release(tr, lock, wlock)



def _getmetadata(**opts):
    metadata = {}
    date = opts.get('date')
    user = opts.get('user')
    if date:
        metadata['date'] = '%i %i' % util.parsedate(date)
    if user:
        metadata['user'] = user
    return metadata


@command('^prune|obsolete|kill',
    [('n', 'new', [], _("successor changeset (DEPRECATED)")),
     ('s', 'succ', [], _("successor changeset")),
     ('r', 'rev', [], _("revisions to prune")),
     ('k', 'keep', None, _("does not modify working copy during prune")),
     ('', 'biject', False, _("do a 1-1 map between rev and successor ranges")),
     ('', 'fold', False,
        _("record a fold (multiple precursors, one successors)")),
     ('', 'split', False,
        _("record a split (on precursor, multiple successors)")),
     ('B', 'bookmark', [], _("remove revs only reachable from given"
                             " bookmark"))] + metadataopts,
    _('[OPTION] [-r] REV...'))
    # -U  --noupdate option to prevent wc update and or bookmarks update ?
def cmdprune(ui, repo, *revs, **opts):
    """hide changesets by marking them obsolete

    Pruned changesets are obsolete with no successors. If they also have no
    descendants, they are hidden (invisible to all commands).

    Non-obsolete descendants of pruned changesets become "unstable". Use :hg:`evolve`
    to handle this situation.

    When you prune the parent of your working copy, Mercurial updates the working
    copy to a non-obsolete parent.

    You can use ``--succ`` to tell Mercurial that a newer version (successor) of the
    pruned changeset exists. Mercurial records successor revisions in obsolescence
    markers.

    You can use the ``--biject`` option to specify a 1-1 mapping (bijection) between
    revisions to pruned (precursor) and successor changesets. This option may be
    removed in a future release (with the functionality provided automatically).

    If you specify multiple revisions in ``--succ``, you are recording a "split" and
    must acknowledge it by passing ``--split``. Similarly, when you prune multiple
    changesets with a single successor, you must pass the ``--fold`` option.
    """
    revs = scmutil.revrange(repo, list(revs) + opts.get('rev'))
    succs = opts['new'] + opts['succ']
    bookmarks = set(opts.get('bookmark'))
    metadata = _getmetadata(**opts)
    biject = opts.get('biject')
    fold = opts.get('fold')
    split = opts.get('split')

    options = [o for o in ('biject', 'fold', 'split') if opts.get(o)]
    if 1 < len(options):
        raise error.Abort(_("can only specify one of %s") % ', '.join(options))

    if bookmarks:
        repomarks, revs = _reachablefrombookmark(repo, revs, bookmarks)
        if not revs:
            # no revisions to prune - delete bookmark immediately
            _deletebookmark(repo, repomarks, bookmarks)

    if not revs:
        raise error.Abort(_('nothing to prune'))

    wlock = lock = tr = None
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        tr = repo.transaction('prune')
        # defines pruned changesets
        precs = []
        revs.sort()
        for p in revs:
            cp = repo[p]
            if not cp.mutable():
                # note: createmarkers() would have raised something anyway
                raise error.Abort('cannot prune immutable changeset: %s' % cp,
                                 hint='see "hg help phases" for details')
            precs.append(cp)
        if not precs:
            raise error.Abort('nothing to prune')

        if not obsolete.isenabled(repo, obsolete.allowunstableopt):
            if repo.revs("(%ld::) - %ld", revs, revs):
                raise error.Abort(_("cannot prune in the middle of a stack"))

        # defines successors changesets
        sucs = scmutil.revrange(repo, succs)
        sucs.sort()
        sucs = tuple(repo[n] for n in sucs)
        if not biject and len(sucs) > 1 and len(precs) > 1:
            msg = "Can't use multiple successors for multiple precursors"
            raise error.Abort(msg)
        elif biject and len(sucs) != len(precs):
            msg = "Can't use %d successors for %d precursors" \
                % (len(sucs), len(precs))
            raise error.Abort(msg)
        elif (len(precs) == 1 and len(sucs) > 1) and not split:
            msg = "please add --split if you want to do a split"
            raise error.Abort(msg)
        elif len(sucs) == 1 and len(precs) > 1 and not fold:
            msg = "please add --fold if you want to do a fold"
            raise error.Abort(msg)
        elif biject:
            relations = [(p, (s,)) for p, s in zip(precs, sucs)]
        else:
            relations = [(p, sucs) for p in precs]

        wdp = repo['.']

        if len(sucs) == 1 and len(precs) == 1 and wdp in precs:
            # '.' killed, so update to the successor
            newnode = sucs[0]
        else:
            # update to an unkilled parent
            newnode = wdp

            while newnode in precs or newnode.obsolete():
                newnode = newnode.parents()[0]


        if newnode.node() != wdp.node():
            if opts.get('keep', False):
                # This is largely the same as the implementation in
                # strip.stripcmd(). We might want to refactor this somewhere
                # common at some point.

                # only reset the dirstate for files that would actually change
                # between the working context and uctx
                descendantrevs = repo.revs("%d::." % newnode.rev())
                changedfiles = []
                for rev in descendantrevs:
                    # blindly reset the files, regardless of what actually
                    # changed
                    changedfiles.extend(repo[rev].files())

                # reset files that only changed in the dirstate too
                dirstate = repo.dirstate
                dirchanges = [f for f in dirstate if dirstate[f] != 'n']
                changedfiles.extend(dirchanges)
                repo.dirstate.rebuild(newnode.node(), newnode.manifest(),
                                      changedfiles)
                writedirstate(dirstate, tr)
            else:
                bookactive = bmactive(repo)
                # Active bookmark that we don't want to delete (with -B option)
                # we deactivate and move it before the update and reactivate it
                # after
                movebookmark = bookactive and not bookmarks
                if movebookmark:
                    bmdeactivate(repo)
                    repo._bookmarks[bookactive] = newnode.node()
                    repo._bookmarks.recordchange(tr)
                commands.update(ui, repo, newnode.rev())
                ui.status(_('working directory now at %s\n') % newnode)
                if movebookmark:
                    bmactivate(repo, bookactive)

        # update bookmarks
        if bookmarks:
            _deletebookmark(repo, repomarks, bookmarks)

        # create markers
        obsolete.createmarkers(repo, relations, metadata=metadata)

        # informs that changeset have been pruned
        ui.status(_('%i changesets pruned\n') % len(precs))

        for ctx in repo.unfiltered().set('bookmark() and %ld', precs):
            # used to be:
            #
            #   ldest = list(repo.set('max((::%d) - obsolete())', ctx))
            #   if ldest:
            #      c = ldest[0]
            #
            # but then revset took a lazy arrow in the knee and became much
            # slower. The new forms makes as much sense and a much faster.
            for dest in ctx.ancestors():
                if not dest.obsolete():
                    updatebookmarks = _bookmarksupdater(repo, ctx.node(), tr)
                    updatebookmarks(dest.node())
                    break

        tr.close()
    finally:
        lockmod.release(tr, lock, wlock)

@command('amend|refresh',
    [('A', 'addremove', None,
     _('mark new/missing files as added/removed before committing')),
    ('e', 'edit', False, _('invoke editor on commit messages')),
    ('', 'close-branch', None,
     _('mark a branch as closed, hiding it from the branch list')),
    ('s', 'secret', None, _('use the secret phase for committing')),
    ] + walkopts + commitopts + commitopts2 + commitopts3 + interactiveopt,
    _('[OPTION]... [FILE]...'))
def amend(ui, repo, *pats, **opts):
    """combine a changeset with updates and replace it with a new one

    Commits a new changeset incorporating both the changes to the given files
    and all the changes from the current parent changeset into the repository.

