changegroup: replace changegroupsubset with makechangegroup

As part of getting rid of all the permutations of changegroup creation, let's
remove changegroupsubset and call makechangegroup instead. This moves the
responsibility of creating the outgoing set to the caller, but that seems like a
relatively reasonable unit of functionality for the caller to have to care about
(i.e. what commits should be bundled).

Differential Revision: https://phab.mercurial-scm.org/D665

hgext/shelve.py

@@ -1,1043 +1,1047
 # shelve.py - save/restore working directory state
 #
 # Copyright 2013 Facebook, Inc.
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 """save and restore changes to the working directory
 
 The "hg shelve" command saves changes made to the working directory
 and reverts those changes, resetting the working directory to a clean
 state.
 
 Later on, the "hg unshelve" command restores the changes saved by "hg
 shelve". Changes can be restored even after updating to a different
 parent, in which case Mercurial's merge machinery will resolve any
 conflicts if necessary.
 
 You can have more than one shelved change outstanding at a time; each
 shelved change has a distinct name. For details, see the help for "hg
 shelve".
 """
 from __future__ import absolute_import
 
 import collections
 import errno
 import itertools
 
 from mercurial.i18n import _
 from mercurial import (
     bookmarks,
     bundle2,
     bundlerepo,
     changegroup,
     cmdutil,
+    discovery,
     error,
     exchange,
     hg,
     lock as lockmod,
     mdiff,
     merge,
     node as nodemod,
     patch,
     phases,
     registrar,
     repair,
     scmutil,
     templatefilters,
     util,
     vfs as vfsmod,
 )
 
 from . import (
     rebase,
 )
 
 cmdtable = {}
 command = registrar.command(cmdtable)
 # Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for
 # extensions which SHIP WITH MERCURIAL. Non-mainline extensions should
 # be specifying the version(s) of Mercurial they are tested with, or
 # leave the attribute unspecified.
 testedwith = 'ships-with-hg-core'
 
 backupdir = 'shelve-backup'
 shelvedir = 'shelved'
 shelvefileextensions = ['hg', 'patch', 'oshelve']
 # universal extension is present in all types of shelves
 patchextension = 'patch'
 
 # we never need the user, so we use a
 # generic user for all shelve operations
 shelveuser = 'shelve@localhost'
 
 class shelvedfile(object):
     """Helper for the file storing a single shelve
 
     Handles common functions on shelve files (.hg/.patch) using
     the vfs layer"""
     def __init__(self, repo, name, filetype=None):
         self.repo = repo
         self.name = name
         self.vfs = vfsmod.vfs(repo.vfs.join(shelvedir))
         self.backupvfs = vfsmod.vfs(repo.vfs.join(backupdir))
         self.ui = self.repo.ui
         if filetype:
             self.fname = name + '.' + filetype
         else:
             self.fname = name
 
     def exists(self):
         return self.vfs.exists(self.fname)
 
     def filename(self):
         return self.vfs.join(self.fname)
 
     def backupfilename(self):
         def gennames(base):
             yield base
             base, ext = base.rsplit('.', 1)
             for i in itertools.count(1):
                 yield '%s-%d.%s' % (base, i, ext)
 
         name = self.backupvfs.join(self.fname)
         for n in gennames(name):
             if not self.backupvfs.exists(n):
                 return n
 
     def movetobackup(self):
         if not self.backupvfs.isdir():
             self.backupvfs.makedir()
         util.rename(self.filename(), self.backupfilename())
 
     def stat(self):
         return self.vfs.stat(self.fname)
 
     def opener(self, mode='rb'):
         try:
             return self.vfs(self.fname, mode)
         except IOError as err:
             if err.errno != errno.ENOENT:
                 raise
             raise error.Abort(_("shelved change '%s' not found") % self.name)
 
     def applybundle(self):
         fp = self.opener()
         try:
             gen = exchange.readbundle(self.repo.ui, fp, self.fname, self.vfs)
             bundle2.applybundle(self.repo, gen, self.repo.currenttransaction(),
                                 source='unshelve',
                                 url='bundle:' + self.vfs.join(self.fname),
                                 targetphase=phases.secret)
         finally:
             fp.close()
 
     def bundlerepo(self):
         return bundlerepo.bundlerepository(self.repo.baseui, self.repo.root,
                                            self.vfs.join(self.fname))
     def writebundle(self, bases, node):
         cgversion = changegroup.safeversion(self.repo)
         if cgversion == '01':
             btype = 'HG10BZ'
             compression = None
         else:
             btype = 'HG20'
             compression = 'BZ'
 
-        cg = changegroup.changegroupsubset(self.repo, bases, [node], 'shelve',
-                                           version=cgversion)
+        outgoing = discovery.outgoing(self.repo, missingroots=bases,
+                                      missingheads=[node])
+        cg = changegroup.makechangegroup(self.repo, outgoing, cgversion,
+                                         'shelve')
+
         bundle2.writebundle(self.ui, cg, self.fname, btype, self.vfs,
                                 compression=compression)
 
     def writeobsshelveinfo(self, info):
         scmutil.simplekeyvaluefile(self.vfs, self.fname).write(info)
 
     def readobsshelveinfo(self):
         return scmutil.simplekeyvaluefile(self.vfs, self.fname).read()
 
 class shelvedstate(object):
     """Handle persistence during unshelving operations.
 
     Handles saving and restoring a shelved state. Ensures that different
     versions of a shelved state are possible and handles them appropriately.
     """
     _version = 2
     _filename = 'shelvedstate'
     _keep = 'keep'
     _nokeep = 'nokeep'
     # colon is essential to differentiate from a real bookmark name
     _noactivebook = ':no-active-bookmark'
 
     @classmethod
     def _verifyandtransform(cls, d):
         """Some basic shelvestate syntactic verification and transformation"""
         try:
             d['originalwctx'] = nodemod.bin(d['originalwctx'])
             d['pendingctx'] = nodemod.bin(d['pendingctx'])
             d['parents'] = [nodemod.bin(h)
                             for h in d['parents'].split(' ')]
             d['nodestoremove'] = [nodemod.bin(h)
                                   for h in d['nodestoremove'].split(' ')]
         except (ValueError, TypeError, KeyError) as err:
             raise error.CorruptedState(str(err))
 
     @classmethod
     def _getversion(cls, repo):
         """Read version information from shelvestate file"""
         fp = repo.vfs(cls._filename)
         try:
             version = int(fp.readline().strip())
         except ValueError as err:
             raise error.CorruptedState(str(err))
         finally:
             fp.close()
         return version
 
     @classmethod
     def _readold(cls, repo):
         """Read the old position-based version of a shelvestate file"""
         # Order is important, because old shelvestate file uses it
         # to detemine values of fields (i.g. name is on the second line,
         # originalwctx is on the third and so forth). Please do not change.
         keys = ['version', 'name', 'originalwctx', 'pendingctx', 'parents',
                 'nodestoremove', 'branchtorestore', 'keep', 'activebook']
         # this is executed only seldomly, so it is not a big deal
         # that we open this file twice
         fp = repo.vfs(cls._filename)
         d = {}
         try:
             for key in keys:
                 d[key] = fp.readline().strip()
         finally:
             fp.close()
         return d
 
     @classmethod
     def load(cls, repo):
         version = cls._getversion(repo)
         if version < cls._version:
             d = cls._readold(repo)
         elif version == cls._version:
             d = scmutil.simplekeyvaluefile(repo.vfs, cls._filename)\
                        .read(firstlinenonkeyval=True)
         else:
             raise error.Abort(_('this version of shelve is incompatible '
                                 'with the version used in this repo'))
 
         cls._verifyandtransform(d)
         try:
             obj = cls()
             obj.name = d['name']
             obj.wctx = repo[d['originalwctx']]
             obj.pendingctx = repo[d['pendingctx']]
             obj.parents = d['parents']
             obj.nodestoremove = d['nodestoremove']
             obj.branchtorestore = d.get('branchtorestore', '')
             obj.keep = d.get('keep') == cls._keep
             obj.activebookmark = ''
             if d.get('activebook', '') != cls._noactivebook:
                 obj.activebookmark = d.get('activebook', '')
         except (error.RepoLookupError, KeyError) as err:
             raise error.CorruptedState(str(err))
 
         return obj
 
     @classmethod
     def save(cls, repo, name, originalwctx, pendingctx, nodestoremove,
              branchtorestore, keep=False, activebook=''):
         info = {
             "name": name,
             "originalwctx": nodemod.hex(originalwctx.node()),
             "pendingctx": nodemod.hex(pendingctx.node()),
             "parents": ' '.join([nodemod.hex(p)
                                  for p in repo.dirstate.parents()]),
             "nodestoremove": ' '.join([nodemod.hex(n)
                                       for n in nodestoremove]),
             "branchtorestore": branchtorestore,
             "keep": cls._keep if keep else cls._nokeep,
             "activebook": activebook or cls._noactivebook
         }
         scmutil.simplekeyvaluefile(repo.vfs, cls._filename)\
                .write(info, firstline=str(cls._version))
 
     @classmethod
     def clear(cls, repo):
         repo.vfs.unlinkpath(cls._filename, ignoremissing=True)
 
 def cleanupoldbackups(repo):
     vfs = vfsmod.vfs(repo.vfs.join(backupdir))
     maxbackups = repo.ui.configint('shelve', 'maxbackups', 10)
     hgfiles = [f for f in vfs.listdir()
                if f.endswith('.' + patchextension)]
     hgfiles = sorted([(vfs.stat(f).st_mtime, f) for f in hgfiles])
     if 0 < maxbackups and maxbackups < len(hgfiles):
         bordermtime = hgfiles[-maxbackups][0]
     else:
         bordermtime = None
     for mtime, f in hgfiles[:len(hgfiles) - maxbackups]:
         if mtime == bordermtime:
             # keep it, because timestamp can't decide exact order of backups
             continue
         base = f[:-(1 + len(patchextension))]
         for ext in shelvefileextensions:
             vfs.tryunlink(base + '.' + ext)
 
 def _backupactivebookmark(repo):
     activebookmark = repo._activebookmark
     if activebookmark:
         bookmarks.deactivate(repo)
     return activebookmark
 
 def _restoreactivebookmark(repo, mark):
     if mark:
         bookmarks.activate(repo, mark)
 
 def _aborttransaction(repo):
     '''Abort current transaction for shelve/unshelve, but keep dirstate
     '''
     tr = repo.currenttransaction()
     backupname = 'dirstate.shelve'
     repo.dirstate.savebackup(tr, backupname)
     tr.abort()
     repo.dirstate.restorebackup(None, backupname)
 
 def createcmd(ui, repo, pats, opts):
     """subcommand that creates a new shelve"""
     with repo.wlock():
         cmdutil.checkunfinished(repo)
         return _docreatecmd(ui, repo, pats, opts)
 
 def getshelvename(repo, parent, opts):
     """Decide on the name this shelve is going to have"""
     def gennames():
         yield label
         for i in itertools.count(1):
             yield '%s-%02d' % (label, i)
     name = opts.get('name')
     label = repo._activebookmark or parent.branch() or 'default'
     # slashes aren't allowed in filenames, therefore we rename it
     label = label.replace('/', '_')
     label = label.replace('\\', '_')
     # filenames must not start with '.' as it should not be hidden
     if label.startswith('.'):
         label = label.replace('.', '_', 1)
 
     if name:
         if shelvedfile(repo, name, patchextension).exists():
             e = _("a shelved change named '%s' already exists") % name
             raise error.Abort(e)
 
         # ensure we are not creating a subdirectory or a hidden file
         if '/' in name or '\\' in name:
             raise error.Abort(_('shelved change names can not contain slashes'))
         if name.startswith('.'):
             raise error.Abort(_("shelved change names can not start with '.'"))
 
     else:
         for n in gennames():
             if not shelvedfile(repo, n, patchextension).exists():
                 name = n
                 break
 
     return name
 
 def mutableancestors(ctx):
     """return all mutable ancestors for ctx (included)
 
     Much faster than the revset ancestors(ctx) & draft()"""
     seen = {nodemod.nullrev}
     visit = collections.deque()
     visit.append(ctx)
     while visit:
         ctx = visit.popleft()
         yield ctx.node()
         for parent in ctx.parents():
             rev = parent.rev()
             if rev not in seen:
                 seen.add(rev)
                 if parent.mutable():
                     visit.append(parent)
 
 def getcommitfunc(extra, interactive, editor=False):
     def commitfunc(ui, repo, message, match, opts):
         hasmq = util.safehasattr(repo, 'mq')
         if hasmq:
             saved, repo.mq.checkapplied = repo.mq.checkapplied, False
         overrides = {('phases', 'new-commit'): phases.secret}
         try:
             editor_ = False
             if editor:
                 editor_ = cmdutil.getcommiteditor(editform='shelve.shelve',
                                                   **opts)
             with repo.ui.configoverride(overrides):
                 return repo.commit(message, shelveuser, opts.get('date'),
                                    match, editor=editor_, extra=extra)
         finally:
             if hasmq:
                 repo.mq.checkapplied = saved
 
     def interactivecommitfunc(ui, repo, *pats, **opts):
         match = scmutil.match(repo['.'], pats, {})
         message = opts['message']
         return commitfunc(ui, repo, message, match, opts)
 
     return interactivecommitfunc if interactive else commitfunc
 
 def _nothingtoshelvemessaging(ui, repo, pats, opts):
     stat = repo.status(match=scmutil.match(repo[None], pats, opts))
     if stat.deleted:
         ui.status(_("nothing changed (%d missing files, see "
                     "'hg status')\n") % len(stat.deleted))
     else:
         ui.status(_("nothing changed\n"))
 
 def _shelvecreatedcommit(repo, node, name):
     bases = list(mutableancestors(repo[node]))
     shelvedfile(repo, name, 'hg').writebundle(bases, node)
     cmdutil.export(repo, [node],
                    fp=shelvedfile(repo, name, patchextension).opener('wb'),
                    opts=mdiff.diffopts(git=True))
 
 def _includeunknownfiles(repo, pats, opts, extra):
     s = repo.status(match=scmutil.match(repo[None], pats, opts),
                     unknown=True)
     if s.unknown:
         extra['shelve_unknown'] = '\0'.join(s.unknown)
         repo[None].add(s.unknown)
 
 def _finishshelve(repo):
     _aborttransaction(repo)
 
 def _docreatecmd(ui, repo, pats, opts):
     wctx = repo[None]
     parents = wctx.parents()
     if len(parents) > 1:
         raise error.Abort(_('cannot shelve while merging'))
     parent = parents[0]
     origbranch = wctx.branch()
 
     if parent.node() != nodemod.nullid:
         desc = "changes to: %s" % parent.description().split('\n', 1)[0]
     else:
         desc = '(changes in empty repository)'
 
     if not opts.get('message'):
         opts['message'] = desc
 
     lock = tr = activebookmark = None
     try:
         lock = repo.lock()
 
         # use an uncommitted transaction to generate the bundle to avoid
         # pull races. ensure we don't print the abort message to stderr.
         tr = repo.transaction('commit', report=lambda x: None)
 
         interactive = opts.get('interactive', False)
         includeunknown = (opts.get('unknown', False) and
                           not opts.get('addremove', False))
 
         name = getshelvename(repo, parent, opts)
         activebookmark = _backupactivebookmark(repo)
         extra = {}
         if includeunknown:
             _includeunknownfiles(repo, pats, opts, extra)
 
         if _iswctxonnewbranch(repo) and not _isbareshelve(pats, opts):
             # In non-bare shelve we don't store newly created branch
             # at bundled commit
             repo.dirstate.setbranch(repo['.'].branch())
 
         commitfunc = getcommitfunc(extra, interactive, editor=True)
         if not interactive:
             node = cmdutil.commit(ui, repo, commitfunc, pats, opts)
         else:
             node = cmdutil.dorecord(ui, repo, commitfunc, None,
                                     False, cmdutil.recordfilter, *pats,
                                     **opts)
         if not node:
             _nothingtoshelvemessaging(ui, repo, pats, opts)
             return 1
 
         _shelvecreatedcommit(repo, node, name)
 
         if ui.formatted():
             desc = util.ellipsis(desc, ui.termwidth())
         ui.status(_('shelved as %s\n') % name)
         hg.update(repo, parent.node())
         if origbranch != repo['.'].branch() and not _isbareshelve(pats, opts):
             repo.dirstate.setbranch(origbranch)
 
         _finishshelve(repo)
     finally:
         _restoreactivebookmark(repo, activebookmark)
         lockmod.release(tr, lock)
 
 def _isbareshelve(pats, opts):
     return (not pats
             and not opts.get('interactive', False)
             and not opts.get('include', False)
             and not opts.get('exclude', False))
 
 def _iswctxonnewbranch(repo):
     return repo[None].branch() != repo['.'].branch()
 
 def cleanupcmd(ui, repo):
     """subcommand that deletes all shelves"""
 
     with repo.wlock():
         for (name, _type) in repo.vfs.readdir(shelvedir):
             suffix = name.rsplit('.', 1)[-1]
             if suffix in shelvefileextensions:
                 shelvedfile(repo, name).movetobackup()
             cleanupoldbackups(repo)
 
 def deletecmd(ui, repo, pats):
     """subcommand that deletes a specific shelve"""
     if not pats:
         raise error.Abort(_('no shelved changes specified!'))
     with repo.wlock():
         try:
             for name in pats:
                 for suffix in shelvefileextensions:
                     shfile = shelvedfile(repo, name, suffix)
                     # patch file is necessary, as it should
                     # be present for any kind of shelve,
                     # but the .hg file is optional as in future we
                     # will add obsolete shelve with does not create a
                     # bundle
                     if shfile.exists() or suffix == patchextension:
                         shfile.movetobackup()
             cleanupoldbackups(repo)
         except OSError as err:
             if err.errno != errno.ENOENT:
                 raise
             raise error.Abort(_("shelved change '%s' not found") % name)
 
 def listshelves(repo):
     """return all shelves in repo as list of (time, filename)"""
     try:
         names = repo.vfs.readdir(shelvedir)
     except OSError as err:
         if err.errno != errno.ENOENT:
             raise
         return []
     info = []
     for (name, _type) in names:
         pfx, sfx = name.rsplit('.', 1)
         if not pfx or sfx != patchextension:
             continue
         st = shelvedfile(repo, name).stat()
         info.append((st.st_mtime, shelvedfile(repo, pfx).filename()))
     return sorted(info, reverse=True)
 
 def listcmd(ui, repo, pats, opts):
     """subcommand that displays the list of shelves"""
     pats = set(pats)
     width = 80
     if not ui.plain():
         width = ui.termwidth()
     namelabel = 'shelve.newest'
     ui.pager('shelve')
     for mtime, name in listshelves(repo):
         sname = util.split(name)[1]
         if pats and sname not in pats:
             continue
         ui.write(sname, label=namelabel)
         namelabel = 'shelve.name'
         if ui.quiet:
             ui.write('\n')
             continue
         ui.write(' ' * (16 - len(sname)))
         used = 16
         age = '(%s)' % templatefilters.age(util.makedate(mtime), abbrev=True)
         ui.write(age, label='shelve.age')
         ui.write(' ' * (12 - len(age)))
         used += 12
         with open(name + '.' + patchextension, 'rb') as fp:
             while True:
                 line = fp.readline()
                 if not line:
                     break
                 if not line.startswith('#'):
                     desc = line.rstrip()
                     if ui.formatted():
                         desc = util.ellipsis(desc, width - used)
                     ui.write(desc)
                     break
             ui.write('\n')
             if not (opts['patch'] or opts['stat']):
                 continue
             difflines = fp.readlines()
             if opts['patch']:
                 for chunk, label in patch.difflabel(iter, difflines):
                     ui.write(chunk, label=label)
             if opts['stat']:
                 for chunk, label in patch.diffstatui(difflines, width=width):
                     ui.write(chunk, label=label)
 
 def patchcmds(ui, repo, pats, opts, subcommand):
     """subcommand that displays shelves"""
     if len(pats) == 0:
         raise error.Abort(_("--%s expects at least one shelf") % subcommand)
 
     for shelfname in pats:
         if not shelvedfile(repo, shelfname, patchextension).exists():
             raise error.Abort(_("cannot find shelf %s") % shelfname)
 
     listcmd(ui, repo, pats, opts)
 
 def checkparents(repo, state):
     """check parent while resuming an unshelve"""
     if state.parents != repo.dirstate.parents():
         raise error.Abort(_('working directory parents do not match unshelve '
                            'state'))
 
 def pathtofiles(repo, files):
     cwd = repo.getcwd()
     return [repo.pathto(f, cwd) for f in files]
 
 def unshelveabort(ui, repo, state, opts):
     """subcommand that abort an in-progress unshelve"""
     with repo.lock():
         try:
             checkparents(repo, state)
 
             repo.vfs.rename('unshelverebasestate', 'rebasestate')
             try:
                 rebase.rebase(ui, repo, **{
                     'abort' : True
                 })
             except Exception:
                 repo.vfs.rename('rebasestate', 'unshelverebasestate')
                 raise
 
             mergefiles(ui, repo, state.wctx, state.pendingctx)
             repair.strip(ui, repo, state.nodestoremove, backup=False,
                          topic='shelve')
         finally:
             shelvedstate.clear(repo)
             ui.warn(_("unshelve of '%s' aborted\n") % state.name)
 
 def mergefiles(ui, repo, wctx, shelvectx):
     """updates to wctx and merges the changes from shelvectx into the
     dirstate."""
     with ui.configoverride({('ui', 'quiet'): True}):
         hg.update(repo, wctx.node())
         files = []
         files.extend(shelvectx.files())
         files.extend(shelvectx.parents()[0].files())
 