    See :hg:`commit` for details about committing changes.

    If you don't specify -m, the parent's message will be reused.

    Behind the scenes, Mercurial first commits the update as a regular child
    of the current parent. Then it creates a new commit on the parent's parents
    with the updated contents. Then it changes the working copy parent to this
    new combined changeset. Finally, the old changeset and its update are hidden
    from :hg:`log` (unless you use --hidden with log).

    Returns 0 on success, 1 if nothing changed.
    """
    opts = opts.copy()
    edit = opts.pop('edit', False)
    log = opts.get('logfile')
    opts['amend'] = True
    if not (edit or opts['message'] or log):
        opts['message'] = repo['.'].description()
    _resolveoptions(ui, opts)
    _alias, commitcmd = cmdutil.findcmd('commit', commands.table)
    return commitcmd[0](ui, repo, *pats, **opts)


def _touchedbetween(repo, source, dest, match=None):
    touched = set()
    for files in repo.status(source, dest, match=match)[:3]:
        touched.update(files)
    return touched

def _commitfiltered(repo, ctx, match, target=None):
    """Recommit ctx with changed files not in match. Return the new
    node identifier, or None if nothing changed.
    """
    base = ctx.p1()
    if target is None:
        target = base
    # ctx
    initialfiles = _touchedbetween(repo, base, ctx)
    if base == target:
        affected = set(f for f in initialfiles if match(f))
        newcontent = set()
    else:
        affected = _touchedbetween(repo, target, ctx, match=match)
        newcontent = _touchedbetween(repo, target, base, match=match)
    # The commit touchs all existing files
    # + all file that needs a new content
    # - the file affected bny uncommit with the same content than base.
    files = (initialfiles - affected) | newcontent
    if not newcontent and files == initialfiles:
        return None

    # Filter copies
    copied = copies.pathcopies(target, ctx)
    copied = dict((src, dst) for src, dst in copied.iteritems()
                  if dst in files)
    def filectxfn(repo, memctx, path, contentctx=ctx, redirect=newcontent):
        if path in redirect:
            return filectxfn(repo, memctx, path, contentctx=target, redirect=())
        if path not in contentctx:
            return None
        fctx = contentctx[path]
        flags = fctx.flags()
        mctx = memfilectx(repo, fctx.path(), fctx.data(),
                          islink='l' in flags,
                          isexec='x' in flags,
                          copied=copied.get(path))
        return mctx

    new = context.memctx(repo,
                         parents=[base.node(), node.nullid],
                         text=ctx.description(),
                         files=files,
                         filectxfn=filectxfn,
                         user=ctx.user(),
                         date=ctx.date(),
                         extra=ctx.extra())
    # commitctx always create a new revision, no need to check
    newid = repo.commitctx(new)
    return newid

def _uncommitdirstate(repo, oldctx, match):
    """Fix the dirstate after switching the working directory from
    oldctx to a copy of oldctx not containing changed files matched by
    match.
    """
    ctx = repo['.']
    ds = repo.dirstate
    copies = dict(ds.copies())
    m, a, r = repo.status(oldctx.p1(), oldctx, match=match)[:3]
    for f in m:
        if ds[f] == 'r':
            # modified + removed -> removed
            continue
        ds.normallookup(f)

    for f in a:
        if ds[f] == 'r':
            # added + removed -> unknown
            ds.drop(f)
        elif ds[f] != 'a':
            ds.add(f)

    for f in r:
        if ds[f] == 'a':
            # removed + added -> normal
            ds.normallookup(f)
        elif ds[f] != 'r':
            ds.remove(f)

    # Merge old parent and old working dir copies
    oldcopies = {}
    for f in (m + a):
        src = oldctx[f].renamed()
        if src:
            oldcopies[f] = src[0]
    oldcopies.update(copies)
    copies = dict((dst, oldcopies.get(src, src))
                  for dst, src in oldcopies.iteritems())
    # Adjust the dirstate copies
    for dst, src in copies.iteritems():
        if (src not in ctx or dst in ctx or ds[dst] != 'a'):
            src = None
        ds.copy(src, dst)

@command('^uncommit',
    [('a', 'all', None, _('uncommit all changes when no arguments given')),
     ('r', 'rev', '', _('revert commit content to REV instead')),
     ] + commands.walkopts,
    _('[OPTION]... [NAME]'))
def uncommit(ui, repo, *pats, **opts):
    """move changes from parent revision to working directory

    Changes to selected files in the checked out revision appear again as
    uncommitted changed in the working directory. A new revision
    without the selected changes is created, becomes the checked out
    revision, and obsoletes the previous one.

    The --include option specifies patterns to uncommit.
    The --exclude option specifies patterns to keep in the commit.

    The --rev argument let you change the commit file to a content of another
    revision. It still does not change the content of your file in the working
    directory.

    Return 0 if changed files are uncommitted.
    """

    wlock = lock = tr = None
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        wctx = repo[None]
        if len(wctx.parents()) <= 0:
            raise error.Abort(_("cannot uncommit null changeset"))
        if len(wctx.parents()) > 1:
            raise error.Abort(_("cannot uncommit while merging"))
        old = repo['.']
        if old.phase() == phases.public:
            raise error.Abort(_("cannot rewrite immutable changeset"))
        if len(old.parents()) > 1:
            raise error.Abort(_("cannot uncommit merge changeset"))
        oldphase = old.phase()


        rev = None
        if opts.get('rev'):
            rev = scmutil.revsingle(repo, opts.get('rev'))
            ctx = repo[None]
            if ctx.p1() == rev or ctx.p2() == rev:
                raise error.Abort(_("cannot uncommit to parent changeset"))

        onahead = old.rev() in repo.changelog.headrevs()
        disallowunstable = not obsolete.isenabled(repo,
                                                  obsolete.allowunstableopt)
        if disallowunstable and not onahead:
            raise error.Abort(_("cannot uncommit in the middle of a stack"))

        # Recommit the filtered changeset
        tr = repo.transaction('uncommit')
        updatebookmarks = _bookmarksupdater(repo, old.node(), tr)
        newid = None
        includeorexclude = opts.get('include') or opts.get('exclude')
        if (pats or includeorexclude or opts.get('all')):
            match = scmutil.match(old, pats, opts)
            newid = _commitfiltered(repo, old, match, target=rev)
        if newid is None:
            raise error.Abort(_('nothing to uncommit'),
                             hint=_("use --all to uncommit all files"))
        # Move local changes on filtered changeset
        obsolete.createmarkers(repo, [(old, (repo[newid],))])
        phases.retractboundary(repo, tr, oldphase, [newid])
        repo.dirstate.beginparentchange()
        repo.dirstate.setparents(newid, node.nullid)
        _uncommitdirstate(repo, old, match)
        repo.dirstate.endparentchange()
        updatebookmarks(newid)
        if not repo[newid].files():
            ui.warn(_("new changeset is empty\n"))
            ui.status(_('(use "hg prune ." to remove it)\n'))
        tr.close()
    finally:
        lockmod.release(tr, lock, wlock)