         # revert will overwrite unknown files, so move them out of the way
         for file in repo.status(unknown=True).unknown:
             if file in files:
                 util.rename(file, scmutil.origpath(ui, repo, file))
         ui.pushbuffer(True)
         cmdutil.revert(ui, repo, shelvectx, repo.dirstate.parents(),
                        *pathtofiles(repo, files),
                        **{'no_backup': True})
         ui.popbuffer()
 
 def restorebranch(ui, repo, branchtorestore):
     if branchtorestore and branchtorestore != repo.dirstate.branch():
         repo.dirstate.setbranch(branchtorestore)
         ui.status(_('marked working directory as branch %s\n')
                   % branchtorestore)
 
 def unshelvecleanup(ui, repo, name, opts):
     """remove related files after an unshelve"""
     if not opts.get('keep'):
         for filetype in shelvefileextensions:
             shfile = shelvedfile(repo, name, filetype)
             if shfile.exists():
                 shfile.movetobackup()
         cleanupoldbackups(repo)
 
 def unshelvecontinue(ui, repo, state, opts):
     """subcommand to continue an in-progress unshelve"""
     # We're finishing off a merge. First parent is our original
     # parent, second is the temporary "fake" commit we're unshelving.
     with repo.lock():
         checkparents(repo, state)
         ms = merge.mergestate.read(repo)
         if list(ms.unresolved()):
             raise error.Abort(
                 _("unresolved conflicts, can't continue"),
                 hint=_("see 'hg resolve', then 'hg unshelve --continue'"))
 
         repo.vfs.rename('unshelverebasestate', 'rebasestate')
         try:
             rebase.rebase(ui, repo, **{
                 'continue' : True
             })
         except Exception:
             repo.vfs.rename('rebasestate', 'unshelverebasestate')
             raise
 
         shelvectx = repo['tip']
         if state.pendingctx not in shelvectx.parents():
             # rebase was a no-op, so it produced no child commit
             shelvectx = state.pendingctx
         else:
             # only strip the shelvectx if the rebase produced it
             state.nodestoremove.append(shelvectx.node())
 
         mergefiles(ui, repo, state.wctx, shelvectx)
         restorebranch(ui, repo, state.branchtorestore)
 
         repair.strip(ui, repo, state.nodestoremove, backup=False,
                      topic='shelve')
         _restoreactivebookmark(repo, state.activebookmark)
         shelvedstate.clear(repo)
         unshelvecleanup(ui, repo, state.name, opts)
         ui.status(_("unshelve of '%s' complete\n") % state.name)
 
 def _commitworkingcopychanges(ui, repo, opts, tmpwctx):
     """Temporarily commit working copy changes before moving unshelve commit"""
     # Store pending changes in a commit and remember added in case a shelve
     # contains unknown files that are part of the pending change
     s = repo.status()
     addedbefore = frozenset(s.added)
     if not (s.modified or s.added or s.removed):
         return tmpwctx, addedbefore
     ui.status(_("temporarily committing pending changes "
                 "(restore with 'hg unshelve --abort')\n"))
     commitfunc = getcommitfunc(extra=None, interactive=False,
                                editor=False)
     tempopts = {}
     tempopts['message'] = "pending changes temporary commit"
     tempopts['date'] = opts.get('date')
     with ui.configoverride({('ui', 'quiet'): True}):
         node = cmdutil.commit(ui, repo, commitfunc, [], tempopts)
     tmpwctx = repo[node]
     return tmpwctx, addedbefore
 
 def _unshelverestorecommit(ui, repo, basename):
     """Recreate commit in the repository during the unshelve"""
     with ui.configoverride({('ui', 'quiet'): True}):
         shelvedfile(repo, basename, 'hg').applybundle()
         shelvectx = repo['tip']
     return repo, shelvectx
 
 def _rebaserestoredcommit(ui, repo, opts, tr, oldtiprev, basename, pctx,
                           tmpwctx, shelvectx, branchtorestore,
                           activebookmark):
     """Rebase restored commit from its original location to a destination"""
     # If the shelve is not immediately on top of the commit
     # we'll be merging with, rebase it to be on top.
     if tmpwctx.node() == shelvectx.parents()[0].node():
         return shelvectx
 
     ui.status(_('rebasing shelved changes\n'))
     try:
         rebase.rebase(ui, repo, **{
             'rev': [shelvectx.rev()],
             'dest': str(tmpwctx.rev()),
             'keep': True,
             'tool': opts.get('tool', ''),
         })
     except error.InterventionRequired:
         tr.close()
 
         nodestoremove = [repo.changelog.node(rev)
                          for rev in xrange(oldtiprev, len(repo))]
         shelvedstate.save(repo, basename, pctx, tmpwctx, nodestoremove,
                           branchtorestore, opts.get('keep'), activebookmark)
 
         repo.vfs.rename('rebasestate', 'unshelverebasestate')
         raise error.InterventionRequired(
             _("unresolved conflicts (see 'hg resolve', then "
               "'hg unshelve --continue')"))
 
     # refresh ctx after rebase completes
     shelvectx = repo['tip']
 
     if tmpwctx not in shelvectx.parents():
         # rebase was a no-op, so it produced no child commit
         shelvectx = tmpwctx
     return shelvectx
 
 def _forgetunknownfiles(repo, shelvectx, addedbefore):
     # Forget any files that were unknown before the shelve, unknown before
     # unshelve started, but are now added.
     shelveunknown = shelvectx.extra().get('shelve_unknown')
     if not shelveunknown:
         return
     shelveunknown = frozenset(shelveunknown.split('\0'))
     addedafter = frozenset(repo.status().added)
     toforget = (addedafter & shelveunknown) - addedbefore
     repo[None].forget(toforget)
 
 def _finishunshelve(repo, oldtiprev, tr, activebookmark):
     _restoreactivebookmark(repo, activebookmark)
     # The transaction aborting will strip all the commits for us,
     # but it doesn't update the inmemory structures, so addchangegroup
     # hooks still fire and try to operate on the missing commits.
     # Clean up manually to prevent this.
     repo.unfiltered().changelog.strip(oldtiprev, tr)
     _aborttransaction(repo)
 
 def _checkunshelveuntrackedproblems(ui, repo, shelvectx):
     """Check potential problems which may result from working
     copy having untracked changes."""
     wcdeleted = set(repo.status().deleted)
     shelvetouched = set(shelvectx.files())
     intersection = wcdeleted.intersection(shelvetouched)
     if intersection:
         m = _("shelved change touches missing files")
         hint = _("run hg status to see which files are missing")
         raise error.Abort(m, hint=hint)
 
 @command('unshelve',
          [('a', 'abort', None,
            _('abort an incomplete unshelve operation')),
           ('c', 'continue', None,
            _('continue an incomplete unshelve operation')),
           ('k', 'keep', None,
            _('keep shelve after unshelving')),
           ('n', 'name', '',
            _('restore shelved change with given name'), _('NAME')),
           ('t', 'tool', '', _('specify merge tool')),
           ('', 'date', '',
            _('set date for temporary commits (DEPRECATED)'), _('DATE'))],
          _('hg unshelve [[-n] SHELVED]'))
 def unshelve(ui, repo, *shelved, **opts):
     """restore a shelved change to the working directory
 
     This command accepts an optional name of a shelved change to
     restore. If none is given, the most recent shelved change is used.
 
     If a shelved change is applied successfully, the bundle that
     contains the shelved changes is moved to a backup location
     (.hg/shelve-backup).
 
     Since you can restore a shelved change on top of an arbitrary
     commit, it is possible that unshelving will result in a conflict
     between your changes and the commits you are unshelving onto. If
     this occurs, you must resolve the conflict, then use
     ``--continue`` to complete the unshelve operation. (The bundle
     will not be moved until you successfully complete the unshelve.)
 
     (Alternatively, you can use ``--abort`` to abandon an unshelve
     that causes a conflict. This reverts the unshelved changes, and
     leaves the bundle in place.)
 
     If bare shelved change(when no files are specified, without interactive,
     include and exclude option) was done on newly created branch it would
     restore branch information to the working directory.
 
     After a successful unshelve, the shelved changes are stored in a
     backup directory. Only the N most recent backups are kept. N
     defaults to 10 but can be overridden using the ``shelve.maxbackups``
     configuration option.
 
     .. container:: verbose
 
        Timestamp in seconds is used to decide order of backups. More
        than ``maxbackups`` backups are kept, if same timestamp
        prevents from deciding exact order of them, for safety.
     """
     with repo.wlock():
         return _dounshelve(ui, repo, *shelved, **opts)
 
 def _dounshelve(ui, repo, *shelved, **opts):
     abortf = opts.get('abort')
     continuef = opts.get('continue')
     if not abortf and not continuef:
         cmdutil.checkunfinished(repo)
     shelved = list(shelved)
     if opts.get("name"):
         shelved.append(opts["name"])
 
     if abortf or continuef:
         if abortf and continuef:
             raise error.Abort(_('cannot use both abort and continue'))
         if shelved:
             raise error.Abort(_('cannot combine abort/continue with '
                                'naming a shelved change'))
         if abortf and opts.get('tool', False):
             ui.warn(_('tool option will be ignored\n'))
 
         try:
             state = shelvedstate.load(repo)
             if opts.get('keep') is None:
                 opts['keep'] = state.keep
         except IOError as err:
             if err.errno != errno.ENOENT:
                 raise
             cmdutil.wrongtooltocontinue(repo, _('unshelve'))
         except error.CorruptedState as err:
             ui.debug(str(err) + '\n')
             if continuef:
                 msg = _('corrupted shelved state file')
                 hint = _('please run hg unshelve --abort to abort unshelve '
                          'operation')
                 raise error.Abort(msg, hint=hint)
             elif abortf:
                 msg = _('could not read shelved state file, your working copy '
                         'may be in an unexpected state\nplease update to some '
                         'commit\n')
                 ui.warn(msg)
                 shelvedstate.clear(repo)
             return
 
         if abortf:
             return unshelveabort(ui, repo, state, opts)
         elif continuef:
             return unshelvecontinue(ui, repo, state, opts)
     elif len(shelved) > 1:
         raise error.Abort(_('can only unshelve one change at a time'))
     elif not shelved:
         shelved = listshelves(repo)
         if not shelved:
             raise error.Abort(_('no shelved changes to apply!'))
         basename = util.split(shelved[0][1])[1]
         ui.status(_("unshelving change '%s'\n") % basename)
     else:
         basename = shelved[0]
 
     if not shelvedfile(repo, basename, patchextension).exists():
         raise error.Abort(_("shelved change '%s' not found") % basename)
 
     lock = tr = None
     try:
         lock = repo.lock()
         tr = repo.transaction('unshelve', report=lambda x: None)
         oldtiprev = len(repo)
 
         pctx = repo['.']
         tmpwctx = pctx
         # The goal is to have a commit structure like so:
         # ...-> pctx -> tmpwctx -> shelvectx
         # where tmpwctx is an optional commit with the user's pending changes
         # and shelvectx is the unshelved changes. Then we merge it all down
         # to the original pctx.
 
         activebookmark = _backupactivebookmark(repo)
         overrides = {('ui', 'forcemerge'): opts.get('tool', '')}
         with ui.configoverride(overrides, 'unshelve'):
             tmpwctx, addedbefore = _commitworkingcopychanges(ui, repo, opts,
                                                              tmpwctx)
             repo, shelvectx = _unshelverestorecommit(ui, repo, basename)
             _checkunshelveuntrackedproblems(ui, repo, shelvectx)
             branchtorestore = ''
             if shelvectx.branch() != shelvectx.p1().branch():
                 branchtorestore = shelvectx.branch()
 
             shelvectx = _rebaserestoredcommit(ui, repo, opts, tr, oldtiprev,
                                               basename, pctx, tmpwctx,
                                               shelvectx, branchtorestore,
                                               activebookmark)
             mergefiles(ui, repo, pctx, shelvectx)
             restorebranch(ui, repo, branchtorestore)
             _forgetunknownfiles(repo, shelvectx, addedbefore)
 
             shelvedstate.clear(repo)
             _finishunshelve(repo, oldtiprev, tr, activebookmark)
             unshelvecleanup(ui, repo, basename, opts)
     finally:
         if tr:
             tr.release()
         lockmod.release(lock)
 
 @command('shelve',
          [('A', 'addremove', None,
            _('mark new/missing files as added/removed before shelving')),
           ('u', 'unknown', None,
            _('store unknown files in the shelve')),
           ('', 'cleanup', None,
            _('delete all shelved changes')),
           ('', 'date', '',
            _('shelve with the specified commit date'), _('DATE')),
           ('d', 'delete', None,
            _('delete the named shelved change(s)')),
           ('e', 'edit', False,
            _('invoke editor on commit messages')),
           ('l', 'list', None,
            _('list current shelves')),
           ('m', 'message', '',
            _('use text as shelve message'), _('TEXT')),
           ('n', 'name', '',
            _('use the given name for the shelved commit'), _('NAME')),
           ('p', 'patch', None,
            _('show patch')),
           ('i', 'interactive', None,
            _('interactive mode, only works while creating a shelve')),
           ('', 'stat', None,
            _('output diffstat-style summary of changes'))] + cmdutil.walkopts,
          _('hg shelve [OPTION]... [FILE]...'))
 def shelvecmd(ui, repo, *pats, **opts):
     '''save and set aside changes from the working directory
 
     Shelving takes files that "hg status" reports as not clean, saves
     the modifications to a bundle (a shelved change), and reverts the
     files so that their state in the working directory becomes clean.
 
     To restore these changes to the working directory, using "hg
     unshelve"; this will work even if you switch to a different
     commit.
 
     When no files are specified, "hg shelve" saves all not-clean
     files. If specific files or directories are named, only changes to
     those files are shelved.
 
     In bare shelve (when no files are specified, without interactive,
     include and exclude option), shelving remembers information if the
     working directory was on newly created branch, in other words working
     directory was on different branch than its first parent. In this
     situation unshelving restores branch information to the working directory.
 
     Each shelved change has a name that makes it easier to find later.
     The name of a shelved change defaults to being based on the active
     bookmark, or if there is no active bookmark, the current named
     branch.  To specify a different name, use ``--name``.
 
     To see a list of existing shelved changes, use the ``--list``
     option. For each shelved change, this will print its name, age,
     and description; use ``--patch`` or ``--stat`` for more details.
 
     To delete specific shelved changes, use ``--delete``. To delete
     all shelved changes, use ``--cleanup``.
     '''
     allowables = [
         ('addremove', {'create'}), # 'create' is pseudo action
         ('unknown', {'create'}),
         ('cleanup', {'cleanup'}),
 #       ('date', {'create'}), # ignored for passing '--date "0 0"' in tests
         ('delete', {'delete'}),
         ('edit', {'create'}),
         ('list', {'list'}),
         ('message', {'create'}),
         ('name', {'create'}),
         ('patch', {'patch', 'list'}),
         ('stat', {'stat', 'list'}),
     ]
     def checkopt(opt):
         if opts.get(opt):
             for i, allowable in allowables:
                 if opts[i] and opt not in allowable:
                     raise error.Abort(_("options '--%s' and '--%s' may not be "
                                        "used together") % (opt, i))
             return True
     if checkopt('cleanup'):
         if pats:
             raise error.Abort(_("cannot specify names when using '--cleanup'"))
         return cleanupcmd(ui, repo)
     elif checkopt('delete'):
         return deletecmd(ui, repo, pats)
     elif checkopt('list'):
         return listcmd(ui, repo, pats, opts)
     elif checkopt('patch'):
         return patchcmds(ui, repo, pats, opts, subcommand='patch')
     elif checkopt('stat'):
         return patchcmds(ui, repo, pats, opts, subcommand='stat')
     else:
         return createcmd(ui, repo, pats, opts)
 
 def extsetup(ui):
     cmdutil.unfinishedstates.append(
         [shelvedstate._filename, False, False,
          _('unshelve already in progress'),
          _("use 'hg unshelve --continue' or 'hg unshelve --abort'")])
     cmdutil.afterresolvedstates.append(
         [shelvedstate._filename, _('hg unshelve --continue')])

mercurial/changegroup.py

@@ -1,1030 +1,1016
 # changegroup.py - Mercurial changegroup manipulation functions
 #
 #  Copyright 2006 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import os
 import struct
 import tempfile
 import weakref
 
 from .i18n import _
 from .node import (
     hex,
     nullrev,
     short,
 )
 
 from . import (
     dagutil,
     discovery,
     error,
     mdiff,
     phases,
     pycompat,
     util,
 )
 
 _CHANGEGROUPV1_DELTA_HEADER = "20s20s20s20s"
 _CHANGEGROUPV2_DELTA_HEADER = "20s20s20s20s20s"
 _CHANGEGROUPV3_DELTA_HEADER = ">20s20s20s20s20sH"
 
 def readexactly(stream, n):
     '''read n bytes from stream.read and abort if less was available'''
     s = stream.read(n)
     if len(s) < n:
         raise error.Abort(_("stream ended unexpectedly"
                            " (got %d bytes, expected %d)")
                           % (len(s), n))
     return s
 
 def getchunk(stream):
     """return the next chunk from stream as a string"""
     d = readexactly(stream, 4)
     l = struct.unpack(">l", d)[0]
     if l <= 4:
         if l:
             raise error.Abort(_("invalid chunk length %d") % l)
         return ""
     return readexactly(stream, l - 4)
 
 def chunkheader(length):
     """return a changegroup chunk header (string)"""
     return struct.pack(">l", length + 4)
 
 def closechunk():
     """return a changegroup chunk header (string) for a zero-length chunk"""
     return struct.pack(">l", 0)
 
 def writechunks(ui, chunks, filename, vfs=None):
     """Write chunks to a file and return its filename.
 
     The stream is assumed to be a bundle file.
     Existing files will not be overwritten.
     If no filename is specified, a temporary file is created.
     """
     fh = None
     cleanup = None
     try:
         if filename:
             if vfs:
                 fh = vfs.open(filename, "wb")
             else:
                 # Increase default buffer size because default is usually
                 # small (4k is common on Linux).
                 fh = open(filename, "wb", 131072)
         else:
             fd, filename = tempfile.mkstemp(prefix="hg-bundle-", suffix=".hg")
             fh = os.fdopen(fd, pycompat.sysstr("wb"))
         cleanup = filename
         for c in chunks:
             fh.write(c)
         cleanup = None
         return filename
     finally:
         if fh is not None:
             fh.close()
         if cleanup is not None:
             if filename and vfs:
                 vfs.unlink(cleanup)
             else:
                 os.unlink(cleanup)
 
 class cg1unpacker(object):
     """Unpacker for cg1 changegroup streams.
 
     A changegroup unpacker handles the framing of the revision data in
     the wire format. Most consumers will want to use the apply()
     method to add the changes from the changegroup to a repository.
 
     If you're forwarding a changegroup unmodified to another consumer,
     use getchunks(), which returns an iterator of changegroup
     chunks. This is mostly useful for cases where you need to know the
     data stream has ended by observing the end of the changegroup.
 
     deltachunk() is useful only if you're applying delta data. Most
     consumers should prefer apply() instead.
 
     A few other public methods exist. Those are used only for
     bundlerepo and some debug commands - their use is discouraged.
     """
     deltaheader = _CHANGEGROUPV1_DELTA_HEADER
     deltaheadersize = struct.calcsize(deltaheader)
     version = '01'
     _grouplistcount = 1 # One list of files after the manifests
 
     def __init__(self, fh, alg, extras=None):
         if alg is None:
             alg = 'UN'
         if alg not in util.compengines.supportedbundletypes:
             raise error.Abort(_('unknown stream compression type: %s')
                              % alg)
         if alg == 'BZ':
             alg = '_truncatedBZ'
 
         compengine = util.compengines.forbundletype(alg)
         self._stream = compengine.decompressorreader(fh)
         self._type = alg
         self.extras = extras or {}
         self.callback = None
 
     # These methods (compressed, read, seek, tell) all appear to only
     # be used by bundlerepo, but it's a little hard to tell.
     def compressed(self):
         return self._type is not None and self._type != 'UN'
     def read(self, l):
         return self._stream.read(l)
     def seek(self, pos):
         return self._stream.seek(pos)
     def tell(self):
         return self._stream.tell()
     def close(self):
         return self._stream.close()
 
     def _chunklength(self):
         d = readexactly(self._stream, 4)
         l = struct.unpack(">l", d)[0]
         if l <= 4:
             if l:
                 raise error.Abort(_("invalid chunk length %d") % l)
             return 0
         if self.callback:
             self.callback()
         return l - 4
 
     def changelogheader(self):
         """v10 does not have a changelog header chunk"""
         return {}
 
     def manifestheader(self):
         """v10 does not have a manifest header chunk"""
         return {}
 
     def filelogheader(self):
         """return the header of the filelogs chunk, v10 only has the filename"""
         l = self._chunklength()
         if not l:
             return {}
         fname = readexactly(self._stream, l)
         return {'filename': fname}
 
     def _deltaheader(self, headertuple, prevnode):
         node, p1, p2, cs = headertuple
         if prevnode is None:
             deltabase = p1
         else:
             deltabase = prevnode
         flags = 0
         return node, p1, p2, deltabase, cs, flags
 
     def deltachunk(self, prevnode):
         l = self._chunklength()
         if not l:
             return {}
         headerdata = readexactly(self._stream, self.deltaheadersize)
         header = struct.unpack(self.deltaheader, headerdata)
         delta = readexactly(self._stream, l - self.deltaheadersize)
         node, p1, p2, deltabase, cs, flags = self._deltaheader(header, prevnode)
         return {'node': node, 'p1': p1, 'p2': p2, 'cs': cs,
                 'deltabase': deltabase, 'delta': delta, 'flags': flags}
 
     def getchunks(self):
         """returns all the chunks contains in the bundle
 
         Used when you need to forward the binary stream to a file or another
         network API. To do so, it parse the changegroup data, otherwise it will
         block in case of sshrepo because it don't know the end of the stream.
         """
         # For changegroup 1 and 2, we expect 3 parts: changelog, manifestlog,
         # and a list of filelogs. For changegroup 3, we expect 4 parts:
         # changelog, manifestlog, a list of tree manifestlogs, and a list of
         # filelogs.
         #
         # Changelog and manifestlog parts are terminated with empty chunks. The
         # tree and file parts are a list of entry sections. Each entry section
         # is a series of chunks terminating in an empty chunk. The list of these
         # entry sections is terminated in yet another empty chunk, so we know
         # we've reached the end of the tree/file list when we reach an empty
         # chunk that was proceeded by no non-empty chunks.
 
         parts = 0
         while parts < 2 + self._grouplistcount:
             noentries = True
             while True:
                 chunk = getchunk(self)
                 if not chunk:
                     # The first two empty chunks represent the end of the
                     # changelog and the manifestlog portions. The remaining
                     # empty chunks represent either A) the end of individual
                     # tree or file entries in the file list, or B) the end of
                     # the entire list. It's the end of the entire list if there
                     # were no entries (i.e. noentries is True).
                     if parts < 2:
                         parts += 1
                     elif noentries:
                         parts += 1
                     break
                 noentries = False
                 yield chunkheader(len(chunk))
                 pos = 0
                 while pos < len(chunk):
                     next = pos + 2**20
                     yield chunk[pos:next]
                     pos = next
             yield closechunk()
 
     def _unpackmanifests(self, repo, revmap, trp, prog, numchanges):
         # We know that we'll never have more manifests than we had
         # changesets.
         self.callback = prog(_('manifests'), numchanges)
         # no need to check for empty manifest group here:
         # if the result of the merge of 1 and 2 is the same in 3 and 4,
         # no new manifest will be created and the manifest group will
         # be empty during the pull
         self.manifestheader()
         repo.manifestlog._revlog.addgroup(self, revmap, trp)
         repo.ui.progress(_('manifests'), None)
         self.callback = None
 
     def apply(self, repo, tr, srctype, url, targetphase=phases.draft,
               expectedtotal=None):
         """Add the changegroup returned by source.read() to this repo.
         srctype is a string like 'push', 'pull', or 'unbundle'.  url is
         the URL of the repo where this changegroup is coming from.
 