@eh.wrapcommand('commit')
def commitwrapper(orig, ui, repo, *arg, **kwargs):
    tr = None
    if kwargs.get('amend', False):
        wlock = lock = None
    else:
        wlock = repo.wlock()
        lock = repo.lock()
    try:
        obsoleted = kwargs.get('obsolete', [])
        if obsoleted:
            obsoleted = repo.set('%lr', obsoleted)
        result = orig(ui, repo, *arg, **kwargs)
        if not result: # commit successed
            new = repo['-1']
            oldbookmarks = []
            markers = []
            for old in obsoleted:
                oldbookmarks.extend(repo.nodebookmarks(old.node()))
                markers.append((old, (new,)))
            if markers:
                obsolete.createmarkers(repo, markers)
            for book in oldbookmarks:
                repo._bookmarks[book] = new.node()
            if oldbookmarks:
                if not wlock:
                    wlock = repo.wlock()
                if not lock:
                    lock = repo.lock()
                tr = repo.transaction('commit')
                repo._bookmarks.recordchange(tr)
                tr.close()
        return result
    finally:
        lockmod.release(tr, lock, wlock)

@command('^split',
    [('r', 'rev', [], _("revision to fold")),
    ] + commitopts + commitopts2,
    _('hg split [OPTION]... [-r] REV'))
def cmdsplit(ui, repo, *revs, **opts):
    """split a changeset into smaller changesets

    By default, split the current revision by prompting for all its hunks to be
    redistributed into new changesets.

    Use --rev to split a given changeset instead.
    """
    tr = wlock = lock = None
    newcommits = []

    revarg = (list(revs) + opts.get('rev')) or ['.']
    if len(revarg) != 1:
        msg = _("more than one revset is given")
        hnt = _("use either `hg split <rs>` or `hg split --rev <rs>`, not both")
        raise error.Abort(msg, hint=hnt)

    rev = scmutil.revsingle(repo, revarg[0])
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        cmdutil.bailifchanged(repo)
        tr = repo.transaction('split')
        ctx = repo[rev]
        r = ctx.rev()
        disallowunstable = not obsolete.isenabled(repo,
                                                  obsolete.allowunstableopt)
        if disallowunstable:
            # XXX We should check head revs
            if repo.revs("(%d::) - %d", rev, rev):
                raise error.Abort(_("cannot split commit: %s not a head") % ctx)

        if len(ctx.parents()) > 1:
            raise error.Abort(_("cannot split merge commits"))
        prev = ctx.p1()
        bmupdate = _bookmarksupdater(repo, ctx.node(), tr)
        bookactive = bmactive(repo)
        if bookactive is not None:
            repo.ui.status(_("(leaving bookmark %s)\n") % bmactive(repo))
        bmdeactivate(repo)
        hg.update(repo, prev)

        commands.revert(ui, repo, rev=r, all=True)
        def haschanges():
            modified, added, removed, deleted = repo.status()[:4]
            return modified or added or removed or deleted
        msg = 'HG: Please, edit the original changeset description.\n\n'
        msg += ctx.description()
        opts['message'] = msg
        opts['edit'] = True
        while haschanges():
            pats = ()
            cmdutil.dorecord(ui, repo, commands.commit, 'commit', False,
                             cmdutil.recordfilter, *pats, **opts)
            # TODO: Does no seem like the best way to do this
            # We should make dorecord return the newly created commit
            newcommits.append(repo['.'])
            if haschanges():
                if ui.prompt('Done splitting? [yN]', default='n') == 'y':
                    commands.commit(ui, repo, **opts)
                    newcommits.append(repo['.'])
                    break
            else:
                ui.status(_("no more change to split\n"))

        tip = repo[newcommits[-1]]
        bmupdate(tip.node())
        if bookactive is not None:
            bmactivate(repo, bookactive)
        obsolete.createmarkers(repo, [(repo[r], newcommits)])
        tr.close()
    finally:
        lockmod.release(tr, lock, wlock)


@eh.wrapcommand('strip', extension='strip', opts=[
    ('', 'bundle', None, _("delete the commit entirely and move it to a "
        "backup bundle")),
    ])
def stripwrapper(orig, ui, repo, *revs, **kwargs):
    if (not ui.configbool('experimental', 'prunestrip') or
        kwargs.get('bundle', False)):
        return orig(ui, repo, *revs, **kwargs)

    if kwargs.get('force'):
        ui.warn(_("warning: --force has no effect during strip with evolve "
                  "enabled\n"))
    if kwargs.get('no_backup', False):
        ui.warn(_("warning: --no-backup has no effect during strips with "
                  "evolve enabled\n"))

    revs = list(revs) + kwargs.pop('rev', [])
    revs = set(scmutil.revrange(repo, revs))
    revs = repo.revs("(%ld)::", revs)
    kwargs['rev'] = []
    kwargs['new'] = []
    kwargs['succ'] = []
    kwargs['biject'] = False
    return cmdprune(ui, repo, *revs, **kwargs)

@command('^touch',
    [('r', 'rev', [], 'revision to update'),
     ('D', 'duplicate', False,
      'do not mark the new revision as successor of the old one'),
     ('A', 'allowdivergence', False,
      'mark the new revision as successor of the old one potentially creating '
      'divergence')],
    # allow to choose the seed ?
    _('[-r] revs'))
def touch(ui, repo, *revs, **opts):
    """create successors that are identical to their predecessors except
    for the changeset ID

    This is used to "resurrect" changesets
    """
    duplicate = opts['duplicate']
    allowdivergence = opts['allowdivergence']
    revs = list(revs)
    revs.extend(opts['rev'])
    if not revs:
        revs = ['.']
    revs = scmutil.revrange(repo, revs)
    if not revs:
        ui.write_err('no revision to touch\n')
        return 1
    if not duplicate and repo.revs('public() and %ld', revs):
        raise error.Abort("can't touch public revision")
    displayer = cmdutil.show_changeset(ui, repo, {'template': shorttemplate})
    wlock = lock = tr = None
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        tr = repo.transaction('touch')
        revs.sort() # ensure parent are run first
        newmapping = {}
        for r in revs:
            ctx = repo[r]
            extra = ctx.extra().copy()
            extra['__touch-noise__'] = random.randint(0, 0xffffffff)
            # search for touched parent
            p1 = ctx.p1().node()
            p2 = ctx.p2().node()
            p1 = newmapping.get(p1, p1)
            p2 = newmapping.get(p2, p2)

            if not (duplicate or allowdivergence):
                # The user hasn't yet decided what to do with the revived
                # cset, let's ask
                sset = obsolete.successorssets(repo, ctx.node())
                nodivergencerisk = len(sset) == 0 or (
                                    len(sset) == 1 and
                                    len(sset[0]) == 1 and
                                    repo[sset[0][0]].rev() == ctx.rev()
                                   )
                if nodivergencerisk:
                    duplicate = False
                else:
                    displayer.show(ctx)
                    index = ui.promptchoice(
                        _("reviving this changeset will create divergence"
                        " unless you make a duplicate.\n(a)llow divergence or"
                        " (d)uplicate the changeset? $$ &Allowdivergence $$ "
                        "&Duplicate"), 0)
                    choice = ['allowdivergence', 'duplicate'][index]
                    if choice == 'allowdivergence':
                        duplicate = False
                    else:
                        duplicate = True

            new, unusedvariable = rewrite(repo, ctx, [], ctx,
                                          [p1, p2],
                                          commitopts={'extra': extra})
            # store touched version to help potential children
            newmapping[ctx.node()] = new

            if not duplicate:
                obsolete.createmarkers(repo, [(ctx, (repo[new],))])
            phases.retractboundary(repo, tr, ctx.phase(), [new])
            if ctx in repo[None].parents():
                repo.dirstate.beginparentchange()
                repo.dirstate.setparents(new, node.nullid)
                repo.dirstate.endparentchange()
        tr.close()
    finally:
        lockmod.release(tr, lock, wlock)