         Return an integer summarizing the change to this repo:
         - nothing changed or no source: 0
         - more heads than before: 1+added heads (2..n)
         - fewer heads than before: -1-removed heads (-2..-n)
         - number of heads stays the same: 1
         """
         repo = repo.unfiltered()
         def csmap(x):
             repo.ui.debug("add changeset %s\n" % short(x))
             return len(cl)
 
         def revmap(x):
             return cl.rev(x)
 
         changesets = files = revisions = 0
 
         try:
             # The transaction may already carry source information. In this
             # case we use the top level data. We overwrite the argument
             # because we need to use the top level value (if they exist)
             # in this function.
             srctype = tr.hookargs.setdefault('source', srctype)
             url = tr.hookargs.setdefault('url', url)
             repo.hook('prechangegroup',
                       throw=True, **pycompat.strkwargs(tr.hookargs))
 
             # write changelog data to temp files so concurrent readers
             # will not see an inconsistent view
             cl = repo.changelog
             cl.delayupdate(tr)
             oldheads = set(cl.heads())
 
             trp = weakref.proxy(tr)
             # pull off the changeset group
             repo.ui.status(_("adding changesets\n"))
             clstart = len(cl)
             class prog(object):
                 def __init__(self, step, total):
                     self._step = step
                     self._total = total
                     self._count = 1
                 def __call__(self):
                     repo.ui.progress(self._step, self._count, unit=_('chunks'),
                                      total=self._total)
                     self._count += 1
             self.callback = prog(_('changesets'), expectedtotal)
 
             efiles = set()
             def onchangelog(cl, node):
                 efiles.update(cl.readfiles(node))
 
             self.changelogheader()
             cgnodes = cl.addgroup(self, csmap, trp, addrevisioncb=onchangelog)
             efiles = len(efiles)
 
             if not cgnodes:
                 repo.ui.develwarn('applied empty changegroup',
                                   config='empty-changegroup')
             clend = len(cl)
             changesets = clend - clstart
             repo.ui.progress(_('changesets'), None)
             self.callback = None
 
             # pull off the manifest group
             repo.ui.status(_("adding manifests\n"))
             self._unpackmanifests(repo, revmap, trp, prog, changesets)
 
             needfiles = {}
             if repo.ui.configbool('server', 'validate'):
                 cl = repo.changelog
                 ml = repo.manifestlog
                 # validate incoming csets have their manifests
                 for cset in xrange(clstart, clend):
                     mfnode = cl.changelogrevision(cset).manifest
                     mfest = ml[mfnode].readdelta()
                     # store file cgnodes we must see
                     for f, n in mfest.iteritems():
                         needfiles.setdefault(f, set()).add(n)
 
             # process the files
             repo.ui.status(_("adding file changes\n"))
             newrevs, newfiles = _addchangegroupfiles(
                 repo, self, revmap, trp, efiles, needfiles)
             revisions += newrevs
             files += newfiles
 
             deltaheads = 0
             if oldheads:
                 heads = cl.heads()
                 deltaheads = len(heads) - len(oldheads)
                 for h in heads:
                     if h not in oldheads and repo[h].closesbranch():
                         deltaheads -= 1
             htext = ""
             if deltaheads:
                 htext = _(" (%+d heads)") % deltaheads
 
             repo.ui.status(_("added %d changesets"
                              " with %d changes to %d files%s\n")
                              % (changesets, revisions, files, htext))
             repo.invalidatevolatilesets()
 
             if changesets > 0:
                 if 'node' not in tr.hookargs:
                     tr.hookargs['node'] = hex(cl.node(clstart))
                     tr.hookargs['node_last'] = hex(cl.node(clend - 1))
                     hookargs = dict(tr.hookargs)
                 else:
                     hookargs = dict(tr.hookargs)
                     hookargs['node'] = hex(cl.node(clstart))
                     hookargs['node_last'] = hex(cl.node(clend - 1))
                 repo.hook('pretxnchangegroup',
                           throw=True, **pycompat.strkwargs(hookargs))
 
             added = [cl.node(r) for r in xrange(clstart, clend)]
             phaseall = None
             if srctype in ('push', 'serve'):
                 # Old servers can not push the boundary themselves.
                 # New servers won't push the boundary if changeset already
                 # exists locally as secret
                 #
                 # We should not use added here but the list of all change in
                 # the bundle
                 if repo.publishing():
                     targetphase = phaseall = phases.public
                 else:
                     # closer target phase computation
 
                     # Those changesets have been pushed from the
                     # outside, their phases are going to be pushed
                     # alongside. Therefor `targetphase` is
                     # ignored.
                     targetphase = phaseall = phases.draft
             if added:
                 phases.registernew(repo, tr, targetphase, added)
             if phaseall is not None:
                 phases.advanceboundary(repo, tr, phaseall, cgnodes)
 
             if changesets > 0:
 
                 def runhooks():
                     # These hooks run when the lock releases, not when the
                     # transaction closes. So it's possible for the changelog
                     # to have changed since we last saw it.
                     if clstart >= len(repo):
                         return
 
                     repo.hook("changegroup", **pycompat.strkwargs(hookargs))
 
                     for n in added:
                         args = hookargs.copy()
                         args['node'] = hex(n)
                         del args['node_last']
                         repo.hook("incoming", **pycompat.strkwargs(args))
 
                     newheads = [h for h in repo.heads()
                                 if h not in oldheads]
                     repo.ui.log("incoming",
                                 "%s incoming changes - new heads: %s\n",
                                 len(added),
                                 ', '.join([hex(c[:6]) for c in newheads]))
 
                 tr.addpostclose('changegroup-runhooks-%020i' % clstart,
                                 lambda tr: repo._afterlock(runhooks))
         finally:
             repo.ui.flush()
         # never return 0 here:
         if deltaheads < 0:
             ret = deltaheads - 1
         else:
             ret = deltaheads + 1
         return ret
 
 class cg2unpacker(cg1unpacker):
     """Unpacker for cg2 streams.
 
     cg2 streams add support for generaldelta, so the delta header
     format is slightly different. All other features about the data
     remain the same.
     """
     deltaheader = _CHANGEGROUPV2_DELTA_HEADER
     deltaheadersize = struct.calcsize(deltaheader)
     version = '02'
 
     def _deltaheader(self, headertuple, prevnode):
         node, p1, p2, deltabase, cs = headertuple
         flags = 0
         return node, p1, p2, deltabase, cs, flags
 
 class cg3unpacker(cg2unpacker):
     """Unpacker for cg3 streams.
 
     cg3 streams add support for exchanging treemanifests and revlog
     flags. It adds the revlog flags to the delta header and an empty chunk
     separating manifests and files.
     """
     deltaheader = _CHANGEGROUPV3_DELTA_HEADER
     deltaheadersize = struct.calcsize(deltaheader)
     version = '03'
     _grouplistcount = 2 # One list of manifests and one list of files
 
     def _deltaheader(self, headertuple, prevnode):
         node, p1, p2, deltabase, cs, flags = headertuple
         return node, p1, p2, deltabase, cs, flags
 
     def _unpackmanifests(self, repo, revmap, trp, prog, numchanges):
         super(cg3unpacker, self)._unpackmanifests(repo, revmap, trp, prog,
                                                   numchanges)
         for chunkdata in iter(self.filelogheader, {}):
             # If we get here, there are directory manifests in the changegroup
             d = chunkdata["filename"]
             repo.ui.debug("adding %s revisions\n" % d)
             dirlog = repo.manifestlog._revlog.dirlog(d)
             if not dirlog.addgroup(self, revmap, trp):
                 raise error.Abort(_("received dir revlog group is empty"))
 
 class headerlessfixup(object):
     def __init__(self, fh, h):
         self._h = h
         self._fh = fh
     def read(self, n):
         if self._h:
             d, self._h = self._h[:n], self._h[n:]
             if len(d) < n:
                 d += readexactly(self._fh, n - len(d))
             return d
         return readexactly(self._fh, n)
 
 class cg1packer(object):
     deltaheader = _CHANGEGROUPV1_DELTA_HEADER
     version = '01'
     def __init__(self, repo, bundlecaps=None):
         """Given a source repo, construct a bundler.
 
         bundlecaps is optional and can be used to specify the set of
         capabilities which can be used to build the bundle. While bundlecaps is
         unused in core Mercurial, extensions rely on this feature to communicate
         capabilities to customize the changegroup packer.
         """
         # Set of capabilities we can use to build the bundle.
         if bundlecaps is None:
             bundlecaps = set()
         self._bundlecaps = bundlecaps
         # experimental config: bundle.reorder
         reorder = repo.ui.config('bundle', 'reorder')
         if reorder == 'auto':
             reorder = None
         else:
             reorder = util.parsebool(reorder)
         self._repo = repo
         self._reorder = reorder
         self._progress = repo.ui.progress
         if self._repo.ui.verbose and not self._repo.ui.debugflag:
             self._verbosenote = self._repo.ui.note
         else:
             self._verbosenote = lambda s: None
 
     def close(self):
         return closechunk()
 
     def fileheader(self, fname):
         return chunkheader(len(fname)) + fname
 
     # Extracted both for clarity and for overriding in extensions.
     def _sortgroup(self, revlog, nodelist, lookup):
         """Sort nodes for change group and turn them into revnums."""
         # for generaldelta revlogs, we linearize the revs; this will both be
         # much quicker and generate a much smaller bundle
         if (revlog._generaldelta and self._reorder is None) or self._reorder:
             dag = dagutil.revlogdag(revlog)
             return dag.linearize(set(revlog.rev(n) for n in nodelist))
         else:
             return sorted([revlog.rev(n) for n in nodelist])
 
     def group(self, nodelist, revlog, lookup, units=None):
         """Calculate a delta group, yielding a sequence of changegroup chunks
         (strings).
 
         Given a list of changeset revs, return a set of deltas and
         metadata corresponding to nodes. The first delta is
         first parent(nodelist[0]) -> nodelist[0], the receiver is
         guaranteed to have this parent as it has all history before
         these changesets. In the case firstparent is nullrev the
         changegroup starts with a full revision.
 
         If units is not None, progress detail will be generated, units specifies
         the type of revlog that is touched (changelog, manifest, etc.).
         """
         # if we don't have any revisions touched by these changesets, bail
         if len(nodelist) == 0:
             yield self.close()
             return
 
         revs = self._sortgroup(revlog, nodelist, lookup)
 
         # add the parent of the first rev
         p = revlog.parentrevs(revs[0])[0]
         revs.insert(0, p)
 
         # build deltas
         total = len(revs) - 1
         msgbundling = _('bundling')
         for r in xrange(len(revs) - 1):
             if units is not None:
                 self._progress(msgbundling, r + 1, unit=units, total=total)
             prev, curr = revs[r], revs[r + 1]
             linknode = lookup(revlog.node(curr))
             for c in self.revchunk(revlog, curr, prev, linknode):
                 yield c
 
         if units is not None:
             self._progress(msgbundling, None)
         yield self.close()
 
     # filter any nodes that claim to be part of the known set
     def prune(self, revlog, missing, commonrevs):
         rr, rl = revlog.rev, revlog.linkrev
         return [n for n in missing if rl(rr(n)) not in commonrevs]
 
     def _packmanifests(self, dir, mfnodes, lookuplinknode):
         """Pack flat manifests into a changegroup stream."""
         assert not dir
         for chunk in self.group(mfnodes, self._repo.manifestlog._revlog,
                                 lookuplinknode, units=_('manifests')):
             yield chunk
 
     def _manifestsdone(self):
         return ''
 
     def generate(self, commonrevs, clnodes, fastpathlinkrev, source):
         '''yield a sequence of changegroup chunks (strings)'''
         repo = self._repo
         cl = repo.changelog
 
         clrevorder = {}
         mfs = {} # needed manifests
         fnodes = {} # needed file nodes
         changedfiles = set()
 
         # Callback for the changelog, used to collect changed files and manifest
         # nodes.
         # Returns the linkrev node (identity in the changelog case).
         def lookupcl(x):
             c = cl.read(x)
             clrevorder[x] = len(clrevorder)
             n = c[0]
             # record the first changeset introducing this manifest version
             mfs.setdefault(n, x)
             # Record a complete list of potentially-changed files in
             # this manifest.
             changedfiles.update(c[3])
             return x
 
         self._verbosenote(_('uncompressed size of bundle content:\n'))
         size = 0
         for chunk in self.group(clnodes, cl, lookupcl, units=_('changesets')):
             size += len(chunk)
             yield chunk
         self._verbosenote(_('%8.i (changelog)\n') % size)
 
         # We need to make sure that the linkrev in the changegroup refers to
         # the first changeset that introduced the manifest or file revision.
         # The fastpath is usually safer than the slowpath, because the filelogs
         # are walked in revlog order.
         #
         # When taking the slowpath with reorder=None and the manifest revlog
         # uses generaldelta, the manifest may be walked in the "wrong" order.
         # Without 'clrevorder', we would get an incorrect linkrev (see fix in
         # cc0ff93d0c0c).
         #
         # When taking the fastpath, we are only vulnerable to reordering
         # of the changelog itself. The changelog never uses generaldelta, so
         # it is only reordered when reorder=True. To handle this case, we
         # simply take the slowpath, which already has the 'clrevorder' logic.
         # This was also fixed in cc0ff93d0c0c.
         fastpathlinkrev = fastpathlinkrev and not self._reorder
         # Treemanifests don't work correctly with fastpathlinkrev
         # either, because we don't discover which directory nodes to
         # send along with files. This could probably be fixed.
         fastpathlinkrev = fastpathlinkrev and (
             'treemanifest' not in repo.requirements)
 
         for chunk in self.generatemanifests(commonrevs, clrevorder,
                 fastpathlinkrev, mfs, fnodes):
             yield chunk
         mfs.clear()
         clrevs = set(cl.rev(x) for x in clnodes)
 
         if not fastpathlinkrev:
             def linknodes(unused, fname):
                 return fnodes.get(fname, {})
         else:
             cln = cl.node
             def linknodes(filerevlog, fname):
                 llr = filerevlog.linkrev
                 fln = filerevlog.node
                 revs = ((r, llr(r)) for r in filerevlog)
                 return dict((fln(r), cln(lr)) for r, lr in revs if lr in clrevs)
 
         for chunk in self.generatefiles(changedfiles, linknodes, commonrevs,
                                         source):
             yield chunk
 
         yield self.close()
 
         if clnodes:
             repo.hook('outgoing', node=hex(clnodes[0]), source=source)
 
     def generatemanifests(self, commonrevs, clrevorder, fastpathlinkrev, mfs,
                           fnodes):
         repo = self._repo
         mfl = repo.manifestlog
         dirlog = mfl._revlog.dirlog
         tmfnodes = {'': mfs}
 
         # Callback for the manifest, used to collect linkrevs for filelog
         # revisions.
         # Returns the linkrev node (collected in lookupcl).
         def makelookupmflinknode(dir):
             if fastpathlinkrev:
                 assert not dir
                 return mfs.__getitem__
 
             def lookupmflinknode(x):
                 """Callback for looking up the linknode for manifests.
 
                 Returns the linkrev node for the specified manifest.
 
                 SIDE EFFECT:
 
                 1) fclnodes gets populated with the list of relevant
                    file nodes if we're not using fastpathlinkrev
                 2) When treemanifests are in use, collects treemanifest nodes
                    to send
 
                 Note that this means manifests must be completely sent to
                 the client before you can trust the list of files and
                 treemanifests to send.
                 """
                 clnode = tmfnodes[dir][x]
                 mdata = mfl.get(dir, x).readfast(shallow=True)
                 for p, n, fl in mdata.iterentries():
                     if fl == 't': # subdirectory manifest
                         subdir = dir + p + '/'
                         tmfclnodes = tmfnodes.setdefault(subdir, {})
                         tmfclnode = tmfclnodes.setdefault(n, clnode)
                         if clrevorder[clnode] < clrevorder[tmfclnode]:
                             tmfclnodes[n] = clnode
                     else:
                         f = dir + p
                         fclnodes = fnodes.setdefault(f, {})
                         fclnode = fclnodes.setdefault(n, clnode)
                         if clrevorder[clnode] < clrevorder[fclnode]:
                             fclnodes[n] = clnode
                 return clnode
             return lookupmflinknode
 
         size = 0
         while tmfnodes:
             dir = min(tmfnodes)
             nodes = tmfnodes[dir]
             prunednodes = self.prune(dirlog(dir), nodes, commonrevs)
             if not dir or prunednodes:
                 for x in self._packmanifests(dir, prunednodes,
                                              makelookupmflinknode(dir)):
                     size += len(x)
                     yield x
             del tmfnodes[dir]
         self._verbosenote(_('%8.i (manifests)\n') % size)
         yield self._manifestsdone()
 
     # The 'source' parameter is useful for extensions
     def generatefiles(self, changedfiles, linknodes, commonrevs, source):
         repo = self._repo
         progress = self._progress
         msgbundling = _('bundling')
 
         total = len(changedfiles)
         # for progress output
         msgfiles = _('files')
         for i, fname in enumerate(sorted(changedfiles)):
             filerevlog = repo.file(fname)
             if not filerevlog:
                 raise error.Abort(_("empty or missing revlog for %s") % fname)
 
             linkrevnodes = linknodes(filerevlog, fname)
             # Lookup for filenodes, we collected the linkrev nodes above in the
             # fastpath case and with lookupmf in the slowpath case.
             def lookupfilelog(x):
                 return linkrevnodes[x]
 
             filenodes = self.prune(filerevlog, linkrevnodes, commonrevs)
             if filenodes:
                 progress(msgbundling, i + 1, item=fname, unit=msgfiles,
                          total=total)
                 h = self.fileheader(fname)
                 size = len(h)
                 yield h
                 for chunk in self.group(filenodes, filerevlog, lookupfilelog):
                     size += len(chunk)
                     yield chunk
                 self._verbosenote(_('%8.i  %s\n') % (size, fname))
         progress(msgbundling, None)
 
     def deltaparent(self, revlog, rev, p1, p2, prev):
         return prev
 
     def revchunk(self, revlog, rev, prev, linknode):
         node = revlog.node(rev)
         p1, p2 = revlog.parentrevs(rev)
         base = self.deltaparent(revlog, rev, p1, p2, prev)
 
         prefix = ''
         if revlog.iscensored(base) or revlog.iscensored(rev):
             try:
                 delta = revlog.revision(node, raw=True)
             except error.CensoredNodeError as e:
                 delta = e.tombstone
             if base == nullrev:
                 prefix = mdiff.trivialdiffheader(len(delta))
             else:
                 baselen = revlog.rawsize(base)
                 prefix = mdiff.replacediffheader(baselen, len(delta))
         elif base == nullrev:
             delta = revlog.revision(node, raw=True)
             prefix = mdiff.trivialdiffheader(len(delta))
         else:
             delta = revlog.revdiff(base, rev)
         p1n, p2n = revlog.parents(node)
         basenode = revlog.node(base)
         flags = revlog.flags(rev)
         meta = self.builddeltaheader(node, p1n, p2n, basenode, linknode, flags)
         meta += prefix
         l = len(meta) + len(delta)
         yield chunkheader(l)
         yield meta
         yield delta
     def builddeltaheader(self, node, p1n, p2n, basenode, linknode, flags):
         # do nothing with basenode, it is implicitly the previous one in HG10
         # do nothing with flags, it is implicitly 0 for cg1 and cg2
         return struct.pack(self.deltaheader, node, p1n, p2n, linknode)
 
 class cg2packer(cg1packer):
     version = '02'
     deltaheader = _CHANGEGROUPV2_DELTA_HEADER
 
     def __init__(self, repo, bundlecaps=None):
         super(cg2packer, self).__init__(repo, bundlecaps)
         if self._reorder is None:
             # Since generaldelta is directly supported by cg2, reordering
             # generally doesn't help, so we disable it by default (treating
             # bundle.reorder=auto just like bundle.reorder=False).
             self._reorder = False
 
     def deltaparent(self, revlog, rev, p1, p2, prev):
         dp = revlog.deltaparent(rev)
         if dp == nullrev and revlog.storedeltachains:
             # Avoid sending full revisions when delta parent is null. Pick prev
             # in that case. It's tempting to pick p1 in this case, as p1 will
             # be smaller in the common case. However, computing a delta against
             # p1 may require resolving the raw text of p1, which could be
             # expensive. The revlog caches should have prev cached, meaning
             # less CPU for changegroup generation. There is likely room to add
             # a flag and/or config option to control this behavior.
             return prev
         elif dp == nullrev:
             # revlog is configured to use full snapshot for a reason,
             # stick to full snapshot.
             return nullrev
         elif dp not in (p1, p2, prev):
             # Pick prev when we can't be sure remote has the base revision.
             return prev
         else:
             return dp
 
     def builddeltaheader(self, node, p1n, p2n, basenode, linknode, flags):
         # Do nothing with flags, it is implicitly 0 in cg1 and cg2
         return struct.pack(self.deltaheader, node, p1n, p2n, basenode, linknode)
 
 class cg3packer(cg2packer):
     version = '03'
     deltaheader = _CHANGEGROUPV3_DELTA_HEADER
 
     def _packmanifests(self, dir, mfnodes, lookuplinknode):
         if dir:
             yield self.fileheader(dir)
 
         dirlog = self._repo.manifestlog._revlog.dirlog(dir)
         for chunk in self.group(mfnodes, dirlog, lookuplinknode,
                                 units=_('manifests')):
             yield chunk
 
     def _manifestsdone(self):
         return self.close()
 
     def builddeltaheader(self, node, p1n, p2n, basenode, linknode, flags):
         return struct.pack(
             self.deltaheader, node, p1n, p2n, basenode, linknode, flags)
 
 _packermap = {'01': (cg1packer, cg1unpacker),
              # cg2 adds support for exchanging generaldelta
              '02': (cg2packer, cg2unpacker),
              # cg3 adds support for exchanging revlog flags and treemanifests
              '03': (cg3packer, cg3unpacker),
 }
 
 def allsupportedversions(repo):
     versions = set(_packermap.keys())
     if not (repo.ui.configbool('experimental', 'changegroup3') or
             repo.ui.configbool('experimental', 'treemanifest') or
             'treemanifest' in repo.requirements):
         versions.discard('03')
     return versions
 
 # Changegroup versions that can be applied to the repo
 def supportedincomingversions(repo):
     return allsupportedversions(repo)
 
 # Changegroup versions that can be created from the repo
 def supportedoutgoingversions(repo):
     versions = allsupportedversions(repo)
     if 'treemanifest' in repo.requirements:
         # Versions 01 and 02 support only flat manifests and it's just too
         # expensive to convert between the flat manifest and tree manifest on
         # the fly. Since tree manifests are hashed differently, all of history
         # would have to be converted. Instead, we simply don't even pretend to
         # support versions 01 and 02.
         versions.discard('01')
         versions.discard('02')
     return versions
 
 def safeversion(repo):
     # Finds the smallest version that it's safe to assume clients of the repo
     # will support. For example, all hg versions that support generaldelta also
     # support changegroup 02.
     versions = supportedoutgoingversions(repo)
     if 'generaldelta' in repo.requirements:
         versions.discard('01')
     assert versions
     return min(versions)
 
 def getbundler(version, repo, bundlecaps=None):
     assert version in supportedoutgoingversions(repo)
     return _packermap[version][0](repo, bundlecaps)
 
 def getunbundler(version, fh, alg, extras=None):
     return _packermap[version][1](fh, alg, extras=extras)
 
 def _changegroupinfo(repo, nodes, source):
     if repo.ui.verbose or source == 'bundle':
         repo.ui.status(_("%d changesets found\n") % len(nodes))
     if repo.ui.debugflag:
         repo.ui.debug("list of changesets:\n")
         for node in nodes:
             repo.ui.debug("%s\n" % hex(node))
 
 def makestream(repo, outgoing, version, source, fastpath=False,
                bundlecaps=None):
     bundler = getbundler(version, repo, bundlecaps=bundlecaps)
     return getsubsetraw(repo, outgoing, bundler, source, fastpath=fastpath)
 
 def makechangegroup(repo, outgoing, version, source, fastpath=False,
                     bundlecaps=None):
     cgstream = makestream(repo, outgoing, version, source,
                           fastpath=fastpath, bundlecaps=bundlecaps)
     return getunbundler(version, util.chunkbuffer(cgstream), None,
                         {'clcount': len(outgoing.missing) })
 
 def getsubsetraw(repo, outgoing, bundler, source, fastpath=False):
     repo = repo.unfiltered()
     commonrevs = outgoing.common
     csets = outgoing.missing
     heads = outgoing.missingheads
     # We go through the fast path if we get told to, or if all (unfiltered
     # heads have been requested (since we then know there all linkrevs will
     # be pulled by the client).
     heads.sort()
     fastpathlinkrev = fastpath or (
             repo.filtername is None and heads == sorted(repo.heads()))
 
     repo.hook('preoutgoing', throw=True, source=source)
     _changegroupinfo(repo, csets, source)
     return bundler.generate(commonrevs, csets, fastpathlinkrev, source)
 
-def changegroupsubset(repo, roots, heads, source, version='01'):
-    """Compute a changegroup consisting of all the nodes that are
-    descendants of any of the roots and ancestors of any of the heads.
-    Return a chunkbuffer object whose read() method will return
-    successive changegroup chunks.
-
-    It is fairly complex as determining which filenodes and which
-    manifest nodes need to be included for the changeset to be complete
-    is non-trivial.
-
-    Another wrinkle is doing the reverse, figuring out which changeset in
-    the changegroup a particular filenode or manifestnode belongs to.
-    """
-    outgoing = discovery.outgoing(repo, missingroots=roots, missingheads=heads)
-    return makechangegroup(repo, outgoing, version, source)
-
 def getlocalchangegroupraw(repo, source, outgoing, bundlecaps=None,
                            version='01'):
     """Like getbundle, but taking a discovery.outgoing as an argument.
 