@command('^fold|squash',
    [('r', 'rev', [], _("revision to fold")),
     ('', 'exact', None, _("only fold specified revisions"))
    ] + commitopts + commitopts2,
    _('hg fold [OPTION]... [-r] REV'))
def fold(ui, repo, *revs, **opts):
    """fold multiple revisions into a single one

    By default, folds all the revisions linearly between the given revisions
    and the parent of the working directory.

    Use --exact for folding only the specified revisions while ignoring the
    parent of the working directory. In this case, the given revisions must
    form a linear unbroken chain.

    .. container:: verbose

     Some examples:

     - Fold the current revision with its parent::

         hg fold .^

     - Fold all draft revisions with working directory parent::

         hg fold 'draft()'

       See :hg:`help phases` for more about draft revisions and
       :hg:`help revsets` for more about the `draft()` keyword

     - Fold revisions between 3 and 6 with the working directory parent::

         hg fold 3::6

     - Fold revisions 3 and 4:

        hg fold "3 + 4" --exact

     - Only fold revisions linearly between foo and @::

         hg fold foo::@ --exact
    """
    revs = list(revs)
    revs.extend(opts['rev'])
    if not revs:
        raise error.Abort(_('no revisions specified'))

    revs = scmutil.revrange(repo, revs)

    if not opts['exact']:
        # Try to extend given revision starting from the working directory
        extrevs = repo.revs('(%ld::.) or (.::%ld)', revs, revs)
        discardedrevs = [r for r in revs if r not in extrevs]
        if discardedrevs:
            raise error.Abort(_("cannot fold non-linear revisions"),
                               hint=_("given revisions are unrelated to parent "
                                      "of working directory"))
        revs = extrevs

    if len(revs) == 1:
        ui.write_err(_('single revision specified, nothing to fold\n'))
        return 1

    roots = repo.revs('roots(%ld)', revs)
    if len(roots) > 1:
        raise error.Abort(_("cannot fold non-linear revisions "
                           "(multiple roots given)"))
    root = repo[roots.first()]
    if root.phase() <= phases.public:
        raise error.Abort(_("cannot fold public revisions"))
    heads = repo.revs('heads(%ld)', revs)
    if len(heads) > 1:
        raise error.Abort(_("cannot fold non-linear revisions "
                           "(multiple heads given)"))
    head = repo[heads.first()]
    disallowunstable = not obsolete.isenabled(repo, obsolete.allowunstableopt)
    if disallowunstable:
        if repo.revs("(%ld::) - %ld", revs, revs):
            raise error.Abort(_("cannot fold chain not ending with a head "\
                               "or with branching"))
    wlock = lock = None
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        tr = repo.transaction('touch')
        try:
            commitopts = opts.copy()
            allctx = [repo[r] for r in revs]
            targetphase = max(c.phase() for c in allctx)

            if commitopts.get('message') or commitopts.get('logfile'):
                commitopts['edit'] = False
            else:
                msgs = ["HG: This is a fold of %d changesets." % len(allctx)]
                msgs += ["HG: Commit message of changeset %s.\n\n%s\n" %
                         (c.rev(), c.description()) for c in allctx]
                commitopts['message'] =  "\n".join(msgs)
                commitopts['edit'] = True

            newid, unusedvariable = rewrite(repo, root, allctx, head,
                                            [root.p1().node(),
                                             root.p2().node()],
                                            commitopts=commitopts)
            phases.retractboundary(repo, tr, targetphase, [newid])
            obsolete.createmarkers(repo, [(ctx, (repo[newid],))
                                 for ctx in allctx])
            tr.close()
        finally:
            tr.release()
        ui.status('%i changesets folded\n' % len(revs))
        if repo['.'].rev() in revs:
            hg.update(repo, newid)
    finally:
        lockmod.release(lock, wlock)



@eh.wrapcommand('graft')
def graftwrapper(orig, ui, repo, *revs, **kwargs):
    kwargs = dict(kwargs)
    revs = list(revs) + kwargs.get('rev', [])
    kwargs['rev'] = []
    obsoleted = kwargs.setdefault('obsolete', [])

    wlock = lock = None
    try:
        wlock = repo.wlock()
        lock = repo.lock()
        if kwargs.get('old_obsolete'):
            if kwargs.get('continue'):
                obsoleted.extend(repo.opener.read('graftstate').splitlines())
            else:
                obsoleted.extend(revs)
        # convert obsolete target into revs to avoid alias joke
        obsoleted[:] = [str(i) for i in repo.revs('%lr', obsoleted)]
        if obsoleted and len(revs) > 1:

            raise error.Abort(_('cannot graft multiple revisions while '
                                'obsoleting (for now).'))

        return commitwrapper(orig, ui, repo,*revs, **kwargs)
    finally:
        lockmod.release(lock, wlock)

@eh.extsetup
def oldevolveextsetup(ui):
    for cmd in ['kill', 'uncommit', 'touch', 'fold']:
        try:
            entry = extensions.wrapcommand(cmdtable, cmd,
                                           warnobserrors)
        except error.UnknownCommand:
            # Commands may be disabled
            continue

    entry = cmdutil.findcmd('commit', commands.table)[1]
    entry[1].append(('o', 'obsolete', [],
                     _("make commit obsolete this revision (DEPRECATED)")))
    entry = cmdutil.findcmd('graft', commands.table)[1]
    entry[1].append(('o', 'obsolete', [],
                     _("make graft obsoletes this revision (DEPRECATED)")))
    entry[1].append(('O', 'old-obsolete', False,
                     _("make graft obsoletes its source (DEPRECATED)")))

#####################################################################
### Obsolescence marker exchange experimenation                   ###
#####################################################################

def obsexcmsg(ui, message, important=False):
    verbose = ui.configbool('experimental', 'verbose-obsolescence-exchange',
                             False)
    if verbose:
        message = 'OBSEXC: ' + message
    if important or verbose:
        ui.status(message)

def obsexcprg(ui, *args, **kwargs):
    topic = 'obsmarkers exchange'
    if ui.configbool('experimental', 'verbose-obsolescence-exchange', False):
        topic = 'OBSEXC'
    ui.progress(topic, *args, **kwargs)

@eh.wrapfunction(exchange, '_pushdiscoveryobsmarkers')
def _pushdiscoveryobsmarkers(orig, pushop):
    if (obsolete.isenabled(pushop.repo, obsolete.exchangeopt)
        and pushop.repo.obsstore
        and 'obsolete' in pushop.remote.listkeys('namespaces')):
        repo = pushop.repo
        obsexcmsg(repo.ui, "computing relevant nodes\n")
        revs = list(repo.revs('::%ln', pushop.futureheads))
        unfi = repo.unfiltered()
        cl = unfi.changelog
        if not pushop.remote.capable('_evoext_obshash_0'):
            # do not trust core yet
            # return orig(pushop)
            nodes = [cl.node(r) for r in revs]
            if nodes:
                obsexcmsg(repo.ui, "computing markers relevant to %i nodes\n"
                                   % len(nodes))
                pushop.outobsmarkers = repo.obsstore.relevantmarkers(nodes)
            else:
                obsexcmsg(repo.ui, "markers already in sync\n")
                pushop.outobsmarkers = []
                pushop.outobsmarkers = repo.obsstore.relevantmarkers(nodes)
            return

        common = []
        obsexcmsg(repo.ui, "looking for common markers in %i nodes\n"
                           % len(revs))
        commonrevs = list(unfi.revs('::%ln', pushop.outgoing.commonheads))
        common = findcommonobsmarkers(pushop.ui, unfi, pushop.remote,
                                      commonrevs)

        revs = list(unfi.revs('%ld - (::%ln)', revs, common))
        nodes = [cl.node(r) for r in revs]
        if nodes:
            obsexcmsg(repo.ui, "computing markers relevant to %i nodes\n"
                               % len(nodes))
            pushop.outobsmarkers = repo.obsstore.relevantmarkers(nodes)
        else:
            obsexcmsg(repo.ui, "markers already in sync\n")
            pushop.outobsmarkers = []

@eh.wrapfunction(wireproto, 'capabilities')
def discocapabilities(orig, repo, proto):
    """wrapper to advertise new capability"""
    caps = orig(repo, proto)
    if obsolete.isenabled(repo, obsolete.exchangeopt):
        caps += ' _evoext_obshash_0'
    return caps

@eh.extsetup
def _installobsmarkersdiscovery(ui):
    hgweb_mod.perms['evoext_obshash'] = 'pull'
    hgweb_mod.perms['evoext_obshash1'] = 'pull'
    # wrap command content
    oldcap, args = wireproto.commands['capabilities']
    def newcap(repo, proto):
        return discocapabilities(oldcap, repo, proto)
    wireproto.commands['capabilities'] = (newcap, args)
    wireproto.commands['evoext_obshash'] = (srv_obshash, 'nodes')
    wireproto.commands['evoext_obshash1'] = (srv_obshash1, 'nodes')
    if getattr(exchange, '_pushdiscoveryobsmarkers', None) is None:
        ui.warn(_('evolve: your mercurial version is too old\n'
                  'evolve: (running in degraded mode, push will '
                  'includes all markers)\n'))
    else:
        olddisco = exchange.pushdiscoverymapping['obsmarker']
        def newdisco(pushop):
            _pushdiscoveryobsmarkers(olddisco, pushop)
        exchange.pushdiscoverymapping['obsmarker'] = newdisco

### Set discovery START

from mercurial import dagutil
from mercurial import setdiscovery

def _obshash(repo, nodes, version=0):
    if version == 0:
        hashs = _obsrelsethashtreefm0(repo)
    elif version ==1:
        hashs = _obsrelsethashtreefm1(repo)
    else:
        assert False
    nm = repo.changelog.nodemap
    revs = [nm.get(n) for n in nodes]
    return [r is None and nullid or hashs[r][1] for r in revs]

def srv_obshash(repo, proto, nodes):
    return wireproto.encodelist(_obshash(repo, wireproto.decodelist(nodes)))

def srv_obshash1(repo, proto, nodes):
    return wireproto.encodelist(_obshash(repo, wireproto.decodelist(nodes),
                                version=1))

@eh.addattr(localrepo.localpeer, 'evoext_obshash')
def local_obshash(peer, nodes):
    return _obshash(peer._repo, nodes)

@eh.addattr(localrepo.localpeer, 'evoext_obshash1')
def local_obshash1(peer, nodes):
    return _obshash(peer._repo, nodes, version=1)

@eh.addattr(wireproto.wirepeer, 'evoext_obshash')
def peer_obshash(self, nodes):
    d = self._call("evoext_obshash", nodes=wireproto.encodelist(nodes))
    try:
        return wireproto.decodelist(d)
    except ValueError:
        self._abort(error.ResponseError(_("unexpected response:"), d))

@eh.addattr(wireproto.wirepeer, 'evoext_obshash1')
def peer_obshash1(self, nodes):
    d = self._call("evoext_obshash1", nodes=wireproto.encodelist(nodes))
    try:
        return wireproto.decodelist(d)
    except ValueError:
        self._abort(error.ResponseError(_("unexpected response:"), d))

def findcommonobsmarkers(ui, local, remote, probeset,
                         initialsamplesize=100,
                         fullsamplesize=200):
    # from discovery
    roundtrips = 0
    cl = local.changelog
    dag = dagutil.revlogdag(cl)
    missing = set()
    common = set()
    undecided = set(probeset)
    totalnb = len(undecided)
    ui.progress(_("comparing with other"), 0, total=totalnb)
    _takefullsample = setdiscovery._takefullsample
    if remote.capable('_evoext_obshash_1'):
        getremotehash = remote.evoext_obshash1
        localhash = _obsrelsethashtreefm1(local)
    else:
        getremotehash = remote.evoext_obshash
        localhash = _obsrelsethashtreefm0(local)

    while undecided:

        ui.note(_("sampling from both directions\n"))
        if len(undecided) < fullsamplesize:
            sample = set(undecided)
        else:
            sample = _takefullsample(dag, undecided, size=fullsamplesize)

        roundtrips += 1
        ui.progress(_("comparing with other"), totalnb - len(undecided),
                    total=totalnb)
        ui.debug("query %i; still undecided: %i, sample size is: %i\n"
                 % (roundtrips, len(undecided), len(sample)))
        # indices between sample and externalized version must match
        sample = list(sample)
        remotehash = getremotehash(dag.externalizeall(sample))

        yesno = [localhash[ix][1] == remotehash[si]
                 for si, ix in enumerate(sample)]

        commoninsample = set(n for i, n in enumerate(sample) if yesno[i])
        common.update(dag.ancestorset(commoninsample, common))

        missinginsample = [n for i, n in enumerate(sample) if not yesno[i]]
        missing.update(dag.descendantset(missinginsample, missing))

        undecided.difference_update(missing)
        undecided.difference_update(common)


    ui.progress(_("comparing with other"), None, total=totalnb)
    result = dag.headsetofconnecteds(common)
    ui.debug("%d total queries\n" % roundtrips)

    if not result:
        return set([nullid])
    return dag.externalizeall(result)


_pushkeyescape = getattr(obsolete, '_pushkeyescape', None)

class pushobsmarkerStringIO(StringIO):
    """hacky string io for progress"""

    @util.propertycache
    def length(self):
        return len(self.getvalue())

    def read(self, size=None):
        obsexcprg(self.ui, self.tell(), unit="bytes", total=self.length)
        return StringIO.read(self, size)

    def __iter__(self):
        d = self.read(4096)
        while d:
            yield d
            d = self.read(4096)