     This is only implemented for local repos and reuses potentially
     precomputed sets in outgoing. Returns a raw changegroup generator."""
     if not outgoing.missing:
         return None
     bundler = getbundler(version, repo, bundlecaps)
     return getsubsetraw(repo, outgoing, bundler, source)
 
 def getchangegroup(repo, source, outgoing, bundlecaps=None,
                    version='01'):
     """Like getbundle, but taking a discovery.outgoing as an argument.
 
     This is only implemented for local repos and reuses potentially
     precomputed sets in outgoing."""
     if not outgoing.missing:
         return None
     return makechangegroup(repo, outgoing, version, source,
                            bundlecaps=bundlecaps)
 
 def getlocalchangegroup(repo, *args, **kwargs):
     repo.ui.deprecwarn('getlocalchangegroup is deprecated, use getchangegroup',
                        '4.3')
     return getchangegroup(repo, *args, **kwargs)
 
 def changegroup(repo, basenodes, source):
     # to avoid a race we use changegroupsubset() (issue1320)
-    return changegroupsubset(repo, basenodes, repo.heads(), source)
+    outgoing = discovery.outgoing(repo, missingroots=basenodes,
+                                  missingheads=repo.heads())
+    return makechangegroup(repo, outgoing, '01', source)
 
 def _addchangegroupfiles(repo, source, revmap, trp, expectedfiles, needfiles):
     revisions = 0
     files = 0
     for chunkdata in iter(source.filelogheader, {}):
         files += 1
         f = chunkdata["filename"]
         repo.ui.debug("adding %s revisions\n" % f)
         repo.ui.progress(_('files'), files, unit=_('files'),
                          total=expectedfiles)
         fl = repo.file(f)
         o = len(fl)
         try:
             if not fl.addgroup(source, revmap, trp):
                 raise error.Abort(_("received file revlog group is empty"))
         except error.CensoredBaseError as e:
             raise error.Abort(_("received delta base is censored: %s") % e)
         revisions += len(fl) - o
         if f in needfiles:
             needs = needfiles[f]
             for new in xrange(o, len(fl)):
                 n = fl.node(new)
                 if n in needs:
                     needs.remove(n)
                 else:
                     raise error.Abort(
                         _("received spurious file revlog entry"))
             if not needs:
                 del needfiles[f]
     repo.ui.progress(_('files'), None)
 
     for f, needs in needfiles.iteritems():
         fl = repo.file(f)
         for n in needs:
             try:
                 fl.rev(n)
             except error.LookupError:
                 raise error.Abort(
                     _('missing file data for %s:%s - run hg verify') %
                     (f, hex(n)))
 
     return revisions, files

mercurial/localrepo.py

@@ -1,2294 +1,2297
 # localrepo.py - read/write repository class for mercurial
 #
 # Copyright 2005-2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import errno
 import hashlib
 import inspect
 import os
 import random
 import time
 import weakref
 
 from .i18n import _
 from .node import (
     hex,
     nullid,
     short,
 )
 from . import (
     bookmarks,
     branchmap,
     bundle2,
     changegroup,
     changelog,
     color,
     context,
     dirstate,
     dirstateguard,
+    discovery,
     encoding,
     error,
     exchange,
     extensions,
     filelog,
     hook,
     lock as lockmod,
     manifest,
     match as matchmod,
     merge as mergemod,
     mergeutil,
     namespaces,
     obsolete,
     pathutil,
     peer,
     phases,
     pushkey,
     pycompat,
     repository,
     repoview,
     revset,
     revsetlang,
     scmutil,
     sparse,
     store,
     subrepo,
     tags as tagsmod,
     transaction,
     txnutil,
     util,
     vfs as vfsmod,
 )
 
 release = lockmod.release
 urlerr = util.urlerr
 urlreq = util.urlreq
 
 # set of (path, vfs-location) tuples. vfs-location is:
 # - 'plain for vfs relative paths
 # - '' for svfs relative paths
 _cachedfiles = set()
 
 class _basefilecache(scmutil.filecache):
     """All filecache usage on repo are done for logic that should be unfiltered
     """
     def __get__(self, repo, type=None):
         if repo is None:
             return self
         return super(_basefilecache, self).__get__(repo.unfiltered(), type)
     def __set__(self, repo, value):
         return super(_basefilecache, self).__set__(repo.unfiltered(), value)
     def __delete__(self, repo):
         return super(_basefilecache, self).__delete__(repo.unfiltered())
 
 class repofilecache(_basefilecache):
     """filecache for files in .hg but outside of .hg/store"""
     def __init__(self, *paths):
         super(repofilecache, self).__init__(*paths)
         for path in paths:
             _cachedfiles.add((path, 'plain'))
 
     def join(self, obj, fname):
         return obj.vfs.join(fname)
 
 class storecache(_basefilecache):
     """filecache for files in the store"""
     def __init__(self, *paths):
         super(storecache, self).__init__(*paths)
         for path in paths:
             _cachedfiles.add((path, ''))
 
     def join(self, obj, fname):
         return obj.sjoin(fname)
 
 def isfilecached(repo, name):
     """check if a repo has already cached "name" filecache-ed property
 
     This returns (cachedobj-or-None, iscached) tuple.
     """
     cacheentry = repo.unfiltered()._filecache.get(name, None)
     if not cacheentry:
         return None, False
     return cacheentry.obj, True
 
 class unfilteredpropertycache(util.propertycache):
     """propertycache that apply to unfiltered repo only"""
 
     def __get__(self, repo, type=None):
         unfi = repo.unfiltered()
         if unfi is repo:
             return super(unfilteredpropertycache, self).__get__(unfi)
         return getattr(unfi, self.name)
 
 class filteredpropertycache(util.propertycache):
     """propertycache that must take filtering in account"""
 
     def cachevalue(self, obj, value):
         object.__setattr__(obj, self.name, value)
 
 
 def hasunfilteredcache(repo, name):
     """check if a repo has an unfilteredpropertycache value for <name>"""
     return name in vars(repo.unfiltered())
 
 def unfilteredmethod(orig):
     """decorate method that always need to be run on unfiltered version"""
     def wrapper(repo, *args, **kwargs):
         return orig(repo.unfiltered(), *args, **kwargs)
     return wrapper
 
 moderncaps = {'lookup', 'branchmap', 'pushkey', 'known', 'getbundle',
               'unbundle'}
 legacycaps = moderncaps.union({'changegroupsubset'})
 
 class localpeer(repository.peer):
     '''peer for a local repo; reflects only the most recent API'''
 
     def __init__(self, repo, caps=None):
         super(localpeer, self).__init__()
 
         if caps is None:
             caps = moderncaps.copy()
         self._repo = repo.filtered('served')
         self._ui = repo.ui
         self._caps = repo._restrictcapabilities(caps)
 
     # Begin of _basepeer interface.
 
     @util.propertycache
     def ui(self):
         return self._ui
 
     def url(self):
         return self._repo.url()
 
     def local(self):
         return self._repo
 
     def peer(self):
         return self
 
     def canpush(self):
         return True
 
     def close(self):
         self._repo.close()
 
     # End of _basepeer interface.
 
     # Begin of _basewirecommands interface.
 
     def branchmap(self):
         return self._repo.branchmap()
 
     def capabilities(self):
         return self._caps
 
     def debugwireargs(self, one, two, three=None, four=None, five=None):
         """Used to test argument passing over the wire"""
         return "%s %s %s %s %s" % (one, two, three, four, five)
 
     def getbundle(self, source, heads=None, common=None, bundlecaps=None,
                   **kwargs):
         chunks = exchange.getbundlechunks(self._repo, source, heads=heads,
                                           common=common, bundlecaps=bundlecaps,
                                           **kwargs)
         cb = util.chunkbuffer(chunks)
 
         if exchange.bundle2requested(bundlecaps):
             # When requesting a bundle2, getbundle returns a stream to make the
             # wire level function happier. We need to build a proper object
             # from it in local peer.
             return bundle2.getunbundler(self.ui, cb)
         else:
             return changegroup.getunbundler('01', cb, None)
 
     def heads(self):
         return self._repo.heads()
 
     def known(self, nodes):
         return self._repo.known(nodes)
 
     def listkeys(self, namespace):
         return self._repo.listkeys(namespace)
 
     def lookup(self, key):
         return self._repo.lookup(key)
 
     def pushkey(self, namespace, key, old, new):
         return self._repo.pushkey(namespace, key, old, new)
 
     def stream_out(self):
         raise error.Abort(_('cannot perform stream clone against local '
                             'peer'))
 
     def unbundle(self, cg, heads, url):
         """apply a bundle on a repo
 
         This function handles the repo locking itself."""
         try:
             try:
                 cg = exchange.readbundle(self.ui, cg, None)
                 ret = exchange.unbundle(self._repo, cg, heads, 'push', url)
                 if util.safehasattr(ret, 'getchunks'):
                     # This is a bundle20 object, turn it into an unbundler.
                     # This little dance should be dropped eventually when the
                     # API is finally improved.
                     stream = util.chunkbuffer(ret.getchunks())
                     ret = bundle2.getunbundler(self.ui, stream)
                 return ret
             except Exception as exc:
                 # If the exception contains output salvaged from a bundle2
                 # reply, we need to make sure it is printed before continuing
                 # to fail. So we build a bundle2 with such output and consume
                 # it directly.
                 #
                 # This is not very elegant but allows a "simple" solution for
                 # issue4594
                 output = getattr(exc, '_bundle2salvagedoutput', ())
                 if output:
                     bundler = bundle2.bundle20(self._repo.ui)
                     for out in output:
                         bundler.addpart(out)
                     stream = util.chunkbuffer(bundler.getchunks())
                     b = bundle2.getunbundler(self.ui, stream)
                     bundle2.processbundle(self._repo, b)
                 raise
         except error.PushRaced as exc:
             raise error.ResponseError(_('push failed:'), str(exc))
 
     # End of _basewirecommands interface.
 
     # Begin of peer interface.
 
     def iterbatch(self):
         return peer.localiterbatcher(self)
 
     # End of peer interface.
 
 class locallegacypeer(repository.legacypeer, localpeer):
     '''peer extension which implements legacy methods too; used for tests with
     restricted capabilities'''
 
     def __init__(self, repo):
         super(locallegacypeer, self).__init__(repo, caps=legacycaps)
 
     # Begin of baselegacywirecommands interface.
 
     def between(self, pairs):
         return self._repo.between(pairs)
 
     def branches(self, nodes):
         return self._repo.branches(nodes)
 
     def changegroup(self, basenodes, source):
         return changegroup.changegroup(self._repo, basenodes, source)
 
     def changegroupsubset(self, bases, heads, source):
-        return changegroup.changegroupsubset(self._repo, bases, heads, source)
+        outgoing = discovery.outgoing(self._repo, missingroots=bases,
+                                      missingheads=heads)
+        return changegroup.makechangegroup(self._repo, outgoing, '01', source)
 
     # End of baselegacywirecommands interface.
 
 # Increment the sub-version when the revlog v2 format changes to lock out old
 # clients.
 REVLOGV2_REQUIREMENT = 'exp-revlogv2.0'
 
 class localrepository(object):
 
     supportedformats = {
         'revlogv1',
         'generaldelta',
         'treemanifest',
         'manifestv2',
         REVLOGV2_REQUIREMENT,
     }
     _basesupported = supportedformats | {
         'store',
         'fncache',
         'shared',
         'relshared',
         'dotencode',
         'exp-sparse',
     }
     openerreqs = {
         'revlogv1',
         'generaldelta',
         'treemanifest',
         'manifestv2',
     }
 
     # a list of (ui, featureset) functions.
     # only functions defined in module of enabled extensions are invoked
     featuresetupfuncs = set()
 
     # list of prefix for file which can be written without 'wlock'
     # Extensions should extend this list when needed
     _wlockfreeprefix = {
         # We migh consider requiring 'wlock' for the next
         # two, but pretty much all the existing code assume
         # wlock is not needed so we keep them excluded for
         # now.
         'hgrc',
         'requires',
         # XXX cache is a complicatged business someone
         # should investigate this in depth at some point
         'cache/',
         # XXX shouldn't be dirstate covered by the wlock?
         'dirstate',
         # XXX bisect was still a bit too messy at the time
         # this changeset was introduced. Someone should fix
         # the remainig bit and drop this line
         'bisect.state',
     }
 
     def __init__(self, baseui, path, create=False):
         self.requirements = set()
         self.filtername = None
         # wvfs: rooted at the repository root, used to access the working copy
         self.wvfs = vfsmod.vfs(path, expandpath=True, realpath=True)
         # vfs: rooted at .hg, used to access repo files outside of .hg/store
         self.vfs = None
         # svfs: usually rooted at .hg/store, used to access repository history
         # If this is a shared repository, this vfs may point to another
         # repository's .hg/store directory.
         self.svfs = None
         self.root = self.wvfs.base
         self.path = self.wvfs.join(".hg")
         self.origroot = path
         # These auditor are not used by the vfs,
         # only used when writing this comment: basectx.match
         self.auditor = pathutil.pathauditor(self.root, self._checknested)
         self.nofsauditor = pathutil.pathauditor(self.root, self._checknested,
                                                 realfs=False, cached=True)
         self.baseui = baseui
         self.ui = baseui.copy()
         self.ui.copy = baseui.copy # prevent copying repo configuration
         self.vfs = vfsmod.vfs(self.path, cacheaudited=True)
         if (self.ui.configbool('devel', 'all-warnings') or
             self.ui.configbool('devel', 'check-locks')):
             self.vfs.audit = self._getvfsward(self.vfs.audit)
         # A list of callback to shape the phase if no data were found.
         # Callback are in the form: func(repo, roots) --> processed root.
         # This list it to be filled by extension during repo setup
         self._phasedefaults = []
         try:
             self.ui.readconfig(self.vfs.join("hgrc"), self.root)
             self._loadextensions()
         except IOError:
             pass
 
         if self.featuresetupfuncs:
             self.supported = set(self._basesupported) # use private copy
             extmods = set(m.__name__ for n, m
                           in extensions.extensions(self.ui))
             for setupfunc in self.featuresetupfuncs:
                 if setupfunc.__module__ in extmods:
                     setupfunc(self.ui, self.supported)
         else:
             self.supported = self._basesupported
         color.setup(self.ui)
 
         # Add compression engines.
         for name in util.compengines:
             engine = util.compengines[name]
             if engine.revlogheader():
                 self.supported.add('exp-compression-%s' % name)
 
         if not self.vfs.isdir():
             if create:
                 self.requirements = newreporequirements(self)
 
                 if not self.wvfs.exists():
                     self.wvfs.makedirs()
                 self.vfs.makedir(notindexed=True)
 
                 if 'store' in self.requirements:
                     self.vfs.mkdir("store")
 
                     # create an invalid changelog
                     self.vfs.append(
                         "00changelog.i",
                         '\0\0\0\2' # represents revlogv2
                         ' dummy changelog to prevent using the old repo layout'
                     )
             else:
                 raise error.RepoError(_("repository %s not found") % path)
         elif create:
             raise error.RepoError(_("repository %s already exists") % path)
         else:
             try:
                 self.requirements = scmutil.readrequires(
                         self.vfs, self.supported)
             except IOError as inst:
                 if inst.errno != errno.ENOENT:
                     raise
 
         cachepath = self.vfs.join('cache')
         self.sharedpath = self.path
         try:
             sharedpath = self.vfs.read("sharedpath").rstrip('\n')
             if 'relshared' in self.requirements:
                 sharedpath = self.vfs.join(sharedpath)
             vfs = vfsmod.vfs(sharedpath, realpath=True)
             cachepath = vfs.join('cache')
             s = vfs.base
             if not vfs.exists():
                 raise error.RepoError(
                     _('.hg/sharedpath points to nonexistent directory %s') % s)
             self.sharedpath = s
         except IOError as inst:
             if inst.errno != errno.ENOENT:
                 raise
 
         if 'exp-sparse' in self.requirements and not sparse.enabled:
             raise error.RepoError(_('repository is using sparse feature but '
                                     'sparse is not enabled; enable the '
                                     '"sparse" extensions to access'))
 
         self.store = store.store(
             self.requirements, self.sharedpath,
             lambda base: vfsmod.vfs(base, cacheaudited=True))
         self.spath = self.store.path
         self.svfs = self.store.vfs
         self.sjoin = self.store.join
         self.vfs.createmode = self.store.createmode
         self.cachevfs = vfsmod.vfs(cachepath, cacheaudited=True)
         self.cachevfs.createmode = self.store.createmode
         if (self.ui.configbool('devel', 'all-warnings') or
             self.ui.configbool('devel', 'check-locks')):
             if util.safehasattr(self.svfs, 'vfs'): # this is filtervfs
                 self.svfs.vfs.audit = self._getsvfsward(self.svfs.vfs.audit)
             else: # standard vfs
                 self.svfs.audit = self._getsvfsward(self.svfs.audit)
         self._applyopenerreqs()
         if create:
             self._writerequirements()
 
         self._dirstatevalidatewarned = False
 
         self._branchcaches = {}
         self._revbranchcache = None
         self.filterpats = {}
         self._datafilters = {}
         self._transref = self._lockref = self._wlockref = None
 
         # A cache for various files under .hg/ that tracks file changes,
         # (used by the filecache decorator)
         #
         # Maps a property name to its util.filecacheentry
         self._filecache = {}
 
         # hold sets of revision to be filtered
         # should be cleared when something might have changed the filter value:
         # - new changesets,
         # - phase change,
         # - new obsolescence marker,
         # - working directory parent change,
         # - bookmark changes
         self.filteredrevcache = {}
 
         # post-dirstate-status hooks
         self._postdsstatus = []
 
         # Cache of types representing filtered repos.
         self._filteredrepotypes = weakref.WeakKeyDictionary()
 
         # generic mapping between names and nodes
         self.names = namespaces.namespaces()
 
         # Key to signature value.
         self._sparsesignaturecache = {}
         # Signature to cached matcher instance.
         self._sparsematchercache = {}
 
     def _getvfsward(self, origfunc):
         """build a ward for self.vfs"""
         rref = weakref.ref(self)
         def checkvfs(path, mode=None):
             ret = origfunc(path, mode=mode)
             repo = rref()
             if (repo is None
                 or not util.safehasattr(repo, '_wlockref')
                 or not util.safehasattr(repo, '_lockref')):
                 return
             if mode in (None, 'r', 'rb'):
                 return
             if path.startswith(repo.path):
                 # truncate name relative to the repository (.hg)
                 path = path[len(repo.path) + 1:]
             if path.startswith('cache/'):
                 msg = 'accessing cache with vfs instead of cachevfs: "%s"'
                 repo.ui.develwarn(msg % path, stacklevel=2, config="cache-vfs")
             if path.startswith('journal.'):
                 # journal is covered by 'lock'
                 if repo._currentlock(repo._lockref) is None:
                     repo.ui.develwarn('write with no lock: "%s"' % path,
                                       stacklevel=2, config='check-locks')
             elif repo._currentlock(repo._wlockref) is None:
                 # rest of vfs files are covered by 'wlock'
                 #
                 # exclude special files
                 for prefix in self._wlockfreeprefix:
                     if path.startswith(prefix):
                         return
                 repo.ui.develwarn('write with no wlock: "%s"' % path,
                                   stacklevel=2, config='check-locks')
             return ret
         return checkvfs
 
     def _getsvfsward(self, origfunc):
         """build a ward for self.svfs"""
         rref = weakref.ref(self)
         def checksvfs(path, mode=None):
             ret = origfunc(path, mode=mode)
             repo = rref()
             if repo is None or not util.safehasattr(repo, '_lockref'):
                 return
             if mode in (None, 'r', 'rb'):
                 return
             if path.startswith(repo.sharedpath):
                 # truncate name relative to the repository (.hg)
                 path = path[len(repo.sharedpath) + 1:]
             if repo._currentlock(repo._lockref) is None:
                 repo.ui.develwarn('write with no lock: "%s"' % path,
                                   stacklevel=3)
             return ret
         return checksvfs
 
     def close(self):
         self._writecaches()
 
     def _loadextensions(self):
         extensions.loadall(self.ui)
 
     def _writecaches(self):
         if self._revbranchcache:
             self._revbranchcache.write()
 
     def _restrictcapabilities(self, caps):
         if self.ui.configbool('experimental', 'bundle2-advertise'):
             caps = set(caps)
             capsblob = bundle2.encodecaps(bundle2.getrepocaps(self))
             caps.add('bundle2=' + urlreq.quote(capsblob))
         return caps
 
     def _applyopenerreqs(self):
         self.svfs.options = dict((r, 1) for r in self.requirements
                                            if r in self.openerreqs)
         # experimental config: format.chunkcachesize
         chunkcachesize = self.ui.configint('format', 'chunkcachesize')
         if chunkcachesize is not None:
             self.svfs.options['chunkcachesize'] = chunkcachesize
         # experimental config: format.maxchainlen
         maxchainlen = self.ui.configint('format', 'maxchainlen')
         if maxchainlen is not None:
             self.svfs.options['maxchainlen'] = maxchainlen
         # experimental config: format.manifestcachesize
         manifestcachesize = self.ui.configint('format', 'manifestcachesize')
         if manifestcachesize is not None:
             self.svfs.options['manifestcachesize'] = manifestcachesize
         # experimental config: format.aggressivemergedeltas
         aggressivemergedeltas = self.ui.configbool('format',
                                                    'aggressivemergedeltas')
         self.svfs.options['aggressivemergedeltas'] = aggressivemergedeltas
         self.svfs.options['lazydeltabase'] = not scmutil.gddeltaconfig(self.ui)
         chainspan = self.ui.configbytes('experimental', 'maxdeltachainspan', -1)
         if 0 <= chainspan:
             self.svfs.options['maxdeltachainspan'] = chainspan
 
         for r in self.requirements:
             if r.startswith('exp-compression-'):
                 self.svfs.options['compengine'] = r[len('exp-compression-'):]
 