@eh.wrapfunction(exchange, '_pushobsolete')
def _pushobsolete(orig, pushop):
    """utility function to push obsolete markers to a remote"""
    stepsdone = getattr(pushop, 'stepsdone', None)
    if stepsdone is not None:
        if 'obsmarkers' in stepsdone:
            return
        stepsdone.add('obsmarkers')
    if util.safehasattr(pushop, 'cgresult'):
        cgresult = pushop.cgresult
    else:
        cgresult = pushop.ret
    if cgresult == 0:
        return
    pushop.ui.debug('try to push obsolete markers to remote\n')
    repo = pushop.repo
    remote = pushop.remote
    if (obsolete.isenabled(repo, obsolete.exchangeopt) and repo.obsstore and
        'obsolete' in remote.listkeys('namespaces')):
        markers = pushop.outobsmarkers
        if not markers:
            obsexcmsg(repo.ui, "no marker to push\n")
        elif remote.capable('_evoext_pushobsmarkers_0'):
            obsdata = pushobsmarkerStringIO()
            for chunk in obsolete.encodemarkers(markers, True):
                obsdata.write(chunk)
            obsdata.seek(0)
            obsdata.ui = repo.ui
            obsexcmsg(repo.ui, "pushing %i obsolescence markers (%i bytes)\n"
                               % (len(markers), len(obsdata.getvalue())),
                      True)
            remote.evoext_pushobsmarkers_0(obsdata)
            obsexcprg(repo.ui, None)
        else:
            rslts = []
            remotedata = _pushkeyescape(markers).items()
            totalbytes = sum(len(d) for k, d in remotedata)
            sentbytes = 0
            obsexcmsg(repo.ui, "pushing %i obsolescence markers in %i "
                               "pushkey payload (%i bytes)\n"
                               % (len(markers), len(remotedata), totalbytes),
                      True)
            for key, data in remotedata:
                obsexcprg(repo.ui, sentbytes, item=key, unit="bytes",
                          total=totalbytes)
                rslts.append(remote.pushkey('obsolete', key, '', data))
                sentbytes += len(data)
                obsexcprg(repo.ui, sentbytes, item=key, unit="bytes",
                          total=totalbytes)
            obsexcprg(repo.ui, None)
            if [r for r in rslts if not r]:
                msg = _('failed to push some obsolete markers!\n')
                repo.ui.warn(msg)
        obsexcmsg(repo.ui, "DONE\n")


@eh.addattr(wireproto.wirepeer, 'evoext_pushobsmarkers_0')
def client_pushobsmarkers(self, obsfile):
    """wireprotocol peer method"""
    self.requirecap('_evoext_pushobsmarkers_0',
                    _('push obsolete markers faster'))
    ret, output = self._callpush('evoext_pushobsmarkers_0', obsfile)
    for l in output.splitlines(True):
        self.ui.status(_('remote: '), l)
    return ret

@eh.addattr(httppeer.httppeer, 'evoext_pushobsmarkers_0')
def httpclient_pushobsmarkers(self, obsfile):
    """httpprotocol peer method
    (Cannot simply use _callpush as http is doing some special handling)"""
    self.requirecap('_evoext_pushobsmarkers_0',
                    _('push obsolete markers faster'))
    try:
        r = self._call('evoext_pushobsmarkers_0', data=obsfile)
        vals = r.split('\n', 1)
        if len(vals) < 2:
            raise error.ResponseError(_("unexpected response:"), r)

        for l in vals[1].splitlines(True):
            if l.strip():
                self.ui.status(_('remote: '), l)
        return vals[0]
    except socket.error as err:
        if err.args[0] in (errno.ECONNRESET, errno.EPIPE):
            raise error.Abort(_('push failed: %s') % err.args[1])
        raise error.Abort(err.args[1])

@eh.wrapfunction(localrepo.localrepository, '_restrictcapabilities')
def local_pushobsmarker_capabilities(orig, repo, caps):
    caps = orig(repo, caps)
    caps.add('_evoext_pushobsmarkers_0')
    return caps

def _pushobsmarkers(repo, data):
    tr = lock = None
    try:
        lock = repo.lock()
        tr = repo.transaction('pushkey: obsolete markers')
        new = repo.obsstore.mergemarkers(tr, data)
        if new is not None:
            obsexcmsg(repo.ui, "%i obsolescence markers added\n" % new, True)
        tr.close()
    finally:
        lockmod.release(tr, lock)
    repo.hook('evolve_pushobsmarkers')

@eh.addattr(localrepo.localpeer, 'evoext_pushobsmarkers_0')
def local_pushobsmarkers(peer, obsfile):
    data = obsfile.read()
    _pushobsmarkers(peer._repo, data)

def srv_pushobsmarkers(repo, proto):
    """wireprotocol command"""
    fp = StringIO()
    proto.redirect()
    proto.getfile(fp)
    data = fp.getvalue()
    fp.close()
    _pushobsmarkers(repo, data)
    return wireproto.pushres(0)

def _buildpullobsmarkersboundaries(pullop):
    """small funtion returning the argument for pull markers call
    may to contains 'heads' and 'common'. skip the key for None.

    Its a separed functio to play around with strategy for that."""
    repo = pullop.repo
    remote = pullop.remote
    unfi = repo.unfiltered()
    revs = unfi.revs('::(%ln - null)', pullop.common)
    common = [nullid]
    if remote.capable('_evoext_obshash_0'):
        obsexcmsg(repo.ui, "looking for common markers in %i nodes\n"
                           % len(revs))
        common = findcommonobsmarkers(repo.ui, repo, remote, revs)
    return {'heads': pullop.pulledsubset, 'common': common}

@eh.uisetup
def addgetbundleargs(self):
    gboptsmap['evo_obscommon'] = 'nodes'

@eh.wrapfunction(exchange, '_pullbundle2extraprepare')
def _addobscommontob2pull(orig, pullop, kwargs):
    ret = orig(pullop, kwargs)
    if ('obsmarkers' in kwargs and
        pullop.remote.capable('_evoext_getbundle_obscommon')):
        boundaries = _buildpullobsmarkersboundaries(pullop)
        common = boundaries['common']
        if common != [nullid]:
            kwargs['evo_obscommon'] = common
    return ret

@eh.wrapfunction(exchange, '_getbundleobsmarkerpart')
def _getbundleobsmarkerpart(orig, bundler, repo, source, **kwargs):
    if 'evo_obscommon' not in kwargs:
        return orig(bundler, repo, source, **kwargs)

    heads = kwargs.get('heads')
    if kwargs.get('obsmarkers', False):
        if heads is None:
            heads = repo.heads()
        obscommon = kwargs.get('evo_obscommon', ())
        assert obscommon
        obsset = repo.unfiltered().set('::%ln - ::%ln', heads, obscommon)
        subset = [c.node() for c in obsset]
        markers = repo.obsstore.relevantmarkers(subset)
        exchange.buildobsmarkerspart(bundler, markers)

@eh.uisetup
def installgetbundlepartgen(ui):
    origfunc = exchange.getbundle2partsmapping['obsmarkers']
    def newfunc(*args, **kwargs):
        return _getbundleobsmarkerpart(origfunc, *args, **kwargs)
    exchange.getbundle2partsmapping['obsmarkers'] = newfunc