         # TODO move "revlogv2" to openerreqs once finalized.
         if REVLOGV2_REQUIREMENT in self.requirements:
             self.svfs.options['revlogv2'] = True
 
     def _writerequirements(self):
         scmutil.writerequires(self.vfs, self.requirements)
 
     def _checknested(self, path):
         """Determine if path is a legal nested repository."""
         if not path.startswith(self.root):
             return False
         subpath = path[len(self.root) + 1:]
         normsubpath = util.pconvert(subpath)
 
         # XXX: Checking against the current working copy is wrong in
         # the sense that it can reject things like
         #
         #   $ hg cat -r 10 sub/x.txt
         #
         # if sub/ is no longer a subrepository in the working copy
         # parent revision.
         #
         # However, it can of course also allow things that would have
         # been rejected before, such as the above cat command if sub/
         # is a subrepository now, but was a normal directory before.
         # The old path auditor would have rejected by mistake since it
         # panics when it sees sub/.hg/.
         #
         # All in all, checking against the working copy seems sensible
         # since we want to prevent access to nested repositories on
         # the filesystem *now*.
         ctx = self[None]
         parts = util.splitpath(subpath)
         while parts:
             prefix = '/'.join(parts)
             if prefix in ctx.substate:
                 if prefix == normsubpath:
                     return True
                 else:
                     sub = ctx.sub(prefix)
                     return sub.checknested(subpath[len(prefix) + 1:])
             else:
                 parts.pop()
         return False
 
     def peer(self):
         return localpeer(self) # not cached to avoid reference cycle
 
     def unfiltered(self):
         """Return unfiltered version of the repository
 
         Intended to be overwritten by filtered repo."""
         return self
 
     def filtered(self, name):
         """Return a filtered version of a repository"""
         # Python <3.4 easily leaks types via __mro__. See
         # https://bugs.python.org/issue17950. We cache dynamically
         # created types so this method doesn't leak on every
         # invocation.
 
         key = self.unfiltered().__class__
         if key not in self._filteredrepotypes:
             # Build a new type with the repoview mixin and the base
             # class of this repo. Give it a name containing the
             # filter name to aid debugging.
             bases = (repoview.repoview, key)
             cls = type(r'%sfilteredrepo' % name, bases, {})
             self._filteredrepotypes[key] = cls
 
         return self._filteredrepotypes[key](self, name)
 
     @repofilecache('bookmarks', 'bookmarks.current')
     def _bookmarks(self):
         return bookmarks.bmstore(self)
 
     @property
     def _activebookmark(self):
         return self._bookmarks.active
 
     # _phaserevs and _phasesets depend on changelog. what we need is to
     # call _phasecache.invalidate() if '00changelog.i' was changed, but it
     # can't be easily expressed in filecache mechanism.
     @storecache('phaseroots', '00changelog.i')
     def _phasecache(self):
         return phases.phasecache(self, self._phasedefaults)
 
     @storecache('obsstore')
     def obsstore(self):
         return obsolete.makestore(self.ui, self)
 
     @storecache('00changelog.i')
     def changelog(self):
         return changelog.changelog(self.svfs,
                                    trypending=txnutil.mayhavepending(self.root))
 
     def _constructmanifest(self):
         # This is a temporary function while we migrate from manifest to
         # manifestlog. It allows bundlerepo and unionrepo to intercept the
         # manifest creation.
         return manifest.manifestrevlog(self.svfs)
 
     @storecache('00manifest.i')
     def manifestlog(self):
         return manifest.manifestlog(self.svfs, self)
 
     @repofilecache('dirstate')
     def dirstate(self):
         sparsematchfn = lambda: sparse.matcher(self)
 
         return dirstate.dirstate(self.vfs, self.ui, self.root,
                                  self._dirstatevalidate, sparsematchfn)
 
     def _dirstatevalidate(self, node):
         try:
             self.changelog.rev(node)
             return node
         except error.LookupError:
             if not self._dirstatevalidatewarned:
                 self._dirstatevalidatewarned = True
                 self.ui.warn(_("warning: ignoring unknown"
                                " working parent %s!\n") % short(node))
             return nullid
 
     def __getitem__(self, changeid):
         if changeid is None:
             return context.workingctx(self)
         if isinstance(changeid, slice):
             # wdirrev isn't contiguous so the slice shouldn't include it
             return [context.changectx(self, i)
                     for i in xrange(*changeid.indices(len(self)))
                     if i not in self.changelog.filteredrevs]
         try:
             return context.changectx(self, changeid)
         except error.WdirUnsupported:
             return context.workingctx(self)
 
     def __contains__(self, changeid):
         """True if the given changeid exists
 
         error.LookupError is raised if an ambiguous node specified.
         """
         try:
             self[changeid]
             return True
         except error.RepoLookupError:
             return False
 
     def __nonzero__(self):
         return True
 
     __bool__ = __nonzero__
 
     def __len__(self):
         return len(self.changelog)
 
     def __iter__(self):
         return iter(self.changelog)
 
     def revs(self, expr, *args):
         '''Find revisions matching a revset.
 
         The revset is specified as a string ``expr`` that may contain
         %-formatting to escape certain types. See ``revsetlang.formatspec``.
 
         Revset aliases from the configuration are not expanded. To expand
         user aliases, consider calling ``scmutil.revrange()`` or
         ``repo.anyrevs([expr], user=True)``.
 
         Returns a revset.abstractsmartset, which is a list-like interface
         that contains integer revisions.
         '''
         expr = revsetlang.formatspec(expr, *args)
         m = revset.match(None, expr)
         return m(self)
 
     def set(self, expr, *args):
         '''Find revisions matching a revset and emit changectx instances.
 
         This is a convenience wrapper around ``revs()`` that iterates the
         result and is a generator of changectx instances.
 
         Revset aliases from the configuration are not expanded. To expand
         user aliases, consider calling ``scmutil.revrange()``.
         '''
         for r in self.revs(expr, *args):
             yield self[r]
 
     def anyrevs(self, specs, user=False, localalias=None):
         '''Find revisions matching one of the given revsets.
 
         Revset aliases from the configuration are not expanded by default. To
         expand user aliases, specify ``user=True``. To provide some local
         definitions overriding user aliases, set ``localalias`` to
         ``{name: definitionstring}``.
         '''
         if user:
             m = revset.matchany(self.ui, specs, repo=self,
                                 localalias=localalias)
         else:
             m = revset.matchany(None, specs, localalias=localalias)
         return m(self)
 
     def url(self):
         return 'file:' + self.root
 
     def hook(self, name, throw=False, **args):
         """Call a hook, passing this repo instance.
 
         This a convenience method to aid invoking hooks. Extensions likely
         won't call this unless they have registered a custom hook or are
         replacing code that is expected to call a hook.
         """
         return hook.hook(self.ui, self, name, throw, **args)
 
     @filteredpropertycache
     def _tagscache(self):
         '''Returns a tagscache object that contains various tags related
         caches.'''
 
         # This simplifies its cache management by having one decorated
         # function (this one) and the rest simply fetch things from it.
         class tagscache(object):
             def __init__(self):
                 # These two define the set of tags for this repository. tags
                 # maps tag name to node; tagtypes maps tag name to 'global' or
                 # 'local'. (Global tags are defined by .hgtags across all
                 # heads, and local tags are defined in .hg/localtags.)
                 # They constitute the in-memory cache of tags.
                 self.tags = self.tagtypes = None
 
                 self.nodetagscache = self.tagslist = None
 
         cache = tagscache()
         cache.tags, cache.tagtypes = self._findtags()
 
         return cache
 
     def tags(self):
         '''return a mapping of tag to node'''
         t = {}
         if self.changelog.filteredrevs:
             tags, tt = self._findtags()
         else:
             tags = self._tagscache.tags
         for k, v in tags.iteritems():
             try:
                 # ignore tags to unknown nodes
                 self.changelog.rev(v)
                 t[k] = v
             except (error.LookupError, ValueError):
                 pass
         return t
 
     def _findtags(self):
         '''Do the hard work of finding tags.  Return a pair of dicts
         (tags, tagtypes) where tags maps tag name to node, and tagtypes
         maps tag name to a string like \'global\' or \'local\'.
         Subclasses or extensions are free to add their own tags, but
         should be aware that the returned dicts will be retained for the
         duration of the localrepo object.'''
 
         # XXX what tagtype should subclasses/extensions use?  Currently
         # mq and bookmarks add tags, but do not set the tagtype at all.
         # Should each extension invent its own tag type?  Should there
         # be one tagtype for all such "virtual" tags?  Or is the status
         # quo fine?
 
 
         # map tag name to (node, hist)
         alltags = tagsmod.findglobaltags(self.ui, self)
         # map tag name to tag type
         tagtypes = dict((tag, 'global') for tag in alltags)
 
         tagsmod.readlocaltags(self.ui, self, alltags, tagtypes)
 
         # Build the return dicts.  Have to re-encode tag names because
         # the tags module always uses UTF-8 (in order not to lose info
         # writing to the cache), but the rest of Mercurial wants them in
         # local encoding.
         tags = {}
         for (name, (node, hist)) in alltags.iteritems():
             if node != nullid:
                 tags[encoding.tolocal(name)] = node
         tags['tip'] = self.changelog.tip()
         tagtypes = dict([(encoding.tolocal(name), value)
                          for (name, value) in tagtypes.iteritems()])
         return (tags, tagtypes)
 
     def tagtype(self, tagname):
         '''
         return the type of the given tag. result can be:
 
         'local'  : a local tag
         'global' : a global tag
         None     : tag does not exist
         '''
 
         return self._tagscache.tagtypes.get(tagname)
 
     def tagslist(self):
         '''return a list of tags ordered by revision'''
         if not self._tagscache.tagslist:
             l = []
             for t, n in self.tags().iteritems():
                 l.append((self.changelog.rev(n), t, n))
             self._tagscache.tagslist = [(t, n) for r, t, n in sorted(l)]
 
         return self._tagscache.tagslist
 
     def nodetags(self, node):
         '''return the tags associated with a node'''
         if not self._tagscache.nodetagscache:
             nodetagscache = {}
             for t, n in self._tagscache.tags.iteritems():
                 nodetagscache.setdefault(n, []).append(t)
             for tags in nodetagscache.itervalues():
                 tags.sort()
             self._tagscache.nodetagscache = nodetagscache
         return self._tagscache.nodetagscache.get(node, [])
 
     def nodebookmarks(self, node):
         """return the list of bookmarks pointing to the specified node"""
         marks = []
         for bookmark, n in self._bookmarks.iteritems():
             if n == node:
                 marks.append(bookmark)
         return sorted(marks)
 
     def branchmap(self):
         '''returns a dictionary {branch: [branchheads]} with branchheads
         ordered by increasing revision number'''
         branchmap.updatecache(self)
         return self._branchcaches[self.filtername]
 
     @unfilteredmethod
     def revbranchcache(self):
         if not self._revbranchcache:
             self._revbranchcache = branchmap.revbranchcache(self.unfiltered())
         return self._revbranchcache
 
     def branchtip(self, branch, ignoremissing=False):
         '''return the tip node for a given branch
 
         If ignoremissing is True, then this method will not raise an error.
         This is helpful for callers that only expect None for a missing branch
         (e.g. namespace).
 
         '''
         try:
             return self.branchmap().branchtip(branch)
         except KeyError:
             if not ignoremissing:
                 raise error.RepoLookupError(_("unknown branch '%s'") % branch)
             else:
                 pass
 
     def lookup(self, key):
         return self[key].node()
 
     def lookupbranch(self, key, remote=None):
         repo = remote or self
         if key in repo.branchmap():
             return key
 
         repo = (remote and remote.local()) and remote or self
         return repo[key].branch()
 
     def known(self, nodes):
         cl = self.changelog
         nm = cl.nodemap
         filtered = cl.filteredrevs
         result = []
         for n in nodes:
             r = nm.get(n)
             resp = not (r is None or r in filtered)
             result.append(resp)
         return result
 
     def local(self):
         return self
 
     def publishing(self):
         # it's safe (and desirable) to trust the publish flag unconditionally
         # so that we don't finalize changes shared between users via ssh or nfs
         return self.ui.configbool('phases', 'publish', untrusted=True)
 
     def cancopy(self):
         # so statichttprepo's override of local() works
         if not self.local():
             return False
         if not self.publishing():
             return True
         # if publishing we can't copy if there is filtered content
         return not self.filtered('visible').changelog.filteredrevs
 
     def shared(self):
         '''the type of shared repository (None if not shared)'''
         if self.sharedpath != self.path:
             return 'store'
         return None
 
     def wjoin(self, f, *insidef):
         return self.vfs.reljoin(self.root, f, *insidef)
 
     def file(self, f):
         if f[0] == '/':
             f = f[1:]
         return filelog.filelog(self.svfs, f)
 
     def changectx(self, changeid):
         return self[changeid]
 
     def setparents(self, p1, p2=nullid):
         with self.dirstate.parentchange():
             copies = self.dirstate.setparents(p1, p2)
             pctx = self[p1]
             if copies:
                 # Adjust copy records, the dirstate cannot do it, it
                 # requires access to parents manifests. Preserve them
                 # only for entries added to first parent.
                 for f in copies:
                     if f not in pctx and copies[f] in pctx:
                         self.dirstate.copy(copies[f], f)
             if p2 == nullid:
                 for f, s in sorted(self.dirstate.copies().items()):
                     if f not in pctx and s not in pctx:
                         self.dirstate.copy(None, f)
 
     def filectx(self, path, changeid=None, fileid=None):
         """changeid can be a changeset revision, node, or tag.
            fileid can be a file revision or node."""
         return context.filectx(self, path, changeid, fileid)
 
     def getcwd(self):
         return self.dirstate.getcwd()
 
     def pathto(self, f, cwd=None):
         return self.dirstate.pathto(f, cwd)
 
     def _loadfilter(self, filter):
         if filter not in self.filterpats:
             l = []
             for pat, cmd in self.ui.configitems(filter):
                 if cmd == '!':
                     continue
                 mf = matchmod.match(self.root, '', [pat])
                 fn = None
                 params = cmd
                 for name, filterfn in self._datafilters.iteritems():
                     if cmd.startswith(name):
                         fn = filterfn
                         params = cmd[len(name):].lstrip()
                         break
                 if not fn:
                     fn = lambda s, c, **kwargs: util.filter(s, c)
                 # Wrap old filters not supporting keyword arguments
                 if not inspect.getargspec(fn)[2]:
                     oldfn = fn
                     fn = lambda s, c, **kwargs: oldfn(s, c)
                 l.append((mf, fn, params))
             self.filterpats[filter] = l
         return self.filterpats[filter]
 
     def _filter(self, filterpats, filename, data):
         for mf, fn, cmd in filterpats:
             if mf(filename):
                 self.ui.debug("filtering %s through %s\n" % (filename, cmd))
                 data = fn(data, cmd, ui=self.ui, repo=self, filename=filename)
                 break
 
         return data
 
     @unfilteredpropertycache
     def _encodefilterpats(self):
         return self._loadfilter('encode')
 
     @unfilteredpropertycache
     def _decodefilterpats(self):
         return self._loadfilter('decode')
 
     def adddatafilter(self, name, filter):
         self._datafilters[name] = filter
 
     def wread(self, filename):
         if self.wvfs.islink(filename):
             data = self.wvfs.readlink(filename)
         else:
             data = self.wvfs.read(filename)
         return self._filter(self._encodefilterpats, filename, data)
 
     def wwrite(self, filename, data, flags, backgroundclose=False):
         """write ``data`` into ``filename`` in the working directory
 
         This returns length of written (maybe decoded) data.
         """
         data = self._filter(self._decodefilterpats, filename, data)
         if 'l' in flags:
             self.wvfs.symlink(data, filename)
         else:
             self.wvfs.write(filename, data, backgroundclose=backgroundclose)
             if 'x' in flags:
                 self.wvfs.setflags(filename, False, True)
         return len(data)
 
     def wwritedata(self, filename, data):
         return self._filter(self._decodefilterpats, filename, data)
 
     def currenttransaction(self):
         """return the current transaction or None if non exists"""
         if self._transref:
             tr = self._transref()
         else:
             tr = None
 
         if tr and tr.running():
             return tr
         return None
 
     def transaction(self, desc, report=None):
         if (self.ui.configbool('devel', 'all-warnings')
                 or self.ui.configbool('devel', 'check-locks')):
             if self._currentlock(self._lockref) is None:
                 raise error.ProgrammingError('transaction requires locking')
         tr = self.currenttransaction()
         if tr is not None:
             scmutil.registersummarycallback(self, tr, desc)
             return tr.nest()
 
         # abort here if the journal already exists
         if self.svfs.exists("journal"):
             raise error.RepoError(
                 _("abandoned transaction found"),
                 hint=_("run 'hg recover' to clean up transaction"))
 
         idbase = "%.40f#%f" % (random.random(), time.time())
         ha = hex(hashlib.sha1(idbase).digest())
         txnid = 'TXN:' + ha
         self.hook('pretxnopen', throw=True, txnname=desc, txnid=txnid)
 
         self._writejournal(desc)
         renames = [(vfs, x, undoname(x)) for vfs, x in self._journalfiles()]
         if report:
             rp = report
         else:
             rp = self.ui.warn
         vfsmap = {'plain': self.vfs} # root of .hg/
         # we must avoid cyclic reference between repo and transaction.
         reporef = weakref.ref(self)
         # Code to track tag movement
         #
         # Since tags are all handled as file content, it is actually quite hard
         # to track these movement from a code perspective. So we fallback to a
         # tracking at the repository level. One could envision to track changes
         # to the '.hgtags' file through changegroup apply but that fails to
         # cope with case where transaction expose new heads without changegroup
         # being involved (eg: phase movement).
         #
         # For now, We gate the feature behind a flag since this likely comes
         # with performance impacts. The current code run more often than needed
         # and do not use caches as much as it could.  The current focus is on
         # the behavior of the feature so we disable it by default. The flag
         # will be removed when we are happy with the performance impact.
         #
         # Once this feature is no longer experimental move the following
         # documentation to the appropriate help section:
         #
         # The ``HG_TAG_MOVED`` variable will be set if the transaction touched
         # tags (new or changed or deleted tags). In addition the details of
         # these changes are made available in a file at:
         #     ``REPOROOT/.hg/changes/tags.changes``.
         # Make sure you check for HG_TAG_MOVED before reading that file as it
         # might exist from a previous transaction even if no tag were touched
         # in this one. Changes are recorded in a line base format::
         #
         #     <action> <hex-node> <tag-name>\n
         #
         # Actions are defined as follow:
         #   "-R": tag is removed,
         #   "+A": tag is added,
         #   "-M": tag is moved (old value),
         #   "+M": tag is moved (new value),
         tracktags = lambda x: None
         # experimental config: experimental.hook-track-tags
         shouldtracktags = self.ui.configbool('experimental', 'hook-track-tags')
         if desc != 'strip' and shouldtracktags:
             oldheads = self.changelog.headrevs()
             def tracktags(tr2):
                 repo = reporef()
                 oldfnodes = tagsmod.fnoderevs(repo.ui, repo, oldheads)
                 newheads = repo.changelog.headrevs()
                 newfnodes = tagsmod.fnoderevs(repo.ui, repo, newheads)
                 # notes: we compare lists here.
                 # As we do it only once buiding set would not be cheaper
                 changes = tagsmod.difftags(repo.ui, repo, oldfnodes, newfnodes)
                 if changes:
                     tr2.hookargs['tag_moved'] = '1'
                     with repo.vfs('changes/tags.changes', 'w',
                                   atomictemp=True) as changesfile:
                         # note: we do not register the file to the transaction
                         # because we needs it to still exist on the transaction
                         # is close (for txnclose hooks)
                         tagsmod.writediff(changesfile, changes)
         def validate(tr2):
             """will run pre-closing hooks"""
             # XXX the transaction API is a bit lacking here so we take a hacky
             # path for now
             #
             # We cannot add this as a "pending" hooks since the 'tr.hookargs'
             # dict is copied before these run. In addition we needs the data
             # available to in memory hooks too.
             #
             # Moreover, we also need to make sure this runs before txnclose
             # hooks and there is no "pending" mechanism that would execute
             # logic only if hooks are about to run.
             #
             # Fixing this limitation of the transaction is also needed to track
             # other families of changes (bookmarks, phases, obsolescence).
             #
             # This will have to be fixed before we remove the experimental
             # gating.
             tracktags(tr2)
             reporef().hook('pretxnclose', throw=True,
                            txnname=desc, **pycompat.strkwargs(tr.hookargs))
         def releasefn(tr, success):
             repo = reporef()
             if success:
                 # this should be explicitly invoked here, because
                 # in-memory changes aren't written out at closing
                 # transaction, if tr.addfilegenerator (via
                 # dirstate.write or so) isn't invoked while
                 # transaction running
                 repo.dirstate.write(None)
             else:
                 # discard all changes (including ones already written
                 # out) in this transaction
                 repo.dirstate.restorebackup(None, 'journal.dirstate')
 
                 repo.invalidate(clearfilecache=True)
 
         tr = transaction.transaction(rp, self.svfs, vfsmap,
                                      "journal",
                                      "undo",
                                      aftertrans(renames),
                                      self.store.createmode,
                                      validator=validate,
                                      releasefn=releasefn,
                                      checkambigfiles=_cachedfiles)
         tr.changes['revs'] = set()
         tr.changes['obsmarkers'] = set()
         tr.changes['phases'] = {}
         tr.changes['bookmarks'] = {}
 
         tr.hookargs['txnid'] = txnid
         # note: writing the fncache only during finalize mean that the file is
         # outdated when running hooks. As fncache is used for streaming clone,
         # this is not expected to break anything that happen during the hooks.
         tr.addfinalize('flush-fncache', self.store.write)
         def txnclosehook(tr2):
             """To be run if transaction is successful, will schedule a hook run
             """
             # Don't reference tr2 in hook() so we don't hold a reference.
             # This reduces memory consumption when there are multiple
             # transactions per lock. This can likely go away if issue5045
             # fixes the function accumulation.
             hookargs = tr2.hookargs
 
             def hook():
                 reporef().hook('txnclose', throw=False, txnname=desc,
                                **pycompat.strkwargs(hookargs))
             reporef()._afterlock(hook)
         tr.addfinalize('txnclose-hook', txnclosehook)
         tr.addpostclose('warms-cache', self._buildcacheupdater(tr))
         def txnaborthook(tr2):
             """To be run if transaction is aborted
             """
             reporef().hook('txnabort', throw=False, txnname=desc,
                            **tr2.hookargs)
         tr.addabort('txnabort-hook', txnaborthook)
         # avoid eager cache invalidation. in-memory data should be identical
         # to stored data if transaction has no error.
         tr.addpostclose('refresh-filecachestats', self._refreshfilecachestats)
         self._transref = weakref.ref(tr)
         scmutil.registersummarycallback(self, tr, desc)
         return tr
 
     def _journalfiles(self):
         return ((self.svfs, 'journal'),
                 (self.vfs, 'journal.dirstate'),
                 (self.vfs, 'journal.branch'),
                 (self.vfs, 'journal.desc'),
                 (self.vfs, 'journal.bookmarks'),
                 (self.svfs, 'journal.phaseroots'))
 
     def undofiles(self):
         return [(vfs, undoname(x)) for vfs, x in self._journalfiles()]
 
     @unfilteredmethod
     def _writejournal(self, desc):
         self.dirstate.savebackup(None, 'journal.dirstate')
         self.vfs.write("journal.branch",
                           encoding.fromlocal(self.dirstate.branch()))
         self.vfs.write("journal.desc",
                           "%d\n%s\n" % (len(self), desc))
         self.vfs.write("journal.bookmarks",
                           self.vfs.tryread("bookmarks"))
         self.svfs.write("journal.phaseroots",
                            self.svfs.tryread("phaseroots"))
 
     def recover(self):
         with self.lock():
             if self.svfs.exists("journal"):
                 self.ui.status(_("rolling back interrupted transaction\n"))
                 vfsmap = {'': self.svfs,
                           'plain': self.vfs,}
                 transaction.rollback(self.svfs, vfsmap, "journal",
                                      self.ui.warn,
                                      checkambigfiles=_cachedfiles)
                 self.invalidate()
                 return True
             else:
                 self.ui.warn(_("no interrupted transaction available\n"))
                 return False
 
     def rollback(self, dryrun=False, force=False):
         wlock = lock = dsguard = None
         try:
             wlock = self.wlock()
             lock = self.lock()
             if self.svfs.exists("undo"):
                 dsguard = dirstateguard.dirstateguard(self, 'rollback')
 
                 return self._rollback(dryrun, force, dsguard)
             else:
                 self.ui.warn(_("no rollback information available\n"))
                 return 1
         finally:
             release(dsguard, lock, wlock)
 
     @unfilteredmethod # Until we get smarter cache management
     def _rollback(self, dryrun, force, dsguard):
         ui = self.ui
         try:
             args = self.vfs.read('undo.desc').splitlines()
             (oldlen, desc, detail) = (int(args[0]), args[1], None)
             if len(args) >= 3:
                 detail = args[2]
             oldtip = oldlen - 1
 
             if detail and ui.verbose:
                 msg = (_('repository tip rolled back to revision %d'
                          ' (undo %s: %s)\n')
                        % (oldtip, desc, detail))
             else:
                 msg = (_('repository tip rolled back to revision %d'
                          ' (undo %s)\n')
                        % (oldtip, desc))
         except IOError:
             msg = _('rolling back unknown transaction\n')
             desc = None
 
         if not force and self['.'] != self['tip'] and desc == 'commit':
             raise error.Abort(
                 _('rollback of last commit while not checked out '
                   'may lose data'), hint=_('use -f to force'))
 
         ui.status(msg)
         if dryrun:
             return 0
 
         parents = self.dirstate.parents()
         self.destroying()
         vfsmap = {'plain': self.vfs, '': self.svfs}
         transaction.rollback(self.svfs, vfsmap, 'undo', ui.warn,
                              checkambigfiles=_cachedfiles)
         if self.vfs.exists('undo.bookmarks'):
             self.vfs.rename('undo.bookmarks', 'bookmarks', checkambig=True)
         if self.svfs.exists('undo.phaseroots'):
             self.svfs.rename('undo.phaseroots', 'phaseroots', checkambig=True)
         self.invalidate()
 
         parentgone = (parents[0] not in self.changelog.nodemap or
                       parents[1] not in self.changelog.nodemap)
         if parentgone:
             # prevent dirstateguard from overwriting already restored one
             dsguard.close()
 
             self.dirstate.restorebackup(None, 'undo.dirstate')
             try:
                 branch = self.vfs.read('undo.branch')
                 self.dirstate.setbranch(encoding.tolocal(branch))
             except IOError:
                 ui.warn(_('named branch could not be reset: '
                           'current branch is still \'%s\'\n')
                         % self.dirstate.branch())
 
             parents = tuple([p.rev() for p in self[None].parents()])
             if len(parents) > 1:
                 ui.status(_('working directory now based on '
                             'revisions %d and %d\n') % parents)
             else:
                 ui.status(_('working directory now based on '
                             'revision %d\n') % parents)
             mergemod.mergestate.clean(self, self['.'].node())
 
         # TODO: if we know which new heads may result from this rollback, pass
         # them to destroy(), which will prevent the branchhead cache from being
         # invalidated.
         self.destroyed()
         return 0
 
     def _buildcacheupdater(self, newtransaction):
         """called during transaction to build the callback updating cache
 
         Lives on the repository to help extension who might want to augment
         this logic. For this purpose, the created transaction is passed to the
         method.
         """
         # we must avoid cyclic reference between repo and transaction.
         reporef = weakref.ref(self)
         def updater(tr):
             repo = reporef()
             repo.updatecaches(tr)
         return updater
 
     @unfilteredmethod
     def updatecaches(self, tr=None):
         """warm appropriate caches
 
         If this function is called after a transaction closed. The transaction
         will be available in the 'tr' argument. This can be used to selectively
         update caches relevant to the changes in that transaction.
         """
         if tr is not None and tr.hookargs.get('source') == 'strip':
             # During strip, many caches are invalid but
             # later call to `destroyed` will refresh them.
             return
 
         if tr is None or tr.changes['revs']:
             # updating the unfiltered branchmap should refresh all the others,
             self.ui.debug('updating the branch cache\n')
             branchmap.updatecache(self.filtered('served'))
 
     def invalidatecaches(self):
 
         if '_tagscache' in vars(self):
             # can't use delattr on proxy
             del self.__dict__['_tagscache']
 
         self.unfiltered()._branchcaches.clear()
         self.invalidatevolatilesets()
         self._sparsesignaturecache.clear()
 
     def invalidatevolatilesets(self):
         self.filteredrevcache.clear()
         obsolete.clearobscaches(self)
 
     def invalidatedirstate(self):
         '''Invalidates the dirstate, causing the next call to dirstate
         to check if it was modified since the last time it was read,
         rereading it if it has.
 
         This is different to dirstate.invalidate() that it doesn't always
         rereads the dirstate. Use dirstate.invalidate() if you want to
         explicitly read the dirstate again (i.e. restoring it to a previous
         known good state).'''
         if hasunfilteredcache(self, 'dirstate'):
             for k in self.dirstate._filecache:
                 try:
                     delattr(self.dirstate, k)
                 except AttributeError:
                     pass
             delattr(self.unfiltered(), 'dirstate')
 
     def invalidate(self, clearfilecache=False):
         '''Invalidates both store and non-store parts other than dirstate
 
         If a transaction is running, invalidation of store is omitted,
         because discarding in-memory changes might cause inconsistency
         (e.g. incomplete fncache causes unintentional failure, but
         redundant one doesn't).
         '''
         unfiltered = self.unfiltered() # all file caches are stored unfiltered
         for k in list(self._filecache.keys()):
             # dirstate is invalidated separately in invalidatedirstate()
             if k == 'dirstate':
                 continue
             if (k == 'changelog' and
                 self.currenttransaction() and
                 self.changelog._delayed):
                 # The changelog object may store unwritten revisions. We don't
                 # want to lose them.
                 # TODO: Solve the problem instead of working around it.
                 continue
 
             if clearfilecache:
                 del self._filecache[k]
             try:
                 delattr(unfiltered, k)
             except AttributeError:
                 pass
         self.invalidatecaches()
         if not self.currenttransaction():
             # TODO: Changing contents of store outside transaction
             # causes inconsistency. We should make in-memory store
             # changes detectable, and abort if changed.
             self.store.invalidatecaches()
 
     def invalidateall(self):
         '''Fully invalidates both store and non-store parts, causing the
         subsequent operation to reread any outside changes.'''
         # extension should hook this to invalidate its caches
         self.invalidate()
         self.invalidatedirstate()
 
     @unfilteredmethod
     def _refreshfilecachestats(self, tr):
         """Reload stats of cached files so that they are flagged as valid"""
         for k, ce in self._filecache.items():
             if k == 'dirstate' or k not in self.__dict__:
                 continue
             ce.refresh()
 
     def _lock(self, vfs, lockname, wait, releasefn, acquirefn, desc,
               inheritchecker=None, parentenvvar=None):
         parentlock = None
         # the contents of parentenvvar are used by the underlying lock to
         # determine whether it can be inherited
         if parentenvvar is not None:
             parentlock = encoding.environ.get(parentenvvar)
         try:
             l = lockmod.lock(vfs, lockname, 0, releasefn=releasefn,
                              acquirefn=acquirefn, desc=desc,
                              inheritchecker=inheritchecker,
                              parentlock=parentlock)
         except error.LockHeld as inst:
             if not wait:
                 raise
             # show more details for new-style locks
             if ':' in inst.locker:
                 host, pid = inst.locker.split(":", 1)
                 self.ui.warn(
                     _("waiting for lock on %s held by process %r "
                       "on host %r\n") % (desc, pid, host))
             else:
                 self.ui.warn(_("waiting for lock on %s held by %r\n") %
                              (desc, inst.locker))
             # default to 600 seconds timeout
             l = lockmod.lock(vfs, lockname,
                              int(self.ui.config("ui", "timeout")),
                              releasefn=releasefn, acquirefn=acquirefn,
                              desc=desc)
             self.ui.warn(_("got lock after %s seconds\n") % l.delay)
         return l
 
     def _afterlock(self, callback):
         """add a callback to be run when the repository is fully unlocked
 
         The callback will be executed when the outermost lock is released
         (with wlock being higher level than 'lock')."""
         for ref in (self._wlockref, self._lockref):
             l = ref and ref()
             if l and l.held:
                 l.postrelease.append(callback)
                 break
         else: # no lock have been found.
             callback()
 
     def lock(self, wait=True):
         '''Lock the repository store (.hg/store) and return a weak reference
         to the lock. Use this before modifying the store (e.g. committing or
         stripping). If you are opening a transaction, get a lock as well.)
 
         If both 'lock' and 'wlock' must be acquired, ensure you always acquires
         'wlock' first to avoid a dead-lock hazard.'''
         l = self._currentlock(self._lockref)
         if l is not None:
             l.lock()
             return l
 
         l = self._lock(self.svfs, "lock", wait, None,
                        self.invalidate, _('repository %s') % self.origroot)
         self._lockref = weakref.ref(l)
         return l
 
     def _wlockchecktransaction(self):
         if self.currenttransaction() is not None:
             raise error.LockInheritanceContractViolation(
                 'wlock cannot be inherited in the middle of a transaction')
 
     def wlock(self, wait=True):
         '''Lock the non-store parts of the repository (everything under
         .hg except .hg/store) and return a weak reference to the lock.
 
         Use this before modifying files in .hg.
 
         If both 'lock' and 'wlock' must be acquired, ensure you always acquires
         'wlock' first to avoid a dead-lock hazard.'''
         l = self._wlockref and self._wlockref()
         if l is not None and l.held:
             l.lock()
             return l
 
         # We do not need to check for non-waiting lock acquisition.  Such
         # acquisition would not cause dead-lock as they would just fail.
         if wait and (self.ui.configbool('devel', 'all-warnings')
                      or self.ui.configbool('devel', 'check-locks')):
             if self._currentlock(self._lockref) is not None:
                 self.ui.develwarn('"wlock" acquired after "lock"')
 
         def unlock():
             if self.dirstate.pendingparentchange():
                 self.dirstate.invalidate()
             else:
                 self.dirstate.write(None)
 
             self._filecache['dirstate'].refresh()
 
         l = self._lock(self.vfs, "wlock", wait, unlock,
                        self.invalidatedirstate, _('working directory of %s') %
                        self.origroot,
                        inheritchecker=self._wlockchecktransaction,
                        parentenvvar='HG_WLOCK_LOCKER')
         self._wlockref = weakref.ref(l)
         return l
 
     def _currentlock(self, lockref):
         """Returns the lock if it's held, or None if it's not."""
         if lockref is None:
             return None
         l = lockref()
         if l is None or not l.held:
             return None
         return l
 
     def currentwlock(self):
         """Returns the wlock if it's held, or None if it's not."""
         return self._currentlock(self._wlockref)
 
     def _filecommit(self, fctx, manifest1, manifest2, linkrev, tr, changelist):
         """
         commit an individual file as part of a larger transaction
         """
 
         fname = fctx.path()
         fparent1 = manifest1.get(fname, nullid)
         fparent2 = manifest2.get(fname, nullid)
         if isinstance(fctx, context.filectx):
             node = fctx.filenode()
             if node in [fparent1, fparent2]:
                 self.ui.debug('reusing %s filelog entry\n' % fname)
                 if manifest1.flags(fname) != fctx.flags():
                     changelist.append(fname)
                 return node
 
         flog = self.file(fname)
         meta = {}
         copy = fctx.renamed()
         if copy and copy[0] != fname:
             # Mark the new revision of this file as a copy of another
             # file.  This copy data will effectively act as a parent
             # of this new revision.  If this is a merge, the first
             # parent will be the nullid (meaning "look up the copy data")
             # and the second one will be the other parent.  For example:
             #
             # 0 --- 1 --- 3   rev1 changes file foo
             #   \       /     rev2 renames foo to bar and changes it
             #    \- 2 -/      rev3 should have bar with all changes and
             #                      should record that bar descends from
             #                      bar in rev2 and foo in rev1
             #
             # this allows this merge to succeed:
             #
             # 0 --- 1 --- 3   rev4 reverts the content change from rev2
             #   \       /     merging rev3 and rev4 should use bar@rev2
             #    \- 2 --- 4        as the merge base
             #
 
             cfname = copy[0]
             crev = manifest1.get(cfname)
             newfparent = fparent2
 
             if manifest2: # branch merge
                 if fparent2 == nullid or crev is None: # copied on remote side
                     if cfname in manifest2:
                         crev = manifest2[cfname]
                         newfparent = fparent1
 
             # Here, we used to search backwards through history to try to find
             # where the file copy came from if the source of a copy was not in
             # the parent directory. However, this doesn't actually make sense to
             # do (what does a copy from something not in your working copy even
             # mean?) and it causes bugs (eg, issue4476). Instead, we will warn
             # the user that copy information was dropped, so if they didn't
             # expect this outcome it can be fixed, but this is the correct
             # behavior in this circumstance.
 
             if crev:
                 self.ui.debug(" %s: copy %s:%s\n" % (fname, cfname, hex(crev)))
                 meta["copy"] = cfname
                 meta["copyrev"] = hex(crev)
                 fparent1, fparent2 = nullid, newfparent
             else:
                 self.ui.warn(_("warning: can't find ancestor for '%s' "
                                "copied from '%s'!\n") % (fname, cfname))
 
         elif fparent1 == nullid:
             fparent1, fparent2 = fparent2, nullid
         elif fparent2 != nullid:
             # is one parent an ancestor of the other?
             fparentancestors = flog.commonancestorsheads(fparent1, fparent2)
             if fparent1 in fparentancestors:
                 fparent1, fparent2 = fparent2, nullid
             elif fparent2 in fparentancestors:
                 fparent2 = nullid
 
         # is the file changed?
         text = fctx.data()
         if fparent2 != nullid or flog.cmp(fparent1, text) or meta:
             changelist.append(fname)
             return flog.add(text, meta, tr, linkrev, fparent1, fparent2)
         # are just the flags changed during merge?
         elif fname in manifest1 and manifest1.flags(fname) != fctx.flags():
             changelist.append(fname)
 
         return fparent1
 
     def checkcommitpatterns(self, wctx, vdirs, match, status, fail):
         """check for commit arguments that aren't committable"""
         if match.isexact() or match.prefix():
             matched = set(status.modified + status.added + status.removed)
 
             for f in match.files():
                 f = self.dirstate.normalize(f)
                 if f == '.' or f in matched or f in wctx.substate:
                     continue
                 if f in status.deleted:
                     fail(f, _('file not found!'))
                 if f in vdirs: # visited directory
                     d = f + '/'
                     for mf in matched:
                         if mf.startswith(d):
                             break
                     else:
                         fail(f, _("no match under directory!"))
                 elif f not in self.dirstate:
                     fail(f, _("file not tracked!"))
 
     @unfilteredmethod
     def commit(self, text="", user=None, date=None, match=None, force=False,
                editor=False, extra=None):
         """Add a new revision to current repository.
 
         Revision information is gathered from the working directory,
         match can be used to filter the committed files. If editor is
         supplied, it is called to get a commit message.
         """
         if extra is None:
             extra = {}
 
         def fail(f, msg):
             raise error.Abort('%s: %s' % (f, msg))
 
         if not match:
             match = matchmod.always(self.root, '')
 
         if not force:
             vdirs = []
             match.explicitdir = vdirs.append
             match.bad = fail
 
         wlock = lock = tr = None
         try:
             wlock = self.wlock()
             lock = self.lock() # for recent changelog (see issue4368)
 
             wctx = self[None]
             merge = len(wctx.parents()) > 1
 
             if not force and merge and not match.always():
                 raise error.Abort(_('cannot partially commit a merge '
                                    '(do not specify files or patterns)'))
 
             status = self.status(match=match, clean=force)
             if force:
                 status.modified.extend(status.clean) # mq may commit clean files
 
             # check subrepos
             subs = []
             commitsubs = set()
             newstate = wctx.substate.copy()
             # only manage subrepos and .hgsubstate if .hgsub is present
             if '.hgsub' in wctx:
                 # we'll decide whether to track this ourselves, thanks
                 for c in status.modified, status.added, status.removed:
                     if '.hgsubstate' in c:
                         c.remove('.hgsubstate')
 
                 # compare current state to last committed state
                 # build new substate based on last committed state
                 oldstate = wctx.p1().substate
                 for s in sorted(newstate.keys()):
                     if not match(s):
                         # ignore working copy, use old state if present
                         if s in oldstate:
                             newstate[s] = oldstate[s]
                             continue
                         if not force:
                             raise error.Abort(
                                 _("commit with new subrepo %s excluded") % s)
                     dirtyreason = wctx.sub(s).dirtyreason(True)
                     if dirtyreason:
                         if not self.ui.configbool('ui', 'commitsubrepos'):
                             raise error.Abort(dirtyreason,
                                 hint=_("use --subrepos for recursive commit"))
                         subs.append(s)
                         commitsubs.add(s)
                     else:
                         bs = wctx.sub(s).basestate()
                         newstate[s] = (newstate[s][0], bs, newstate[s][2])
                         if oldstate.get(s, (None, None, None))[1] != bs:
                             subs.append(s)
 
                 # check for removed subrepos
                 for p in wctx.parents():
                     r = [s for s in p.substate if s not in newstate]
                     subs += [s for s in r if match(s)]
                 if subs:
                     if (not match('.hgsub') and
                         '.hgsub' in (wctx.modified() + wctx.added())):
                         raise error.Abort(
                             _("can't commit subrepos without .hgsub"))
                     status.modified.insert(0, '.hgsubstate')
 
             elif '.hgsub' in status.removed:
                 # clean up .hgsubstate when .hgsub is removed
                 if ('.hgsubstate' in wctx and
                     '.hgsubstate' not in (status.modified + status.added +
                                           status.removed)):
                     status.removed.insert(0, '.hgsubstate')
 
             # make sure all explicit patterns are matched
             if not force:
                 self.checkcommitpatterns(wctx, vdirs, match, status, fail)
 
             cctx = context.workingcommitctx(self, status,
                                             text, user, date, extra)
 
             # internal config: ui.allowemptycommit
             allowemptycommit = (wctx.branch() != wctx.p1().branch()
                                 or extra.get('close') or merge or cctx.files()
                                 or self.ui.configbool('ui', 'allowemptycommit'))
             if not allowemptycommit:
                 return None
 
             if merge and cctx.deleted():
                 raise error.Abort(_("cannot commit merge with missing files"))
 
             ms = mergemod.mergestate.read(self)
             mergeutil.checkunresolved(ms)
 
             if editor:
                 cctx._text = editor(self, cctx, subs)
             edited = (text != cctx._text)
 
             # Save commit message in case this transaction gets rolled back
             # (e.g. by a pretxncommit hook).  Leave the content alone on
             # the assumption that the user will use the same editor again.
             msgfn = self.savecommitmessage(cctx._text)
 
             # commit subs and write new state
             if subs:
                 for s in sorted(commitsubs):
                     sub = wctx.sub(s)
                     self.ui.status(_('committing subrepository %s\n') %
                         subrepo.subrelpath(sub))
                     sr = sub.commit(cctx._text, user, date)
                     newstate[s] = (newstate[s][0], sr)
                 subrepo.writestate(self, newstate)
 
             p1, p2 = self.dirstate.parents()
             hookp1, hookp2 = hex(p1), (p2 != nullid and hex(p2) or '')
             try:
                 self.hook("precommit", throw=True, parent1=hookp1,
                           parent2=hookp2)
                 tr = self.transaction('commit')
                 ret = self.commitctx(cctx, True)
             except: # re-raises
                 if edited:
                     self.ui.write(
                         _('note: commit message saved in %s\n') % msgfn)
                 raise
             # update bookmarks, dirstate and mergestate
             bookmarks.update(self, [p1, p2], ret)
             cctx.markcommitted(ret)
             ms.reset()
             tr.close()
 
         finally:
             lockmod.release(tr, lock, wlock)
 
         def commithook(node=hex(ret), parent1=hookp1, parent2=hookp2):
             # hack for command that use a temporary commit (eg: histedit)
             # temporary commit got stripped before hook release
             if self.changelog.hasnode(ret):
                 self.hook("commit", node=node, parent1=parent1,
                           parent2=parent2)
         self._afterlock(commithook)
         return ret
 
     @unfilteredmethod
     def commitctx(self, ctx, error=False):
         """Add a new revision to current repository.
         Revision information is passed via the context argument.
         """
 
         tr = None
         p1, p2 = ctx.p1(), ctx.p2()
         user = ctx.user()
 
         lock = self.lock()
         try:
             tr = self.transaction("commit")
             trp = weakref.proxy(tr)
 
             if ctx.manifestnode():
                 # reuse an existing manifest revision
                 mn = ctx.manifestnode()
                 files = ctx.files()
             elif ctx.files():
                 m1ctx = p1.manifestctx()
                 m2ctx = p2.manifestctx()
                 mctx = m1ctx.copy()
 
                 m = mctx.read()
                 m1 = m1ctx.read()
                 m2 = m2ctx.read()
 
                 # check in files
                 added = []
                 changed = []
                 removed = list(ctx.removed())
                 linkrev = len(self)
                 self.ui.note(_("committing files:\n"))
                 for f in sorted(ctx.modified() + ctx.added()):
                     self.ui.note(f + "\n")
                     try:
                         fctx = ctx[f]
                         if fctx is None:
                             removed.append(f)
                         else:
                             added.append(f)
                             m[f] = self._filecommit(fctx, m1, m2, linkrev,
                                                     trp, changed)
                             m.setflag(f, fctx.flags())
                     except OSError as inst:
                         self.ui.warn(_("trouble committing %s!\n") % f)
                         raise
                     except IOError as inst:
                         errcode = getattr(inst, 'errno', errno.ENOENT)
                         if error or errcode and errcode != errno.ENOENT:
                             self.ui.warn(_("trouble committing %s!\n") % f)
                         raise
 
                 # update manifest
                 self.ui.note(_("committing manifest\n"))
                 removed = [f for f in sorted(removed) if f in m1 or f in m2]
                 drop = [f for f in removed if f in m]
                 for f in drop:
                     del m[f]
                 mn = mctx.write(trp, linkrev,
                                 p1.manifestnode(), p2.manifestnode(),
                                 added, drop)
                 files = changed + removed
             else:
                 mn = p1.manifestnode()
                 files = []
 
             # update changelog
             self.ui.note(_("committing changelog\n"))
             self.changelog.delayupdate(tr)
             n = self.changelog.add(mn, files, ctx.description(),
                                    trp, p1.node(), p2.node(),
                                    user, ctx.date(), ctx.extra().copy())
             xp1, xp2 = p1.hex(), p2 and p2.hex() or ''
             self.hook('pretxncommit', throw=True, node=hex(n), parent1=xp1,
                       parent2=xp2)
             # set the new commit is proper phase
             targetphase = subrepo.newcommitphase(self.ui, ctx)
             if targetphase:
                 # retract boundary do not alter parent changeset.
                 # if a parent have higher the resulting phase will
                 # be compliant anyway
                 #
                 # if minimal phase was 0 we don't need to retract anything
                 phases.registernew(self, tr, targetphase, [n])
             tr.close()
             return n
         finally:
             if tr:
                 tr.release()
             lock.release()
 
     @unfilteredmethod
     def destroying(self):
         '''Inform the repository that nodes are about to be destroyed.
         Intended for use by strip and rollback, so there's a common
         place for anything that has to be done before destroying history.
 