@eh.wrapfunction(exchange, '_pullobsolete')
def _pullobsolete(orig, pullop):
    if not obsolete.isenabled(pullop.repo, obsolete.exchangeopt):
        return None
    if 'obsmarkers' not in getattr(pullop, 'todosteps', ['obsmarkers']):
        return None
    if 'obsmarkers' in getattr(pullop, 'stepsdone', []):
        return None
    wirepull = pullop.remote.capable('_evoext_pullobsmarkers_0')
    if not wirepull:
        return orig(pullop)
    if 'obsolete' not in pullop.remote.listkeys('namespaces'):
        return None # remote opted out of obsolescence marker exchange
    tr = None
    ui = pullop.repo.ui
    boundaries = _buildpullobsmarkersboundaries(pullop)
    if not set(boundaries['heads']) - set(boundaries['common']):
        obsexcmsg(ui, "nothing to pull\n")
        return None

    obsexcmsg(ui, "pull obsolescence markers\n", True)
    new = 0

    if wirepull:
        obsdata = pullop.remote.evoext_pullobsmarkers_0(**boundaries)
        obsdata = obsdata.read()
        if len(obsdata) > 5:
            obsexcmsg(ui, "merging obsolescence markers (%i bytes)\n"
                           % len(obsdata))
            tr = pullop.gettransaction()
            old = len(pullop.repo.obsstore._all)
            pullop.repo.obsstore.mergemarkers(tr, obsdata)
            new = len(pullop.repo.obsstore._all) - old
            obsexcmsg(ui, "%i obsolescence markers added\n" % new, True)
        else:
            obsexcmsg(ui, "no unknown remote markers\n")
        obsexcmsg(ui, "DONE\n")
    if new:
        pullop.repo.invalidatevolatilesets()
    return tr

def _getobsmarkersstream(repo, heads=None, common=None):
    revset = ''
    args = []
    repo = repo.unfiltered()
    if heads is None:
        revset = 'all()'
    elif heads:
        revset += "(::%ln)"
        args.append(heads)
    else:
        assert False, 'pulling no heads?'
    if common:
        revset += ' - (::%ln)'
        args.append(common)
    nodes = [c.node() for c in repo.set(revset, *args)]
    markers = repo.obsstore.relevantmarkers(nodes)
    obsdata = StringIO()
    for chunk in obsolete.encodemarkers(markers, True):
        obsdata.write(chunk)
    obsdata.seek(0)
    return obsdata

@eh.addattr(wireproto.wirepeer, 'evoext_pullobsmarkers_0')
def client_pullobsmarkers(self, heads=None, common=None):
    self.requirecap('_evoext_pullobsmarkers_0', _('look up remote obsmarkers'))
    opts = {}
    if heads is not None:
        opts['heads'] = wireproto.encodelist(heads)
    if common is not None:
        opts['common'] = wireproto.encodelist(common)
    if util.safehasattr(self, '_callcompressable'):
        f = self._callcompressable("evoext_pullobsmarkers_0", **opts)
    else:
        f = self._callstream("evoext_pullobsmarkers_0", **opts)
        f = self._decompress(f)
    length = int(f.read(20))
    chunk = 4096
    current = 0
    data = StringIO()
    ui = self.ui
    obsexcprg(ui, current, unit="bytes", total=length)
    while current < length:
        readsize = min(length - current, chunk)
        data.write(f.read(readsize))
        current += readsize
        obsexcprg(ui, current, unit="bytes", total=length)
    obsexcprg(ui, None)
    data.seek(0)
    return data

@eh.addattr(localrepo.localpeer, 'evoext_pullobsmarkers_0')
def local_pullobsmarkers(self, heads=None, common=None):
    return _getobsmarkersstream(self._repo, heads=heads, common=common)

def srv_pullobsmarkers(repo, proto, others):
    opts = wireproto.options('', ['heads', 'common'], others)
    for k, v in opts.iteritems():
        if k in ('heads', 'common'):
            opts[k] = wireproto.decodelist(v)
    obsdata = _getobsmarkersstream(repo, **opts)
    finaldata = StringIO()
    obsdata = obsdata.getvalue()
    finaldata.write('%20i' % len(obsdata))
    finaldata.write(obsdata)
    finaldata.seek(0)
    return wireproto.streamres(proto.groupchunks(finaldata))

def _obsrelsethashtreefm0(repo):
    return _obsrelsethashtree(repo, obsolete._fm0encodeonemarker)

def _obsrelsethashtreefm1(repo):
    return _obsrelsethashtree(repo, obsolete._fm1encodeonemarker)

def _obsrelsethashtree(repo, encodeonemarker):
    cache = []
    unfi = repo.unfiltered()
    markercache = {}
    repo.ui.progress(_("preparing locally"), 0, total=len(unfi))
    for i in unfi:
        ctx = unfi[i]
        entry = 0
        sha = util.sha1()
        # add data from p1
        for p in ctx.parents():
            p = p.rev()
            if p < 0:
                p = nullid
            else:
                p = cache[p][1]
            if p != nullid:
                entry += 1
                sha.update(p)
        tmarkers = repo.obsstore.relevantmarkers([ctx.node()])
        if tmarkers:
            bmarkers = []
            for m in tmarkers:
                if not m in markercache:
                    markercache[m] = encodeonemarker(m)
                bmarkers.append(markercache[m])
            bmarkers.sort()
            for m in bmarkers:
                entry += 1
                sha.update(m)
        if entry:
            cache.append((ctx.node(), sha.digest()))
        else:
            cache.append((ctx.node(), nullid))
        repo.ui.progress(_("preparing locally"), i, total=len(unfi))
    repo.ui.progress(_("preparing locally"), None)
    return cache

@command('debugobsrelsethashtree',
        [('', 'v0', None, 'hash on marker format "0"'),
         ('', 'v1', None, 'hash on marker format "1" (default)')] , _(''))
def debugobsrelsethashtree(ui, repo, v0=False, v1=False):
    """display Obsolete markers, Relevant Set, Hash Tree
    changeset-node obsrelsethashtree-node

    It computed form the "orsht" of its parent and markers
    relevant to the changeset itself."""
    if v0 and v1:
        raise error.Abort('cannot only specify one format')
    elif v0:
        treefunc = _obsrelsethashtreefm0
    else:
        treefunc = _obsrelsethashtreefm1

    for chg, obs in treefunc(repo):
        ui.status('%s %s\n' % (node.hex(chg), node.hex(obs)))

_bestformat = max(obsolete.formats.keys())


@eh.wrapfunction(obsolete, '_checkinvalidmarkers')
def _checkinvalidmarkers(orig, markers):
    """search for marker with invalid data and raise error if needed

    Exist as a separated function to allow the evolve extension for a more
    subtle handling.
    """
    if 'debugobsconvert' in sys.argv:
        return
    for mark in markers:
        if node.nullid in mark[1]:
            raise error.Abort(_('bad obsolescence marker detected: '
                               'invalid successors nullid'),
                             hint=_('You should run `hg debugobsconvert`'))