         This is mostly useful for saving state that is in memory and waiting
         to be flushed when the current lock is released. Because a call to
         destroyed is imminent, the repo will be invalidated causing those
         changes to stay in memory (waiting for the next unlock), or vanish
         completely.
         '''
         # When using the same lock to commit and strip, the phasecache is left
         # dirty after committing. Then when we strip, the repo is invalidated,
         # causing those changes to disappear.
         if '_phasecache' in vars(self):
             self._phasecache.write()
 
     @unfilteredmethod
     def destroyed(self):
         '''Inform the repository that nodes have been destroyed.
         Intended for use by strip and rollback, so there's a common
         place for anything that has to be done after destroying history.
         '''
         # When one tries to:
         # 1) destroy nodes thus calling this method (e.g. strip)
         # 2) use phasecache somewhere (e.g. commit)
         #
         # then 2) will fail because the phasecache contains nodes that were
         # removed. We can either remove phasecache from the filecache,
         # causing it to reload next time it is accessed, or simply filter
         # the removed nodes now and write the updated cache.
         self._phasecache.filterunknown(self)
         self._phasecache.write()
 
         # refresh all repository caches
         self.updatecaches()
 
         # Ensure the persistent tag cache is updated.  Doing it now
         # means that the tag cache only has to worry about destroyed
         # heads immediately after a strip/rollback.  That in turn
         # guarantees that "cachetip == currenttip" (comparing both rev
         # and node) always means no nodes have been added or destroyed.
 
         # XXX this is suboptimal when qrefresh'ing: we strip the current
         # head, refresh the tag cache, then immediately add a new head.
         # But I think doing it this way is necessary for the "instant
         # tag cache retrieval" case to work.
         self.invalidate()
 
     def walk(self, match, node=None):
         '''
         walk recursively through the directory tree or a given
         changeset, finding all files matched by the match
         function
         '''
         self.ui.deprecwarn('use repo[node].walk instead of repo.walk', '4.3')
         return self[node].walk(match)
 
     def status(self, node1='.', node2=None, match=None,
                ignored=False, clean=False, unknown=False,
                listsubrepos=False):
         '''a convenience method that calls node1.status(node2)'''
         return self[node1].status(node2, match, ignored, clean, unknown,
                                   listsubrepos)
 
     def addpostdsstatus(self, ps):
         """Add a callback to run within the wlock, at the point at which status
         fixups happen.
 
         On status completion, callback(wctx, status) will be called with the
         wlock held, unless the dirstate has changed from underneath or the wlock
         couldn't be grabbed.
 
         Callbacks should not capture and use a cached copy of the dirstate --
         it might change in the meanwhile. Instead, they should access the
         dirstate via wctx.repo().dirstate.
 
         This list is emptied out after each status run -- extensions should
         make sure it adds to this list each time dirstate.status is called.
         Extensions should also make sure they don't call this for statuses
         that don't involve the dirstate.
         """
 
         # The list is located here for uniqueness reasons -- it is actually
         # managed by the workingctx, but that isn't unique per-repo.
         self._postdsstatus.append(ps)
 
     def postdsstatus(self):
         """Used by workingctx to get the list of post-dirstate-status hooks."""
         return self._postdsstatus
 
     def clearpostdsstatus(self):
         """Used by workingctx to clear post-dirstate-status hooks."""
         del self._postdsstatus[:]
 
     def heads(self, start=None):
         if start is None:
             cl = self.changelog
             headrevs = reversed(cl.headrevs())
             return [cl.node(rev) for rev in headrevs]
 
         heads = self.changelog.heads(start)
         # sort the output in rev descending order
         return sorted(heads, key=self.changelog.rev, reverse=True)
 
     def branchheads(self, branch=None, start=None, closed=False):
         '''return a (possibly filtered) list of heads for the given branch
 
         Heads are returned in topological order, from newest to oldest.
         If branch is None, use the dirstate branch.
         If start is not None, return only heads reachable from start.
         If closed is True, return heads that are marked as closed as well.
         '''
         if branch is None:
             branch = self[None].branch()
         branches = self.branchmap()
         if branch not in branches:
             return []
         # the cache returns heads ordered lowest to highest
         bheads = list(reversed(branches.branchheads(branch, closed=closed)))
         if start is not None:
             # filter out the heads that cannot be reached from startrev
             fbheads = set(self.changelog.nodesbetween([start], bheads)[2])
             bheads = [h for h in bheads if h in fbheads]
         return bheads
 
     def branches(self, nodes):
         if not nodes:
             nodes = [self.changelog.tip()]
         b = []
         for n in nodes:
             t = n
             while True:
                 p = self.changelog.parents(n)
                 if p[1] != nullid or p[0] == nullid:
                     b.append((t, n, p[0], p[1]))
                     break
                 n = p[0]
         return b
 
     def between(self, pairs):
         r = []
 
         for top, bottom in pairs:
             n, l, i = top, [], 0
             f = 1
 
             while n != bottom and n != nullid:
                 p = self.changelog.parents(n)[0]
                 if i == f:
                     l.append(n)
                     f = f * 2
                 n = p
                 i += 1
 
             r.append(l)
 
         return r
 
     def checkpush(self, pushop):
         """Extensions can override this function if additional checks have
         to be performed before pushing, or call it if they override push
         command.
         """
         pass
 
     @unfilteredpropertycache
     def prepushoutgoinghooks(self):
         """Return util.hooks consists of a pushop with repo, remote, outgoing
         methods, which are called before pushing changesets.
         """
         return util.hooks()
 
     def pushkey(self, namespace, key, old, new):
         try:
             tr = self.currenttransaction()
             hookargs = {}
             if tr is not None:
                 hookargs.update(tr.hookargs)
             hookargs['namespace'] = namespace
             hookargs['key'] = key
             hookargs['old'] = old
             hookargs['new'] = new
             self.hook('prepushkey', throw=True, **hookargs)
         except error.HookAbort as exc:
             self.ui.write_err(_("pushkey-abort: %s\n") % exc)
             if exc.hint:
                 self.ui.write_err(_("(%s)\n") % exc.hint)
             return False
         self.ui.debug('pushing key for "%s:%s"\n' % (namespace, key))
         ret = pushkey.push(self, namespace, key, old, new)
         def runhook():
             self.hook('pushkey', namespace=namespace, key=key, old=old, new=new,
                       ret=ret)
         self._afterlock(runhook)
         return ret
 
     def listkeys(self, namespace):
         self.hook('prelistkeys', throw=True, namespace=namespace)
         self.ui.debug('listing keys for "%s"\n' % namespace)
         values = pushkey.list(self, namespace)
         self.hook('listkeys', namespace=namespace, values=values)
         return values
 
     def debugwireargs(self, one, two, three=None, four=None, five=None):
         '''used to test argument passing over the wire'''
         return "%s %s %s %s %s" % (one, two, three, four, five)
 
     def savecommitmessage(self, text):
         fp = self.vfs('last-message.txt', 'wb')
         try:
             fp.write(text)
         finally:
             fp.close()
         return self.pathto(fp.name[len(self.root) + 1:])
 
 # used to avoid circular references so destructors work
 def aftertrans(files):
     renamefiles = [tuple(t) for t in files]
     def a():
         for vfs, src, dest in renamefiles:
             # if src and dest refer to a same file, vfs.rename is a no-op,
             # leaving both src and dest on disk. delete dest to make sure
             # the rename couldn't be such a no-op.
             vfs.tryunlink(dest)
             try:
                 vfs.rename(src, dest)
             except OSError: # journal file does not yet exist
                 pass
     return a
 
 def undoname(fn):
     base, name = os.path.split(fn)
     assert name.startswith('journal')
     return os.path.join(base, name.replace('journal', 'undo', 1))
 
 def instance(ui, path, create):
     return localrepository(ui, util.urllocalpath(path), create)
 
 def islocal(path):
     return True
 
 def newreporequirements(repo):
     """Determine the set of requirements for a new local repository.
 
     Extensions can wrap this function to specify custom requirements for
     new repositories.
     """
     ui = repo.ui
     requirements = {'revlogv1'}
     if ui.configbool('format', 'usestore'):
         requirements.add('store')
         if ui.configbool('format', 'usefncache'):
             requirements.add('fncache')
             if ui.configbool('format', 'dotencode'):
                 requirements.add('dotencode')
 
     compengine = ui.config('experimental', 'format.compression')
     if compengine not in util.compengines:
         raise error.Abort(_('compression engine %s defined by '
                             'experimental.format.compression not available') %
                           compengine,
                           hint=_('run "hg debuginstall" to list available '
                                  'compression engines'))
 
     # zlib is the historical default and doesn't need an explicit requirement.
     if compengine != 'zlib':
         requirements.add('exp-compression-%s' % compengine)
 
     if scmutil.gdinitconfig(ui):
         requirements.add('generaldelta')
     if ui.configbool('experimental', 'treemanifest'):
         requirements.add('treemanifest')
     if ui.configbool('experimental', 'manifestv2'):
         requirements.add('manifestv2')
 
     revlogv2 = ui.config('experimental', 'revlogv2')
     if revlogv2 == 'enable-unstable-format-and-corrupt-my-data':
         requirements.remove('revlogv1')
         # generaldelta is implied by revlogv2.
         requirements.discard('generaldelta')
         requirements.add(REVLOGV2_REQUIREMENT)
 
     return requirements

mercurial/wireproto.py

@@ -1,1060 +1,1063
 # wireproto.py - generic wire protocol support functions
 #
 # Copyright 2005-2010 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import hashlib
 import os
 import tempfile
 
 from .i18n import _
 from .node import (
     bin,
     hex,
     nullid,
 )
 
 from . import (
     bundle2,
     changegroup as changegroupmod,
+    discovery,
     encoding,
     error,
     exchange,
     peer,
     pushkey as pushkeymod,
     pycompat,
     repository,
     streamclone,
     util,
 )
 
 urlerr = util.urlerr
 urlreq = util.urlreq
 
 bundle2requiredmain = _('incompatible Mercurial client; bundle2 required')
 bundle2requiredhint = _('see https://www.mercurial-scm.org/wiki/'
                         'IncompatibleClient')
 bundle2required = '%s\n(%s)\n' % (bundle2requiredmain, bundle2requiredhint)
 
 class abstractserverproto(object):
     """abstract class that summarizes the protocol API
 
     Used as reference and documentation.
     """
 
     def getargs(self, args):
         """return the value for arguments in <args>
 
         returns a list of values (same order as <args>)"""
         raise NotImplementedError()
 
     def getfile(self, fp):
         """write the whole content of a file into a file like object
 
         The file is in the form::
 
             (<chunk-size>\n<chunk>)+0\n
 
         chunk size is the ascii version of the int.
         """
         raise NotImplementedError()
 
     def redirect(self):
         """may setup interception for stdout and stderr
 
         See also the `restore` method."""
         raise NotImplementedError()
 
     # If the `redirect` function does install interception, the `restore`
     # function MUST be defined. If interception is not used, this function
     # MUST NOT be defined.
     #
     # left commented here on purpose
     #
     #def restore(self):
     #    """reinstall previous stdout and stderr and return intercepted stdout
     #    """
     #    raise NotImplementedError()
 
 class remoteiterbatcher(peer.iterbatcher):
     def __init__(self, remote):
         super(remoteiterbatcher, self).__init__()
         self._remote = remote
 
     def __getattr__(self, name):
         # Validate this method is batchable, since submit() only supports
         # batchable methods.
         fn = getattr(self._remote, name)
         if not getattr(fn, 'batchable', None):
             raise error.ProgrammingError('Attempted to batch a non-batchable '
                                          'call to %r' % name)
 
         return super(remoteiterbatcher, self).__getattr__(name)
 
     def submit(self):
         """Break the batch request into many patch calls and pipeline them.
 
         This is mostly valuable over http where request sizes can be
         limited, but can be used in other places as well.
         """
         # 2-tuple of (command, arguments) that represents what will be
         # sent over the wire.
         requests = []
 
         # 4-tuple of (command, final future, @batchable generator, remote
         # future).
         results = []
 
         for command, args, opts, finalfuture in self.calls:
             mtd = getattr(self._remote, command)
             batchable = mtd.batchable(mtd.im_self, *args, **opts)
 
             commandargs, fremote = next(batchable)
             assert fremote
             requests.append((command, commandargs))
             results.append((command, finalfuture, batchable, fremote))
 
         if requests:
             self._resultiter = self._remote._submitbatch(requests)
 
         self._results = results
 
     def results(self):
         for command, finalfuture, batchable, remotefuture in self._results:
             # Get the raw result, set it in the remote future, feed it
             # back into the @batchable generator so it can be decoded, and
             # set the result on the final future to this value.
             remoteresult = next(self._resultiter)
             remotefuture.set(remoteresult)
             finalfuture.set(next(batchable))
 
             # Verify our @batchable generators only emit 2 values.
             try:
                 next(batchable)
             except StopIteration:
                 pass
             else:
                 raise error.ProgrammingError('%s @batchable generator emitted '
                                              'unexpected value count' % command)
 
             yield finalfuture.value
 
 # Forward a couple of names from peer to make wireproto interactions
 # slightly more sensible.
 batchable = peer.batchable
 future = peer.future
 
 # list of nodes encoding / decoding
 
 def decodelist(l, sep=' '):
     if l:
         return map(bin, l.split(sep))
     return []
 
 def encodelist(l, sep=' '):
     try:
         return sep.join(map(hex, l))
     except TypeError:
         raise
 
 # batched call argument encoding
 
 def escapearg(plain):
     return (plain
             .replace(':', ':c')
             .replace(',', ':o')
             .replace(';', ':s')
             .replace('=', ':e'))
 
 def unescapearg(escaped):
     return (escaped
             .replace(':e', '=')
             .replace(':s', ';')
             .replace(':o', ',')
             .replace(':c', ':'))
 
 def encodebatchcmds(req):
     """Return a ``cmds`` argument value for the ``batch`` command."""
     cmds = []
     for op, argsdict in req:
         # Old servers didn't properly unescape argument names. So prevent
         # the sending of argument names that may not be decoded properly by
         # servers.
         assert all(escapearg(k) == k for k in argsdict)
 
         args = ','.join('%s=%s' % (escapearg(k), escapearg(v))
                         for k, v in argsdict.iteritems())
         cmds.append('%s %s' % (op, args))
 
     return ';'.join(cmds)
 
 # mapping of options accepted by getbundle and their types
 #
 # Meant to be extended by extensions. It is extensions responsibility to ensure
 # such options are properly processed in exchange.getbundle.
 #
 # supported types are:
 #
 # :nodes: list of binary nodes
 # :csv:   list of comma-separated values
 # :scsv:  list of comma-separated values return as set
 # :plain: string with no transformation needed.
 gboptsmap = {'heads':  'nodes',
              'common': 'nodes',
              'obsmarkers': 'boolean',
              'bundlecaps': 'scsv',
              'listkeys': 'csv',
              'cg': 'boolean',
              'cbattempted': 'boolean'}
 
 # client side
 
 class wirepeer(repository.legacypeer):
     """Client-side interface for communicating with a peer repository.
 
     Methods commonly call wire protocol commands of the same name.
 
     See also httppeer.py and sshpeer.py for protocol-specific
     implementations of this interface.
     """
     # Begin of basewirepeer interface.
 
     def iterbatch(self):
         return remoteiterbatcher(self)
 
     @batchable
     def lookup(self, key):
         self.requirecap('lookup', _('look up remote revision'))
         f = future()
         yield {'key': encoding.fromlocal(key)}, f
         d = f.value
         success, data = d[:-1].split(" ", 1)
         if int(success):
             yield bin(data)
         else:
             self._abort(error.RepoError(data))
 
     @batchable
     def heads(self):
         f = future()
         yield {}, f
         d = f.value
         try:
             yield decodelist(d[:-1])
         except ValueError:
             self._abort(error.ResponseError(_("unexpected response:"), d))
 
     @batchable
     def known(self, nodes):
         f = future()
         yield {'nodes': encodelist(nodes)}, f
         d = f.value
         try:
             yield [bool(int(b)) for b in d]
         except ValueError:
             self._abort(error.ResponseError(_("unexpected response:"), d))
 
     @batchable
     def branchmap(self):
         f = future()
         yield {}, f
         d = f.value
         try:
             branchmap = {}
             for branchpart in d.splitlines():
                 branchname, branchheads = branchpart.split(' ', 1)
                 branchname = encoding.tolocal(urlreq.unquote(branchname))
                 branchheads = decodelist(branchheads)
                 branchmap[branchname] = branchheads
             yield branchmap
         except TypeError:
             self._abort(error.ResponseError(_("unexpected response:"), d))
 
     @batchable
     def listkeys(self, namespace):
         if not self.capable('pushkey'):
             yield {}, None
         f = future()
         self.ui.debug('preparing listkeys for "%s"\n' % namespace)
         yield {'namespace': encoding.fromlocal(namespace)}, f
         d = f.value
         self.ui.debug('received listkey for "%s": %i bytes\n'
                       % (namespace, len(d)))
         yield pushkeymod.decodekeys(d)
 
     @batchable
     def pushkey(self, namespace, key, old, new):
         if not self.capable('pushkey'):
             yield False, None
         f = future()
         self.ui.debug('preparing pushkey for "%s:%s"\n' % (namespace, key))
         yield {'namespace': encoding.fromlocal(namespace),
                'key': encoding.fromlocal(key),
                'old': encoding.fromlocal(old),
                'new': encoding.fromlocal(new)}, f
         d = f.value
         d, output = d.split('\n', 1)
         try:
             d = bool(int(d))
         except ValueError:
             raise error.ResponseError(
                 _('push failed (unexpected response):'), d)
         for l in output.splitlines(True):
             self.ui.status(_('remote: '), l)
         yield d
 
     def stream_out(self):
         return self._callstream('stream_out')
 
     def getbundle(self, source, **kwargs):
         self.requirecap('getbundle', _('look up remote changes'))
         opts = {}
         bundlecaps = kwargs.get('bundlecaps')
         if bundlecaps is not None:
             kwargs['bundlecaps'] = sorted(bundlecaps)
         else:
             bundlecaps = () # kwargs could have it to None
         for key, value in kwargs.iteritems():
             if value is None:
                 continue
             keytype = gboptsmap.get(key)
             if keytype is None:
                 assert False, 'unexpected'
             elif keytype == 'nodes':
                 value = encodelist(value)
             elif keytype in ('csv', 'scsv'):
                 value = ','.join(value)
             elif keytype == 'boolean':
                 value = '%i' % bool(value)
             elif keytype != 'plain':
                 raise KeyError('unknown getbundle option type %s'
                                % keytype)
             opts[key] = value
         f = self._callcompressable("getbundle", **opts)
         if any((cap.startswith('HG2') for cap in bundlecaps)):
             return bundle2.getunbundler(self.ui, f)
         else:
             return changegroupmod.cg1unpacker(f, 'UN')
 
     def unbundle(self, cg, heads, url):
         '''Send cg (a readable file-like object representing the
         changegroup to push, typically a chunkbuffer object) to the
         remote server as a bundle.
 
         When pushing a bundle10 stream, return an integer indicating the
         result of the push (see changegroup.apply()).
 
         When pushing a bundle20 stream, return a bundle20 stream.
 
         `url` is the url the client thinks it's pushing to, which is
         visible to hooks.
         '''
 
         if heads != ['force'] and self.capable('unbundlehash'):
             heads = encodelist(['hashed',
                                 hashlib.sha1(''.join(sorted(heads))).digest()])
         else:
             heads = encodelist(heads)
 
         if util.safehasattr(cg, 'deltaheader'):
             # this a bundle10, do the old style call sequence
             ret, output = self._callpush("unbundle", cg, heads=heads)
             if ret == "":
                 raise error.ResponseError(
                     _('push failed:'), output)
             try:
                 ret = int(ret)
             except ValueError:
                 raise error.ResponseError(
                     _('push failed (unexpected response):'), ret)
 
             for l in output.splitlines(True):
                 self.ui.status(_('remote: '), l)
         else:
             # bundle2 push. Send a stream, fetch a stream.
             stream = self._calltwowaystream('unbundle', cg, heads=heads)
             ret = bundle2.getunbundler(self.ui, stream)
         return ret
 
     # End of basewirepeer interface.
 
     # Begin of baselegacywirepeer interface.
 
     def branches(self, nodes):
         n = encodelist(nodes)
         d = self._call("branches", nodes=n)
         try:
             br = [tuple(decodelist(b)) for b in d.splitlines()]
             return br
         except ValueError:
             self._abort(error.ResponseError(_("unexpected response:"), d))
 
     def between(self, pairs):
         batch = 8 # avoid giant requests
         r = []
         for i in xrange(0, len(pairs), batch):
             n = " ".join([encodelist(p, '-') for p in pairs[i:i + batch]])
             d = self._call("between", pairs=n)
             try:
                 r.extend(l and decodelist(l) or [] for l in d.splitlines())
             except ValueError:
                 self._abort(error.ResponseError(_("unexpected response:"), d))
         return r
 
     def changegroup(self, nodes, kind):
         n = encodelist(nodes)
         f = self._callcompressable("changegroup", roots=n)
         return changegroupmod.cg1unpacker(f, 'UN')
 
     def changegroupsubset(self, bases, heads, kind):
         self.requirecap('changegroupsubset', _('look up remote changes'))
         bases = encodelist(bases)
         heads = encodelist(heads)
         f = self._callcompressable("changegroupsubset",
                                    bases=bases, heads=heads)
         return changegroupmod.cg1unpacker(f, 'UN')
 
     # End of baselegacywirepeer interface.
 
     def _submitbatch(self, req):
         """run batch request <req> on the server
 
         Returns an iterator of the raw responses from the server.
         """
         rsp = self._callstream("batch", cmds=encodebatchcmds(req))
         chunk = rsp.read(1024)
         work = [chunk]
         while chunk:
             while ';' not in chunk and chunk:
                 chunk = rsp.read(1024)
                 work.append(chunk)
             merged = ''.join(work)
             while ';' in merged:
                 one, merged = merged.split(';', 1)
                 yield unescapearg(one)
             chunk = rsp.read(1024)
             work = [merged, chunk]
         yield unescapearg(''.join(work))
 
     def _submitone(self, op, args):
         return self._call(op, **args)
 
     def debugwireargs(self, one, two, three=None, four=None, five=None):
         # don't pass optional arguments left at their default value
         opts = {}
         if three is not None:
             opts['three'] = three
         if four is not None:
             opts['four'] = four
         return self._call('debugwireargs', one=one, two=two, **opts)
 
     def _call(self, cmd, **args):
         """execute <cmd> on the server
 
         The command is expected to return a simple string.
 
         returns the server reply as a string."""
         raise NotImplementedError()
 
     def _callstream(self, cmd, **args):
         """execute <cmd> on the server
 
         The command is expected to return a stream. Note that if the
         command doesn't return a stream, _callstream behaves
         differently for ssh and http peers.
 
         returns the server reply as a file like object.
         """
         raise NotImplementedError()
 
     def _callcompressable(self, cmd, **args):
         """execute <cmd> on the server
 
         The command is expected to return a stream.
 
         The stream may have been compressed in some implementations. This
         function takes care of the decompression. This is the only difference
         with _callstream.
 
         returns the server reply as a file like object.
         """
         raise NotImplementedError()
 
     def _callpush(self, cmd, fp, **args):
         """execute a <cmd> on server
 
         The command is expected to be related to a push. Push has a special
         return method.
 
         returns the server reply as a (ret, output) tuple. ret is either
         empty (error) or a stringified int.
         """
         raise NotImplementedError()
 
     def _calltwowaystream(self, cmd, fp, **args):
         """execute <cmd> on server
 
         The command will send a stream to the server and get a stream in reply.
         """
         raise NotImplementedError()
 
     def _abort(self, exception):
         """clearly abort the wire protocol connection and raise the exception
         """
         raise NotImplementedError()
 
 # server side
 
 # wire protocol command can either return a string or one of these classes.
 class streamres(object):
     """wireproto reply: binary stream
 
     The call was successful and the result is a stream.
 
     Accepts either a generator or an object with a ``read(size)`` method.
 
     ``v1compressible`` indicates whether this data can be compressed to
     "version 1" clients (technically: HTTP peers using
     application/mercurial-0.1 media type). This flag should NOT be used on
     new commands because new clients should support a more modern compression
     mechanism.
     """
     def __init__(self, gen=None, reader=None, v1compressible=False):
         self.gen = gen
         self.reader = reader
         self.v1compressible = v1compressible
 
 class pushres(object):
     """wireproto reply: success with simple integer return
 
     The call was successful and returned an integer contained in `self.res`.
     """
     def __init__(self, res):
         self.res = res
 
 class pusherr(object):
     """wireproto reply: failure
 
     The call failed. The `self.res` attribute contains the error message.
     """
     def __init__(self, res):
         self.res = res
 
 class ooberror(object):
     """wireproto reply: failure of a batch of operation
 
     Something failed during a batch call. The error message is stored in
     `self.message`.
     """
     def __init__(self, message):
         self.message = message
 
 def getdispatchrepo(repo, proto, command):
     """Obtain the repo used for processing wire protocol commands.
 
     The intent of this function is to serve as a monkeypatch point for
     extensions that need commands to operate on different repo views under
     specialized circumstances.
     """
     return repo.filtered('served')
 
 def dispatch(repo, proto, command):
     repo = getdispatchrepo(repo, proto, command)
     func, spec = commands[command]
     args = proto.getargs(spec)
     return func(repo, proto, *args)
 
 def options(cmd, keys, others):
     opts = {}
     for k in keys:
         if k in others:
             opts[k] = others[k]
             del others[k]
     if others:
         util.stderr.write("warning: %s ignored unexpected arguments %s\n"
                           % (cmd, ",".join(others)))
     return opts
 
 def bundle1allowed(repo, action):
     """Whether a bundle1 operation is allowed from the server.
 
     Priority is:
 
     1. server.bundle1gd.<action> (if generaldelta active)
     2. server.bundle1.<action>
     3. server.bundle1gd (if generaldelta active)
     4. server.bundle1
     """
     ui = repo.ui
     gd = 'generaldelta' in repo.requirements
 
     if gd:
         v = ui.configbool('server', 'bundle1gd.%s' % action, None)
         if v is not None:
             return v
 
     v = ui.configbool('server', 'bundle1.%s' % action, None)
     if v is not None:
         return v
 
     if gd:
         v = ui.configbool('server', 'bundle1gd')
         if v is not None:
             return v
 
     return ui.configbool('server', 'bundle1')
 
 def supportedcompengines(ui, proto, role):
     """Obtain the list of supported compression engines for a request."""
     assert role in (util.CLIENTROLE, util.SERVERROLE)
 
     compengines = util.compengines.supportedwireengines(role)
 
     # Allow config to override default list and ordering.
     if role == util.SERVERROLE:
         configengines = ui.configlist('server', 'compressionengines')
         config = 'server.compressionengines'
     else:
         # This is currently implemented mainly to facilitate testing. In most
         # cases, the server should be in charge of choosing a compression engine
         # because a server has the most to lose from a sub-optimal choice. (e.g.
         # CPU DoS due to an expensive engine or a network DoS due to poor
         # compression ratio).
         configengines = ui.configlist('experimental',
                                       'clientcompressionengines')
         config = 'experimental.clientcompressionengines'
 
     # No explicit config. Filter out the ones that aren't supposed to be
     # advertised and return default ordering.
     if not configengines:
         attr = 'serverpriority' if role == util.SERVERROLE else 'clientpriority'
         return [e for e in compengines
                 if getattr(e.wireprotosupport(), attr) > 0]
 
     # If compression engines are listed in the config, assume there is a good
     # reason for it (like server operators wanting to achieve specific
     # performance characteristics). So fail fast if the config references
     # unusable compression engines.
     validnames = set(e.name() for e in compengines)
     invalidnames = set(e for e in configengines if e not in validnames)
     if invalidnames:
         raise error.Abort(_('invalid compression engine defined in %s: %s') %
                           (config, ', '.join(sorted(invalidnames))))
 
     compengines = [e for e in compengines if e.name() in configengines]
     compengines = sorted(compengines,
                          key=lambda e: configengines.index(e.name()))
 
     if not compengines:
         raise error.Abort(_('%s config option does not specify any known '
                             'compression engines') % config,
                           hint=_('usable compression engines: %s') %
                           ', '.sorted(validnames))
 
     return compengines
 
 # list of commands
 commands = {}
 
 def wireprotocommand(name, args=''):
     """decorator for wire protocol command"""
     def register(func):
         commands[name] = (func, args)
         return func
     return register
 
 @wireprotocommand('batch', 'cmds *')
 def batch(repo, proto, cmds, others):
     repo = repo.filtered("served")
     res = []
     for pair in cmds.split(';'):
         op, args = pair.split(' ', 1)
         vals = {}
         for a in args.split(','):
             if a:
                 n, v = a.split('=')
                 vals[unescapearg(n)] = unescapearg(v)
         func, spec = commands[op]
         if spec:
             keys = spec.split()
             data = {}
             for k in keys:
                 if k == '*':
                     star = {}
                     for key in vals.keys():
                         if key not in keys:
                             star[key] = vals[key]
                     data['*'] = star
                 else:
                     data[k] = vals[k]
             result = func(repo, proto, *[data[k] for k in keys])
         else:
             result = func(repo, proto)
         if isinstance(result, ooberror):
             return result
         res.append(escapearg(result))
     return ';'.join(res)
 
 @wireprotocommand('between', 'pairs')
 def between(repo, proto, pairs):
     pairs = [decodelist(p, '-') for p in pairs.split(" ")]
     r = []
     for b in repo.between(pairs):
         r.append(encodelist(b) + "\n")
     return "".join(r)
 
 @wireprotocommand('branchmap')
 def branchmap(repo, proto):
     branchmap = repo.branchmap()
     heads = []
     for branch, nodes in branchmap.iteritems():
         branchname = urlreq.quote(encoding.fromlocal(branch))
         branchnodes = encodelist(nodes)
         heads.append('%s %s' % (branchname, branchnodes))
     return '\n'.join(heads)
 
 @wireprotocommand('branches', 'nodes')
 def branches(repo, proto, nodes):
     nodes = decodelist(nodes)
     r = []
     for b in repo.branches(nodes):
         r.append(encodelist(b) + "\n")
     return "".join(r)
 
 @wireprotocommand('clonebundles', '')
 def clonebundles(repo, proto):
     """Server command for returning info for available bundles to seed clones.
 
     Clients will parse this response and determine what bundle to fetch.
 
     Extensions may wrap this command to filter or dynamically emit data
     depending on the request. e.g. you could advertise URLs for the closest
     data center given the client's IP address.
     """
     return repo.vfs.tryread('clonebundles.manifest')
 
 wireprotocaps = ['lookup', 'changegroupsubset', 'branchmap', 'pushkey',
                  'known', 'getbundle', 'unbundlehash', 'batch']
 
 def _capabilities(repo, proto):
     """return a list of capabilities for a repo
 
     This function exists to allow extensions to easily wrap capabilities
     computation
 
     - returns a lists: easy to alter
     - change done here will be propagated to both `capabilities` and `hello`
       command without any other action needed.
     """
     # copy to prevent modification of the global list
     caps = list(wireprotocaps)
     if streamclone.allowservergeneration(repo):
         if repo.ui.configbool('server', 'preferuncompressed'):
             caps.append('stream-preferred')
         requiredformats = repo.requirements & repo.supportedformats
         # if our local revlogs are just revlogv1, add 'stream' cap
         if not requiredformats - {'revlogv1'}:
             caps.append('stream')
         # otherwise, add 'streamreqs' detailing our local revlog format
         else:
             caps.append('streamreqs=%s' % ','.join(sorted(requiredformats)))
     if repo.ui.configbool('experimental', 'bundle2-advertise'):
         capsblob = bundle2.encodecaps(bundle2.getrepocaps(repo))
         caps.append('bundle2=' + urlreq.quote(capsblob))
     caps.append('unbundle=%s' % ','.join(bundle2.bundlepriority))
 
     if proto.name == 'http':
         caps.append('httpheader=%d' %
                     repo.ui.configint('server', 'maxhttpheaderlen'))
         if repo.ui.configbool('experimental', 'httppostargs'):
             caps.append('httppostargs')
 
         # FUTURE advertise 0.2rx once support is implemented
         # FUTURE advertise minrx and mintx after consulting config option
         caps.append('httpmediatype=0.1rx,0.1tx,0.2tx')
 
         compengines = supportedcompengines(repo.ui, proto, util.SERVERROLE)
         if compengines:
             comptypes = ','.join(urlreq.quote(e.wireprotosupport().name)
                                  for e in compengines)
             caps.append('compression=%s' % comptypes)
 
     return caps
 
 # If you are writing an extension and consider wrapping this function. Wrap
 # `_capabilities` instead.
 @wireprotocommand('capabilities')
 def capabilities(repo, proto):
     return ' '.join(_capabilities(repo, proto))
 
 @wireprotocommand('changegroup', 'roots')
 def changegroup(repo, proto, roots):
     nodes = decodelist(roots)
     cg = changegroupmod.changegroup(repo, nodes, 'serve')
     return streamres(reader=cg, v1compressible=True)
 
 @wireprotocommand('changegroupsubset', 'bases heads')
 def changegroupsubset(repo, proto, bases, heads):
     bases = decodelist(bases)
     heads = decodelist(heads)
-    cg = changegroupmod.changegroupsubset(repo, bases, heads, 'serve')
+    outgoing = discovery.outgoing(repo, missingroots=bases,
+                                  missingheads=heads)
+    cg = changegroupmod.makechangegroup(repo, outgoing, '01', 'serve')
     return streamres(reader=cg, v1compressible=True)
 
 @wireprotocommand('debugwireargs', 'one two *')
 def debugwireargs(repo, proto, one, two, others):
     # only accept optional args from the known set
     opts = options('debugwireargs', ['three', 'four'], others)
     return repo.debugwireargs(one, two, **opts)
 
 @wireprotocommand('getbundle', '*')
 def getbundle(repo, proto, others):
     opts = options('getbundle', gboptsmap.keys(), others)
     for k, v in opts.iteritems():
         keytype = gboptsmap[k]
         if keytype == 'nodes':
             opts[k] = decodelist(v)
         elif keytype == 'csv':
             opts[k] = list(v.split(','))
         elif keytype == 'scsv':
             opts[k] = set(v.split(','))
         elif keytype == 'boolean':
             # Client should serialize False as '0', which is a non-empty string
             # so it evaluates as a True bool.
             if v == '0':
                 opts[k] = False
             else:
                 opts[k] = bool(v)
         elif keytype != 'plain':
             raise KeyError('unknown getbundle option type %s'
                            % keytype)
 
     if not bundle1allowed(repo, 'pull'):
         if not exchange.bundle2requested(opts.get('bundlecaps')):
             if proto.name == 'http':
                 return ooberror(bundle2required)
             raise error.Abort(bundle2requiredmain,
                               hint=bundle2requiredhint)
 
     try:
         if repo.ui.configbool('server', 'disablefullbundle'):
             # Check to see if this is a full clone.
             clheads = set(repo.changelog.heads())
             heads = set(opts.get('heads', set()))
             common = set(opts.get('common', set()))
             common.discard(nullid)
             if not common and clheads == heads:
                 raise error.Abort(
                     _('server has pull-based clones disabled'),
                     hint=_('remove --pull if specified or upgrade Mercurial'))
 
         chunks = exchange.getbundlechunks(repo, 'serve', **opts)
     except error.Abort as exc:
         # cleanly forward Abort error to the client
         if not exchange.bundle2requested(opts.get('bundlecaps')):
             if proto.name == 'http':
                 return ooberror(str(exc) + '\n')
             raise # cannot do better for bundle1 + ssh
         # bundle2 request expect a bundle2 reply
         bundler = bundle2.bundle20(repo.ui)
         manargs = [('message', str(exc))]
         advargs = []
         if exc.hint is not None:
             advargs.append(('hint', exc.hint))
         bundler.addpart(bundle2.bundlepart('error:abort',
                                            manargs, advargs))
         return streamres(gen=bundler.getchunks(), v1compressible=True)
     return streamres(gen=chunks, v1compressible=True)
 
 @wireprotocommand('heads')
 def heads(repo, proto):
     h = repo.heads()
     return encodelist(h) + "\n"
 
 @wireprotocommand('hello')
 def hello(repo, proto):
     '''the hello command returns a set of lines describing various
     interesting things about the server, in an RFC822-like format.
     Currently the only one defined is "capabilities", which
     consists of a line in the form:
 
     capabilities: space separated list of tokens
     '''
     return "capabilities: %s\n" % (capabilities(repo, proto))
 
 @wireprotocommand('listkeys', 'namespace')
 def listkeys(repo, proto, namespace):
     d = repo.listkeys(encoding.tolocal(namespace)).items()
     return pushkeymod.encodekeys(d)
 
 @wireprotocommand('lookup', 'key')
 def lookup(repo, proto, key):
     try:
         k = encoding.tolocal(key)
         c = repo[k]
         r = c.hex()
         success = 1
     except Exception as inst:
         r = str(inst)
         success = 0
     return "%s %s\n" % (success, r)
 
 @wireprotocommand('known', 'nodes *')
 def known(repo, proto, nodes, others):
     return ''.join(b and "1" or "0" for b in repo.known(decodelist(nodes)))
 
 @wireprotocommand('pushkey', 'namespace key old new')
 def pushkey(repo, proto, namespace, key, old, new):
     # compatibility with pre-1.8 clients which were accidentally
     # sending raw binary nodes rather than utf-8-encoded hex
     if len(new) == 20 and util.escapestr(new) != new:
         # looks like it could be a binary node
         try:
             new.decode('utf-8')
             new = encoding.tolocal(new) # but cleanly decodes as UTF-8
         except UnicodeDecodeError:
             pass # binary, leave unmodified
     else:
         new = encoding.tolocal(new) # normal path
 
     if util.safehasattr(proto, 'restore'):
 
         proto.redirect()
 
         try:
             r = repo.pushkey(encoding.tolocal(namespace), encoding.tolocal(key),
                              encoding.tolocal(old), new) or False
         except error.Abort:
             r = False
 
         output = proto.restore()
 
         return '%s\n%s' % (int(r), output)
 
     r = repo.pushkey(encoding.tolocal(namespace), encoding.tolocal(key),
                      encoding.tolocal(old), new)
     return '%s\n' % int(r)
 
 @wireprotocommand('stream_out')
 def stream(repo, proto):
     '''If the server supports streaming clone, it advertises the "stream"
     capability with a value representing the version and flags of the repo
     it is serving. Client checks to see if it understands the format.
     '''
     if not streamclone.allowservergeneration(repo):
         return '1\n'
 
     def getstream(it):
         yield '0\n'
         for chunk in it:
             yield chunk
 
     try:
         # LockError may be raised before the first result is yielded. Don't
         # emit output until we're sure we got the lock successfully.
         it = streamclone.generatev1wireproto(repo)
         return streamres(gen=getstream(it))
     except error.LockError:
         return '2\n'
 
 @wireprotocommand('unbundle', 'heads')
 def unbundle(repo, proto, heads):
     their_heads = decodelist(heads)
 
     try:
         proto.redirect()
 
         exchange.check_heads(repo, their_heads, 'preparing changes')
 
         # write bundle data to temporary file because it can be big
         fd, tempname = tempfile.mkstemp(prefix='hg-unbundle-')
         fp = os.fdopen(fd, pycompat.sysstr('wb+'))
         r = 0
         try:
             proto.getfile(fp)
             fp.seek(0)
             gen = exchange.readbundle(repo.ui, fp, None)
             if (isinstance(gen, changegroupmod.cg1unpacker)
                 and not bundle1allowed(repo, 'push')):
                 if proto.name == 'http':
                     # need to special case http because stderr do not get to
                     # the http client on failed push so we need to abuse some
                     # other error type to make sure the message get to the
                     # user.
                     return ooberror(bundle2required)
                 raise error.Abort(bundle2requiredmain,
                                   hint=bundle2requiredhint)
 
             r = exchange.unbundle(repo, gen, their_heads, 'serve',
                                   proto._client())
             if util.safehasattr(r, 'addpart'):
                 # The return looks streamable, we are in the bundle2 case and
                 # should return a stream.
                 return streamres(gen=r.getchunks())
             return pushres(r)
 
         finally:
             fp.close()
             os.unlink(tempname)
 
     except (error.BundleValueError, error.Abort, error.PushRaced) as exc:
         # handle non-bundle2 case first
         if not getattr(exc, 'duringunbundle2', False):
             try:
                 raise
             except error.Abort:
                 # The old code we moved used util.stderr directly.
                 # We did not change it to minimise code change.
                 # This need to be moved to something proper.
                 # Feel free to do it.
                 util.stderr.write("abort: %s\n" % exc)
                 if exc.hint is not None:
                     util.stderr.write("(%s)\n" % exc.hint)
                 return pushres(0)
             except error.PushRaced:
                 return pusherr(str(exc))
 
         bundler = bundle2.bundle20(repo.ui)
         for out in getattr(exc, '_bundle2salvagedoutput', ()):
             bundler.addpart(out)
         try:
             try:
                 raise
             except error.PushkeyFailed as exc:
                 # check client caps
                 remotecaps = getattr(exc, '_replycaps', None)
                 if (remotecaps is not None
                         and 'pushkey' not in remotecaps.get('error', ())):
                     # no support remote side, fallback to Abort handler.
                     raise
                 part = bundler.newpart('error:pushkey')
                 part.addparam('in-reply-to', exc.partid)
                 if exc.namespace is not None:
                     part.addparam('namespace', exc.namespace, mandatory=False)
                 if exc.key is not None:
                     part.addparam('key', exc.key, mandatory=False)
                 if exc.new is not None:
                     part.addparam('new', exc.new, mandatory=False)
                 if exc.old is not None:
                     part.addparam('old', exc.old, mandatory=False)
                 if exc.ret is not None:
                     part.addparam('ret', exc.ret, mandatory=False)
         except error.BundleValueError as exc:
             errpart = bundler.newpart('error:unsupportedcontent')
             if exc.parttype is not None:
                 errpart.addparam('parttype', exc.parttype)
             if exc.params:
                 errpart.addparam('params', '\0'.join(exc.params))
         except error.Abort as exc:
             manargs = [('message', str(exc))]
             advargs = []
             if exc.hint is not None:
                 advargs.append(('hint', exc.hint))
             bundler.addpart(bundle2.bundlepart('error:abort',
                                                manargs, advargs))
         except error.PushRaced as exc:
             bundler.newpart('error:pushraced', [('message', str(exc))])
         return streamres(gen=bundler.getchunks())