@command(
    'debugobsconvert',
    [('', 'new-format', _bestformat, _('Destination format for markers.'))],
    '')
def debugobsconvert(ui, repo, new_format):
    origmarkers = repo.obsstore._all  # settle version
    if new_format == repo.obsstore._version:
        msg = _('New format is the same as the old format, not upgrading!')
        raise error.Abort(msg)
    f = repo.svfs('obsstore', 'wb', atomictemp=True)
    known = set()
    markers = []
    for m in origmarkers:
        # filter out invalid markers
        if nullid in m[1]:
            m = list(m)
            m[1] = tuple(s for s in m[1] if s != nullid)
            m = tuple(m)
        if m in known:
            continue
        known.add(m)
        markers.append(m)
    ui.write(_('Old store is version %d, will rewrite in version %d\n') % (
        repo.obsstore._version, new_format))
    map(f.write, obsolete.encodemarkers(markers, True, new_format))
    f.close()
    ui.write(_('Done!\n'))


@eh.wrapfunction(wireproto, 'capabilities')
def capabilities(orig, repo, proto):
    """wrapper to advertise new capability"""
    caps = orig(repo, proto)
    if obsolete.isenabled(repo, obsolete.exchangeopt):
        caps += ' _evoext_pushobsmarkers_0'
        caps += ' _evoext_pullobsmarkers_0'
        caps += ' _evoext_obshash_0'
        caps += ' _evoext_obshash_1'
        caps += ' _evoext_getbundle_obscommon'
    return caps


@eh.extsetup
def _installwireprotocol(ui):
    localrepo.moderncaps.add('_evoext_pullobsmarkers_0')
    hgweb_mod.perms['evoext_pushobsmarkers_0'] = 'push'
    hgweb_mod.perms['evoext_pullobsmarkers_0'] = 'pull'
    wireproto.commands['evoext_pushobsmarkers_0'] = (srv_pushobsmarkers, '')
    wireproto.commands['evoext_pullobsmarkers_0'] = (srv_pullobsmarkers, '*')
    # wrap command content
    oldcap, args = wireproto.commands['capabilities']
    def newcap(repo, proto):
        return capabilities(oldcap, repo, proto)
    wireproto.commands['capabilities'] = (newcap, args)

# Mercurial >= 3.6 passes ui
def _helploader(ui=None):
    return help.gettext(evolutionhelptext)

@eh.uisetup
def _setuphelp(ui):
    for entry in help.helptable:
        if entry[0] == "evolution":
            break
    else:
        help.helptable.append((["evolution"], _("Safely Rewriting History"),
                      _helploader))
        help.helptable.sort()

def _relocatecommit(repo, orig, commitmsg):
    if commitmsg is None:
        commitmsg = orig.description()
    extra = dict(orig.extra())
    if 'branch' in extra:
        del extra['branch']
    extra['rebase_source'] = orig.hex()

    backup = repo.ui.backupconfig('phases', 'new-commit')
    try:
        targetphase = max(orig.phase(), phases.draft)
        repo.ui.setconfig('phases', 'new-commit', targetphase, 'rebase')
        # Commit might fail if unresolved files exist
        nodenew = repo.commit(text=commitmsg, user=orig.user(),
                              date=orig.date(), extra=extra)
    finally:
        repo.ui.restoreconfig(backup)
    return nodenew

def _finalizerelocate(repo, orig, dest, nodenew, tr):
    destbookmarks = repo.nodebookmarks(dest.node())
    nodesrc = orig.node()
    destphase = repo[nodesrc].phase()
    oldbookmarks = repo.nodebookmarks(nodesrc)
    if nodenew is not None:
        phases.retractboundary(repo, tr, destphase, [nodenew])
        obsolete.createmarkers(repo, [(repo[nodesrc], (repo[nodenew],))])
        for book in oldbookmarks:
            repo._bookmarks[book] = nodenew
    else:
        obsolete.createmarkers(repo, [(repo[nodesrc], ())])
        # Behave like rebase, move bookmarks to dest
        for book in oldbookmarks:
            repo._bookmarks[book] = dest.node()
    for book in destbookmarks: # restore bookmark that rebase move
        repo._bookmarks[book] = dest.node()
    if oldbookmarks or destbookmarks:
        repo._bookmarks.recordchange(tr)

evolvestateversion = 0

@eh.uisetup
def setupevolveunfinished(ui):
    data = ('evolvestate', True, False, _('evolve in progress'),
           _("use 'hg evolve --continue' or 'hg update' to abort"))
    cmdutil.unfinishedstates.append(data)

@eh.wrapfunction(hg, 'clean')
def clean(orig, repo, *args, **kwargs):
    ret = orig(repo, *args, **kwargs)
    util.unlinkpath(repo.join('evolvestate'), ignoremissing=True)
    return ret

def _evolvestatewrite(repo, state):
    # [version]
    # [type][length][content]
    #
    # `version` is a 4 bytes integer (handled at higher level)
    # `type` is a single character, `length` is a 4 byte integer, and
    # `content` is an arbitrary byte sequence of length `length`.
    f = repo.vfs('evolvestate', 'w')
    try:
        f.write(_pack('>I', evolvestateversion))
        current = state['current']
        key = 'C' # as in 'current'
        format = '>sI%is' % len(current)
        f.write(_pack(format, key, len(current), current))
    finally:
        f.close()

def _evolvestateread(repo):
    try:
        f = repo.vfs('evolvestate')
    except IOError, err:
        if err.errno != errno.ENOENT:
            raise
        return None
    try:
        versionblob = f.read(4)
        if len(versionblob) < 4:
            repo.ui.debug('ignoring corrupted evolvestte (file contains %i bits)'
                          % len(versionblob))
            return None
        version = _unpack('>I', versionblob)[0]
        if version != evolvestateversion:
            raise error.Abort(_('unknown evolvestate version %i')
                                % version, hint=_('upgrade your evolve'))
        records = []
        data = f.read()
        off = 0
        end = len(data)
        while off < end:
            rtype = data[off]
            off += 1
            length = _unpack('>I', data[off:(off + 4)])[0]
            off += 4
            record = data[off:(off + length)]
            off += length
            if rtype == 't':
                rtype, record = record[0], record[1:]
            records.append((rtype, record))
        state = {}
        for rtype, rdata in records:
            if rtype == 'C':
                state['current'] = rdata
            elif rtype.lower():
                repo.ui.debug('ignore evolve state record type %s' % rtype)
            else:
                raise error.Abort(_('unknown evolvestate field type %r')
                                  % rtype, hint=_('upgrade your evolve'))
        return state
    finally:
        f.close()

def _evolvestatedelete(repo):
    util.unlinkpath(repo.join('evolvestate'), ignoremissing=True)

def _evolvemerge(repo, orig, dest, pctx, keepbranch):
    """Used by the evolve function to merge dest on top of pctx.
    return the same tuple as merge.graft"""
    if repo['.'].rev() != dest.rev():
       merge.update(repo, dest, False, True, False)
    if bmactive(repo):
       repo.ui.status(_("(leaving bookmark %s)\n") % bmactive(repo))
    bmdeactivate(repo)
    if keepbranch:
       repo.dirstate.setbranch(orig.branch())

    try:
       r = merge.graft(repo, orig, pctx, ['local', 'graft'], True)
    except TypeError:
       # not using recent enough mercurial
       if len(orig.parents()) == 2:
           raise error.Abort(
               _("no support for evolving merge changesets yet"),
               hint=_("Redo the merge and use `hg prune <old> --succ "
                      "<new>` to obsolete the old one"))

       r = merge.graft(repo, orig, pctx, ['local', 'graft'])
    return r